1. Introduction
Replacement of dysfunctional tissues or organs with healthy counterparts remains the ultimate goal in medicine. Organoids have emerged as versatile
in vitro platforms for modeling human development, disease pathology, and personalized therapies [
1]. Modern biofabrication techniques such as three-dimensional (3D) bioprinting [
2], microfluidic systems [
3], and bioengineered scaffolds [
4] are increasingly being integrated into organoid culture systems in order to create more physiologically relevant microenvironments and enhance the scalability and reproducibility of tissue constructs. These innovations have opened new avenues for generating biological substitutes for tissue repair and organ replacement [
5]. Functional organoids are primarily derived from human pluripotent stem cells (PSCs) [
[6],
[7],
[8]] or adult tissue-derived stem cells (ASCs) [
9,
10] through intricate
in vitro cell manipulation steps that typically encompass cell isolation, expansion, and genetic processing to achieve the quality and quantity requirements. Subsequently, dissociated stem cells are cultured in 3D environments and sequentially exposed to a defined combination of signaling cues that provide appropriate biochemical and biophysical stimuli targeting key regulatory pathways. These processes promote stem cell differentiation into diverse cell types and facilitate their assembly into multicellular organoids. Despite significant achievements, biofabrication-intensive approaches face substantial challenges in clinical translation owing to their labor-intensive nature, high cost, and time-consuming protocols. Additionally, cell-level strategies are hindered by the limited availability of transplantable stem cells, difficulty in controlling cell fate, ethical concerns, risk of immune rejection, and tumorigenicity. These limitations therefore highlight the urgent need for a more efficient and clinically viable strategy to provide an unlimited regenerative source for generating therapeutically relevant organoids.
Direct generation of functional organoids from readily accessible adult tissues offers a promising solution to overcome these barriers and expand the clinical potential of organoid-based therapies. Most adult tissues harbor resident ASCs within specialized microenvironments known as niches [
11]. The dynamic interplay between ASCs and niche components, including extracellular matrix (ECM) molecules, growth factors, cytokines, and mechanical cues, plays a crucial role in regard to maintaining tissue homeostasis and enabling regeneration following injury [
12]. Emerging evidence has demonstrated the remarkable plasticity of ASCs in transdifferentiation across germ layer boundaries both
in vivo and
in vitro, challenging the traditional lineage restrictions [
13,
14]. These findings highlight the untapped potential of adult tissues as direct sources of functional organoids.
In this study, we have presented a simple, safe, and scalable strategy for the direct generation of multilineage functional organoids with significant implications for disease modelling and regenerative medicine. By culturing human adult adipose tissue in a specialized suspension culture system, we developed reaggregated microfat (RMF) tissues that could directly differentiate into functional organoids in all three germ layers. These include mesodermal bone marrow organoids capable of reconstituting human hematopoiesis in immunodeficient mice, endodermal islet organoids that reverse hyperglycemia in streptozotocin (STZ)-induced diabetic mice, and ectodermal neural-like tissue containing neuronal or neuroglial cells.
2. Methods and materials
2.1. Data and code availability
Data are available upon reasonable request. All data relevant to the study are included in the article or uploaded as supplementary information, except for the raw data for single-cell RNA sequencing (scRNA-seq) reported in this publication can be accessed under the Gene Expression Omnibus (GSE280107) on request. This paper analyses existing data.
Any additional information is available upon request.
2.2. Animal models
All the animal experimental procedures were approved by the Animal Care and Use Committee of Shanghai Ninth People’s Hospital, Shanghai Jiao Tong University School of Medicine, China, with project license numbers of SH9H-2021-A79-1 (bone marrow organoid associated animal experiments) and SH9H-2023-A778-1 (islet organoid associated animal experiments). Mice were housed in the specific pathogen-free facility at Shanghai Ninth People’s Hospital under a 12-h light/12-h dark cycle (06:00 to 18:00) at room temperature of (22 ± 1) °C, with ad libitum access to food and water.
For subcutaneous implantation of RMF-derived chondroid pellets, NOD/LtJ-PrkdcscidIl2rgem1/Shjh (NPSG) female mice aged four weeks were used. For secondary transplantation of human cluster of differentiation 34 positive (hCD34+) cells, NPSG female mice aged 8–10 weeks were used. For kidney subscapular implantation of RMF-derived islet organoids, NOD.Cg-Prkdcscid/Shjh (NOD/SCID) mice aged 8–10 weeks were used. Mice of the same sex were randomly assigned to experimental groups. Both strains of mice were purchased from Shanghai Jihui Laboratory Animal Care Co., Ltd. (China).
2.3. Human subjects
Human adult adipose tissue samples were collected from individuals who underwent liposuction surgery. The criteria for adipose tissue donor enrolment were: ① adult (18–60 years) undergoing liposuction or abdomen plasticity surgeries; ② body mass index (BMI) ranging from 19.5 to 38 kg∙m–2 (indicating normal to obese); ③ exclusion of infectious/genetic/psychiatric conditions. Donor demographics: age: 20–50 years (median 31 years); BMI: 19.9–36.7 kg∙m–2 (normal (19.5–24.9 kg∙m–2): 38.5%, overweight (25–29.9 kg∙m–2): 51.7%, obese (30–38 kg∙m–2): 13.8%); depot distribution: abdomen (79.3%), thigh (17.2%), and mixed (3.5%). Donor information was shown in Table S1 in Appendix A. The collection, processing, and experimental use of human adult adipose tissue samples were performed in full accordance with national regulations and institutional guidelines for human biomedical research in China. All protocols involving human tissue were reviewed and approved by the Institutional Review Board of Shanghai Ninth People’s Hospital, Shanghai Jiao Tong University School of Medicine (SH9H-2023-TK189-1). Written informed consent was obtained from all donors. Data were processed in compliance with applicable data privacy regulations.
Human umbilical cord blood samples were collected from healthy full-term births at Shanghai Ninth People’s Hospital following Institutional Review Board (IRB)-approved protocols (IRB No. SH9H-2023-TK189-1), or purchased from the Shandong Qilu Stemcell Engineering Co., Ltd. (China). All blood samples were received as de-identified, therefore, the information on the age and/or gender of the donors is not available.
2.4. Quantification and statistical analysis
All data are presented as means ± standard deviation (SD) and were analyzed by GraphPad Prism 9 (GraphPad Software Inc., USA). Paired or unpaired two-tailed Student’s t tests were performed when two groups of samples were compared. One-way or two-way analysis of variance (ANOVA) with Tukey’s or Dunnett’s tests were performed when multiple groups were compared. Log-rank (Mantel–Cox) test is used for comparison of survival rate between multiple groups. p values are indicated in the figures as: not significant (ns), p > 0.05; ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001. Statistical details for each experiment can be found in the figures and the legends.
3. Results
3.1. The in vitro self-reaggregation process enhances the multipotent differentiation capacity of microfat tissues
We first established a protocol for generating RMF tissues from human adult adipose tissue, as outlined in
Fig. 1(a) and Fig. S1(a) in Appendix A. Subcutaneous adipose tissue was obtained from patients who underwent liposuction surgery. The harvested lipoaspirates were washed, minced, and mechanically emulsified in order to produce microfat, which was subsequently cultured in a growth medium on ultra-low attachment plates for three weeks to induce self-reaggregation. During the suspension culture period, the microfat progressively condensed into discrete microtissue clusters (
Fig. 1(a)). Scanning electron microscopy (SEM) and hematoxylin and eosin (HE) staining revealed abundant fibroblast-like cells, particularly on the surface of RMF tissues (
Figs. 1(b) and
(c)). This observation was supported by strong the Ki67 staining (
Fig. 1(d)), which indicated active cell proliferation throughout the culture period. In parallel, the ECM phenotype gradually shifted from adipose to stromal, as evidenced by the increased laminin and fibronectin expression, particularly in regions enriched with proliferating cells (
Fig. 1(e)). Laminin and fibronectin are key ECM components of stem cell niches and are known to support the regenerative capacity [
15,
16]. Quantitative analysis demonstrated a significant increase in the DNA content over the three week culture period (
Fig. 1(f)), suggesting substantial cellular proliferation within the RMF tissues. To further compact the tissues, RMF tissues were punched into 4 mm pellets using a biopsy punch and then cultured for an additional week, forming smooth, spherical RMF pellets, each containing approximately 5.92 × 10
4 cells (
Fig. 1(g)). After an additional week of culture, fibroblast-like cells became more prominent both on the surface and within the RMF pellets (
Figs. 1(b) and
(c)).
To characterize the cellular composition of the RMF pellets, they were enzymatically dissociated into single cells (hereafter referred to as RMF cells) and subjected to flow cytometry. For comparison, we included three control groups: stromal vascular fraction cells (SVFs) obtained from native adipose tissue, primary adipose tissue-derived stem cells (P0 ADSCs; cultured for 5–7 d without passaging), and expanded ADSCs (maintained through regular passaging over a four week period). The detailed procedures for cell isolation and group design are provided in Fig. S1(b) in Appendix A. Flow cytometry revealed that RMF pellets consisted of approximately 66.1% mesenchymal stem cells (MSCs; CD45
–CD73
+CD90
+), 11.15% pericytes (CD45
–CD31
–CD146
+), and 5.95% endothelial progenitor cells (EPCs; CD45
–CD31
+CD34
+). Compared with native adipose tissue, RMF pellets exhibited a significantly higher proportion of MSCs, whereas the percentages of pericytes and EPCs were comparable. Relative to expanded ADSCs, RMF pellets showed a lower MSC fraction but significantly greater proportions of pericytes and EPCs (
Fig. 1(h)). We further assessed the self-renewal and differentiation potentials of the RMF. As shown in Fig. S1(c) in Appendix A, RMF cells exhibited a proliferative capacity comparable to that of SVFs and were significantly greater than that of both P0 and expanded ADSCs. Similarly, RMF cells exhibited colony-forming efficiency on par with SVFs and they were superior to that of expanded ADSCs (Fig. S1(d) in Appendix A). Furthermore, RMF retained a mesoderm trilineage differentiation potential similar to that of freshly harvested SVFs and substantially greater than that of either P0 or expanded ADSCs (Fig. S1(e) in Appendix A).
In order to further elucidate the cell population dynamics during
in vitro self-reaggregation, we performed scRNA-Seq on cells derived from native adipose tissues, RMF pellets (4 mm), and expanded ADSCs using a 10× Genomics platform. The resulting atlas consisted of 65 501 cells, which were clustered based on the differential expression of cell type-specific marker genes and visualized using Uniform Manifold Approximation and Projection (UMAP). Cluster analysis identified eight distinct cell clusters across the three sample types, categorized into five major cell types based on the spatial distribution within the UMAP: adipose-derived stem and progenitor cell (ASPC;ASPC-1 through ASPC-4) clusters, preadipocyte clusters, endothelial cell clusters, smooth muscle cell (SMC) clusters, and immune cell clusters (
Fig. 2(a), Fig. S2(a) in Appendix A). Cell type assignment was performed using established markers and transcriptional signatures previously reported for adipose tissue subpopulations [
17,
18] (Figs. S2(b) and (c) in Appendix A). Analysis of the relative abundance of these five major cell types revealed that the proportion of ASPCs increased markedly after four weeks of
in vitro culture, from 24.57% in native fat to 97.17% in the RMF pellets, and further to 99.90% in expanded ADSCs (
Fig. 2(a)). These results were consistent with our earlier flow cytometry findings when comparing RMF cells, SVFs (native fat), and expanded ADSCs (
Fig. 1(h)). To visualize the inter-sample transcription relationships, we generated a ternary plot that positioned each cell according to the expression of known signature genes from native fat, RMF, and expanded ADSCs. RMF cells exhibited a transcriptional profile that more closely assembled native fat than expanded ADSC samples (
Fig. 2(b)). Collectively, these results indicate that
in vitro self-reaggregation preserves the cellular phenotype more faithfully than conventional two-dimensional (2D) expansion while maintaining greater transcriptional similarity to native fat tissue.
In the present study, we identified four distinct subpopulations of ASPCs across all three sample types. To evaluate the stemness status of these ASPC subpopulations, we performed pseudotime trajectory analysis. Our results revealed that the ASPC-1 cluster, comprised primarily of cells from native fat samples (96.04%), occupied the earliest point along the developmental trajectory. This was followed by the ASPC-2 and ASPC-3 clusters, which were predominantly derived from the RMF pellets (94.91% and 96.55%, respectively). Conversely, the ASPC-4 cluster, which mainly originated from the expanded ADSCs, was positioned at the terminal end of the trajectory (
Fig. 2(c)). Pseudotime analysis further demonstrated that ASPC-2 and ASPC-3 clusters from RMF pellets represented intermediate differentiation states, positioned between native fat-derived APSCs and those from expanded ADSCs (
Fig. 2(d)). Furthermore, the performed cell-to-cell correlation analysis showed that clusters within RMF pellets displayed stronger similarity to their counterparts in native fat tissue than clusters from expanded ADSCs (
Fig. 2(e), Fig. S2(d) in Appendix A). The trajectory position of ASPC-4 is further detailed in
Fig. 2(f). Notably, both ASPC-2 and ASPC-3 clusters exhibited high expression of genes associated with stemness (
SPON2, HTRA1, TIMP1, STMN2, PRDX2, PRDX4, CCND1, and
NUMB), cell cycle progression (
MMP2, MMP14, and
TWIST1), and mesodermal lineage differentiation (
ADH1B, CTSK, LRP1, TIMP2, and
FGF7) (Fig. S2(e) in Appendix A). These observations therefore suggest that RMF-derived ASPC subpopulations maintain an intermediate differentiated phenotype, reflecting a transitional state between the native fat tissue and expanded ADSCs.
3.2. Generation of humanized bone marrow organoids from RMF pellets
To further evaluate the mesodermal differentiation potential of RMF tissues, we sought to transform the RMF pellets into functional humanized bone marrow organoids that faithfully recapitulated the structural and functional complexity of native human bone marrow in an animal model [
19].
Fig. 3(a) illustrates our two-stage protocol for generating humanized bone marrow organoids from 4 mm RMF pellets. Initially, the RMF pellets were the chondrogenically primed in ultra-low attachment plates for four weeks, followed by an additional period of two weeks of hypertrophic introduction. Histological analysis revealed that chondrogenically primed RMF pellets exhibited typical cartilage-like features, including cuboidal cells within large lacunae and strong positive staining for Safranin O (Saf-O) and collagen type II (Col II). After two weeks of hypertrophic induction, the RMF pellets displayed a mature hypertrophic cartilage phenotype characterized by an increased expression of Col X, a specific marker of hypertrophic cartilage (
Fig. 3(b)). ECM remodeling from a chondrogenic to a hypertrophic profile was confirmed by elevated expression of matrix metalloprotein 9 (MMP9) and MMP13. Osteogenic and angiogenic potentials were also indicated by the presence of Col I and bone sialoprotein (BSP) (Fig. S3(a) in Appendix A), indicating the readiness for endochondral ossification upon
in vivo implantation. To assess the reproducibility of the generation of hypertrophic chondroid pellets, we processed 240 pellets from six independent adipose tissue donors (40 pellets per donor) and evaluated them using Saf-O staining. Of these, 44 were graded as low-quality, 98 as moderate-quality, and 98 as high-quality (Fig. S3(b) in Appendix A).
Hypertrophic chondroid pellets were subcutaneously implanted into the dorsal region of NPSG mice (
Fig. 3(a)) and harvested 2, 4, 6, 8, and 12 weeks post-transplantation (Fig. S3(c) in Appendix A). Movat pentachrome staining demonstrated progressive remodeling in the presence of five different tissue types: bone, cartilage, bone marrow, adipose tissue, and fibrotic tissue. At week 2, the explants were predominantly cartilaginous, with minimal bone or hematopoietic tissue formation observed at the periphery. At week 4, the levels of the bone and hematopoietic components increased. By 6–12 weeks, samples darkened macroscopically and histologically revealed gradual replacement of the cartilage by bone, cortical shell formation, and expansion of the bone marrow cavity (Figs. S3(c)–(f) in Appendix A). Micro-computed tomography (CT) analysis confirmed progressive ossicle maturation with an average ossicle volume of ∼30 mm
3 and a declining bone volume (BV)/total volume (TV) ratio (Figs. S3(g) and (h) in Appendix A). To assess reproducibility, we analyzed 160 implanted chondroid pellets from four independent donors (40 per donor) by HE staining. After 12 weeks, 155 explants contained both bone and hematopoietic tissues. Of these, 48 were graded as low-quality, 76 as moderate-quality, and 31 as high-quality (Fig. S3(i) in Appendix A), indicating high reproducibility.
To characterize the architecture of the ossicles, we examined the samples retrieved at 12 weeks. Bone marrow organoids exhibited both endosteal and perivascular niches, with the former containing osteoblasts, osteoclasts, and osteocytes adjacent to trabeculae and the latter located near sinusoids and composed of mesenchymal and hematopoietic cells [
20]. Morphologically, the ossicles mimicked the native bone marrow, with cortical bone shells surrounding the trabecular bone, vascular channels, and diverse marrow cell lineages, including erythroid, myeloid, and megakaryocytic cells (
Fig. 3(c)). Each ossicle yielded an average of 4.75 × 10
6 bone marrow cells (
Fig. 3(d)). Immunostaining confirmed the presence of osteocalcin (OCN) positive bone matrix, osteorix-positive osteoblasts and osteocytes, and tartrate-resistant acidic phosphatase (TRAP) positive osteoclasts, reflecting active bone remodeling (
Fig. 3(e)). CD146
+ stromal cells and leptin receptor-positive MSCs (LepR
+) (
Fig. 3(f)) were also observed. Laminin staining revealed a dense vascular network within the marrow cavity, including sinusoidal and arteriole-like vessels. Neuro-glial antigen 2 positive (NG2
+) pericytes and α-smooth muscle actin positive (α-SMA
+) mural cells surrounded these structures (
Fig. 3(g)). Strikingly, human-specific NuMA (hNuMA) positive cells were detected in the critical bone, trabecular bone, vessels, stroma, and adipose components (
Fig. 3(h)), indicating successful humanization of the bone marrow microenvironment. Hereafter, these ossicles are hereafter referred to as humanized ossicles (hOssicles).
Immunostaining for murine CD45 (mCD45) positive cells revealed that most hematopoietic cells within the hOssicles were host-derived (
Fig. 3(i)). Additionally, rare hematopoietic stem cells (HSCs; CD48
–CD150
+) were also detected (
Fig. 3(j)). Flow cytometry analysis confirmed that the composition of murine hematopoietic stem and progenitor cells (HSPCs) in hOssicles resembled that in mouse femurs (mFemurs), including HSCs (lineage
– sca-1
+ c-kit
+ (L
–S
+K
+)), short-term HSCs (ST-HSCs; L
–S
+K
+CD34
+CD135
–), long-term HSCs (LT-HSCs; L
–S
+K
+CD34
–CD135
–), multipotent progenitors (MPPs; L
–S
+K
+CD34
+CD135
+), hematopoietic progenitor cells (HPCs; lineage
– sca-1
− c-kit
+ progenitors (L
–S
–K
+)), megakaryocyte-erythroid progenitors (MEPs; L
–S
–K
+CD34
–CD16/32
–), common myeloid progenitors (CMPs; L
–S
–K
+CD34
+CD16/32
–), granulocyte-monocyte progenitors (GMPs; L
–S
–K
+CD34
+CD16/32
+), and common lymphoid progenitors (CLPs; Lin
–IL-7R
+CD135
+) (Fig. S4 in Appendix A).
Collectively, these results have demonstrated that RMF pellets can be successfully converted into hOssicles through endochondral ossification, supporting the establishment of a functional bone marrow niche of host origin.
3.3. RMF-derived bone marrow organoids support human hematopoiesis in immunodeficient mice
To determine whether hOssicles could support human hematopoiesis and sustain functional hematopoietic cells
in vivo, NPSG mice bearing hOssicles for eight weeks were sublethally irradiated and intravenously transplanted with 5 × 10
5 human umbilical cord blood-derived CD34
+ (hUCB-CD34
+) cells after a 24-h interval, following a standardized protocol [
21]. Engraftment, self-renewal, and differentiation of the transplanted cells were evaluated using flow cytometry, colony-forming unit (CFU) assays, and immunohistochemistry (1st xenotransplantation;
Fig. 4(a)).
Human cell engraftment was detected in all hOssicles at 8 and 16 weeks post-transplantation, as evidenced by the presence of hNuMA
+ cells in the bone marrow cavities (
Fig. 4(b)). Engraftment levels of human CD45
+ cells (as a percentage of total CD45
+ cells, including humans and mice) increased significantly over time in hOssicles, with a rate of 30.72% at week 8 and 67.12% at week 16, far exceeding those in mFemurs (12.67% and 24.55%, respectively;
Fig. 4(c)), which is consistent with previous studies [
22,
23]. These results suggest that hOssicles offer a more favorable environment for human HSPC (hHSPC) engraftment than murine bone marrow. Additionally, immunohistochemical analysis confirmed the presence of diverse human hematopoietic cells in 12 week hOssicles, including hCD45
+ total hematopoietic cells, hCD34
+ HSCs, hCD19
+ B lymphoid cells, hCD33
+ myeloid cells, hCD14
+ monocytes, and hCD15
+ granulocytes. Notably, hCD34
+ cells were localized near the sinusoidal regions, while hCD45
+ cells were broadly distributed in the marrow cavity (Fig. S5 in Appendix A). Quantitative analysis revealed that the hOssicles maintained significantly higher levels of human HSCs (hHSCs; CD45
+Lin
–CD34
+CD38
–) and human HPCs (hHPCs; CD45
+Lin
–CD34
+CD38
+) than mFemurs at both 8 and 16 weeks post-transplantation (
Fig. 4(d), Figs. S6(a) and (b) in Appendix A). The proportions of mature human T cells (CD45
+CD3
+), B cells (CD45
+CD3
–CD19
+), and myeloid cells (CD45
+CD3
–CD33
+) in hOssicles were comparable to those in mFemurs at both time points (
Fig. 4(e), Figs. S6(a) and (b)). To assess the functionality of the engrafted hHSPCs, human cells isolated from the hOssicles and mFemurs at 16 weeks were subjected to
in vitro CFU assays. Both sources yielded burst-forming unit-erythroid (BFU-E); CFU-granulocyte and macrophage (CFU-GM); and CFU-granulocyte, erythrocyte, monocyte, and megakaryocyte (CFU-GEMM) colonies. However, hOssicle-derived cells produced significantly more BFU-E and CFU-GM colonies than mFemur-derived cells (
Fig. 4(f)), suggesting that hOssicles provide a superior niche for hHSPC homing and hematopoietic output.
To further assess their self-renewal capacity, hCD34
+ cells isolated from hOssicles and mFemur were transplanted into secondary NPSG recipients that did not bear hOssicles (2nd xenotransplantation). The cells were enriched by magnetic-activated cell sorting (MACS) and intravenously infused into sublethally irradiated mice (
Fig. 4(a)). At 8 and 16 weeks post-the 2nd transplantation, mice receiving hOssicle-derived hCD34
+ cells exhibited higher engraftment levels of hCD45
+ cells and greater frequencies of immature hHSPCs (gated as hCD34
+ cells) than those receiving mFemurs-derived cells (
Figs. 4(g) and
(h), Figs. S6(c) and (d) in Appendix A). The proportions of mature human lymphoid and myeloid subsets were comparable between the two groups (
Fig. 4(i), Figs. S6(c) and (d)). Together, these results demonstrate that RMF-derived hOssicles provide a supportive humanized microenvironment capable of sustaining long-term engraftment, self-renewal, and multilineage differentiation of hHSPCs, thereby recapitulating functional human hematopoiesis in mice.
3.4. Generation of insulin (INS)-producing organoids from RMF pellets
To investigate whether RMF tissues could give rise to functional endodermal-lineage organoids, we developed a protocol for differentiating RMF pellets into INS-producing islet organoids. Our primary objective was to establish a differentiation protocol with high efficiency and good reproducibility that yields sufficient cell quantity and quality for preclinical applications. We observed that generating dense RMF pellets with high cellularity was crucial, likely because native pancreatic islets are densely packed mini-organs with high metabolic demands. To this end, we optimized a modified protocol that produced smaller RMF pellets with a higher cellular density (Fig. S7(a) in Appendix A). This protocol generated compact, spherical 1 mm RMF pellets after four weeks of suspension culture (Figs. S7(b) and (c) in Appendix A). As the pellet size decreased, we observed a significant increase in DNA content per cubic millimeter and a corresponding increase in cell density (Figs. S7(d) and (e) in Appendix A). The 1 mm RMF pellets contained approximately 2.77 × 104 cells (Fig. S7(e)), with proliferating cells predominantly located at the periphery (Fig. S7(f) in Appendix A). Additionally, reducing the pellet size significantly improved cell viability (Figs. S7(g) and (h) in Appendix A). Owing to their superior density and viability, we used 1 mm RMF pellets for subsequent islet organoid generation.
Conventional protocols for the transdifferentiation of non-pancreatic ASCs into INS-producing cells or islet-like organoids typically involve two stages: induction into pancreatic progenitors followed by β-cell maturation, often without endodermal differentiation [
24]. Here, we present a refined four-stage protocol that mimics
in vivo islet development, guiding RMF cells through the definitive endoderm, pancreatic progenitor, endocrine progenitor, and β-cell stages (
Fig. 5(a)) based on prior studies and pilot optimization [
9,
[25],
[26],
[27]]. After the first stage, most cells were found to have expressed the definitive endodermal markers forkhead box protein A2 (FOXA2) and SRY-box transcription factor 17 (SOX17), as confirmed by immunofluorescence staining (
Fig. 5(b)). In stage two, most cells co-expressed pancreatic progenitor markers pancreatic and duodenal homeobox 1 (PDX1) and homo sapiens NK6 homeobox 1 (NXK6.1) (
Fig. 5(c)). In the third stage, subsets of cells began expressing C-peptide (CP), a byproduct of INS synthesis (
Fig. 5(d)). Specifically, 11.93% were CP
+PDX1
+ and 15.12% were CP
+NKX6.1
+. These cells co-expressed the endocrine progenitor markers neurogenin 3 (NGN3) and musculoaponeurotic fibrosarcoma oncogene homolog B (MAFB) (
Fig. 5(d), Fig. S8(a) in Appendix A), thus indicating the presence of islet progenitor cells. CP expression was markedly increased and the majority of the CP
+ cells co-expressed PDX1 and NXK6.1, with 15.78% CP
+PDX1
+ and 20.15% CP
+NXK6.1
+ (
Figs. 5(e)–(g), Fig. S8(a)). At the end of stage four, despite a ∼55.94% cell death rate, each pellet retained approximately 1.94 × 10
4 viable cells (
Fig. 5(g), Fig. S8(b) in Appendix A).
Based on these values, we estimated an average of ∼3000 CP
+NXK6.1
+ cells per organoid (
Fig. 5(g)). Notably, we also detected the expression of mature β-cell markers musculoaponeurotic fibrosarcoma oncogene homolog A (MAFA) and homeobox protein NK 2 homolog B (NKX2.2), along with glucagon (GCG)
+ α-like cells and polyhormonal CP
+GCG
+ cells (
Fig. 5(e)). Quantitative real-time polymerase chain reaction (qRT-PCR) further confirmed the stage-specific pancreatic gene expression patterns (Fig. S8(c) in Appendix A).
In order to assess functionality, we challenged the final stage organoids with three sequential high glucose exposures followed by KCl-induced depolarization. Under low-glucose conditions, human INS secretion was minimal (0.74 μIU per 1 × 10
3 cells), but rose significantly, by 3.14-fold (2.31 μIU per 1 × 10
3 cells), after the first high-glucose challenge (
Fig. 5(h)). The organoids continued to respond to the subsequent glucose challenges with a stable INS output, and KCl stimulation triggered maximal INS release (3.14 μIU per 1 × 10
3 cells) (
Fig. 5(h)). Transmission electron microscopy revealed the presence of INS granules within organoid cells, averaging (47 ± 17.8) granules per cell, including both immature (diffused gray) and mature (dense core) granules (
Figs. 5(i) and
(j)). These findings confirmed the ability of RMF-derived islet organoids to store and secrete INS in response to glucose stimulation.
To further characterize the cellular composition, we performed scRNA-seq on 8715 cells from two RMF islet organoid donors (
Fig. 5(k), Figs. S9(a)–(c) in Appendix A) and compared them with published human islet datasets [
28] (
Fig. 5(k), Figs. S9(d)–(f) in Appendix A). Four endocrine cell types were identified: β-, α-, δ-, or γ-like cells, each expressing canonical markers:
INS for β-cells,
GCG for α-cells, somatostatin (
SST) for δ-cells, and pancreatic polypeptide (
PPY) for γ-cells (Figs. S9(b) and (e)). These endocrine cells also expressed classical markers including proprotein convertase subtilisin/kexin type 2 (
PCSK2), urocortin 3 (
UCN3), paired box 6 (
PAX6), solute carrier family 30 member 8 (
SLC30A8), chromogranin A (
CHGA), and islet amyloid polypeptide (
IAPP) (Figs. S9(c) and (f)). Integrated clustering revealed that the RMF-derived β-like cells shared similar identity scores and transcriptional profiles with native islet β-cells, indicating genes associated with β-cell maturation, metabolism, and INS secretion (
Fig. 5(l), Fig. S9(g) in Appendix A). Additionally, scRNA-seq analysis revealed the presence of pancreatic exocrine-like cells, including ductal and acinar-like cells, as well as a small population of mesenchymal cells, consistent with previous reports on stem cell-derived islet organoids using both scRNA-seq [
[29],
[30],
[31],
[32],
[33]] and immunohistochemistry [
34,
35]. Collectively, these results demonstrate that RMF-derived islet organoids contain a population of transcriptionally mature, functional β-like cells closely resembling their native human counterparts.
3.5. RMF-derived islet organoids acquire rapid graft vascularization and reverse diabetes in mice
To further validate the longevity and functionality of RMF-derived islet organoids
in vivo, we performed two sets of transplantation experiments. Previous studies have emphasized the importance of early vascular integration in the maturation and sustained function of stem cell-derived islet organoids [
36,
37]. In our initial experiments, we transplanted RMF-islet organoids derived from three independent donors under the kidney capsules of nondiabetic immunodeficient NOD/SCID mice (three mice per donor, 40–50 pellets per mouse) to assess early vascularization and graft integration (Fig. S10(a) in Appendix A). One week post-transplantation, the performed histological analysis revealed limited peri-graft vascularization and minimal vessel ingrowth into the center of the grafts (Fig. S10(b) in Appendix A). Immunofluorescence staining confirmed the presence of INS-positive (INS
+) β-cells and abundant CD31
+ endothelial cells, though most retained as isolate cells without tube formation (Fig. S10(c) in Appendix A). By weeks 2 and 4, perfused blood vessels containing red blood cells were observed within and around the grafts (Figs. S10(b) and (c)). Moreover, the number of hypoxia-induced factor 1α (HIF1α, hypoxia marker) positive and terminal deoxynucleotidyltransferase-mediated dUTP-biotin nick end labeling (TUNEL;apoptotic) positive cells within the grafts significantly decline over time (Figs. S10(d) and (e) in Appendix A). Qualitative analysis revealed progressively increased vascular areas in both HE and CD31-stained immunofluorescence images (Figs. S10(f) and (g) in Appendix A), alongside progressively decreased hypoxic stress in HIF1α and TUNEL-stained images (Figs. S10(h) and (i) in Appendix A). These results indicate rapid neovascularization, which alleviated hypoxic stress and reduced cell death. Overall, these findings have demonstrated that RMF-islet organoids promote rapid graft vascularization and early engraftment post-transplantation.
In the second experiment, RMF-islet organoids from six independent donors were transplanted under the kidney capsule of NOD/SCID mice with STZ-induced diabetes (three mice per donor) to assess long-term glycemic control (
Fig. 6(a), Fig. S11(a) in Appendix A). Based on the number of CP
+NKX6.1
+ cells per organoids, 42–48 RMF-islet organoids were transplanted per mouse, with maximum capacity for one kidney. Undifferentiated RMF pellets from the same donor were used as the controls. All of the mice that received RMF-islet organoids (STZ, organoids) survived until nephrectomy on day 91 post-STZ induction. In contrast, mice transplanted with undifferentiated RMF pellets (STZ, RMF pellets) and non-transplanted controls (STZ, no transplantation (txp)) showed median survival times of 41.5 and 49 d, respectively, although these differences were not statistically significant (Fig. S11(b) in Appendix A). Non-fasting blood glucose levels in the RMF-islet organoid group dropped rapidly after transplantation, stabilized near normoglycemia within two weeks, and remained within the normal range until nephrectomy. Removal of the grafts at week 12 resulted in a rapid return of hyperglycemia (
Fig. 6(b), Fig. S11(c) in Appendix A), confirming that the observed glycemia control was conferred by the transplanted RMF-islet organoids. In contrast, mice receiving undifferentiated RMF pellets remained severely hyperglycemic (∼600 mg∙dL
–1) until death (
Fig. 6(b), Fig. S11(c)).
Body weight in the RMF-organoid group was found to have increased steadily following diabetes reversal, without signs of obesity (
Fig. 6(c), Fig. S11(d) in Appendix A). Human INS was detected in serum as early as two weeks post-transplantation (∼10 μUI∙mL
–1 fasting), with a 1.6-fold increase observed at 30 min after intraperitoneal glucose injections. This INS response was maintained until graft removal (
Figs. 6(d) and
(e), Fig. S11(e) in Appendix A). Intraperitoneal glucose tolerance test (IPGTT) performed at weeks 2 and 12 demonstrated significantly improved glucose tolerance and faster glucose clearance in the organoid groups than in the RMF pellets or untreated control groups (
Fig. 6(f), Fig. S11(f) in Appendix A).
The performed histological analysis of the grafts confirmed preserved islet-like cellular complexity, including INS
+ cells perfused by CD31
+ vasculature, minor populations of GCG
+ and SST
+ cells, and occasional residual perilipin-1
+ adipocytes. INS
+ cells co-expressed PDX1, NKX6.1, and mature β-cell markers MAFA and NKX2.2. Approximately 18.63% of the cells within the grafts were INS
+NKX6.1
+ cells, which was slightly higher than that observed before pre-transplantation. The co-localization of INS and human-specific hNuMA confirmed the human origin of the grafts (
Figs. 7(a) and
(b)).
scRNA-seq comparing the RMF-islet organoids before and 12 weeks after transplantation confirmed the presence of all four major pancreatic endocrine cell types and a stable cellular composition over time (
Figs. 7(c) and
(d), Fig. S12 in Appendix A). Transcriptional maturation was indicated by the upregulation of certain functional genes, including
INS,
NKX2.2,
ENUROD1,
UCN3, CPE, G6PC2, CHGA, PCSK1, IAPP, and
SLC30A8 (
Fig. 7(d)), suggesting
in vivo maturation.
Postmortem histological analysis confirmed complete STZ-induced β-cell destruction (Fig. S13(a) in Appendix A), validating the diabetic model. No evidence of β-cell regeneration was observed in the native pancreas after transplantation (Figs. S13(b)–(d) in Appendix A). Importantly, no signs of tumorigenesis were detected in the major organs at 13 weeks post-transplantation (Fig. S13(e) in Appendix A), demonstrating the efficacy and safety of the RMF-islet organoid therapy. Taken together, these results demonstrated that RMF-derived islet organoids undergo rapid vascularization, achieve long-term glycemic control, and remain functionally stable and safe in diabetic mice.
3.6. Generation of neural-like tissues from RMF pellets in vitro
To assess whether RMF tissues can be converted into ectodermal derivatives, we generated neural-like tissues from the RMF pellets. In this protocol, 1 mm RMF pellets were first differentiated into neurospheres exhibiting a neural stem cell (NSC) phenotype, followed by induction into neuronal and neuroglial lineages using sequential exposure to specialized differentiation media (Fig. S14(a) in Appendix A). After two weeks of culture in serum-free medium supplemented with fibroblast growth factor-2 (FGF-2) and epidermal growth factor (EGF), immunofluorescence analysis revealed widespread expression of Nestin, an intermediate filament protein and canonical NSC marker, within the RMF pellets. A subpopulation of cells also expressed the proliferation marker Ki67 and neurogenic transcription factors including SOX2, PAX6, and FOXG1 (Fig. S14(b) in Appendix A), confirming the acquisition of an NSC-like phenotype. Upon subsequent culture in neuronal differentiation medium, these NSC-like cells expressed mature neuronal markers, such as neuronal nuclei (NeuN), neuron-specific class III β-tubulin (TUJ1), and microtubule-associated protein 2 (MAP2) (Fig. S14(c) in Appendix A). When cultured in a neuroglial differentiation medium, the cells progressively expressed glial-lineage markers including glial fibrillary acidic protein (GFAP) and S100 calcium-binding protein β (S100β) (Fig. S14(d) in Appendix A). qRT-PCR analysis supported these immunofluorescence findings, confirming the upregulation of key transcriptional factors and also lineage-specific genes at various stages. At the NSC stage: PAX6, SOX2, FOXG1, and Nestin were upregulated (Fig. S14(e) in Appendix A). During neurogenesis: neuron-specific enolase (NSE), MAP2, TUJ1, neurofilament light chain (NEFL), and neurofilament medium chain (NEFM) were upregulated (Fig. S14(f) in Appendix A). During gliogenesis: GFAP and S100B were upregulated (Fig. S14(g) in Appendix A). Collectively, these findings demonstrate the successful in vitro conversion of human adult adipose tissues into neural-like tissues using RMF pellet-derived intermediates, supporting the ectodermal differentiation potential of RMF tissues.
4. Discussion
The prevailing strategy for organoid generation involves differentiation of dissociated stem cells through complex in vitro manipulations, which limits their clinical translation. Herein, we have demonstrated a straightforward, safe, and scalable approach for the direct generation of functional organoids and tissues from human adult adipose tissue. These include mesodermal humanized bone marrow organoids that support normal human hematopoiesis in immunodeficient mice following the xenotransplantation of human HSCs, endodermal INS-producing organoids that reverse hyperglycemia in STZ-induced diabetic mice, and neuroectodermal neural-like tissues that exhibit neuronal and neuroglial phenotypes. Notably, these differentiation processes occur entirely at the tissue level in vitro, without intermediate stem cell isolation or genetic manipulation, representing a departure from conventional cell-based methodologies. Collectively, our findings highlight human adult adipose tissue as a simple, abundant, and clinically relevant source of diverse therapeutic organoids.
In this study, we have hypothesized that preserving tissue integrity might help retain ASCs within native-like microenvironments, potentially enhancing their trans-differentiation potential across germ layer boundaries. We selected human adult adipose tissue as the model because of its accessibility, minimal ethical constraints, and known regenerative potential [
38,
39]. Previous studies have suggested that even terminally differentiated adipocytes can dedifferentiate under specific stimuli [
40,
41] and that intact tissue architectures support stem cell heterogeneity and function [
42,
43]. In our system, the microfat tissues cultured in suspension retained their histological architecture, likely mimicking a regenerative niche. This was supported by the enhanced trilineage differentiation capacity and higher pseudo-time values observed in RMF pellets than in conventional 2D-expanded ADSCs. Further studies are required to determine whether this effect is primarily driven by preserved niche cues, paracrine interactions, or intrinsic reprogramming mechanisms.
We also successfully generated humanized bone marrow organoids in mice that recapitulated the critical features of the native human bone marrow niche, including the endosteal, perivascular, and hematopoietic components [
8,
20,
44]. Previous strategies have relied on co-culturing human MSCs within matrices [
19], which can form an ectopic humanized bone marrow niche when transplanted into immunodeficient mice [
22,
44,
45]. We employed a developmental bioengineering approach to generate hOssicles from RMF tissues. In this approach, RMF pellets were first induced into a hypertrophic cartilage phenotype
in vitro, and subsequently transplanted into NPSG mice for further maturation by recapitulating endochondral ossification. By tracing the cell origin, we confirmed that the endosteal niche primarily originated from human cells, supporting the establishment of a humanized bone marrow microenvironment. Upon xenotransplantation of the hUCB-CD34
+ cells, these organoids supported multilineage hematopoiesis. To the best of our knowledge, this is the first report of humanized bone marrow organoids derived directly from adult adipose tissue. These organoids may serve as valuable models for studying hematopoiesis, leukemia, and bone metastasis in solid tumors.
Following mesodermal lineage success, we have hypothesized that altering the niche environment through defined biochemical and biophysical cues could redirect RMF tissues across germ layer boundaries. Indeed, prior studies have shown that mesodermal ADSCs can transdifferentiated into endoderm-derived hepatocytes [
[46],
[47],
[48]] and β-like cells [
49,
50], as well as ectoderm-derived neurons [
51,
52], Schwann cells [
53,
54], keratinocytes [
55,
56], and corneal epithelial cells [
57]. Building on this concept, we were able to successfully convert RMF tissue into INS-producing islet organoids. Currently, islet organoids are primarily derived from PSCs, encompassing both embryonic stem cells [
30,
37,
58] and induced PSCs [
31,
32], or transdifferentiated from various lineages of cells including human fibroblasts [
33,
59,
60], pancreatic exocrine acinar cells [
61], liver cells [
62], and gastric stem cells [
9]. As an alternative approach to this paradigm, we converted RMF pellets into islet organoids that have a cellular composition similar to that of native islets and displayed a biphasic glucose-dependent INS secretion capacity
in vitro. Notably, the conversion of RMF tissues into islet organoids occurred without transitioning through a pluripotent state, a step associated with genomic instability risks, distinguishing them from generation of β-like cells from fibroblasts [
33,
59,
60] or ADSCs [
63,
64] by introducing exogenous lineage-specific transcription factors or exposure to chemical compounds. When transplanted into diabetic mice, they ameliorated hyperglycemia towards normal levels and glycemic improvement was maintained throughout the study period. These findings highlight their potential as a therapeutic platform for the treatment of diabetes.
We further demonstrate the generation of neuroectodermal tissues from RMF tissues. Upon exposure to specific neural-inductive media, the RMF pellets expressed neuronal and glial markers, suggesting their potential utility in neural repair. However, future studies are needed to assess their electrophysiological properties and synaptic functionality to validate their maturation and network integration potential. Such assays are essential for evaluating their application in neural regeneration and disease modeling.
In addition to its biological advantages, the RMF-based strategy offers substantial practical improvements over the conventional ASC/PSC-based organoid generation approaches. Traditional protocols typically involve enzymatic dissociation, long-term 2D expansion, reprogramming, and 3D spheroid induction, which are labour-intensive, expensive, and variable. In contrast, our strategy only required minimal tissue processing without genetic manipulation. This streamlined workflow improves scalability, reproducibility, and clinical applicability. However, as a newly developed organoid generation strategy, this study has several limitations. ① The heterogeneity of human adipose tissue samples from different donors introduces viability, which poses challenges to the generation of consistently functional organoids. Donor screening and standardization measures, including strict inclusion criteria, uniform processing protocols, and biological replicates, ensured the consistent generation of RMF pellets and downstream lineage-specific organoids. Our previous large-cohort analysis [
65] further demonstrated that the key cellular subpopulations relevant to organoid formation remained relatively stable within the demographic range represented in our study. ② The differentiation efficiency of our adipose tissue-based protocol is lower than that of PSC-based approaches. Future studies will optimize the differentiation medium and condition combinations to enhance the differentiation efficiency. ③ Elucidating the mechanisms underlying the role of the adipose niche in RMF tissues during multilineage differentiation is crucial. Dissecting the molecular, cellular, and spatial components of adipose niches during microfat tissue self-reaggregation and ADSC differentiation could significantly advance this technology towards clinical application. In addition to these experimental limitations, several translational challenges remain before RMF-derived organoids can be clinically applied. These include the need for long-term
in vivo validation of both safety and functionality, standardization under GMP-compliant manufacturing, and robust quality control frameworks to ensure batch consistency. Furthermore, navigating the evolving regulatory landscape of tissue-derived biologics, including the classification, licensing, and clinical trial requirements, is critical for moving this technology toward real-world therapeutic use. Addressing these challenges will be the focus of future translational studies.
5. Conclusions
This study presents a robust and versatile strategy for directly generating functional organoids spanning all three germ layers from human adult adipose tissue. Through simple emulsification and suspension culture, microfat tissues self-organize into RMF pellets, preserving the stem/progenitor diversity and multilineage potential. These RMF-derived constructs gave rise to bone marrow organoids that supported human hematopoiesis in mice, INS-secreting islet organoids that reversed diabetes, and neural-like tissues that express neuronal and glial markers. By bypassing cell isolation, prolonged expansion, and genetic reprogramming, our method overcomes the major limitations associated with PSC-based approaches. This strategy offers a scalable, autologous, and clinically translatable platform for regenerative medicine, disease modeling, and cell therapy development. Furthermore, our findings have revealed the remarkable regenerative plasticity of adult adipose tissue and established RMF culture as a powerful tool for exploiting this capacity. Future studies should focus on refining lineage specificity and advancing this platform toward clinical-grade applications.
CRediT authorship contribution statement
Ru-Lin Huang: Writing – original draft, Project administration, Investigation, Data curation, Writing – review & editing, Visualization, Methodology, Funding acquisition, Conceptualization. Jing Yang: Software, Investigation, Methodology. Yuxin Yan: Software, Methodology, Investigation. Xiangqi Liu: Methodology, Investigation. Xiya Yin: Methodology, Investigation. Chuanqi Liu: Methodology, Investigation. Xingran Liu: Methodology, Investigation. Rehanguli Aimaier: Methodology, Investigation. Qiumei Ji: Methodology, Investigation. Gen Li: Software. Tao Zan: Investigation, Methodology. Kang Zhang: Writing – original draft, Investigation, Writing – review & editing, Methodology. Qingfeng Li: Writing – review & editing, Project administration, Conceptualization, Resources, Funding acquisition.
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This study is supported by the National Natural Science Foundation of China (82372535 to Ru-Lin Huang and 82361138568 to Qingfeng Li), the Shanghai Clinical Research Center of Plastic and Reconstructive Surgery supported by Science and Technology Commission of Shanghai Municipality (22MC1940300), and the Shanghai Plastic Surgery Research Center of Shanghai Priority Research Center (2023ZZ02023).
We thank Dr. Haoran Wang for invaluable support.
Appendix A. Supplementary data
Supplementary data to this article can be found online at
https://doi.org/10.1016/j.eng.2025.06.031.
PDF
(7080KB)
