Organoids: Current Applications and Future Directions.

Organoid cultivation relies on the selection of appropriate cell sources (embryonic stem cells (ESCs), iPSCs, adult stem cells, etc.), the optimization of culture media supplemented with growth factors and cytokines combined with 3D culture techniques (including natural/synthetic matrix‐based systems and dynamic microfluidic platforms), and the use of natural or synthetic scaffold materials integrated with ECM components to precisely simulate the in vivo microenvironment, all of which are crucial for constructing functional organoid models. During organoid construction, the source and quality of the cells are the fundamental prerequisites for ensuring the accuracy of the model [ 53 ]. The common cell sources include the following: ESCs, iPSCs, and adult stem cells [ 54 ]. ESCs possess totipotency, meaning that they can be induced to differentiate into various types of tissues and organs in vitro [ 53 ]. Their advantage lies in their high differentiation potential, but they are also associated with certain ethical and clinical transformation‐related controversies and limitations [ 55 ]. iPSCs are obtained by reprogramming somatic cells, avoiding the ethical controversy associated with ESCs, and they possess high differentiation capabilities [ 32 ]. Organoids constructed using iPSCs can not only reflect individual specificity but also enable the construction of personalized disease models [ 56 ]. Adult stem cells and other primary cells have tissue‐specific origins, such as primary cells from organs such as the liver and the intestine, which can have more similar functions and physiological states as the original organs [ 54 ]. Although these cells possess relatively constrained capacity for proliferation and differentiation, they exhibit distinct strengths in disease simulation, drug response testing, and research targeting tissue regenerative mechanisms [ 57 ]. The selection and preparation of culture media play crucial roles in cell differentiation and organ formation during the process of constructing organoids [ 58 ]. Traditional two‐dimensional culture media can no longer meet the demands of complex three‐dimensional structures; thus, three‐dimensional culture techniques have been widely applied [ 59 ]. Culture media usually require the addition of growth factors, cytokines, and auxiliary molecules [ 15 ]. These components not only provide the nutrients needed for cell growth but also regulate the self‐organization process of cells through signal transduction pathways [ 33 ]. In recent years, culture medium optimization methods based on high‐throughput screening have emerged continuously, providing a scientific basis for the construction of functional organoids [ 60 ]. Traditional two‐dimensional culture is limited by cells being in planar contact, whereas three‐dimensional culture can be used to establish a three‐dimensional cellular environment through natural or synthetic matrices [ 61 ]. Common organoid three‐dimensional culture systems include culture methods based on natural matrices such as collagen and Matrigel, as well as synthetic scaffolds such as polylactic‐co‐glycolic acid [ 62 ]. Three‐dimensional cell culture not only facilitates direct interaction between cells but also promotes rich ECM structure and signaling in the in vivo microenvironment [ 63 ]. The dynamic culture platform constructed by using microfluidic technology can simulate in vivo fluid circulation and hemodynamic conditions, further enhancing the physiological relevance of organoid models [ 64 ]. The dynamic culture system can provide real‐time nutrition, allow waste removal, and provide growth factor gradients, promoting the formation and functional development of complex networks within organoids [ 65 ]. The success of organoid cultivation relies on the precise simulation of the physiological microenvironment within the body [ 34 ]. The scaffold material serves as the basis for three‐dimensional cultivation, and its physical, chemical, and biological compatibility directly affect the adhesion, proliferation, and differentiation of cells [ 35 ]. The scaffold materials can be classified into two major categories: natural materials and synthetic materials [ 66 ]. Natural materials such as collagen, fibrin, and gelatin are widely used because of their biocompatibility and bioactivity [ 32 ]; synthetic materials such as polycaprolactone and polylactic acid achieve the desired mechanical properties and stability via changes in structural parameters [ 67 ]. Advanced 3D bioprinting and microprocessing technologies can generate scaffolds with complex pore structures and gradient changes according to specific requirements [ 68 ]. Such scaffolds enable a more natural spatial distribution of cells during the cultivation process and allow precise control of local growth factor concentrations via adjustment of parameters such as pore density and surface morphology [ 68 ]. ECM is important for maintaining organ functions, and its components are complex and diverse. In modern organoid cultivation, some ECM components, such as glycosaminoglycans and elastin, are often incorporated into scaffold materials to simulate the interaction between cells and the matrix in the body, thereby promoting the complete expression of cell functions [ 69 ]. Organoid cultivation relies on the selection of appropriate cell sources (ESCs, iPSCs, adult stem cells, etc.), the optimization of culture media supplemented with growth factors and cytokines combined with 3D culture techniques (including natural/synthetic matrix‐based systems and dynamic microfluidic platforms), and the use of natural or synthetic scaffold materials integrated with ECM components to precisely simulate the in vivo microenvironment, all of which are important for constructing functional organoid models.

Conceptualization; data curation; formal analysis; writing – original draft: Yueqi Leng and Yue Wang. Data curation; formal analysis: Canhui Cao and Feng Deng. Data curation; formal analysis: Weisi Lian, Xingtong Chen, and Mingmei Lin. Investigation: Zhonghong Zeng, Dan Mo, and Shangqi Li. Conceptualization: Yilei He and Yang Yu. Conceptualization; funding acquisition; writing – review and editing: Heng Pan, Ping Zhou, and Rong Li. All authors have read and approved the final version of the manuscript.

The authors have nothing to report.

The next stage of organoid research will be defined less by the generation of additional organoid types and more by whether these systems can become physiologically interpretable, technically reproducible, and clinically useful. Over the past decade, the field has moved from proof‐of‐concept self‐organization toward disease modeling, drug screening, regenerative testing, and patient‐specific applications. However, many organoids remain simplified representations of selected tissue compartments rather than integrated organ systems. Future progress should therefore focus on four interconnected goals: improving physiological fidelity, building system‐level models, integrating perturbation and analytical technologies, and establishing standards that allow organoid‐based findings to inform clinical decision‐making [ 1 , 2 , 3 , 4 , 5 , 6 , 25 , 30 ]. A major priority is to improve the structural and functional completeness of organoid models. Current organoids often reproduce epithelial or parenchymal organization but incompletely capture vasculature, immune cells, neural regulation, stromal remodeling, and mechanical forces. This is especially important because these components are not passive background elements; they actively regulate tissue maturation, disease progression, inflammatory responses, and therapeutic sensitivity. Vascularization is among the most urgent challenges. Endothelial coculture, angiogenic factor induction, microfluidic perfusion, transplantation‐assisted vascularization, and 3D bioprinting have been explored as strategies to create more stable vascular networks [ 5 , 25 , 30 , 36 , 37 ]. Studies in kidney organoids have shown that flow‐enhanced culture can promote vascular maturation, while decellularized extracellular matrices can provide organ‐specific cues for vascular and epithelial development [ 37 , 90 ]. Similarly, vascularized retinal organoids, vascularized endometrium‐on‐a‐chip systems, and engineered vascular organoid platforms illustrate that vascularization is becoming an organ‐specific design problem rather than a generic culture improvement [ 51 , 85 , 133 ]. With respect to endometrial organoids, improved physiological relevance requires more than the addition of endothelial cells. The endometrium is shaped by cyclic hormone exposure, stromal decidualization, immune cell remodeling, vascular adaptation, and embryo‒epithelium interactions. Recent assembloid, endometrium‐on‐chip, and embryo–endometrial interface models have begun to address these features by integrating epithelial organoids with stromal cells, matrix support, perfusion‐like systems, or embryo/blastoid interaction platforms [ 68 , 131 , 132 , 133 ]. These models are particularly valuable because implantation and endometrial repair cannot be understood from the perspective of epithelial behavior alone. In my view, the field should avoid treating “complexity” as an endpoint in itself. A useful organoid model does not need to contain every cell type found in vivo; it needs to contain the components required to answer a specific biological question. For implantation, that may mean epithelial polarity, decidual stromal cells, trophoblast interactions, vascular cues, and uterine immune cells. For fibrosis, stromal fibroblasts, ECM remodeling, inflammatory signals, and mechanical stiffness may be more important. Future models should therefore be designed around mechanism‐driven fidelity rather than maximal cellular complexity. Matrix engineering will also be central to this transition. Matrigel has been indispensable for establishing many organoid systems, but its undefined composition, variable stiffness, and limited clinical compatibility restrict its mechanistic interpretation and translational use [ 32 , 38 ]. Designer matrices, tunable hydrogels, organ‐specific decellularized ECMs, and fully synthetic culture systems provide more controllable alternatives [ 4 , 32 , 90 , 158 ]. These materials should not simply replace Matrigel as a technical upgrade; rather, they should be used to define how matrix stiffness, ligand density, degradation kinetics, and tissue‐specific ECM composition influence organoid fate. This will be particularly important for studying liver fibrosis, kidney maturation, tumor invasion, and endometrial regeneration, where the ECM is itself part of the disease mechanism. Another important direction is the transition from single‐organoid models to multiorgan systems. Human diseases rarely occur in isolation within one tissue. Endocrine disorders, metabolic diseases, cancer metastasis, immune‐mediated diseases, and drug toxicity all involve communication among multiple organs. Multiorganoid chips and linked microphysiological systems offer a way to study these interactions under more controlled conditions than animal models and with more human relevance than standard cell culture [ 25 , 27 , 30 , 99 ]. For example, heart–kidney organoid systems can model cardiorenal interactions, whereas gut–liver or liver–pancreas platforms may be useful for studying nutrient metabolism, drug clearance, and systemic toxicity [ 99 , 112 ]. In reproductive medicine, multiorgan integration has particular potential. The endometrium does not function independently; it is regulated by ovarian steroid hormones, hypothalamic–pituitary signaling, immune status, metabolic state, and embryo‐derived signals. Future models that link ovarian, endometrial, trophoblast, and possibly hypothalamic–pituitary components could provide a more complete platform for studying PCOS, implantation failure, endometriosis‐associated infertility, menopause, and hormone‐dependent endometrial disorders [ 31 , 45 , 46 , 126 , 232 ]. Such systems would be especially valuable for distinguishing whether a disease phenotype originates from intrinsic endometrial dysfunction, abnormal ovarian endocrine input, or disrupted embryo–endometrial communication. This distinction is difficult to achieve in clinical samples and almost impossible in simple epithelial organoids. Cancer research will also benefit from multiorgan platforms. Patient‐derived tumor organoids already preserve important features of tumor heterogeneity and drug response [ 21 , 50 , 130 ]. However, tumor progression, metastasis, and therapeutic resistance are strongly influenced by the organ microenvironment. Pairing primary tumor organoids with liver, lung, vascular, immune, or stromal compartments may allow researchers to study metastatic tropism, immune escape, drug metabolism, and organ‐specific toxicity in a single experimental framework. The key challenge is to ensure that these linked systems remain interpretable. Multiorgan models should not become technically impressive but should be biologically opaque. Their value depends on carefully defined inputs, measurable outputs, and validation against patient data. Future organoid studies will likely combine model engineering with functional perturbation and high‐resolution analysis. Multiomics approaches, especially single‐cell transcriptomics, spatial transcriptomics, epigenomics, proteomics, and metabolomics, can map how different cell states emerge, interact, and respond to disease or treatment [ 26 , 77 , 154 ]. In endometrial research, spatial and single‐cell studies have already shown that epithelial, stromal, immune, endothelial, and metabolic programs change across the menstrual cycle and in disorders such as endometriosis, adenomyosis, thin endometrium, and PCOS [ 134 , 147 , 148 ]. Combining these datasets with organoids will make it possible to test whether disease‐associated signatures are merely correlative or functionally causal. CRISPR‐based perturbation will be especially useful for this purpose. Genome editing in organoids allows researchers to introduce or correct mutations, perform loss‐of‐function screens, validate disease‐associated genes, and identify therapeutic vulnerabilities [ 20 , 39 , 40 , 41 ]. In cancer organoids, this approach can be used to test oncogenic drivers, resistance mechanisms, and synthetic lethal interactions. In endometrial disease, CRISPR‐based tools could help clarify the functional roles of HOX gene methylation, progesterone resistance, PI3K/AKT pathway activation, ferroptosis, epithelial–stromal communication, and fibrosis‐related signaling [ 155 , 156 , 157 , 158 , 168 , 169 , 170 , 171 , 172 , 173 , 174 , 175 , 176 , 177 , 178 , 179 , 180 , 181 , 182 , 183 , 184 , 185 , 186 , 187 ]. The future value of CRISPR‐based organoid systems will depend on moving beyond single‐gene validation toward pathway‐level interpretation, ideally integrated with spatial and functional readouts. AI will become increasingly useful, but its role should be practical rather than ornamental. AI can assist with automated segmentation, organoid morphology scoring, high‐content imaging, culture optimization, and prediction of drug response [ 6 , 8 , 26 ]. It may also help standardize organoid quality control by extracting features that are difficult to assess manually, such as growth kinetics, lumen formation, branching complexity, polarity, cell death, and differentiation state. However, AI models are only as reliable as the datasets used to train them. For organoid research, imaging, culture metadata, passage number, donor background, matrix type, sequencing data, and clinical annotation must be collected in a standardized way. Otherwise, AI will amplify existing variability rather than solving it. In my view, the most useful AI applications will be those that connect organoid morphology and molecular profiles with functional endpoints, such as drug sensitivity, electrophysiology, barrier function, hormone responsiveness, implantation competence, or regenerative capacity. Clinical translation will require a shift from promising individual studies to reproducible, standardized, and prospectively validated platforms. PDO biobanks have already demonstrated value in research on cancer, cystic fibrosis, liver disease, and gynecological disorders by preserving patient‐specific molecular and functional features [ 43 , 44 , 50 , 108 , 130 , 181 , 182 , 183 , 184 , 185 , 186 , 187 ]. However, biobanks are useful only when linked to high‐quality clinical metadata, consistent culture protocols, reliable quality control, and clinically meaningful endpoints. For oncology, this approach involves comparing organoid drug sensitivity with real patient response in prospective studies. For genetic diseases, the use of organoid‐based functional assays as predictors of therapeutic benefit should be validated. For regenerative medicine, it involves demonstrating engraftment, safety, functional integration, and long‐term efficacy in appropriate preclinical and clinical settings. Endometrial organoid biobanks deserve particular attention. Endometrial diseases are heterogeneous and hormonally dynamic, and the diseased tissue is often difficult to sample repeatedly in vivo. Biobanks covering the normal cycling endometrium, endometriosis, adenomyosis, PCOS‐related endometrial dysfunction, thin endometrium, intrauterine adhesion, hyperplasia, and endometrial cancer could provide a unified resource for studying disease mechanisms, hormone responsiveness, receptivity, fibrosis, and treatment response [ 44 , 45 , 130 , 142 , 143 ]. However, researchers in the field should be careful to not equate biobank size with scientific value. A smaller biobank with a well‐defined menstrual cycle phase, hormone exposure, pathology, fertility outcome, treatment history, and omics annotation may be more useful than a large but poorly annotated collection. For reproductive applications, cycle timing and endocrine context should be treated as essential metadata, not optional information. Standardization must also extend to culture reporting. Future studies should routinely report tissue source, donor characteristics, passage number, matrix composition, medium formulation, hormone treatment, culture duration, organoid size distribution, cell‐type composition, genomic stability, and functional assays. For translational or therapeutic use, good manufacturing practice‐compliant workflows and xeno‐free or chemically defined matrices will be needed [ 6 , 9 , 32 ]. Regulatory frameworks will also need to distinguish among the organoids used as research models, diagnostic tools, drug screening platforms, and cell‐based therapeutic products. These categories have different safety, reproducibility, and validation requirements. Without such distinctions, clinical translation may remain fragmented despite rapid technical progress. As organoid models become more sophisticated, ethical and sustainability‐related issues will become increasingly important. Brain organoids and neural assembloids raise questions about neural maturation, sensory input, and the boundaries of ethically acceptable modeling [ 70 , 71 , 72 , 73 , 74 , 75 ]. Embryo–endometrial interface models and postimplantation coculture systems require careful observation because they approach sensitive stages of early human development [ 31 , 46 , 132 ]. Tumor organoid biobanks and patient‐derived disease models also involve genetic data, privacy protection, consent for future use, data sharing, and potential commercialization. These questions should not be treated as obstacles to innovation; rather, they are part of building a trustworthy translational framework. Cost and accessibility are also practical concerns. Many organoid protocols rely on expensive growth factors, specialized matrices, microfluidic devices, sequencing technologies, and advanced imaging systems. If these platforms remain limited to a small number of well‐funded laboratories, their clinical and global impact will be restricted. More affordable defined media, reusable or scalable culture devices, robust cryopreservation, and simplified quality‐control assays will be necessary for broader adoption [ 6 , 9 , 49 ]. This is particularly relevant for reproductive medicine and rare diseases, where patient populations may be geographically dispersed and sample availability is limited. In essence, the future of organoid research should not be framed simply as making organoids larger, more complex, or more visually similar to organs. The more important goal is to make them more faithful to the biological question, more reproducible across laboratories, and more predictive of patient‐relevant outcomes. For some applications, this will require vascularized, immune‐competent, multiorgan systems; for others, a simpler but well‐controlled epithelial organoid may be more informative. The field will mature when the model complexity is matched to the experimental purpose. This principle is especially important in gynecology: endometrial organoids, embryo‒endometrial interface models, and reproductive multiorgan systems provide new opportunities to study hormone‐dependent disease, implantation failure, fibrosis, and regenerative repair, but their long‐term value depends on rigorous standardization, clinical annotation, and careful functional validation. If these challenges are addressed, organoids will become not only experimental models but also practical tools for precision medicine, regenerative therapy, and mechanism‐driven clinical decision‐making.

Although organoid models have substantially improved the physiological relevance of in vitro research, they should not be viewed as complete miniature organs. Their value lies in capturing selected principles of tissue organization, lineage differentiation, disease‐associated phenotypes, and patient‐specific responses rather than reproducing the full anatomical, vascular, immune, neural, and mechanical complexity of native organs. This distinction is important because many current limitations are not simply technical imperfections but reflect the gap between self‐organized tissue fragments and fully integrated organs. The major challenges discussed below are therefore not isolated problems; they are interrelated barriers that determine whether organoids can move from descriptive models toward reproducible and clinically actionable platforms [ 25 , 30 ]. The absence of a stable, perfusable vasculature remains among the most fundamental limitations of organoid systems. In vivo, blood vessels provide oxygen, nutrients, endocrine signals, immune cell trafficking, waste removal, and organ‐specific endothelial cues. In contrast, most organoids depend mainly on passive diffusion. As the organoid size or culture duration increases, oxygen and nutrient gradients develop, and the inner regions may become hypoxic, metabolically stressed, or necrotic, which directly limits long‐term maturation and functional readouts [ 36 , 37 ]. This problem is particularly relevant for brain, cardiac, kidney, liver, retinal, and endometrial organoids, where vascular signals are not merely supportive but actively shape tissue patterning, maturation, and injury responses [ 74 , 85 , 90 ]. Several strategies have been developed to address this limitation. Gao et al. summarized broad vascularization approaches, including endothelial coculture, angiogenic factor stimulation, microfluidic perfusion, transplantation‐induced host vascularization, and bioengineering‐based vessel assembly [ 5 ]. In kidney organoids, Homan et al. reported that fluid flow can enhance vascularization and maturation, indicating that physical perfusion cues may be as important as endothelial cell inclusion itself [ 37 ]. Kim et al. further demonstrated that a decellularized kidney ECM improves the vascularization and maturation of kidney organoids, suggesting that organ‐specific matrix composition can instruct vascular development [ 90 ]. In reproductive models, Ahn et al. established a vascularized endometrium‐on‐a‐chip system, while Shibata et al. and Mol et al. advanced embryo–endometrial interface models that began to incorporate stromal and vascular‐like features [ 31 , 32 , 33 , 34 , 35 , 36 , 37 , 38 , 39 , 40 , 41 , 42 , 43 , 44 , 45 , 46 , 47 , 48 , 49 , 50 , 51 , 52 , 53 , 54 , 55 , 56 , 57 , 58 , 59 , 60 , 61 , 62 , 63 , 64 , 65 , 66 , 67 , 68 , 69 , 70 , 71 , 72 , 73 , 74 , 75 , 76 , 77 , 78 , 79 , 80 , 81 , 82 , 83 , 84 , 85 , 86 , 87 , 88 , 89 , 90 , 91 , 92 , 93 , 94 , 95 , 96 , 97 , 98 , 99 , 100 , 101 , 102 , 103 , 104 , 105 , 106 , 107 , 108 , 109 , 110 , 111 , 112 , 113 , 114 , 115 , 116 , 117 , 118 , 119 , 120 , 121 , 122 , 123 , 124 , 125 , 126 , 127 , 128 , 129 , 130 , 131 , 132 , 133 ]. These studies show that vascularization is moving from a generic technical goal toward an organ‐specific design principle. Nevertheless, current vascularized organoids still rarely contain fully perfusable, stable, hierarchically patterned, and functionally mature vascular networks. Many systems produce endothelial‐like structures but do not reproduce long‐term blood flow, vessel barrier function, pericyte coverage, or organ‐specific endothelial heterogeneity. Research efforts within this field ought to shift focus away from simply verifying vascular formation within organoid models, toward establishing tailored criteria for vascular functionality matched to distinct experimental objectives. Studies examining drug metabolism prioritize sustained perfusion and intact barrier function, investigations of embryo implantation or tissue fibrosis center on signaling interactions between endothelial, stromal and immune cells, while work targeting neural or cardiac tissue maturation places greater weight on trophic factors and mechanical cues originating from vascular compartments. Establishing functional evaluation standards tailored to different research objectives can help avoid relying solely on morphological features to judge vascular structures. Most organoid systems primarily reproduce epithelial or parenchymal compartments and only partially capture immune surveillance, inflammatory remodeling, neural regulation, and stromal crosstalk. This limitation is not trivial because immune and neural components are not accessory elements in many organs. Cross tissue systems display unique multicellular signaling axes: intestinal immune microbial neural crosstalk governs epithelial barrier integrity and inflammatory pathogenesis [ 113 ], brain microglia mediate synaptic remodeling and neuroinflammatory processes [ 71 ], and endometrial immune cells, stromal fibroblasts, and endothelial cells together with neuroangiogenic signals jointly modulate menstruation, embryo implantation, endometriosis progression, and tissue repair [ 145 , 146 , 147 , 148 , 149 , 150 , 151 , 152 , 153 , 154 ]. Different organ systems have addressed this issue with different degrees of progress. Intestinal culture models represent one of the most mature research systems, as coculture setups and organ‐on‐chip devices allow direct characterization of signaling crosstalk between host tissue, resident microbes, and immune populations [ 115 , 116 , 117 , 118 ]. Brain organoid studies have made progress in terms of neuronal patterning and assembloids, but the incorporation of microglia, vascular cells, and peripheral immune components in a developmentally appropriate manner remains difficult [ 70 , 71 , 72 , 73 , 74 , 75 ]. In endometrial research, Rawlings et al. developed endometrial assembloids that combine epithelial organoids with stromal cells, and Gnecco et al. used a synthetic matrix‐based coculture system to study epithelial–stromal crosstalk [ 131 , 158 ]. These models are important because they shift the endometrial field away from epithelial‐only organoids toward more physiologically relevant multicellular systems. However, even these approaches do not yet fully recapitulate the cyclic immune, vascular, stromal, and hormonal microenvironment of the human endometrium. A practical challenge is that adding more cell types does not automatically improve model quality. Each additional component introduces new variables, including cell source, maturation state, donor matching, culture timing, and relative abundance. Poorly controlled coculture can increase biological noise rather than physiological fidelity. Therefore, immune‐competent or neural‐integrated organoids should be constructed according to the mechanism being studied. For example, embryo implantation research relies on uterine natural killer cells, decidual stromal cells, trophoblast communication, and adaptive vascular structures. Models constructed to investigate endometriosis need macrophages, fibroblasts, endothelial cells, and sensory nerve‐derived factors, while platforms for cancer immunotherapy research require tumor responsive immune cells and physiological immune checkpoint backgrounds [ 132 , 133 , 134 , 135 , 136 , 137 , 138 , 139 , 140 , 141 ]. A mature organoid model should not simply contain more cell types but should contain the right cell types in the right state and at the right time. Organoid heterogeneity remains a persistent obstacle for quantitative biology and clinical translation. Organoids can vary in size, shape, polarity, cellular composition, maturation status, gene expression state, and functional output across donors, tissue sources, matrix batches, passages, and laboratories [ 6 , 9 , 26 ]. Some heterogeneity is biologically meaningful, especially in PDOs that preserve interpatient differences and tumor diversity [ 21 , 24 , 130 ]. However, uncontrolled technical heterogeneity makes it difficult to distinguish true disease‐specific phenotypes from culture‐induced variation. This issue is particularly evident when mature organoid systems are compared with emerging systems. Intestinal organoids have become relatively standardized because their stem‐cell niche factors, epithelial architecture, and functional assays are well established [ 15 , 113 , 114 , 115 , 116 , 117 , 118 , 119 , 120 , 121 ]. In contrast, brain, cardiac, kidney, and endometrial organoids still show substantial variability in maturation, lineage composition, and protocol‐dependent outcomes [ 70 , 71 , 72 , 73 , 74 , 75 , 86 , 87 , 88 , 89 , 90 , 91 , 92 , 93 , 94 , 95 , 96 , 97 , 98 , 99 ]. Patient‐derived cancer organoid biobanks have shown the translational value of standardized culture and drug testing, but even in oncology, a consensus on organoid‐based drug sensitivity testing remains under development [ 21 , 50 , 250 ]. These comparisons suggest that reproducibility improves when the field defines not only culture recipes but also minimal quality control standards, functional endpoints, and reporting frameworks. For large‐scale screening, scalability is just as important as biological fidelity. Manual embedding, variable Matrigel droplets, inconsistent organoid size, and subjective imaging analysis limit throughput and reproducibility. Automated culture platforms, high‐content imaging, AI‐assisted phenotyping, and standardized cryopreservation strategies can partly solve these problems [ 6 , 8 , 9 , 26 ]. However, technical automation should not replace biological validation. A scalable organoid platform is meaningful only if its readouts remain linked to clinically relevant phenotypes, such as drug response, barrier function, electrophysiology, hormone responsiveness, implantation competence, metabolic activity, or regenerative potential. In this sense, scalability should be treated as a translational requirement rather than simply an engineering convenience. The microenvironment is both the strength and weakness of organoid culture. Natural extracellular matrices such as Matrigel provide basement‐membrane‐like support and have enabled the establishment of many organoid systems, but they also introduce poorly defined biochemical compositions, variable stiffness, growth factor contamination, and limited clinical compatibility [ 32 , 38 ]. These issues are particularly problematic when organoids are used for drug screening, mechanistic perturbation, or regenerative applications because matrix‐derived signals may confound pathway interpretation and reduce reproducibility. Synthetic and tunable matrices offer important alternatives. Gjorevski et al. reported that designer matrices can support intestinal stem cell expansion and organoid formation by controlling matrix mechanics and biochemical ligands [ 32 ]. Gnecco et al. extended this concept to the human endometrium, using a fully synthetic ECM to study epithelial–stromal crosstalk in a more defined system [ 158 ]. A decellularized ECM represents another strategy, as it preserves tissue‐specific biochemical cues; for example, a kidney decellularized matrix enhances the vascularization and maturation of kidney organoids [ 90 ]. These approaches illustrate two complementary directions: synthetic hydrogels improve controllability, whereas decellularized matrices improve organ‐specific biological relevance. However, neither strategy is perfect. Synthetic matrices may lack the full biochemical complexity of native tissues, whereas decellularized matrices can still vary among donors, organs, and preparation methods [ 4 , 32 , 90 , 158 ]. Organ‐on‐chip systems add another layer of control by introducing flow, gradients, mechanical forces, and spatial organization, but they also increase technical complexity and reduce the ease of standardization [ 25 , 30 , 106 , 133 ]. Therefore, the next step should not be the universal replacement of Matrigel with a single “better” matrix. Instead, the matrix should be selected according to the biological question. Developmental models may require dynamic stiffness and morphogen gradients; cancer models may require tumor‐specific ECM remodeling; liver and kidney models may require perfusion‐compatible matrices; and endometrial models may require hormone‐responsive stromal‐like mechanisms. A more useful standard for future organoid research is not whether the matrix is natural or synthetic but whether its composition, mechanics, and biological signals are defined well enough to support reproducible interpretation.

This study was supported by the National Natural Science Foundation of China (82288102, 8257061581, 82271699), and the Beijing Natural Science Foundation (JQ26036, 7252152, 7254445), and the Key Clinical Projects of Peking University Third Hospital (BYSYZD2023028).

Organoid technology provides a three‐dimensional, self‐organizing experimental platform that bridges the gap between conventional two‐dimensional culture and in vivo physiology [ 1 , 2 , 3 , 4 , 32 ]. Through the integration of suitable cell sources, well defined morphogen and growth factor signals, as well as extracellular matrix (ECM) scaffolds, organoids are capable of reconstructing core features including tissue structural organization, cell fate determination, intercellular signaling, and functional characteristics unique to each organ type [ 33 ]. Compared with monolayer cultures, organoids better preserve spatial organization, cellular heterogeneity, polarity, and patient‐specific genetic or phenotypic characteristics, making them particularly useful for studying human development, disease initiation, drug response, and regenerative potential across multiple organ systems [ 34 , 35 ]. These advantages are especially evident in contexts where primary tissues are scarce, animal models incompletely reproduce human biology, or longitudinal sampling is difficult, such as in the context of neurodevelopmental disorders, cancer precision medicine, host–pathogen interactions, inherited diseases, and reproductive medicine [ 36 , 37 , 38 ].​ Despite these strengths, these organoids should not be regarded as complete miniature organs. Most current systems still capture only selected epithelial, parenchymal, or lineage‐restricted compartments, while vascular perfusion, immune surveillance, neural regulation, stromal remodeling, and mechanical forces are often absent or only partially reconstructed [ 39 , 40 , 41 , 42 ]. As organoid size or culture duration increases, limited diffusion of oxygen and nutrients can lead to hypoxia, necrosis, and incomplete maturation, thereby affecting functional readouts and translational reliability [ 43 , 44 , 45 , 46 ]. Furthermore, numerous existing culture protocols remain dependent on undefined native matrix materials including Matrigel. These substances generate inconsistent outcomes across different production batches and hinder the establishment of standardized systems suitable for clinical applications [ 47 , 48 , 49 , 50 ]. Variations derived from distinct tissue donors, variable cell passage counts, differing culture periods, divergent cell differentiation states, and laboratory exclusive culture protocols collectively impair experimental reproducibility and hinder consistent large scale comparative analysis [ 51 ]. Accordingly, further advancement of organoid research necessitates fully defined matrix materials, coculture systems featuring vascular structures and functional immune components, organ‐on‐chip devices, unified quality control benchmarks, as well as synergistic combinations with multiomics profiling, CRISPR‐mediated gene manipulation, AI tools, and clinical tissue biobanks with complete clinical annotation records [ 34 , 52 ].

Organoid technology has become important for modeling human development, disease mechanisms, drug response, and regenerative potential. By integrating stem cell biology, matrix engineering, microfluidics, genome editing, multiomics, and AI, organoids provide a practical bridge between traditional in vitro models and in vivo physiology. For the brain, retina, kidney, heart, lung, liver, intestine, and endometrium, these systems preserve key features of tissue architecture, cellular diversity, lineage organization, and patient‐specific disease phenotypes, supporting more reliable studies of organ development, injury, infection, fibrosis, cancer, toxicity, and therapeutic response. Despite this progress, compared with complete organs, organoids remain simplified models. Incomplete vascularization, limited immune and neural integration, immature cell states, matrix variability, scalability barriers, and inconsistent standardization continue to limit reproducibility and clinical translation. Future work should focus on building models that are not simply more complex but better matched to specific biological and clinical questions. Defined matrices, vascularized coculture systems, organ‐on‐a‐chip platforms, spatial multiomics, CRISPR‐based perturbation, AI‐assisted phenotyping, and clinically annotated biobanks will be central to this transition. With stronger biological fidelity, technical standardization, and clinical validation, organoids are expected to become practical tools for precision medicine, regenerative therapy, toxicology assessment, and mechanism‐guided drug development (Figure  5 ). Current applications and future directions of organoids in different organs. Organoid systems enable versatile research across multiple human tissues. Representative current applications include neural circuit reconstruction, age‐related macular degeneration disease modeling, CRISPR‐mediated gene editing for cystic kidney disease mutation study, fabrication of cardiac organoid patches, drug high‐throughput screening, in vivo functional tissue engraftment, host–microbe crosstalk analysis, and mechanistic exploration of tissue hormone responses. Future work will focus on improving model fidelity and expanding application scope. Technological convergence and integration will further advance organoid research toward precision medicine.

Organoid technology has become an important tool in modern biomedical research. It has transformed the way researchers investigate human development and disease mechanisms, perform drug discovery, and study regenerative medicine [ 1 , 2 , 3 , 4 ]. Organoids are three‐dimensional structures generated from stem cells or primary tissues under defined culture conditions [ 5 , 6 , 7 ]. They retain many of the structural and functional features of native organs through intrinsic self‐organization [ 8 , 9 ]. The evolution of organoid technology has been driven by a series of landmark advances that have progressively transformed it from a conceptual framework into a versatile platform for biomedical research. The concept of cellular self‐organization originated in 1907, when Wilson demonstrated that dissociated sponge cells could spontaneously reaggregate into organized tissues, providing the theoretical foundation for organoid development [ 10 ]. In 1987, the introduction of Matrigel established a biologically relevant three‐dimensional culture matrix [ 11 ], and in 2006, the development of induced pluripotent stem cell (iPSC) technology further expanded the range of cellular sources available for organoid generation [ 12 , 13 , 14 ]. A defining milestone was achieved in 2009, when Clevers et al. [ 15 ] generated long‐term intestinal organoids from single Lgr5 + adult stem cells, establishing the first modern organoid culture system and providing a universal strategy for epithelial organoid generation [ 16 ]. The following years witnessed the rapid expansion of organoid models: gastric organoids were established in 2010 [ 17 ], human colorectal cancer (CRC) organoids and retinal organoids were reported in 2011 [ 18 , 19 ]; in 2013, vascularized iPSC‐derived liver buds and cerebral organoids further expanded organoid technology from epithelial self‐renewal systems toward organ‐bud engineering and regenerative medicine [ 20 , 21 ]. By 2015, standardized kidney organoids and functional airway and alveolar organoids had been established [ 22 , 23 ], greatly extending the application of organoids to developmental biology and disease modeling. The establishment of the first patient‐derived organoid (PDO) biobank before 2016 marked another major step toward precision oncology and personalized medicine [ 24 , 25 ]. Since 2018, the field has entered an era focused on engineering and translation. Vascularized‐transplanted brain organoids [ 26 ], multiorgan organ‐on‐chip systems [ 27 ], and artificial intelligence(AI)‐enabled organoid platforms have substantially improved the physiological relevance and experimental scalability of this technology. More recently, reproductive organoids, clinically guided PDO‐based drug sensitivity testing, and bioprinted vascularized composite organoids have further accelerated the translation of organoid technology from laboratory research to regenerative medicine and precision therapy (Figure  1 ) [ 28 , 29 , 30 , 31 ]. Timeline of organoid technology development. Key evolutionary milestones cover the construction of 3D extracellular matrix culture platforms, the development of induced pluripotent stem cell technology, the first long‐term epithelial organoid protocols, the generation of diverse tissue‐specific organoids including cerebral, liver, gastric, colorectal, retinal, kidney, lung, and endometrial models together with vascularized organ buds, patient‐derived organoid biobanks, organ‐on‐chip systems, multiorgan integration, and AI‐assisted organoid engineering. This chronological roadmap traces the field from foundational cell self‐organization to modern multitissue, clinically oriented organoid engineering. Although organoid research has advanced rapidly, recent reviews have often been organized around a single organ, disease, or technology, with relatively little emphasis on the common principles that underlie different organoid systems. The aim of this review is to provide a broader perspective by integrating recent advances across multiple organ systems and identifying shared opportunities and challenges in organoid research. Its major strength lies in the comparison of representative organoid models across multiple organ systems to identify common cultivation strategies, universal technical bottlenecks, and emerging engineering solutions, thereby offering a broader framework for the future development and clinical application of organoid technology. To present these advances in a coherent framework, this review is organized according to the developmental progression of organoid technology. We first summarize the overall strengths and current bottlenecks of organoid systems, followed by the fundamental considerations underlying organoid establishment, including cell source selection, culture media, three‐dimensional culture techniques, support materials, and microenvironment simulation. We then examine representative applications across major organ systems and summarize their biological and translational significance. Building on these applications, we discuss the key challenges that continue to limit organoid fidelity and clinical implementation, including insufficient vascularization, the absence of immune and neural components, heterogeneity, and poor standardization. Finally, we outline future directions focused on improving physiological relevance through multiorgan integration, technological convergence with multiomics, CRISPR, and artificial intelligence (AI), as well as standardized biobanks, clinical translation, and ethical considerations.

The authors declare no conflicts of interest.