Introduction
The human gastric epithelium is organized into glandular units composed of spatially arranged stem cells, progenitor populations, and differentiated epithelial lineages [1,2]. Within the corpus region, epithelial homeostasis is maintained through stem cell activity and coordinated differentiation along the gland axis [3,4]. Disruption of this balance contributes to pathological conditions, including atrophy, metaplasia, and tumorigenesis [5,6]. Gastric organoids have emerged as powerful in vitro models that recapitulate key functional characteristics of the gastric epithelium [7]. These 3-dimensional cultures preserve stem cell hierarchies and lineage differentiation programs, thereby enabling investigation of epithelial homeostasis and disease-related mechanisms under controlled experimental conditions [8].
Among the signaling pathways that regulate gastric epithelial maintenance, the WNT signaling pathway plays a critical role in preserving stem cell properties [9,10]. Canonical WNT/β-catenin signaling promotes transcriptional programs essential for stem cell maintenance, and its activity is reinforced by R-spondin (RSPO)–mediated receptor stabilization [11]. Continuous WNT/RSPO signaling is required to sustain stem cell populations within gastric organoids [12]. Aberrant WNT signaling has been implicated in multiple gastric pathologies, including abnormal regeneration, metaplastic changes, and gastric neoplasia [13–15]. These observations underscore the importance of tightly regulated WNT activity in maintaining gastric epithelial homeostasis.
Given the central role of the WNT signaling pathway in regulating stem cell identity, considerable effort has focused on identifying WNT-associated transcriptional signatures [16,17]. However, the temporal dynamics of these signatures remain poorly defined, including when stem cell–associated programs begin to decline and when differentiation-associated gene programs are activated. Moreover, bulk RNA sequencing and real-time polymerase chain reaction (PCR) approaches are limited in their ability to resolve lineage-specific transcriptional transitions. In contrast, single-cell RNA sequencing (scRNA-seq) captures the heterogeneity of epithelial populations and enables precise identification of stem cell–specific and differentiation-associated transcriptional programs over time.
Existing methods for inducing differentiation in gastric organoids have primarily involved modulation of epidermal growth factor and bone morphogenetic protein signaling [18]. However, these approaches often yield modest or incomplete differentiation phenotypes. Because WNT signaling plays a central role in maintaining gastric stemness, perturbation of WNT/RSPO signaling may provide a more robust strategy to induce epithelial differentiation [17,19]. Notably, prior studies investigating WNT perturbation in gastric organoids have frequently used tissues adjacent to gastric cancer, which may not fully represent normal gastric epithelium [19]. In contrast, organoids derived from histologically normal human gastric corpus tissue enable more accurate characterization of physiological WNT-dependent epithelial dynamics.
In this study, we performed a temporal single-cell transcriptomic analysis of normal human gastric corpus organoids following WNT/RSPO perturbation. By systematically profiling transcriptional changes across defined time points, we aimed to characterize the temporal dynamics of stemness and differentiation and to clarify WNT signaling–associated transcriptional signatures at single-cell resolution. Temporal perturbation of WNT and RSPO enables identification of differentiation-associated transcriptional programs in a time-dependent manner, thereby providing a single-cell framework for defining WNT pathway–associated signatures in normal human gastric organoids.
Materials and Methods
Ethics statement: Human gastric corpus tissues were obtained during gastrointestinal endoscopy at Seoul National University Hospital (SNUH). The study was approved by the Institutional Review Board of SNUH (IRB No. 2010-178-1171).
1. In vitro human gastric corpus organoid cultures
Gastric tissues were cut into pieces smaller than 5 mm and washed in ice-cold phosphate-buffered saline (PBS) to remove residual mucus and blood cells. Tissue fragments were dissociated into single cells using a gentleMACS Octo Dissociator with Heater (Miltenyi Biotec) according to the manufacturer’s protocol at the Genomic Medicine Institute Research Service Center. Cell pellets were resuspended in 1 mL of red blood cell lysis buffer (Miltenyi Biotec) for 10 minutes at room temperature. Cells were then embedded in growth factor–reduced Matrigel (Corning) and plated in prewarmed 24-well tissue culture plates. Plates were incubated at 37°C for 15 minutes, followed by addition of 500 μL of prewarmed gastric complete medium containing 10 μM Y-27632 (Tocris) for the first 3 days [20]. Detailed medium composition is provided in Table 1. The medium was changed every 3 days. Established 3-dimensional gastric corpus organoids were passaged at a 1:10 ratio every 1–2 weeks.
2. WNT/RSPO perturbation
Following mechanical dissociation and passaging, organoids were cultured in complete gastric medium for 3 days to allow stabilization. The organoids were then washed twice with PBS and cultured in medium lacking WNT and RSPO to induce WNT/RSPO perturbation.
3. scRNA-seq library preparation
Single-cell suspensions were prepared from gastric organoids at the indicated time points following WNT/RSPO perturbation. Organoids were dissociated into single cells and washed with ice-cold PBS without calcium or magnesium. Cells were fixed using the ScaleBio FixKit (Scale Biosciences) according to the manufacturer’s protocol. scRNA-seq libraries were then prepared using the QuantumScale Single Cell RNA–Medium Kit (Scale Biosciences). Briefly, fixed single cells were subjected to in situ reverse transcription using barcoded poly(dT) primers containing molecular barcodes and partial sequencing adapters. Cells labeled with distinct reverse-transcription barcodes were pooled and loaded onto the Quantum Barcoding Plate, followed by loading of barcoded beads. Complementary DNA (cDNA) molecules were released from cells and captured on barcoded beads through ligation. After bead oligo blocking and second-strand synthesis, amplified cDNA was generated by semi-suppressive PCR amplification.
Indexed RNA libraries compatible with Illumina sequencing were constructed using the QuantumScale Library Index Kit-I-1 (Scale Biosciences), which adds Illumina P5/P7 adapters and i5 index sequences during the final index PCR step. Final libraries were purified by SPRI bead cleanup, quantified using the Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific), and quality assessed with an Agilent TapeStation before sequencing. Libraries were sequenced on an Illumina NovaSeq 6000 system (Illumina) in paired-end mode.
4. Preprocessing and quality control
Raw sequencing data were processed using ScaleRna (v2.1.0; Scale Biosciences Inc.), and raw gene expression matrices were analyzed using Scanpy (v1.10.0). Quality-control filtering was performed independently for each sample. Cells were identified as outliers based on median absolute deviation (MAD)-based thresholds (5 MADs) applied to log-transformed total counts, numbers of detected genes, and the percentage of counts in the top 20 expressed genes. Cells with high mitochondrial gene content (>20%) were removed. Additional filtering excluded cells with fewer than 50 detected genes and cells with total unique molecular identifier (UMI) counts exceeding 10,000. Doublet detection was performed using SOLO implemented in scvi-tools (v1.1.2). Before integration, cells with fewer than 100 detected genes were excluded, and genes expressed in fewer than 2 cells were removed.
5. Data integration and cell type annotation
Batch correction and data integration were performed using the SCVI model implemented in scvi-tools (v1.1.2), incorporating technical batch identity (library) as a categorical covariate and technical covariates (total UMI counts and numbers of detected genes), while preserving biological variation across time points. The latent representation learned by SCVI was used for downstream clustering and visualization. Leiden clustering was performed, and Uniform Manifold Approximation and Projection (UMAP) was generated on the basis of the SCVI latent space. Cell types were annotated using canonical marker gene expression. All computational analyses were performed on the computing server at the Genomic Medicine Institute Research Service Center.
6. Differential gene expression analysis and pathway enrichment analysis
Differential gene expression analysis was performed using the rank_genes_groups function in Scanpy (v1.10.0). For each time point, gene expression was compared with that at baseline (0 hours) using the t-test. Results were visualized using volcano plots showing log₂ fold change versus −log₁₀ adjusted p-value. Genes identified as significantly differentially expressed were subjected to pathway enrichment analysis using the Enrichr interface in GSEApy (v1.1.11). Enrichment analysis was performed using the Gene Ontology Biological Process database. Enriched pathways were ranked by combined score, and the results were visualized using bar plots.
7. Pseudotime trajectory analysis
Diffusion pseudotime (DPT) analysis was performed using Scanpy (v1.10.0) to infer temporal progression of cellular states. Gene expression data were normalized to 10,000 counts per cell and log transformed before analysis. Highly variable genes (n=1,500) were selected using the Cell Ranger–based method in Scanpy. Principal component analysis was performed, and a neighborhood graph was constructed using the top 10 principal components. The root cell was defined as the cell with the minimum value along the selected diffusion component, and DPT values were calculated using the sc.tl.dpt function.
8. Pseudobulk analysis
Pseudobulk expression profiles were generated by aggregating raw counts across cells within each sample using the get_pseudobulk function implemented in decoupler (v1.7.0). Counts were summed for each sample. Pseudobulk data were normalized to counts per million using total-count normalization, followed by log transformation.
Results
1. Response of human gastric corpus organoids to WNT/RSPO perturbation
To investigate the dynamic transcriptional response to WNT pathway perturbation, we used organoids derived from patients without gastric disease. Human gastric corpus organoids were stabilized in complete medium for 3 days and subsequently exposed to WNT/RSPO perturbation (Fig. 1A). Fig. 1B shows bright-field images of gastric corpus organoids under different culture conditions. Control organoids were maintained in complete medium for 7 days and exhibited robust growth while preserving a spherical morphology throughout the culture period. Under the Stabilization/-WR condition, organoids were cultured in complete medium for 3 days and then maintained for 4 days under -WR conditions. These organoids remained viable and preserved overall structural integrity comparable to that of controls. In contrast, No-stabilization/-WR organoids exposed to -WR conditions immediately after plating failed to expand and rapidly lost structural integrity, indicating impaired survival in the absence of prior stabilization. According to previous studies, immediate WNT withdrawal without prior stabilization demonstrates the dependence of normal gastric organoids on WNT signaling [20]. Conversely, withdrawal of WNT signaling after stabilization in intestinal metaplasia gastric organoids promotes differentiation and suppresses stem cell programs [19]. Consistent with these previous reports, simultaneous withdrawal of WNT and RSPO in our system yielded similar findings, supporting the dependence of gastric organoids on WNT signaling.
2. WNT/RSPO perturbation induces time-dependent epithelial cell state transitions
Organoids were first stabilized in complete medium for 3 days, followed by WNT/RSPO perturbation. Cells were harvested at 0, 6, 12, 24, 48, and 72 hours after perturbation and subjected to scRNA-seq (Fig. 1A). Integration of single-cell transcriptomes identified distinct epithelial cell populations (Fig. 1C). Cell type annotation was based on canonical marker gene expression. Gastric stem cells were defined by expression of STMN1, MKI67, and IQGAP3, whereas pit mucous cells were identified by expression of MUC5AC, TFF1, GKN1, and GKN2. MUC4 expression defined a subset of mucous cells that have been associated with inflammatory or stress-responsive epithelial states [21]. Tuft cells were annotated on the basis of POU2F3 and AVIL expression (Fig. 1D and 1E). The relative proportions of each cell population changed over time following WNT/RSPO perturbation, indicating dynamic remodeling of epithelial cell states (Fig. 1F). At baseline (0 hours), the epithelial population was composed predominantly of stem cells, which accounted for 89.8% of all cells, whereas differentiated mucous populations were present at lower proportions (MUC4+ mucous cells, 3.6%; pit mucous cells, 6.0%). A time-dependent shift in cellular composition was observed. Stem cell populations gradually declined over 72 hours, whereas tuft cells and mucous cell populations expanded following WNT/RSPO perturbation. Together, these observations support the conclusion that WNT/RSPO perturbation after stabilization induces a shift from stem cell states toward differentiated epithelial lineages.
3. WNT/RSPO perturbation induces dynamic cell state transitions
To further investigate temporal cellular dynamics following WNT/RSPO perturbation, we analyzed the distribution of cells across time points on the UMAP embedding (Fig. 2A). Cells showed gradual redistribution across the 72-hour time course, indicating progressive transcriptional changes. DPT analysis was performed to infer the trajectory of cellular state transitions. Projection of pseudotime onto the UMAP revealed continuous progression along a differentiation axis (Fig. 2B). When pseudotime distributions were compared across annotated cell types, gastric stem cells were enriched at earlier pseudotime states, whereas tuft cells and mucous cells were distributed predominantly at later pseudotime values (Fig. 2C). These findings support a time-dependent transition from stem cell states toward differentiated epithelial lineages following WNT/RSPO perturbation.
4. Transcriptional remodeling following WNT/RSPO perturbation
To characterize transcriptional changes induced by WNT/RSPO perturbation, differential gene expression analysis was performed at each time point using the 0-hour data as the reference (Fig. 3A). The number of differentially expressed genes (DEGs) increased markedly after 12 hours of WNT/RSPO perturbation. Whereas only 596 DEGs were detected at 6 hours, the number increased sharply at 12 hours (13,047 DEGs) and remained elevated at 24 hours (11,551 DEGs) and 48 hours (17,014 DEGs). These findings suggest the onset of large-scale transcriptomic remodeling after 12 hours; however, because WNT/RSPO perturbation also changes cell-type composition over time, the increased DEG counts may reflect both within-cell-state changes and compositional shifts. Comparison of the 0-hour and 24-hour time points revealed substantial transcriptional alterations (Fig. 3B). Several genes related to mitochondrial respiration/stress (e.g., MT-CO1) and cytoskeletal remodeling/WNT-associated processes (e.g., LPP) were identified among the significantly upregulated DEGs. In addition, gastric epithelial differentiation markers such as MUC5AC were also prominently upregulated at later time points [16,22,23]. Gene Ontology enrichment analysis of upregulated DEGs identified at 24 hours revealed significant enrichment of pathways related to epithelial differentiation, cytoskeletal organization, and cellular remodeling (Fig. 3C).
5. Phase-specific transcriptional dynamics following WNT/RSPO perturbation
To further investigate lineage-specific transcriptional dynamics, pseudobulk expression analysis was performed (Fig. 4). Cycling gastric stem cell–associated genes (LGR5, STMN1, and TNFRSF19), canonical WNT target genes (ZNRF3, RNF43, and AXIN2), and WNT signaling–related genes (SLC7A2, SLC16A10, and MNS1) progressively decreased over time following WNT/RSPO perturbation (Fig. 4A–C). Whereas MUC6 expression gradually decreased over time, MUC5AC, a marker of differentiated pit mucous cells, peaked at 48 hours and subsequently declined. In contrast, MUC4 expression progressively increased and reached its maximum level at 72 hours (Fig. 4D). The increase in MUC4 expression is consistent with differentiation toward a mucous lineage, with MUC4 marking an inflammatory-associated state within differentiated pit mucous cells. Similarly, gastric mucous cell markers exhibited distinct temporal patterns (Fig. 4E). TFF2 expression peaked early at 24 hours and decreased thereafter, whereas TFF1 expression reached its maximal level at 48 hours before declining at later time points. Notably, most differentiation-associated genes showed peak expression at or after 48 hours following WNT/RSPO perturbation. These findings indicate that early transcriptional responses (≤24 hours) primarily reflect direct effects of WNT signaling perturbation, whereas later time points (≥48 hours) capture progressive differentiation-associated transcriptional programs.
Discussion
In this study, we performed a temporal single-cell transcriptomic analysis of human gastric corpus organoids following WNT/RSPO perturbation. We demonstrate that gastric organoids show strong dependence on WNT signaling and that stabilization before withdrawal influences subsequent transcriptional remodeling. Importantly, our temporal analysis reveals distinct transcriptional phases following WNT/RSPO perturbation, separating early WNT signaling responses from later differentiation-associated programs.
Our data indicate that transcriptional remodeling occurs in a temporally structured manner. Minimal transcriptional alterations were observed at 6 hours following WNT/RSPO withdrawal. A marked increase in DEGs emerged at 12 hours, indicating a major transcriptional transition. Notably, early time points (≤24 hours) were characterized predominantly by suppression of canonical WNT target genes and stem cell–associated programs. In contrast, later time points (≥48 hours) were dominated by activation of differentiation-associated gene signatures, including mucous and tuft cell markers. These findings demonstrate that WNT withdrawal triggers a sequential transcriptional cascade in which primary WNT-dependent responses precede lineage differentiation programs. This temporal separation provides a framework for distinguishing direct WNT signaling effects from secondary differentiation-driven transcriptional changes (Fig. 5).
Importantly, the observed shift from stem cell–associated programs toward mucous and other differentiated epithelial lineages following WNT/RSPO withdrawal is consistent with the physiological organization of the gastric epithelium. In the gastric gland, WNT signaling is enriched within the stem cell niche, whereas differentiated epithelial populations occupy regions of comparatively lower WNT activity [17,18,24]. Our data recapitulate this gradient-dependent organization in an organoid model, supporting the concept that reduced WNT signaling promotes differentiation toward mucous lineages.
The lack of temporal resolution in previous studies may explain inconsistencies in defining WNT-associated transcriptional signatures in gastric organoids. Many studies have relied on endpoint analyses, often performed several days after WNT withdrawal [20]. At such late time points, differentiation-associated transcriptional programs are already established, potentially confounding identification of direct WNT-responsive genes. Moreover, bulk RNA sequencing averages gene expression across heterogeneous epithelial populations. Because WNT withdrawal induces shifts in cellular composition, bulk profiling may capture the combined effects of signaling suppression and lineage redistribution. This cellular heterogeneity likely contributes to the difficulty of defining a WNT signaling signature. By performing scRNA-seq across defined temporal intervals, our study delineates the transcriptional dynamics that most accurately reflect primary WNT pathway perturbation.
Several limitations should be considered. Tissue-derived epithelial organoids do not fully recapitulate the cellular complexity of the native gastric niche. In vivo, gastric epithelial stem cells are regulated through interactions with mesenchymal, immune, and vascular components [25]. In contrast, organoid cultures primarily capture epithelial-intrinsic responses. Therefore, the transcriptional changes observed following WNT/RSPO withdrawal may not fully reflect the niche-dependent regulatory mechanisms operating within the gastric gland.
In addition, this study involved simultaneous perturbation of WNT and RSPO to model disruption of the stem cell niche. WNT ligands bind Frizzled and LRP receptors to initiate β-catenin stabilization and transcriptional activation of target genes [26]. RSPO proteins stabilize Frizzled receptors at the cell surface and enhance WNT pathway signaling [27]. Consequently, this study reflects perturbation of the integrated WNT niche rather than the isolated contribution of individual components. Defining the respective roles of WNT and RSPO will require experiments involving independent perturbation of each component.
In summary, our study provides a temporal framework for interpreting transcriptional responses to WNT/RSPO perturbation in human gastric corpus organoids. By analyzing transcriptional dynamics at single-cell resolution, we distinguish early WNT-dependent gene suppression from later differentiation-associated programs. These findings highlight the importance of temporal resolution when defining pathway-specific transcriptional signatures and demonstrate that the timing of analysis influences biological interpretation. Collectively, this work advances understanding of WNT-dependent epithelial remodeling and provides a methodological foundation for future studies investigating signaling-driven lineage transitions in the human gastric epithelium.
PDF Links
PubReader
ePub Link
Full text via DOI
Download Citation
Print
