Functional human salivary gland organoids for tissue regeneration in chemically defined culture systems

Summary

Regenerative therapies for salivary gland dysfunction remain unavailable. We establish a fully chemically defined, xeno-free three-dimensional culture system to generate functional human salivary gland organoids. These organoids recapitulate the typical glandular architecture and display multilineage cellular composition, supporting long-term expansion with high transcriptomic fidelity to the primary tissue. The organoids exhibit key salivary functions, including glycoprotein secretion, amylase expression, and calcium flux in response to cholinergic stimulation. Single-cell transcriptomic analysis reveals preserved epithelial heterogeneity within the organoids and indicates a basal-to-ductal-to-acinar bifurcated differentiation trajectory. In the orthotopic transplantation model using non-obese diabetic mice with Sjögren syndrome, the organoids significantly improve salivary secretion. In the ectopic subrenal capsule transplantation model using immunodeficient mice, the organoids achieve glandular tissue reconstruction and vascularization. This study establishes a robust, functionally mature, and clinically translatable human salivary gland organoid system as a platform for tissue regeneration therapies targeting salivary hypofunction disorders such as Sjögren syndrome.

Graphical abstract

Highlights

• •Chemically defined culture generates functional human salivary gland organoids
• •Organoids preserve epithelial heterogeneity and multilineage differentiation
• •Organoids exhibit salivary secretion, amylase expression, and calcium signaling
• •Transplanted organoids restore salivary secretion and reconstruct glandular tissue

Zhang et al. establish a chemically defined system that generates functional human salivary gland organoids. These organoids preserve epithelial diversity, perform key secretory functions, and restore salivary secretion and glandular structure after transplantation, providing a translational platform for treating salivary hypofunction.

Introduction

As a vital exocrine organ in the human body, the salivary glands perform essential functions, including lubricating the oral cavity, cleansing the mouth, aiding in digestion, and contributing to immune defense.¹ However, due to head and neck radiation therapy, salivary gland tumor resection surgery, autoimmune diseases, and aging, among other factors, salivary gland function is frequently compromised or diminished, leading to the occurrence of xerostomia.^1,2,3,4,5,6 Xerostomia is often accompanied by complications, such as difficulty chewing and swallowing, speech impediments, and taste loss, which significantly impact patients’ quality of life and physical health.⁷ Currently, treatment methods for xerostomia can only provide temporary relief with limited efficacy.^4,5,8

With the advancement of organoid culture techniques, salivary gland organoids have emerged as a promising approach for salivary gland tissue regeneration.^{9,10,11,12,13} Previous studies have demonstrated that salivary gland organoids can be generated from various types of stem cells, including embryonic stem cells,¹⁴ induced pluripotent stem cells,¹⁵ and salivary-gland-derived stem/progenitor cells.^16,17,18 However, the use of embryonic and induced pluripotent stem cells may pose potential risks such as genetic mutations, teratogenicity, and immune rejection, which are not conducive to clinical translation.¹³ Therefore, using salivary-gland-derived stem/progenitor cells from adult tissues as a foundation for salivary gland organoid culture has distinct advantages.

In previous studies,^16,17,18 culture medium quality was found to be a critical factor when utilizing stem/progenitor cells derived from human or animal salivary glands for organoid culture. The currently employed culture media commonly include basic media such as Dulbecco’s modified Eagle medium (DMEM) and RPMI-1640, along with supplements like bovine pituitary extract, fetal bovine serum, epidermal growth factor (EGF), fibroblast growth factor (FGF), transforming growth factor β, Wnt3a, R-spondin1, and other cytokines.^{13,14,15,16,17} However, challenges persist regarding the accuracy of growth factor concentrations and combinations in culture medium, along with concerns regarding the undefined components of some animal-derived reagents, which render the resultant organoids less suitable for translational research. Furthermore, three-dimensional (3D) cell culture typically relies on scaffold-dependent methods, with commonly used support structures including Matrigel, fibronectin, collagen types I and IV, laminin, and gelatin, among others. This method is widely employed, as it effectively mimics the physicochemical properties of the extracellular matrix, providing a suitable environment for cell aggregation, proliferation, and differentiation. Optimal pore size also enhances the efficacy of growth factors in culture medium.¹⁹ Nevertheless, challenges persist regarding the compositional clarity of these materials, thereby limiting the application of salivary gland organoids in tissue regeneration and translational research. Furthermore, researchers have explored the use of artificially synthesized hydrogels as scaffolds for salivary gland tissue engineering, using materials such as polyethylene glycol, polyglycolic acid, and polylactic acid to culture cell lines and bioprint cells.^20,21 However, challenges remain in ensuring cell viability, aggregation, growth, and polarization, thus necessitating further investigation.²² Furthermore, some studies have attempted scaffold-free 3D culture methods, aiming to facilitate the self-organization of cell aggregates without scaffold support, using techniques such as rotating flasks, biopolymer-coated plates, hanging drop microplates, ultra-low attachment plates, and organ chips.^22,23,24,25 However, to date, such research remains lacking in the field of salivary gland organoids.

In this study, we aimed to optimize the culture conditions to establish a long-term culture system for human salivary gland organoids (hSGOs) with a well-defined chemical composition. We also characterized the obtained organoids and transplanted them to determine whether they could generate structurally mature and functional salivary glands and restore damaged salivary gland function in vivo.

Results

Optimization of culture conditions enables hSGO formation

In a prior study, human submandibular gland (SMG) epithelial stem/progenitor cells were isolated and differentiated in vitro into primary salivary gland organoids using a keratinocyte medium with supplements (KM-S).¹⁶ To establish a robust 3D system for hSGOs, we compared KM-S with a chemically defined complete medium (CM) under two culture strategies (Matrigel-embedded vs. Matrigel-free) (Figure 1A). Organoids grown in CM consistently displayed more cohesive spheroid morphology and clearer epithelial organization than those in KM-S, with the Matrigel-free CM condition showing the most uniform structural appearance (Figure 1B). Hematoxylin and eosin (H&E) staining further demonstrated that CM-cultured organoids exhibited well-defined luminal spaces and more organized epithelial layers, whereas KM-S-derived organoids tended to show looser and less organized tissue architecture (Figure 1C). To complement these qualitative observations, we quantified organoid size and lineage marker intensities across conditions (Figures S1A and S1B). These observations indicate that CM substantially improves epithelial organization and that the absence of Matrigel further enhances acinar-like architecture and luminal clarity. Consistent with this, immunofluorescence analysis showed stronger expression of lineage markers (AQP5, K19, EpCAM, and vimentin) under CM culture (Figure 1D). Notably, AQP5 displayed sharper apical enrichment in the Matrigel-free CM condition, as further illustrated by 3D reconstruction from confocal z stack imaging (Videos S1 and S2), supporting improved polarization and acinar maturation in the absence of Matrigel. Overall, these results demonstrate that CM provides optimal growth support, while the Matrigel-free CM condition yields the most physiologically relevant epithelial polarity and differentiation and thus represents the most suitable strategy for generating functional hSGOs.

Figure 1.

Establishment and optimization of the hSGO culture system

(A) Schematic illustration of the culture workflow. Human salivary gland tissues were enzymatically dissociated and seeded into either keratinocyte medium (KM) or complete medium (CM) with or without Matrigel.

(B) Bright-field images of organoids cultured under four conditions. Scale bars, 200 μm.

(C) Hematoxylin and eosin (H&E) staining showing histological features of organoids under different conditions. Scale bars, 100 μm.

(D) Immunofluorescence staining for AQP5, K19, EPCAM, and Vimentin in organoids. Scale bars, 50 μm.

(E) Bright-field images of organoids grown in complete CM or CM lacking single components. Scale bars, 500 μm.

(F) Organoid formation efficiency on day 7 in each condition relative to the CM control (n = 3).

Data are presented as mean ± SD. Statistical significance was determined using one-way ANOVA. ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001.

See also Figures S1 and S2; Tables S1 and S2.

Optimization of culture medium components for salivary gland organoid formation and differentiation

To refine the chemically defined CM, we generated a series of single-factor omission variants to evaluate the role of each component. Brightfield imaging showed that removal of A83-01, EGF, B27, Wnt-3a, Noggin, or Niacinamide resulted in fewer or morphologically abnormal organoids, whereas omission of DEX, FGF10, R-spondin 1, or N-acetylcysteine produced milder but still detectable changes (Figure 1E). Quantification of organoid-forming efficiency showed significantly reduced organoid formation in each of these omission groups compared with that in the complete CM condition (Figure 1F), suggesting that these factors are indispensable for organoid growth and morphogenesis. Time course analysis further demonstrated sustained impairment in growth dynamics when any of these components was removed, based on normalized viability measurements across 14 days (Figure S1C).

Further, quantitative reverse transcription PCR (RT-qPCR) analysis revealed marked alterations in the expression of multiple developmental and functional marker genes in salivary gland organoids upon removing specific factors from CM (Figure S2A). The key developmental transcription factors SOX9 and SOX10 were significantly downregulated in the absence of FGF10 (p < 0.05), consistent with the established role of the FGF10-FGFR2b axis in maintaining distal epithelial identity and supporting acinar-lineage maturation. R-spondin1 depletion similarly reduced SOX9 and SOX10 expression, indicating that canonical Wnt signaling is required to sustain epithelial lineage programs in salivary gland cultures.²⁶ Notably, removal of the antioxidant NAC also suppressed SOX9 and SOX10 (p < 0.05), suggesting that oxidative stress disrupts the stability of SOX9⁺/SOX10⁺ distal epithelial lineages during culture.

The basal/stem cell markers KRT5 and KRT14 were highly sensitive to changes in Wnt signaling and growth factor conditions. In the NAC group, KRT5 expression increased by approximately 4-fold, while KRT14 expression was significantly elevated (p < 0.05), indicating that oxidative stress may suppress differentiation cues or selectively promote the survival of KRT5/14⁺ basal-like cells. Conversely, FGF10 removal significantly reduced KRT5 expression (p < 0.05), suggesting its role in maintaining the basal progenitor cell pool.^14,26 The proliferation marker KI67 followed a similar trend, showing expression changes consistent with the basal cell marker dynamics.

Among differentiation-associated markers, the mature acinar cell marker MIST1 showed the greatest sensitivity to culture factor removal. Its expression was the most significantly reduced under FGF10-deficient conditions (p < 0.001), in line with the known role of FGF10 in driving acinar differentiation. Consistently, AQP5 transcription was also markedly downregulated when FGF10 or R-spondin1 was omitted, as shown in Figure S2A, further supporting the essential roles of these two factors in maintaining acinar identity. In contrast, the ductal marker KRT7 and myoepithelial marker ACTA2 (alpha-smooth muscle actin) were not significantly affected by the removal of any single factor, indicating a more restricted dependency on these signals.^16,17

Together, these findings demonstrate that each CM component plays a specific role in regulating lineage commitment, proliferation, and maturation. Disruption of any key signal skews the balance between progenitor maintenance and differentiation, leading to disorganized organoid development and compromised functionality. A coordinated signaling milieu is thus essential for preserving hSGO integrity in vitro.

Salivary gland organoids closely recapitulate the structural and multicellular architecture of native tissue

H&E staining showed that hSGOs derived from primary human tissue formed compact, glandular-like structures with organized acinar and ductal regions, closely resembling native salivary glands (Figure 2A). Immunofluorescence confirmed the presence of multiple lineages, including EpCAM⁺ epithelial and VIM⁺ stromal-like cells. Within the epithelial compartment, KRT7⁺ ductal cells, αSMA⁺ myoepithelial cells, KRT5⁺ basal cells, and scattered CD133⁺ progenitors were identified, reflecting native-like cellular heterogeneity (Figure 2B). Furthermore, 3D reconstruction from confocal z stack imaging revealed a distinct spatial arrangement, with AQP5⁺ acinar cell clusters surrounding KRT19⁺ ductal epithelial structures, indicating that the organoids self-organize into physiologically relevant architectures (Figure 2C; Videos S1 and S2). Notably, a population of TUBB3⁺/VIM⁺ double-positive cells was also observed, suggesting incorporation of neural-like stromal elements (Figure 2D).

Figure 2.

hSGOs recapitulate major salivary gland cell types and structures

(A) H&E staining of human salivary gland tissues and derived organoids. Scale bars, top: 100 μm; bottom: 50 μm.

(B) Immunofluorescence for EpCAM, Vim, KRT7, αSMA, KRT5, and CD133. Nuclei are stained with 4′,6-diamidino-2-phenylindole (DAPI); red arrows indicate organized epithelial arrangement. Scale bars, 50 μm.

(C) z stack projections show acinar (AQP5⁺) and ductal (KRT19⁺) structures. Scale bars, top: 50 μm; bottom: 100 μm.

(D) Co-expression of neuronal marker TUBB3 and mesenchymal marker Vim. Scale bars, 50 μm.

(E–G) Transmission electron microscopy. (E) Lumenal regions (stars, left) are surrounded by ductal cells bearing apical microvilli (arrows, left), while adjacent myoepithelial cells display flattened nuclei (stars, right) and prominent actin bundles (arrows, right). (F) Submandibular gland (SMG) organoids contain serous granules (electron-dense; arrows, left) and mucous granules (electron-lucent; arrows, right). (G) Typical glandular ultrastructure in parotid gland (PG) and sublingual gland (SLG) tissues. Scale bars, top: 5 μm; bottom: 1 μm.

See also Videos S1 and S2.

Transmission electron microscopy confirmed key ultrastructural features. Ductal-like structures were formed by epithelial cells with microvilli and tight junctions, while surrounding myoepithelial cells exhibited flattened nuclei and cytoplasmic actin bundles encasing acinar domains (Figure 2E). Secretory granules typical of each gland subtype were observed: submandibular hSGOs contained both electron-dense serous and electron-lucent mucous vesicles (Figure 2F), while parotid and sublingual hSGOs exhibited canonical acinar morphology (Figure 2G).

Collectively, histological, immunophenotypic, 3D, and ultrastructural analyses demonstrated that this culture system preserves differentiated acinar populations and enables spatially organized co-existence of ductal, myoepithelial, and stromal-like cells, effectively recapitulating native salivary gland architecture.

Salivary gland organoids recapitulate gland-specific secretory protein expression and functional responsiveness

Immunofluorescence analysis revealed gland-specific secretory protein expression in hSGOs. MUC7, a mucous acinar marker, was detected in SMG- and sublingual gland (SLG)-derived organoids, while AMY1, a serous acinar enzyme, was predominantly expressed in SMG- and parotid gland (PG)-derived organoids (Figure 3A), consistent with their tissue origins. Both MUC7 and AMY1 signals were frequently observed within luminal compartments as well as in acinar-like cells, a pattern that reflects the expected accumulation of secreted proteins and has been similarly reported in native glands and previous organoid studies.¹⁷

Figure 3.

hSGOs exhibit salivary secretory functions

(A) Immunofluorescence staining for MUC7 and AMY1 in tissues and organoids from the SMG, PG, and SLG. Scale bars, 50 μm.

(B) Periodic acid-Schiff (PAS) and Alcian blue-PAS (AB-PAS) staining showing mucin secretion in tissues and organoids. Scale bars, 50 μm.

(C) Time course of AMY1 protein secretion in organoid culture supernatants (n = 4).

(D) Calcium flux assay using Fluo 4-AM following carbachol (CCh) stimulation, recorded over time, with representative fluorescence images shown at indicated time points. Scale bars, 100 μm.

(E) qPCR of KRT7, KRT19, AQP5, and SOX10 after 24 h of CCh stimulation (n = 3).

Data are presented as mean ± SD. Statistical significance was determined using Student’s t test. ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001.

See also Figure S2.

Periodic acid-Schiff (PAS) and Alcian blue-PAS (AB-PAS) staining confirmed abundant mucin-positive cells and luminal secretions, indicating active secretory function (Figure 3B). Enzyme-linked immunosorbent assay further demonstrated time-dependent increases in AMY1 levels in SMG-derived organoid supernatants, confirming sustained secretion during extended culture (Figure 3C).

To evaluate the physiological responsiveness of hSGOs to neuronal stimulation, organoids were acutely treated with the cholinergic agonist carbachol (CCh). This triggered a rapid elevation in intracellular Ca²⁺ levels, reflected by a sharply increased fluorescent intensity, indicating that hSGOs respond to parasympathetic neurotransmitter stimulation by activating calcium-dependent secretory pathways (Figure 3D). Consistent with muscarinic-receptor-mediated signaling, this CCh-induced Ca²⁺ rise was largely abolished by atropine pretreatment, as shown in Figure S2B. To further assess the breadth of physiological responsiveness, we tested additional receptor pathways and found that adrenergic (isoproterenol) and purinergic (ATP) stimulation also induced detectable Ca²⁺ increases in hSGOs (Figure S2C), indicating preserved multi-receptor secretory signaling. After 24 h of CCh stimulation, RT-qPCR revealed upregulation of secretion-related genes, including KRT7, KRT19, AQP5, and SOX10 (Figure 3E).

Together, these results demonstrate that hSGOs maintain tissue-specific secretory protein expression and respond to neuronal stimulation, supporting their utility as a physiologically relevant model for salivary gland function.

hSGOs maintain long-term structural stability and transcriptional consistency

To assess the long-term expansion capacity of hSGOs, we conducted serial passaging and analyzed their morphology, lineage composition, and transcriptomic profiles. Bright-field imaging revealed sustained growth and compact spherical morphology from passage 0 (P0) through passage 4 (P4), with structural stability established by day 14 (Figure 4A). Immunofluorescence confirmed the retention of diverse salivary gland cell populations, including EpCAM⁺, AQP5⁺, K7⁺, K19⁺, αSMA⁺, K5⁺, CD133⁺, and VIM⁺ cells, across passages, indicating multilineage stability (Figure 4B).

Figure 4.

hSGOs maintain stability during long-term expansion

(A) Bright-field images of primary (P0, days 4–14) and serially passaged (P1–P4, day 7) organoids, showing progressively stable morphology. Scale bars, 200 μm.

(B) Immunofluorescence staining of P0 day 14 and P4 day 14 organoids for EpCAM, K7, K19, K5, AQP5, αSMA, vimentin, and CD133. Scale bars, 50 μm.

(C) Principal-component analysis (PCA) of transcriptomes from human salivary gland (SG) tissue, P0, and P4 organoids (n = 3).

(D) Venn diagram of shared or group-specific differentially expressed genes (DEGs) across tissues, P0, and P4.

(E) Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment heatmap highlighting key biological processes related to secretion, cell cycle, and signaling in SG tissue and P0 and P4 organoids.

See also Figure S3.

Principal-component analysis of transcriptomic data indicated that organoids clustered consistently across passages and closely with their tissue of origin. Furthermore, all organoids used for single-cell RNA sequencing (scRNA-seq) and bulk RNA-seq analyses were derived from adult human submandibular glands (SGs) (Figure 4C). Pearson correlation heatmaps further confirmed high transcriptomic similarity between organoids and their corresponding tissues (Figure S3A). Venn diagram analysis revealed that >16,000 genes were commonly expressed among tissues, primary organoids (P0), and fourth-passage organoids (P4), indicating strong preservation of the original tissue transcriptional landscape over time (Figure 4D).

To further evaluate the transcriptional consistency of hSGOs, we compared the transcriptomic profiles of P0 and P4 organoids with those of native salivary gland tissues using scatterplots, volcano plots, gene ontology (GO) enrichment, and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses. Scatterplots showed high overall similarity between organoids and tissues, with relatively few differentially expressed genes (DEGs) (Figure S3B). GO enrichment analysis of these DEGs revealed common enrichment in pathways such as p53 signaling, cell cycle, proteasome, cell adhesion molecules, and bacterial invasion of epithelial cells (Figure S3C). These results suggest that proliferation, stress response, and epithelial-barrier-related functions are retained in both P0 and P4 hSGOs.

Further transcriptomic analysis revealed that P0 and P4 organoids expressed a broad set of key functional genes at levels highly comparable to those in tissue samples. Heatmap clustering highlighted shared expression of genes associated with salivary secretion (MUC7), epithelial adhesion (CDH1 and EPCAM), ductal markers (KRT7 and KRT19), and epithelial integrity (BPIFA2, GSN, and MMRN9) (Figure 4E). Notably, the basal cell marker KRT5 and proliferation-associated gene MKI67 were highly expressed in organoids, suggesting maintained stem/progenitor cell activity during long-term in vitro expansion. Pathway enrichment analysis further supported these findings, showing strong conservation of biological processes such as “cellular community-eukaryotes,” “signal transduction,” and “cell motility” between organoids and tissues. In contrast, certain secretory-function-related pathways, including the “digestive system,” along with key secretory genes such as AQP5 and KLK1, were upregulated in tissues but expressed at lower levels in organoids, suggesting partial loss or delayed maturation of full secretory function (Figure 4E).

Together, these results demonstrate that hSGOs maintain stable morphology, multilineage composition, and transcriptional fidelity over prolonged culture and multiple passages, offering a robust and scalable model system for downstream functional studies.

Single-cell RNA sequencing reveals epithelial heterogeneity and lineage differentiation trajectories in human salivary gland tissues and organoids

To systematically investigate the epithelial heterogeneity and developmental trajectories in human salivary gland tissues and their derived organoids, we performed scRNA-seq on adult SMG tissues (T group) and their corresponding organoids derived from the same donors (O group). In total, 14,385 cells were profiled from tissue samples and 10,670 cells from organoids, among which 5,602 and 9,737 were identified as epithelial cells, respectively (Figures S4A and S4B).

Uniform-manifold-approximation-and-projection-based dimensionality reduction revealed six major epithelial cell clusters in the tissue group (Figure 5A). In contrast, the organoid group displayed three additional proliferative subtypes, resulting in a total of nine distinct epithelial subpopulations (Figure 5B).^17,26,27,28 Based on canonical marker gene expression, these subtypes were annotated as follows: basal cells, proliferative basal cells, ductal cells, proliferative ductal cells, KLK1⁺ ductal cells, and their proliferative counterparts, mucous acinar cells, serous acinar cells, and myoepithelial cells. The identities of these clusters were defined by representative marker genes and further validated using DotPlot analysis (Figures 5C and 5E; Figures S4C–S4I). Specifically, basal lineage cells expressed KRT5 and KRT15, whereas their proliferative subtypes co-expressed MKI67 and PCLAF; ductal cells were characterized by KRT7, KRT19, and KLK1 expression. In acinar lineages, mucous acinar cells highly expressed AQP5, TFF3, and BPIFA2, whereas serous acinar cells expressed ODAM, CRISP3, and PIP.^17,27 Myoepithelial cells specifically expressed MYL9. The relative abundance of each epithelial subtype differed significantly between tissues and organoids (Figure 5D). Organoids were enriched in proliferative and progenitor-like clusters, whereas the proportion of mature acinar cells was notably lower than in tissues. This likely reflects a dynamic equilibrium between stem/progenitor cell expansion and differentiation within the organoid culture environment.

Figure 5.

Single-cell RNA sequencing reveals epithelial heterogeneity and lineage trajectories in salivary gland tissues and hSGOs

(A and B) Uniform manifold approximation and projection (UMAP) plots of epithelial cells from adult SG tissues and hSGOs, identifying major clusters including basal, ductal, KLK1⁺ ductal, mucous and serous acinar, and myoepithelial populations.

(C) Feature plots showing the expression of canonical markers: KRT7/KRT19 (ductal), KRT5/KRT15 (basal), AQP5/TFF3/BPIFB2 (mucous acinar), ODAM/CRISP3/PIP (serous acinar), KLK1/SMIM5 (KLK1⁺ duct), MKI67/PCLAF (proliferative), and MYL9 (myoepithelial).

(D) Relative composition of epithelial subtypes in tissue and organoids.

(E) Dot plot summarizing marker gene expression across clusters (dot size: % of expressing cells; color: expression level).

(F) Pseudotime analysis reveals a differentiation trajectory from basal cells toward ductal and acinar branches.

(G and H) Trajectory trees illustrate bifurcation into ductal and acinar lineages, with KLK1⁺ and myoepithelial cells near transition states.

See also Figures S4 and S5.

We next examined the proliferative clusters in greater depth. Pathway enrichment analysis showed that the basal proliferation, duct proliferation, and KLK1 duct proliferation clusters were strongly enriched in cell cycle and DNA replication pathways (Figure S5A), consistent with a proliferative and incompletely differentiated transcriptional state.

To explore lineage relationships, we performed pseudotime trajectory inference, which revealed a bifurcating transcriptional continuum in which basal-lineage cells occupied early positions, followed by divergence toward ductal-like and acinar-like expression profiles (Figures 5F–5H; Figure S5B). These trajectories represent computationally inferred transcriptional progressions rather than experimentally validated lineage pathways. Within this framework, KLK1⁺ ductal and myoepithelial clusters were positioned in intermediate regions along the trajectory, consistent with transitional transcriptional states.²⁹

CytoTRACE analysis further supported this interpretation. Basal and myoepithelial clusters exhibited lower predicted differentiation scores, indicating relatively immature transcriptional characteristics within the epithelial hierarchy (Figure S5C). In accordance with these findings, PAGA connectivity mapping placed the three proliferative clusters between basal and mature ductal/acinar compartments, suggesting that they occupy intermediate transcriptional positions rather than terminal locations within the lineage continuum (Figure S5D).

In summary, hSGOs faithfully recapitulate the major epithelial lineages of native tissue in terms of marker gene expression and exhibit a more complex epithelial hierarchy with distinct progenitor and transitional populations. These findings highlight the high degree of biomimicry achieved in organoids, while revealing distinct heterogeneity and stem-like features under in vitro conditions, providing a robust model for studying salivary gland epithelial development, regeneration, and disease mechanisms.

Orthotopic transplantation of hSGOs significantly improves salivary gland function and achieves structural integration in non-obese diabetic mice

To evaluate the in vivo regenerative potential of hSGOs, we orthotopically transplanted them into the SMGs of 14-week-old NOD mice, a model of Sjögren syndrome, and conducted systematic assessments at 7 weeks post-transplantation (Figure 6A). Compared with the phosphate-buffered saline-injected control group, hSGO-transplanted mice showed significantly increased stimulated salivary flow (Figure 6B, p < 0.05), indicating restored secretory function of salivary glands. Histological analysis revealed no significant difference in the number of lymphocytic foci between the two groups (Figure 6C), suggesting that the functional improvement was mainly caused by epithelial restoration rather than inflammation resolution. Protein expression analysis further supported these findings: western blot results showed marked upregulation of the acinar water channel protein AQP5 in the hSGO-transplanted group (Figure 6D, p < 0.01), implying enhanced acinar activity. GFP-tracing experiments demonstrated that the transplanted hSGOs could survive in vivo and integrate into the host ductal network (Figure 6E), with GFP⁺/KRT19⁺ human epithelial cells clearly embedded within murine ductal structures, indicating efficient engraftment and structural fusion. To further confirm the presence of human-derived epithelial cells within transplanted glands at the histological level, we performed immunohistochemistry using the human-specific cytoplasmic marker STEM121. STEM121⁺ cells were detectable within ductal epithelial regions of the grafted salivary glands, corroborating the engraftment pattern observed in GFP/KRT19 staining (Figure S6A).To further investigate the molecular basis of this functional integration, proteomic profiling was performed. In the saliva of transplanted mice, a total of 305 human salivary proteins were detected, including 93 human-specific proteins and 212 shared human-mouse proteins (Figure 6F). Mass spectrometry confirmed the presence of multiple human-specific peptides, such as GPSSVEDIK (Figure 6G), demonstrating the secretory activity of transplanted hSGOs.

Figure 6.

Orthotopic transplantation of hSGOs restores salivary function in non-obese diabetic mice

(A) Schematic illustration of the transplantation procedure. hSGOs were transplanted into the SMGs of 14-week-old non-obese diabetic (NOD) mice and analyzed after 7 weeks

(B) Stimulated salivary flow rates significantly increased in the hSGO group compared with the phosphate-buffered saline (PBS) control (n = 6, ∗p < 0.05).

(C) H&E staining and quantification of lymphocytic foci showed no significant difference between the PBS and hSGO groups (n = 6). Scale bars, left: 1 mm; right: 100 μm.

(D) Western blotting revealed upregulation of AQP5 in the hSGO group (n = 6, ∗∗p < 0.01).

(E) Left: GFP⁺ hSGOs in vitro. Scale bars, 200 μm; middle: whole-mount GFP signal in cleared SMG post-transplantation. Scale bars, 2 mm; right: two-photon imaging shows integration of human KRT19⁺ cells into host ductal structures. Scale bars, 20 μm.

(F) Venn diagram of human and mouse salivary proteins detected in transplanted mouse saliva.

(G) Mass spectrometry identified specific human-derived peptide fragments in post-transplantation saliva.

Data are presented as mean with individual data points. Statistical significance was determined using one-way ANOVA (B) and Student’s t test (D). ∗p < 0.05 and ∗∗p < 0.01.

See also Figure S6.

Collectively, these results indicate that hSGOs survive and structurally integrate with host tissues in vivo as well as restore salivary secretion at a functional level, thereby validating their regenerative potential for treating salivary gland dysfunction in disorders such as Sjögren syndrome.

hSGOs form gland-like structures and achieve vascularized reconstruction after subcapsular kidney transplantation in mice

To evaluate the tissue-construction potential of hSGOs, we transplanted them alone or in combination with E12.5 murine salivary gland mesenchyme into the subcapsular space of immunodeficient NPG mice (Figure 7A; Figure S6B). Tissues were harvested at 4-, 8-, and 12-weeks post-transplantation. Macroscopic examination revealed that hSGOs formed visible tissue masses under the kidney capsule, with a notable increase in graft volume observed at 8–12 weeks, suggesting their capacity for sustained growth in vivo (Figure 7B). In contrast, the mesenchyme-only group developed only a small amount of white-fibrous-like tissue without glandular structures (Figure S6C).

Figure 7.

Subcapsular transplantation of hSGOs enables glandular-like morphogenesis and vascularized regeneration

(A) Schematic illustration of transplantation strategy. hSGOs were transplanted alone or with E12.5 mouse salivary mesenchyme under the kidney capsule of NSG mice and analyzed at 4, 8, and 12 weeks

(B) Macroscopic grafts showed progressive growth over time. Scale bars, 500 μm.

(C) At 4 weeks, H&E and AB-PAS staining revealed epithelial-like structures with mucinous secretions. Immunofluorescence confirmed the presence of KRT19⁺, AQP5⁺, αSMA⁺, and CD31⁺ cells. Scale bars, 50 μm.

(D) At 8–12 weeks, co-transplanted grafts exhibited more mature glandular structures and enhanced acinar differentiation (AQP5⁺ and MUC7⁺) compared to hSGO-alone grafts. Scale bars, 50 μm.

See also Figure S6.

Histological analysis further validated these findings. H&E and Masson staining showed that the mesenchyme-only grafts mainly comprised collagenous fibrous tissue lacking epithelial organization (Figure S6D). By 4 weeks post-transplantation, hSGOs had developed organized epithelial-like structures beneath the kidney capsule. H&E and AB-PAS staining revealed the presence of mucin-like secretory material, indicating acquisition of glandular characteristics (Figure 7C). Immunofluorescence staining also revealed diverse epithelial cell types, including KRT19⁺ duct-like cells, AQP5⁺ acinar-like cells, and αSMA⁺/CD31⁺ myoepithelial and endothelial cells, indicating early epithelial organization and vascular integration. Immunohistochemistry for the human-specific marker STEM121 confirmed the presence of human cells in the grafts (Figure S6E).

By 8–12 weeks (Figures 7D; Figure S6F), the grafted tissues showed more mature and complex epithelial architectures. In the hSGO-alone group, KRT19⁺ ductal structures predominated, whereas co-transplantation with mesenchyme markedly promoted the emergence of AQP5⁺ and MUC7⁺ acinar-like clusters, suggesting enhanced secretory differentiation driven by mesenchymal cues. Consistent with this, AMY1-AQP5 co-localization further confirmed the formation of polarized, secretion-competent acinar-like units, supporting the mesenchyme-dependent enhancement of acinar maturation. Notably, in the kidney capsule, a site lacking a natural excretory duct, some secretory acinar-like regions developed fluid-filled luminal cavities, resembling “reservoirs.” This suggested that hSGOs retained secretory activity in ectopic conditions and, supported by surrounding vasculature, continuously produced saliva-like fluids. Upon formation of a drainage connection with the host system, these structures may further mature into fully functional glandular units.

Collectively, these findings indicate that hSGOs can survive stably in vivo as well as undergo multilineage epithelial differentiation under the influence of mesenchymal signals. Moreover, they can generate gland-like architectures with functional secretory and vascular features in ectopic environments, supporting their potential for structural and functional reconstruction in salivary gland tissue engineering and regenerative medicine.

Discussion

In this study, we established a fully chemically defined, xeno-free 3D culture system for generating hSGOs, which exhibits significantly enhanced culture stability and reproducibility, providing a reliable foundation for translational applications. The resulting hSGOs can be expanded long-term while maintaining multilineage composition and functional properties closely resembling those of their tissue of origin, both structurally and transcriptomically. Importantly, this system eliminates the reliance on animal-derived matrices such as Matrigel. Several recent salivary gland organoid studies have relied on Matrigel-based or partially defined culture systems primarily for disease modeling,^17,18 but these approaches still depend on xenogeneic matrices and lack fully controlled biochemical composition. In contrast, our chemically defined and Matrigel-free system provides a standardized and regeneration-oriented platform suitable for mechanistic and translational studies.

Through systematic optimization, FGF10, R-spondin1, and NAC were identified as essential factors that jointly maintain robust SOX9⁺/SOX10⁺ epithelial identity and enhanced acinar differentiation. FGF10, a key regulator of distal epithelial populations, has been well documented in salivary gland development models.^30,31 R-spondin1 acts by potentiating Wnt signaling to preserve epithelial stemness, while NAC alleviates oxidative stress and supports stem cell viability during prolonged culture.^32,33 The synergistic action of these factors fosters a homeostatic niche that balances proliferation and differentiation. In conditions lacking these components, organoid formation efficiency decreased and lineage skewing occurred, highlighting the complex interplay of these pathways in regulating cell fate and tissue organization. Beyond identifying their individual roles, our findings provide a mechanistic framework for further optimizing organoid maturation strategies.

One of the major breakthroughs of this study is the improved maintenance and differentiation of acinar cells. The developed hSGOs retained gland-specific features from different origins—submandibular, parotid, and sublingual—and enabled clear resolution of two distinct acinar lineages (mucous and serous) at the single-cell level. This represents a clear improvement over prior models, in which acinar cells are often lost or incompletely differentiated during long-term culture.^34,35 As acinar cells are the primary functional unit for saliva secretion, preserving their identity and activity is essential for both mechanistic and translational applications.^12,13 Single-cell transcriptomic analysis revealed that hSGOs faithfully recapitulated the epithelial heterogeneity of native glands, encompassing basal, ductal, myoepithelial, and acinar populations. Notably, a bifurcated differentiation trajectory from basal cells toward either ductal or acinar fates was evident. Within this trajectory, KLK1⁺ ductal cells and myoepithelial cells occupied key branching nodes, suggesting latent stemness or plasticity.^36,37,38 Compared with existing studies using animal models or induced pluripotent stem-cell-derived cell,^14,15,39,40 our findings more closely approximate human salivary gland development and provide additional insights into lineage hierarchy and regeneration mechanisms.

We further validated the regenerative potential of hSGOs in both orthotopic (SMG of NOD mice) and ectopic (renal capsule) transplantation models. In vivo, the organoids survived long-term, integrated structurally with host tissue, and restored secretory function. In NOD mice, transplantation resulted in significantly increased stimulated saliva output, elevated AQP5 expression, and the detection of human-specific salivary proteins. GFP tracing and immunostaining revealed integration with the native ductal system, confirming genuine differentiation and reconstruction potential. In the ectopic setting, hSGOs differentiated into gland-like structures with mucous-rich lumens and signs of sustained secretory activity and vascularization, even in the absence of natural ductal connections. Notably, we identified a subset of cells within the regenerated grafts, which co-expressed both ductal and acinar markers, suggesting the presence of transitional or bipotent epithelial populations.²⁹ This phenotype was particularly prominent in the ectopic renal capsule model, where gland-like structures with mucous-filled cavities emerged despite the lack of native salivary microenvironmental cues. This dual expression pattern likely reflects an intermediate differentiation state, recapitulating developmental or regenerative trajectories of the salivary epithelium.^36,37 These findings are consistent with prior reports indicating that ductal and acinar lineages can originate from common progenitor pools.^1,29,38,41 Thus, hSGOs generate mature epithelial cell types and partially mimic endogenous lineage maturation during in vivo integration. This plasticity underscores the ability of these organoids to interact dynamically with the host niche and supports their translational potential for salivary gland repair and regeneration.

This study has some limitations. Although we assessed pilocarpine-stimulated saliva secretion, lymphocytic infiltration, and human protein output after transplantation, we did not evaluate longitudinal water consumption, an important behavioral indicator in Sjögren syndrome models. Longitudinal assessment of water intake may provide additional insight into functional restoration. The proportion of fully mature acinar cells and their secretory output remain lower than that in native tissue. Some key functional pathways also remain underrepresented, indicating a need for further refining the culture microenvironment and differentiation protocols. Additional microenvironmental cues may further promote functional maturation, as suggested by recent reports.^{10,19,21,26,42} Another challenge is establishing an output duct system. Although the organoids produce saliva-like secretions, they lack defined connections to the oral cavity. Establishing physiologic saliva drainage remains an important consideration for future translational applications.

In summary, this study establishes a chemically defined and xeno-free human salivary gland organoid system that supports long-term expansion, preserves epithelial heterogeneity, and enables functional tissue reconstruction in vivo. Despite the inherent limitations of current in vitro and animal models, the robust reproducibility, functional competence, and translational compatibility of this platform provide a solid foundation for future regenerative applications. These findings advance human salivary gland organoid technology and offer a scalable framework for studying gland development, disease mechanisms, and therapeutic regeneration.

Limitations of the study

This study has several limitations. Although the chemically defined culture system supports long-term expansion and multilineage differentiation of human salivary gland organoids, the in vitro environment cannot fully recapitulate the complex biomechanical, neural, and immune cues present in native salivary glands. Functional characterization primarily relied on secretory markers, calcium signaling, and stimulated saliva output, which represent indirect measures of gland function and do not capture all aspects of physiological saliva regulation. In vivo validation was performed using orthotopic transplantation in NOD mice and ectopic renal capsule models, which differ from human salivary glands in anatomy, immune context, and disease progression, thereby limiting direct clinical extrapolation. In addition, transcriptomic and proteomic analyses provide population-level and molecular insights but do not fully resolve dynamic functional interactions at the single-cell or tissue scale. Addressing these limitations will require further integration of advanced functional assays and more physiologically relevant models.

Resource availability

Lead contact

Further information and requests should be directed to and will be fulfilled by the lead contact, Zhigang Cai (c2013xs@163.com).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •Bulk RNA-seq data have been deposited in the NCBI Sequence Read Archive (SRA) under SRA: PRJNA1294407. Single-cell RNA-seq data generated in this study have been deposited in the Gene Expression Omnibus (GEO) database under GEO: GSE303450. Mass-spectrometry-based proteomics data have been deposited to the ProteomeXchange Consortium via the iProX partner repository^43,44 under iProX: PXD066438.
• •This study did not generate new custom code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon reasonable request.

Acknowledgments

This study was supported by the National Key R&D Program of China (grant no. 2022YFC2504200) and the National Natural Science Foundation of China (grant no. 82470994 and 82301023).

Author contributions

Z.C., Y.S., and X.Z. conceived the study. X.Z. and Y.C. performed most experiments and data analyses. J.Q., X.S., S.X., and Z.C. contributed to the experiments, data acquisition, and/or analysis. X.S., S.X., Y.S., and Z.C. supervised the study. X.Z. wrote the manuscript, which the other authors edited and approved.

Declaration of interests

The authors declare no competing interests.

Declaration of generative AI and AI-assisted technologies in the writing process

During the preparation of this work, the authors used ChatGPT (OpenAI) in order to assist with language polishing, improve clarity of expression and support formatting according to journal guidelines. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
anti-AQP5	Abcam	Cat#ab92320; RRID: AB_2049171
anti-K19	Abcam	Cat#ab7754; RRID: AB_306048
anti-EpCAM	Abcam	Cat#ab223582; RRID: AB_2762366
anti-Vimentin	Abcam	Cat#ab8069; RRID: AB_306239
anti-K7	Abcam	Cat#ab181598; RRID: AB_2783822
anti-α-SMA	Abcam	Cat#ab7817; RRID: AB_262054
anti-K5	Abcam	Cat#ab52635; RRID: AB_869890
anti-CD133	Abcam	Cat#ab264538; RRID: Not available
anti-TUBB3	Abcam	Cat#ab52623; RRID: AB_869991
anti-MUC7	Biorbyt	Cat#orb101843; RRID: Not available
anti-AMY1	Santa Cruz Biotechnology	Cat#sc-46657; RRID: AB_626668
anti-CD31	Abcam	Cat#ab182981; RRID: AB_2920881
anti-Ki67	Abcam	Cat#ab279653; RRID: AB_2934265
anti-STEM121	Takara Bio	Cat#Y40410; RRID: AB_2801314
anti-AQP5 (WB)	Santa Cruz Biotechnology	Cat#sc-514022; RRID: AB_2891066
anti-β-Actin	Biorbyt	Cat#orb10033; RRID: AB_10749201
Alexa Fluor 594-goat anti-rabbit IgG	ZSGB-BIO	Cat#ZF-0516; RRID: AB_2936330
Alexa Fluor 488-goat anti-mouse IgG	ZSGB-BIO	Cat#ZF-0512; RRID: AB_3662677
HRP-goat anti-rabbit IgG	ZSGB-BIO	Cat#ZB-2301; RRID: AB_2747412
HRP-goat anti-mouse IgG	ZSGB-BIO	Cat#ZB-2305; RRID: AB_2747415
Biological samples
Human salivary gland tissues (parotid, submandibular, sublingual)	Peking University School and Hospital of Stomatology	Not applicable
Chemicals, peptides, and recombinant proteins
Collagenase Type I	Worthington Biochemical	Cat#LS004194
Collagenase Type II	Worthington Biochemical	Cat#LS004176
Collagenase Type IV	Worthington Biochemical	Cat#LS004186
A83-01	Sigma-Aldrich	Cat#SML0788
Recombinant human Wnt-3a	Novoprotein	Cat#C18K
Recombinant human Noggin	Novoprotein	Cat#CB89
Recombinant human R-spondin 1	Novoprotein	Cat#CX83
Recombinant human EGF	Novoprotein	Cat#C029
Recombinant human FGF10	Novoprotein	Cat#CR11
Matrigel	Corning	Cat#354234
TrypLE Express	Thermo Fisher Scientific	Cat#12604013
4′,6-diamidino-2-phenylindole (DAPI)	Solarbio	Cat#C0065
Critical commercial assays
CellTiter-Glo 3D Cell Viability Assay	Promega	Cat#G9683
Fluo-4 Calcium Imaging Kit	Thermo Fisher Scientific	Cat#F10489
Human AMY1 ELISA Kit	Elabscience	Cat#E-EL-H6292
BCA Protein Assay Kit	Thermo Scientific	Cat#23227
ECL Ultra Detection Kit	New Cell & Molecular Biotech	Cat#P10100
Nuohai Tissue Clearing Kit	Nuohai Life Science	Cat#NH-CR-210701
Carbachol (CCh)	MCE	Cat#HY-B0276
Atropine	MCE	Cat#HY-B1245
Isoprenaline hydrochloride/ Isoproterenol	MCE	Cat#HY-B0666A
ATP (Adenosine 5′-triphosphate disodium salt)	MCE	Cat#HY-B0586
Pilocarpine hydrochloride	MCE	Cat#HY-B0726
Deposited data
RNA-seq datasets	This paper	SRA: PRJNA1294407
Single-cell RNA-seq datasets	This paper	GEO: GSE303450
Proteomics datasets	This paper	iProX: PXD066438
Experimental models: Organisms/strains
NOD/ShiLtJGpt mice	GemPharmatech	No. N000235
NOD.Cg-Prkdcscid Il2rgtm1Vst/Vst (NPG)	Beijing Vitalstar Biotech	No. VS101001
ICR mice	Department of Laboratory Animal Science, Peking University Health Science Center	Not applicable
Oligonucleotides
qPCR primers	This paper	See Table S1
Software and algorithms
fastp (v0.21.0)	OpenGene	https://github.com/OpenGene/fastp
Cell Ranger (v7.1.0)	10× Genomics	https://www.10xgenomics.com/support/software/cell-ranger
Seurat (v4.1.1)	Satija Lab	https://satijalab.org/seurat/
scran (fastMNN, v1.6.2)	Bioconductor	https://bioconductor.org/packages/scran
Monocle 2 (v2.22.0)	Trapnell Lab	https://cole-trapnell-lab.github.io/monocle-release
Monocle 3 (v1.3.1)	Trapnell Lab	https://cole-trapnell-lab.github.io/monocle3
CytoTRACE (v1.0.0)	Stanford University	https://cytotrace.stanford.edu
DIA-NN (v1.8.1)	Demichev Lab	https://github.com/vdemichev/DiaNN
GraphPad Prism	GraphPad Software	https://www.graphpad.com
ImageJ (v1.8.0)	NIH	https://imagej.nih.gov/ij/
Dr.Tom Bioinformatics Platform	BGI Technology	https://www.bgi.com
NovelBrain Cloud Analysis Platform	NovelBio Co., Ltd	https://www.novelbrain.com

Experimental model and study participant details

Human subjects

Human salivary gland tissues were obtained during clinically indicated surgeries with informed consent and approval from the Institutional Review Board of Peking University School and Hospital of Stomatology (PKUSSIRB-2024103163). Gender was recorded based on clinical information. A total of 49 donors were included (male: 32/49; female: 17/49), with a mean age of 35.2 years (range 8–71 years). Clinical diagnoses comprised both benign (32/49) and malignant (17/49) salivary gland or maxillofacial lesions. Samples were allocated to downstream organoid culture and experimental analyses based on tissue availability and experimental requirements. No gender-stratified analyses were performed due to limited sample size.

Animals

NOD/ShiLtJGpt mice (strain No. N000235) were purchased from GemPharmatech (Nanjing, China). Female mice aged 14 weeks were used for orthotopic submandibular gland transplantation experiments. NOD.Cg-Prkdcscid Il2rgtm1Vst/Vst (NPG; catalog No. VS101001) mice were obtained from Beijing Vitalstar Biotech. Male NPG mice aged 6–8 weeks were used for subcapsular renal heterotopic transplantation experiments. All mice were maintained under specific pathogen–free (SPF) conditions with ad libitum access to food and water and a 12 h light/dark cycle. Animals were randomly assigned to experimental groups. Investigators were not blinded during experimentation or outcome assessment. No sex-specific analyses were performed unless otherwise stated.

Embryonic tissues

Embryonic mesenchyme was isolated from submandibular gland germs of embryonic day 12.5 (E12.5) ICR mouse embryos. Pregnant ICR mice were used as embryo donors, and embryonic tissues were collected in accordance with institutional animal welfare and ethics guidelines. Embryonic sex was not determined at this developmental stage.

Primary cultures and organoids

Human salivary gland organoids were established from primary patient tissues as described in the method details. No established or immortalized cell lines were used in this study. All experiments were performed using primary tissue–derived organoids; therefore, cell line authentication was not applicable. Organoid cultures were routinely tested for mycoplasma contamination using PCR-based methods and were confirmed to be negative.

Ethics statement

All animal experiments were approved by the Laboratory Animal Welfare and Ethics Subcommittee of the Peking University Biomedical Ethics Committee (PUIRB-LA2023459) and were conducted in accordance with institutional and national guidelines for the care and use of laboratory animals.

Method details

Salivary gland tissue acquisition

Human salivary gland tissues, including parotid, submandibular, and sublingual glands (minimum size 0.5 × 0.5 × 0.3 cm), were collected during clinically indicated surgeries under informed consent and Institutional Review Board approval (PKUSSIRB-2024103163). Tissue samples were obtained only when sufficient residual glandular tissue was available without compromising clinical treatment. Patients with a history of head and neck radiotherapy, systemic autoimmune diseases (e.g., Sjögren’s syndrome), severe tissue necrosis or active infection, or insufficient tissue volume were excluded. Fresh tissues were immediately placed in DMEM/F12 (Thermo Fisher Scientific, 12634028) supplemented with penicillin–streptomycin (Thermo Fisher Scientific, 15240062) and stored at 4°C until further processing.

Tissue digestion and organoid culture

For digestion, tissues were minced into 1–3 mm³ fragments and incubated at 37°C for 60 min in a shaker with 1 mg/mL of Collagenase Types I, II, and IV (Worthington; LS004194, LS004176, LS004186). The digestion was terminated by immediately diluting the cell suspension 10-fold and placing it on ice, after which the suspension was passed through a 100-μm strainer, centrifuged, and washed twice with DMEM/F12. Cell viability and density were assessed using an automated counter. Cells were either embedded in Matrigel (Corning, 354234) and overlaid with medium after polymerization, or seeded directly into ultra-low attachment plates (Corning, 7007) with the respective medium.

Cultures were maintained at 37°C with 5% CO₂, and media were refreshed every 2–3 days. Two culture conditions were used: keratinocyte medium (Sciencell) supplemented with KM-S (5 μg/mL BSA, 5 ng/mL FGF2, 1 ng/mL EGF, 5 μg/mL insulin, 5 μg/mL transferrin, and 0.5 μg/mL hydrocortisone; Sigma-Aldrich), and complete medium (CM), which included DMEM/F12 with 1% GlutaMAX, 1% HEPES, 1% penicillin-streptomycin (Thermo Fisher), 2% B-27 (Gibco), 1.25 mM N-acetylcysteine, 10 mM niacinamide, 1 μM dexamethasone, 100 nM A83-01 (Sigma-Aldrich), and growth factors: 500 ng/mL Wnt-3a, 100 ng/mL Noggin, 100 ng/mL R-spondin 1, 50 ng/mL EGF, and 100 ng/mL FGF10 (Novoprotein).

Cell viability was measured using CellTiter-Glo 3D reagent (Promega) following the manufacturer’s instructions, with luminescence read by a Varioskan LUX multimode reader (Thermo Fisher). Organoids were passaged every 7–10 days at a 1:2–1:3 ratio via mechanical dissociation. Cells from tissues or organoids were cryopreserved in organoid freezing medium (Biogenous, E238023) at −80°C or in liquid nitrogen.

Histological and immunofluorescence staining

Organoids were fixed in 4% paraformaldehyde, cryoprotected in 10%, 20%, and 30% sucrose, and embedded in OCT compound (SAKURA 4583). Samples were frozen, 6-μm sections were prepared, and stored at −20°C. For paraffin embedding, tissues were fixed, dehydrated through graded ethanol, embedded in paraffin, 3-μm sections were prepared, and stored at 4°C. Both cryosections and paraffin sections were processed for hematoxylin and eosin (H&E), Alcian blue–periodic acid–Schiff (AB-PAS), Masson’s trichrome, and immunofluorescence staining using standard protocols.

Primary antibodies included: AQP5 (1:200, Abcam, ab92320), K19 (1:500, Abcam, ab7754), EpCAM (1:500, Abcam, ab223582), vimentin (1:200, Abcam, ab8069), K7 (1:200, Abcam, ab181598), α-SMA (1:150, Abcam, ab7817), K5 (1:200, Abcam, ab52635), CD133 (1:500, Abcam, ab264538), TUBB3 (1:100, Abcam, ab52623), MUC7 (1:200, Biorbyt, orb101843), AMY1 (1:200, Santa Cruz, sc46657), CD31 (1:500, Abcam, ab182981), Ki67 (1:100, Abcam, ab279653), and STEM121 (1:1000, TAKARA BIO, Y40410). Secondary antibodies included Alexa Fluor® 594-conjugated goat anti-rabbit IgG (1:200, ZSGB-BIO, ZF-0516) and Alexa Fluor® 488-conjugated goat anti-mouse IgG (1:200, ZSGB-BIO, ZF-0512). Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) (1 μg/mL, Solarbio, C0065).

Bright-field images were captured using an inverted phase-contrast microscope (Olympus, Japan), and immunofluorescence images were acquired using a Nikon A1R-si confocal microscope (Nikon, Japan).

Calcium influx assay

Intracellular calcium flux in organoids was assessed using the Fluo-4 Calcium Imaging Kit (Thermo Fisher). To assess receptor-specific responses, the following agents were applied during recording: the muscarinic agonist carbachol (CCh, 10 μM), the muscarinic antagonist atropine (1 μM, 10-min pretreatment), the β-adrenergic agonist isoprenaline (10 μM), and the purinergic agonist ATP (100 μM). Signal changes were recorded using a confocal microscope (Zeiss).

Transmission electron microscopy

Organoids were fixed overnight in 2.5% glutaraldehyde at 4°C, dehydrated, embedded in resin, and sectioned into 60–70 nm ultrathin slices using an ultramicrotome (Leica UC7). Sections were mounted on copper grids, stained, air-dried, and imaged using a transmission electron microscope (Hitachi HT-7800) at 80 kV.

Quantitative reverse transcription polymerase chain reaction (qRT-PCR)

Total RNA was extracted from hSGOs using TRIzol reagent (Invitrogen) according to the manufacturer’s instructions. cDNA was synthesized using the RevertAid™ First Strand cDNA Synthesis Kit (Thermo Fisher). qRT-PCR was performed using FastStart Universal SYBR Green Master (ROX) (Roche) with the following cycling conditions: 95°C for 10 min, followed by 40 cycles of 95°C for 5 s and 60°C for 30 s. Relative mRNA expression levels were calculated using the comparative Ct method.

Enzyme-linked immunosorbent assay (ELISA)

Culture supernatants from salivary gland organoids were collected at various time points. AMY1 (salivary alpha-amylase) levels were quantified using a Human AMY1 ELISA Kit (Elabscience, E-EL-H6292) following the manufacturer’s protocol. Optical density was measured at 450 nm with a microplate reader, and concentrations were determined using a standard curve.

RNA sequencing

Total RNA from tissues and organoids was extracted as previously described. Poly(A)+ mRNA was enriched using oligo(dT)-coated magnetic beads, fragmented, and reverse transcribed with random N6 primers to generate double-stranded cDNA. The cDNA was end-repaired, A-tailed, and ligated to bubble adapters, followed by PCR amplification. The products were denatured and circularized to form single-stranded DNA libraries. Strand-specific libraries were constructed and sequenced on the DNBSEQ platform (BGI Technology). Library preparation and sequencing were performed by BGI. Bioinformatics analyses, including heatmap visualization, GO and KEGG pathway enrichment, were conducted using the Dr. Tom platform (BGI Technology).

Single-cell RNA sequencing

Salivary gland tissues were minced and enzymatically dissociated using a collagenase-based solution at 37°C with gentle pipetting until single-cell suspensions were obtained. Organoids were dissociated into single-cell suspensions using TrypLE (37°C, 15 min). Cell viability was assessed using AO/PI staining and analyzed with a Countstar Fluorescence Cell Analyzer. Single-cell libraries were prepared using the 10× Genomics Chromium Controller and Single Cell 3′ v3 Reagent Kit (10× Genomics, Pleasanton, CA). Approximately 1,000 cells/μL were loaded per channel to generate Gel Bead-in-Emulsions (GEMs). Following reverse transcription, GEMs were broken, and barcoded cDNA was purified, amplified, fragmented, A-tailed, ligated to adapters, and further amplified by indexed PCR. Final libraries were quantified using the Qubit High Sensitivity DNA Assay (Thermo Fisher) and size-checked with a High Sensitivity DNA chip on a Bioanalyzer 2200 (Agilent). Sequencing was performed on an Illumina platform (150 bp paired-end).

Data processing was conducted by NovelBio Co., Ltd. using the NovelBrain Cloud Analysis Platform. Adapter trimming and quality filtering were performed using fastp (v0.21.0).⁴⁵ Reads were aligned to the human reference genome (GRCh38, Ensembl v104) with CellRanger v7.1.0 to generate feature-barcode matrices. Aggregated matrices were constructed by down-sampling based on the number of mapped barcoded reads per cell. Cells expressing >200 genes and with mitochondrial UMI rates <20% (for group O) or <50% (for group T) were retained; mitochondrial genes were excluded from downstream analyses.

Seurat (version:4.1.1) was used for data normalization and scaling, regressing out UMI counts and mitochondrial content. Batch effects were corrected using fastMNN (scran v1.6.2). Principal component analysis (PCA) was performed on the top 2,000 highly variable genes, and the top 10 PCs were used for UMAP and t-SNE visualization. Unsupervised clustering was performed using a graph-based approach. Marker genes were identified using the FindAllMarkers function (Wilcoxon rank-sum test) with thresholds: log₂FC > 0.25, p < 0.05, and min.pct >0.1. Subclustering and re-analysis were conducted to refine cell type annotations.

Cell differentiation trajectories were inferred using Monocle2⁴⁶ (DDRTree algorithm, default parameters) based on marker genes and raw expression data. Branch-specific genes were identified using BEAM. Additionally, trajectory reconstruction was performed using Monocle3 (v1.3.1),⁴⁷ and cell differentiation states were inferred with CytoTRACE2 (v1.0.0)⁴⁸ based on gene count complexity.

Orthotopic transplantation of hSGOs

Mice were anesthetized with ketamine (100 mg/kg) and xylazine (10 mg/kg) administered intraperitoneally. A midline cervical incision was made to expose the submandibular glands. Approximately 3,000 hSGOs were injected bilaterally into the glands in the experimental group, whereas the control animals were injected with PBS. Incisions were closed with sutures. Animals were randomly assigned to ensure no significant differences in body weight or baseline saliva secretion between groups.

Subcapsular renal heterotopic transplantation of hSGOs and embryonic mesenchyme

To assess regenerative potential, three groups were established: (1) hSGOs alone, (2) embryonic mesenchyme alone, and (3) hSGOs combined with embryonic mesenchyme. Embryonic mesenchyme was isolated from submandibular gland germs of E12.5 mouse embryos. The lower jaw was dissected under a stereomicroscope, and gland germs were micro-dissected. Epithelial and mesenchymal components were separated manually using fine needles in cold PBS(−), and the mesenchyme was collected for transplantation.^49,50 NPG mice were anesthetized, and a flank incision was made to expose the kidney. A subcapsular pocket was created using a microsyringe, and the respective graft (hSGOs, mesenchyme, or both) was transplanted. The kidney was repositioned into the peritoneal cavity, and the abdominal wall and skin were sutured. At the experimental endpoints, all animals were euthanized using CO₂ inhalation followed by cervical dislocation, according to institutional animal care and ethics guidelines.

Saliva collection and proteomic analysis

Pilocarpine (0.5 mg/kg, i.p.) was used to induce saliva secretion, which was collected for 10–15 min, centrifuged at 3,000 × g for 10 min at 4°C, and stored at −80°C. Proteins were extracted from 50 mg samples using SDT buffer (4% sodium dodecyl sulfate [SDS], 100 mM Tris-HCl, 0.1 M DTT), followed by sonication, boiling, and centrifugation (14,000 × g, 15 min). Concentrations were measured using the bicinchoninic acid (BCA) assay. FASP-based digestion was performed, and peptides were desalted (C18), resuspended in 0.1% formic acid, and mixed with iRT standards. Peptides (2 μg) with iRT standards were separated by nanoLC (Bruker Nanoelute) using a 30-min gradient at 250 nL/min, then analyzed on a timsTOF Pro mass spectrometer in DIA mode. DIA-NN (v1.8.1) was used for data analysis with standard settings (1 missed cleavage, fixed carbamidomethylation, variable methionine oxidation, and 1% FDR).

Western blot analysis

Mouse submandibular glands were lysed in radioimmunoprecipitation assay buffer (Solarbio, R0010), and total protein was extracted via ultrasonic homogenization. Protein concentration was measured using the BCA protein assay kit (Thermos Scientific, 23227), followed by 15% SDS-polyacrylamide gel electrophoresis (100 V, 1–2 h) and transfer to polyvinylidene fluoride membranes (Millipore, IPVH00010) using transfer buffer (Bio-Rad,10026938). Membranes were blocked for 10 min at 25°C (New Cell & Molecular Biotech, P30500), then incubated overnight at 4°C with the following primary antibodies: anti-AQP5 (mouse, 1:1,000; Santa Cruz, sc-514022) and anti-β-actin (rabbit, 1:1,000; Biorbyt, orb10033). After three additional TBST washes, membranes were incubated with horseradish peroxidase-conjugated secondary antibodies (1:10,000; ZSGB-BIO, ZB-2305 and ZB-2301), and detection was performed using an ECL Ultra Kit (New Cell & Molecular Biotech). Signal was imaged with an Eblot system, and band intensity was quantified using ImageJ (v1.8.0, NIH).

Tissue clearing and imaging

Tissue clearing was performed using the Nuohai Life Science Tissue Clearing Kit (NH-CR-210701) following the manufacturer’s protocol. Samples were fixed in 4% paraformaldehyde, dehydrated, and processed through sequential clearing steps until optically transparent. Cleared tissues were initially imaged using a Leica M205 FA fluorescence stereomicroscope for morphological and fluorescence assessment. Selected regions were further analyzed by high-resolution 3D imaging using a Leica TCS-SP8 DIVE two-photon confocal microscope.

Quantification and Statistical analysis

Data were obtained from three independent experiments and analyzed using GraphPad Prism (GraphPad Software, San Diego, CA, USA). Comparisons between two groups were performed using unpaired two-tailed Student’s t-tests. For comparisons involving multiple groups, one-way analysis of variance followed by Tukey’s post hoc test was used. Statistical significance was set at p < 0.05. Animals were randomly assigned to experimental groups to minimize bias. Investigators were not blinded to group allocation during experimentation or outcome assessment.

Published: February 17, 2026