PDX1+ cell budding morphogenesis in a stem cell-derived islet spheroid system – Nature Communications

Development of a budding-type stem cell-derived islet spheroid system

We and others previously developed protocols for bulk differentiation of hPSCs into hormone-expressing islet cells by timed addition of defined soluble factors^{8,9,10,11,12,13,23,24,25}. This bulk-type differentiation starts from a robust induction of definitive endoderm (DE) cells through activation of TGFβ and Wnt signaling. By fine tuning the doses of Wnt agonists at the DE stage (Stage 1), we established a budding-type differentiation system generating islet cells that are highly enriched in bud structures (Fig. 1a and Supplementary Fig. 1). Specifically, depending on the cell lines used, budding-type differentiation was induced with CHIR99021 concentrations of 0.2–1.5 µM (or MCX-928 at 0.1–0.5 µM; Wnt^low) whereas bulk-type differentiation was induced (i.e., PDX1+ buds are no longer formed) when CHIR99021 was used at 3 µM (or 1 µM MCX-928; Wnt^med) in combination with GDF8 or Activin A (Supplementary Fig. 1a–c). In the presence of TGFβ ligand, medium-level Wnt activation (Wnt^med) was sufficient to drive hPSCs to differentiate into DE cells (>95% FOXA2+/SOX17+ cells) whereas low-level Wnt activation (Wnt^low) generated ~75% FOXA2+/SOX17+ DE cells within Stage 1 cell populations (Supplementary Fig. 1a, d–f), suggesting that the extent of Wnt signaling (as examined by >60 Wnt pathway genes, Supplementary Fig. 2) determined endoderm specification efficiency. In the Wnt^med bulk-type differentiation, an average of 95% PDX1+ cells and 73% PDX1+/NKX6.1+ pancreatic progenitors were generated in Stage 4 cultures, and insulin was expressed throughout Stage 7 clusters (Supplementary Fig. 1b–d, g). The Wnt^low budding-type differentiation produced an average of 32% PDX1+ cells and ~ 5% PDX1+/NKX6.1+ cells at the end of Stage 4, and insulin was largely restricted to a local area of Stage 7 clusters (Supplementary Fig. 1b, c, e, g), indicating that the single-day treatment (for Stage 1 day 1 only) with Wnt^med or Wnt^low condition drastically impacted subsequent differentiation toward the endocrine lineage.

a Schematic of a tunable differentiation strategy to generate two types of islet models from stem cells under static suspension culture. The bulk- or budding-type differentiation is established by simply fine tuning the doses of Wnt agonist at DE stage. Specifically, Wnt^med condition at DE stage induces bulk type whereas Wnt^low condition at DE stage induces budding type. The “bulk” and “budding” are defined as to whether islet cells are broadly developed throughout clusters or enriched in local area of clusters. Created with BioRender.com released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license. b Typical flow cytometry plot of Stage 4 cells from budding differentiation cultures. c Live imaging showing that a small proportion of cells were committed into INS+ cells (indicated by INS-GFP reporter). Of note, INS+ cells were enriched in small bud-like structures rather dispersed in entire clusters over the course of differentiation. Scale bar, 300 μm. d Tracking individual cluster in living cultures shows that the appearance of budding structures preceded the gradual induction of INS+ cells within the buds. Scale bar, 300 μm. e Representative image of Stage 7 islet bud stained for major islet cell types. INS insulin, GCG glucagon, SST somatostatin. Scale bars, 20 μm. Representative flow plots (f) and quantification (g) of CHGA+, INS+, and GCG+ cells within the bud and body compartments after enzymatic isolation (see Methods) from Stage 7 clusters. n = 5 independent differentiations, **p < 0.01 versus body, unpaired two-tailed t-test. Budding differentiation was induced by 100 ng/mL GDF8 plus 1–1.5 μM CHIR99021 for Stage 1 Day 1 using the Mel1 INS^GFP/W line. Source data are provided as a Source Data file.

Full size image

By comparing transcriptomic differences in cells generated by the two protocols (Supplementary Fig. 3a), we find that pancreatic cell transcripts (PDX1, NKX6.1, SOX9) and pro-endocrine cell transcripts (NGN3, NEUROD1) were less abundant in budding-type Stage 4 cells (Supplementary Fig. 3b). The expression levels of most transcripts became comparable between budding-type and bulk-type cells at Stages 5 and 7 (Supplementary Fig. 3c, d). Nevertheless, relative to bulk-type cells, we note higher levels of GHRL, SLC18A1, KCNK1 transcripts and lower levels of GCG, PPY, ARX, NKX2.2, ABCC8 transcripts in budding-type Stage 7 cells relative to cells from bulk-type differentiations (Supplementary Fig. 3d). Although transcripts of functional beta cell markers MAFA, IAPP, GCK, PSCK1, KCNK3 were similar between the two types of differentiation (Supplementary Fig. 3d), insulin secretion and total insulin content of budding-type Stage 7 beta cells was 1.4-fold and 1.7-fold lower than those of bulk-type cells, respectively (Supplementary Fig. 1h), suggesting a less mature phenotype of beta cells generated by the budding protocol. As initial seeding densities were the same and total cell numbers were indistinguishable between the two types of differentiation (Supplementary Fig. 1i), these differences are unlikely attributed to seeding condition, distinct cell survival or growth rate during early-stage culture.

To examine whether the budding-type differentiation could be reproduced with different Wnt inducers and in multiple hPSC lines, three Wnt agonists (CHIR99021, MCX-928 and mWnt3a) were tested and four hPSC lines (one hiPSC and three hESC lines) were differentiated under Wnt^low conditions (Supplementary Fig. 4). When combined with TGFβ ligands, all the three Wnt agonists used at low concentrations induced a budding-type differentiation process, while increasing the concentrations of CHIR99021 and MCX-928 to a medium level resulted in a switch to a more uniform bulk-type differentiation (Supplementary Fig. 4a). Consistent across the four hPSC lines, 70–85% FOXA2+/SOX17+ DE cells at the end of Stage 1 and 20–40% PDX1+ cells at the end of Stage 4 were obtained from the Wnt^low-mediated budding-type differentiation (Supplementary Fig. 4b–d). Dithizone staining of Stage 7 clusters showed zinc-enriched beta cells in bud structures (Supplementary Fig. 4e), whose cells appeared to be more compact and had a darker appearance than main bodies under phase contrast imaging (Supplementary Fig. 4f), again supporting the concept that the Wnt^low conditions used at the DE stage induce a budding-type differentiation in vitro.

Next, we quantified the efficiencies of bud formation (including PDX1+ pancreatic buds and INS+ islet buds) during the differentiation process with our Wnt^low protocol on various hPSC lines (Supplementary Fig. 5). The percentages of clusters that formed PDX1+ bud(s) for the cell lines examined were 98.7% ± 0.2% (HUES4 PDXeG), 93.1% ± 3.1% (Mel1 INS^GFP/W), 88.3% ± 0.9% (H1 hESC), 87.7% ± 1.4% (GCaMP hiPSC) and 97.3% ± 0.4% (HUES8) (Supplementary Fig. 5a, b, d). The percentages of clusters that formed INS+ islet bud(s) quantified at the end of protocol were 92.5% ± 4.1% (HUES4 PDXeG), 85.4% ± 3.2% (Mel1 INS^GFP/W), 77.4% ± 10.2% (H1 hESC), 71.4% ± 12.3% (GCaMP hiPSC) and 84.5% ± 5.9% (HUES8) (Supplementary Fig. 5e, f, h). The number of PDX1+ buds or INS+ islet buds per main body was also quantified using HUES4 PDXeG and Mel1 INS^GFP/W reporter lines, both showing >75% main bodies with only one bud and ~10% main bodies with two or more buds (Supplementary Fig. 5c, g). These results demonstrate the high efficiency and reproducibility of bud formation observed with the Wnt^low protocol in various hPSC lines.

Islet formation through budding morphogenesis is mediated by PDX1+ cell clustering

To characterize the budding process in further detail, we used Mel1 INS^GFP/W hESCs (an insulin reporter line²⁶) to monitor the differentiation process. Consistently, PDX1 was expressed in about one-third of Stage 4 cells generated with the budding-type differentiation protocol (Fig. 1b). After single-cell dissociation and aggregation of Stage 4 cells into clusters followed by transition to static suspension culture for further endocrine cell induction, we noted that a proportion of cells were committed to insulin-positive cells as indicated by expression of the GFP reporter. Interestingly, these insulin-expressing cells were enriched within bud-like structures rather than dispersed throughout clusters (Fig. 1c). To track this process in living cultures, we differentiated individual clusters in ultralow attachment U-bottom 96-wells. Live imaging revealed that bud structures appeared during Stage 5 to early Stage 6 protruding from main body clusters, followed by a gradual induction of insulin-positive cells within the buds (Fig. 1d). Additional characterization showed that these buds were primarily endocrine cells and enriched in all major islet cell types (Fig. 1e–g and Supplementary Fig. 6).

Since the formation of bud structures preceded induction of insulin-positive cells (Fig. 1d), we sought to characterize these initial bud structures through expression of progenitor markers. Based upon observations that insulin-positive cells were specifically derived in the bud niche associated with elevated expression of PDX1 (Fig. 2a), we reasoned that formation of initial bud structures was a result of PDX1+ cell enrichment, either through cell migration or local proliferation of PDX1+ cells. With whole-cluster imaging, we observed a progressive change from an initial scattering of PDX1+ cells throughout entire clusters towards the high enrichment in bud structures, suggesting a migration of PDX1+ cells from main bodies to buds (Fig. 2b). The proportions of PDX1+ cells remained consistent along with the low PDX1+ cell proliferation rate (Fig. 2c and Supplementary Fig. 7) during budding process indicating that formation of PDX1+ cell-enriched buds is unlikely due to local expansion of PDX1+ cells within the buds. Finally, to directly visualize PDX1+ cells in living cultures, we used a PDX1 reporter hESC line (HUES4 PDXeG²⁷) while imaging the process of bud formation (Supplementary Fig. 8a). We noted PDX1+ cells tended to form clustered structures in both planar and suspension cultures (Supplementary Fig. 8b–i) and the clustering process appeared to be independent of differentiation factors in culture media (Supplementary Fig. 8j). By tracking PDXeG-positive cells in individual clusters over the course of differentiation with time-course snapshot imaging, we demonstrated migration of PDX1+ cells forming the initial bud structures protruding from main bodies (Fig. 2d).

a Representative images of Stage 4-6 clusters stained for PDX1 and NKX6.1. INS expression was indicated by INS-GFP in clusters derived from Mel1 INS^GFP/W hESCs. Scale bar, 50 μm. b Representative z-projection images of Stage 4-5 clusters (highlighted with white dotted lines) stained for PDX1 from whole-cluster imaging. Scale bar, 100 μm. c Quantification of the percentage of PDX1+ cells and proliferative PDX1+ cells at indicated stages. n = 3–10 independent differentiations, ns not significant, one-way ANOVA with Dunnett test for multiple comparisons to S4D6 clusters. Created with BioRender.com released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license. d Daily monitoring of individual cluster showing the process of PDX1+ cell clustering. A PDX1 reporter hESC line (HUES4 PDXeG) was used for the time-course snapshot visualization. Scale bar, 100 μm. e Schematic of PDX1+ cell clustering and proposed consequences on following differentiation. Created with BioRender.com released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license. f Representative images of Stage 5-7 clusters stained for PDX1, NKX6.1, and various endocrine markers. NGN3 neurogenin 3. SYP synaptophysin. Scale bar, 50 μm. Representative flow cytometry plots (g) and quantification (h) of Stage 4-5 entire clusters and Stage 6-7 isolated buds examined for pancreatic progenitor markers (PDX1, NKX6.1), endocrine markers (CHGA, NEUROD1) and INS expression. n = 4–5 independent differentiations, ns not significant, *p < 0.05, **p < 0.01, one-way ANOVA with Dunnett test for multiple comparisons to Stage 4 clusters. Budding differentiation was induced by 100 ng/mL GDF8 plus 1–1.5 μM CHIR99021 for Stage 1 Day 1 using the Mel1 INS^GFP/W line (a–c, f–h) and HUES4 PDXeG line (d). Source data are provided as a Source Data file.

Full size image

The PDX1+ cell-enriched bud niche may facilitate local endocrine cell induction (Fig. 2e). In support of this hypothesis, we detected expression of pro-endocrine markers NEUROD1 and NGN3 (transient and sporadic) in Stage 5 bud structures, pan-endocrine marker synaptophysin (SYN) in Stage 6 islet buds, as well as expression of PDX1 and NKX6.1 in INS+ cells of Stage 7 islet buds (Fig. 2a, f). Similarly, flow analysis of entire clusters from Stages 4-5 and isolated islet buds from Stages 6-7 showed sequential expression of PDX1, NKX6.1, NEUROD1, CHGA, and INS over the course of budding-type differentiation (Fig. 2g, h), suggesting that endocrine cell development within bud structures followed a progression similar to that observed in bulk-type differentiation of hPSCs.

Cytoarchitectural rearrangement and functional heterogeneity of differentiating islet buds

Budding-type differentiations produced islet buds comprised of major islet cell types including insulin-expressing beta cells, glucagon-expressing alpha cells and somatostatin-expressing delta cells (Fig. 1e–g). Interestingly, we observed the majority of Stage 7 buds organized into a core-mantle architecture with beta cells in the center and alpha cells in the periphery (Fig. 3a, b). Quantification of islet cell composition revealed that Stage 6 islet buds contained mostly INS+/GCG+ bi-hormonal cells, whereas Stage 7 islet buds had significantly increased proportions of INS+/GCG- and GCG+/INS- monohormonal cells (Fig. 3c). To explore whether biochemical cues affected these changes, we examined the effects of either adding or subtracting key components, which are unique in either Stage 6 or Stage 7 medium, in an extended culture following Stage 6 (Supplementary Fig. 9a). Immunostaining showed that the extended culture promoted a transition to monohormonal cells for all conditions, except the basal medium condition (Supplementary Fig. 9b). Cytoarchitecturally, removal or inclusion of Stage 6-unique compounds (LDN, GSiXX) did not change the intermingled organization of islet cells; by contrast, inclusion of Stage 7-unique compounds (R428, NAC, Trolox) in extended cultures appeared to induce the core-mantle structure (Supplementary Fig. 9b, c). However, by individually supplementing Stage 7-unique components we did not see effects on cytoarchitectural changes (Supplementary Fig. 9), suggesting that there is a synergistic requirement for more than one factor, rather than a single biochemical cue tested here, for the islet cell rearrangement.

a Representative images of typical Stage 6-7 clusters stained for major islet cell types. INS insulin, GCG glucagon, SST somatostatin. Nuclei were counterstained with DAPI (blue). Scale bar, 100 μm. b Quantification of the percentage of different cytoarchitectural types adopted by Stage 6-7 islet buds. n = 30–50 clusters for each from 5-8 independent differentiations. c Flow cytometry quantification showing the percentage of major islet cell composition in Stage 6-7 islet buds. n = 5 independent differentiations, ns not significant, *p < 0.05, **p < 0.01 versus Stage 6 islet buds, unpaired two-tailed t-test. d Schematic of microfluidic chip system (termed as “microperifusion”) to perfuse single islet bud for imaging calcium activities and detecting dynamic insulin secretion. Created with BioRender.com released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license. e Traces of glucose-stimulated calcium responses in representative Stage 6-7 islet buds were categorized by an unsupervised clustering algorithm (see Methods). Glucose stimulation (16.7 mM) was added at the time point of 5-min (indicated by arrows). Quantification showing the number of cluster types (f) and the percentage of synchronized cells per cluster type (g) in Stage 6-7 islet buds. n = 10 clusters for each from 3 to 4 independent differentiations. **p < 0.01 versus Stage 6 islet buds, unpaired two-tailed t-test. h Microperifusion showing dynamic insulin secretion from Stage 6-7 islet buds, which were categorized to high glucose (HG)-responders, low glucose (LG)-responders, and glucose non-responders (Non). Traces were presented as mean value ± SEM (n = 3–5 clusters for each from 4 independent differentiations) and individual traces were also shown with dotted lines. i Quantification showing the percentages of glucose-responder types and architectural types adopted by Stage 6-7 buds. n = 30–35 clusters for each from 4 independent differentiations. Budding differentiation was induced by 100 ng/mL GDF8 plus 1–1.5 μM CHIR99021 for Stage 1 Day 1 using the Mel1 INS^GFP/W line. Source data are provided as a Source Data file.

Full size image

We next examined the function of developing islet buds by utilizing a microfluidic chip-based perifusion system²⁸ (termed “microperifusion”) to perfuse single clusters with simultaneous imaging of intracellular calcium levels (Fig. 3d and Supplementary Fig. 10a). Robust insulin secretion of individual primary human islets from different donors was consistently detected (Supplementary Fig. 10b). With the setup validated, we first performed calcium imaging on islet buds and categorized calcium activities in response to 3.3 mM and 16.7 mM glucose with an unsupervised clustering algorithm²⁹. Three distinct responsive types were identified in Stage 6 buds, with calcium activities synchronized among cells within certain responsive types but unsynchronized among different responsive types. In Stage 7 buds a single cluster type was frequently identified, suggesting that the cells displayed generally synchronized calcium responses in the bud (Fig. 3e and Supplementary Movies 1, 2). Quantification revealed that Stage 6 islet buds were identified with more calcium-response types (4.4 ± 1.4 versus 1.7 ± 0.2 in Stage 7 buds) while Stage 7 islet buds had more synchronized cells per active region (16.2 ± 2.9 versus 6.9 ± 1.4 in Stage 6 buds) (Fig. 3f, g), in keeping with the presence of more homotypic interactions among beta cells in Stage 7 buds. Next, by perfusing individual clusters for insulin secretion measurement we observed high glucose (HG)-responders, low glucose (LG)-responders, and glucose non-responders in both Stage 6-7 islet buds, albeit at different proportions (Fig. 3h, i). LG-responders were more common in Stage 6 buds while HG-responders were observed more frequently in Stage 7 buds. Although all spheroids were responsive to depolarization with KCl, the majority were not responsive to elevated glucose concentrations, indicating the developing islet buds were functionally immature (Fig. 3h, i). To correlate function with cytoarchitecture, we retrieved individual islet buds from chips after microperifusion for immunostaining. The HG-responders in Stage 6-7 islet buds mainly adopted a core-mantle organization with predominant monohormonal islet cells, while LG-responders in Stage 6-7 buds had a mixed organization predominantly consisting of bi-hormonal INS+/GCG+ cells (Fig. 3i), supporting the concept that monohormonal INS+ cells are more faithfully responsive to glucose challenges relative to bi-hormonal cells⁹.

Modeling human pancreas diseases with the tunable differentiation system

The transcription factor PDX1 is a master regulator of pancreas lineage commitment³⁰, with PDX1 deficiency causing pancreatic agenesis seen in various models^31,32,33. Consistent with this, we observed a requirement for PDX1 in eliciting cell budding morphogenesis during pancreatic patterning by applying both bulk- and budding-type differentiations with a PDX1 knockout (KO) hESC line³². The absence of PDX1 did not impact DE formation, but stopped formation of pancreatic progenitors, endocrine cells and major islet cell types as well as insulin production (Fig. 4 and Supplementary Fig. 11). Moreover, while pancreatic buds were consistently formed in wildtype (WT) hESC-derived clusters with our budding differentiation protocol, cell budding (or pancreatic bud formation) was blocked in differentiating clusters derived from the PDX1 KO hESCs regardless of differentiation pipeline applied (Fig. 4e and Supplementary Fig. 11a, b). These results corroborate previous work by recapitulating the pancreatic agenesis phenotype with a stem cell model and again emphasize the importance of PDX1 in pancreatic specification and morphogenesis.

Representative flow plots (a) and quantification of Stage 1 cells examined for FOXA2 and SOX17 (b), Stage 3-4 cells examined for PDX1 and NKX6.1 (c, d) from HUES8 wildtype (WT), PDX1 KO and RFX6 KO hESC-derived cultures. Both bulk and budding differentiation pipelines were applied. n = 4 independent differentiations, ns not significant, **p < 0.01, one-way ANOVA with Dunnett test for multiple comparisons to WT controls. e Representative images of typical Stage 4-7 clusters stained for pancreatic progenitor markers (PDX1, NKX6.1), endocrine marker (CHGA), and islet cell types (INS, GCG, SST, PPY) Nuclei were counterstained with DAPI (blue). Scale bars, 100 μm. f Quantification of CHGA+ cells in Stage 5 clusters. The percentage of CHGA+ cells was normalized to the total number of PDX1+ cells. n = 4 independent differentiations, ns not significant, **p < 0.01, one-way ANOVA with Dunnett test for multiple comparisons to WT controls. g Summary of the presence of islet cell types in WT, PDX1 KO, and RFX6 KO hESC-derived Stage 7 clusters. h Schematic summary of our stem cell modeling with WT, PDX1 KO, and RFX6 KO hESC lines. Created with BioRender.com released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license. Bulk and budding differentiation were induced by 100 ng/mL GDF8 plus 3 μM and 1.5 μM CHIR99021 for Stage 1 Day 1, respectively, using the indicated hESC lines. Source data are provided as a Source Data file.

Full size image

To further explore the utility of our in vitro system in modeling other pancreatic diseases, we examined the impact of RFX6 deficiency using a RFX6 KO hESC line³², given RFX6 gene mutations can cause pancreas hypoplasia³⁴. RFX6 is required for pancreatic progenitor differentiation and maintenance of mature alpha and beta cell function^34,35,36,37. However, it is presently unknown whether RFX6 directly affects early pancreatic morphogenesis. To address this, we assessed the impact of RFX6 deficiency using both differentiation pipelines. While RFX6 KO did not affect DE specification, it significantly compromised the induction of PDX1+ cells (KO: 44.3% ± 1.2% versus WT: 94.1% ± 3.8%) and PDX1+/NKX6.1+ pancreatic progenitors (KO: 36% ± 0.9% versus WT: 75.5% ± 2.1%) in our bulk differentiation protocol (Fig. 4a–d). The reduction of PDX1+ cells to ~44% of RFX6 KO Stage 4 cultures in the bulk pipeline predicted a switch to a budding-type differentiation pattern. Indeed, PDX1+ cell budding morphogenesis occurred (Fig. 4e). Despite the bud formation, subsequent commitment to CHGA+ endocrine precursors was markedly impaired (Fig. 4e, f). In the budding differentiation protocol, the number of PDX1+ cells was further decreased in RFX6 KO Stage 4 cultures (16.3% ± 2.4%). Morphologically, these sporadic PDX1+ cells formed even smaller buds (Fig. 4e and Supplementary Fig. 11b). In both differentiation pipelines, cells with RFX6 deficiency failed to secrete insulin or generate normal islet cell types, interestingly with the exception that PPY+ cells which seemed to develop normally. Rare INS+ and GCG+ cells were detected; however, they were INS+/GCG+ bi-hormonal cells, suggesting functional incompetency (Fig. 4e, g and Supplementary Fig. 11c, d). Taken together, our stem cell modeling not only corroborates previous findings but also uncovers an unexpected role of RFX6 in regulating pancreatic cell patterning and potentially the early pancreas morphology.

Distinct transcriptomic profiles of islet buds and main bodies

PDX1+ cell budding morphogenesis occurs in a heterogeneous cell population in our budding models. As islet cells were predominantly derived within the PDX1+ cell-enriched bud niche, we investigated the cellular identity of the main bodies (i.e., the PDX1-negative compartment) and the role of main body cells in islet bud differentiation. To address these questions, we purified islet buds and main bodies by enzymatic isolation and sorted the two compartments (PDXeG-positive buds and PDXeG-negative bodies) for bulk RNA sequencing (RNA-seq) (Fig. 5a, b). RNA-seq analysis revealed that bud and body cells exhibited very distinct transcriptomic profiles (Fig. 5c–e). Specifically, bud cells were enriched for PDX1 and NKX6.1 transcripts as well as expressed higher levels of genes associated with endocrine cell commitment (NGN3, NEUROD1, CHGA, ISL1, RFX6, MNX1, FEV, PAX4, ARX), islet cell types (INS, GCG, GHRL) and beta cell function (GCK, G6PC2, ABCC8, KCNJ11, SIX2, UCN3) (Fig. 5f). By contrast, main body cells expressed higher levels of genes associated with ductal cell markers (SOX9, KRT18, MUC1, SPP1, TPM1), cell proliferation (PCNA, CDK2, MCM2, MCM3), and pancreatic progenitor state maintenance (YAP1, NOTCH1, NOTCH2, NOTCH3, HES1, WNT3A) (Fig. 5f). Functional enrichment revealed that pathways of cell cycle, Hippo-YAP mechanosensitive signaling, extracellular matrix (ECM)-receptor interaction and axon guidance were up-regulated in main bodies (Fig. 5g, h). We confirmed the selected differentially expressed genes between buds and main bodies using qPCR (Fig. 5i–l).

a Bulk RNA-seq design for characterizing islet bud and main body. Created with BioRender.com released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license. b Representative images before and after enzymatic isolation (see Methods). The HUES4 PDXeG line was used for differentiation and for facilitating the sorting of GFP+ buds and GFP- bodies. Scale bars, 100 μm. Principal component analysis (c) and sample correlation (d) showing overall distinct transcriptomic profiles of islet buds and main bodies. n = 4 independent differentiations. Differential expression analysis showing all differentially expressed genes (DEGs) in (e), with selected DEGs in buds and main bodies highlighted in (f). Functional pathway enrichment analysis in buds (g) and main bodies (h). n = 4 independent batches of differentiations. Over-representation analysis with one-sided Fisher’s exact test, p values adjusted for multiple testing by Benjamini-Hochberg method. i–l qPCR assays validating expression of selected genes in buds and main bodies. n = 4 independent differentiations, ns not significant, *p < 0.05, **p < 0.01 versus GFP- body, unpaired two-tailed t-test. Budding differentiation was induced by 100 ng/mL GDF8 plus 1–1.5 μM CHIR99021 for Stage 1 Day 1 using the HUES4 PDXeG line. Source data are provided as a Source Data file.

Full size image

In addition to the transcripts related to ductal lineage and progenitor state, we noted that main body cells expressed higher levels of pancreatic mesenchyme markers (THY1, VIM, COL1A1, PBX1, PBX3) as well as SLIT1 and SLIT2 (Fig. 5f, j, k). The Robo/Slit pathway is reported to be involved in many aspects of pancreas organogenesis and Slit ligands function as important pro-endocrine mesenchymal signal factors in mouse and human pancreatic tissues^38,39,40. With ROBO2 receptor robustly expressed in bud cells (Fig. 5f, k), the elevated expression of SLIT1 and SLIT2 in main body cells may provide surrounding cues beneficial to the differentiation and/or function of islet buds. However, Stage 6 islet buds separated from main bodies continued to differentiate and acquired comparable function (Supplementary Fig. 12), arguing against the requirement for crosstalk between the bud and main body during late stages, but an earlier role of body cells in bud morphogenesis/differentiation cannot be excluded.

To further characterize main body cells, we immunostained several differentially expressed candidates as informed by RNA-seq and qPCR analysis. Specifically, we revealed (1) remarkably higher expression levels of THY1 and VIM (pancreatic mesenchyme/fibroblast markers) in the main bodies; (2) mutually exclusive expression of the ductal cell marker SOX9 in main body cells and the endocrine cell marker NEUROD1 in bud cells; (3) higher expression level of YAP1 (a mechanosensitive signal for balancing progenitor cell self-renewal and differentiation) in the body cells versus lower expression of YAP1 in the NEUROD1+ endocrine-committed bud cells; and (4) higher expression of the anterior foregut marker SOX2 in body cells. We also examined the exocrine cell marker Trypsin 1/2/3 but did not detect any expression in either bud or body cells, ruling out the exocrine identity of main body cells (Supplementary Fig. 13). Collectively, these results suggest a potential pancreatic mesenchyme/fibroblast phenotype or anterior foregut/ductal lineage commitment but not exocrine lineage of the main body cells. Nevertheless, pathways associated with cell apoptosis, DNA base excision repair, and mismatch repair were highly enriched in main body cells (Fig. 5h), suggesting that the identity of these cells cannot be properly maintained when entering endocrine induction stages.

Involvement of EphB3/4 signaling in the PDX1+ cell budding morphogenesis

Our RNA-seq results identified candidate signaling pathways that could be involved in the PDX1+ cell budding process. Indeed, we noted several pathways that are differentially enriched in the buds versus main bodies and are also known to mediate cell migration or cell compaction. These include TGFβ signaling (epithelial-to-mesenchymal transition, EMT)⁴¹, RhoA/ROCK (cytoskeletal remodeling)⁴², EphA/EphrinA^43,44,45, and Robo/Slit (axon guidance cues)³⁸. However, addition of specific inhibitors or activators targeting these signaling pathways at the budding stage did not impact PDX1+ cell clustering (Supplementary Fig. 14). A recent study reports that Wnt signaling separates the progenitor and endocrine compartments⁴⁶; however, modulation of the Wnt pathway by supplementing cell cultures with Wnt activators/inhibitors at the budding stage did not affect PDX1+ bud formation or bud/body segregation in our spheroid system (Supplementary Fig. 14).

Guided by RNA-seq data, we noted that among axon guidance cues, EphB/EphrinB were differentially expressed in buds and main bodies (Fig. 5f). qPCR and immunostaining confirmed higher expression levels of EphB (EPHB1, EPHB2, EPHB3, EPHB4, EPHB6) and EphrinB (EFNB1, EFNB2, EFNB3) in main bodies relative to islet buds (Fig. 5l and Supplementary Fig. 15). Notably, EphB/EphrinB is known to regulate cell sprouting, cell type segregation and boundary formation during tissue development^47,48,49,50. For example, EphB3b/EphrinB1 signaling is reported to orient hepatoblasts migration and liver bud formation⁵¹. Particularly in the pancreas, EphB signaling is required for proper pancreatic epithelium branching morphogenesis and EphB3 is found to be transiently expressed in delaminating endocrine-committed cells^52,53. To examine whether EphB signaling could affect PDX1+ cell clustering/budding morphogenesis, we added EphB inhibitors during Stage 5 when budding occurred (Fig. 6a). Strikingly, EphB3 and EphB4 specific inhibitors suppressed PDX1+ cell clustering in a dose-dependent manner, indicating that both EphB3 and EphB4 signaling are involved in the budding morphogenesis by preventing the intermixing of PDX1+ cells and non-endocrine main body cells (Supplementary Fig. 16). Indeed, PDX1+ cells displayed the highest extent of scattering when EphB signaling was blocked by pan-EphB inhibition (Supplementary Fig. 14). Compared to the clustered phenotype of PDX1+ cells in vehicle controls, surface plot analysis and live imaging showed interrupted PDX1+ cell clustering and the absence of budding morphogenesis when EphB3/4 signaling was perturbed (Fig. 6b–e).

a Study design for pharmacological perturbation of EphB signaling at budding stage. Created with BioRender.com released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license. b, c Representative images (b) of Stage 5 clusters that were treated with DMSO vehicle or a combination of EphB3/4 inhibitors at budding stage. Surface plot analysis (c) showing PDX1+ cell distribution patterns in representative Stage 5 clusters as indicated by arrowheads in (b). Scale bar, 300 μm. EphB3/4 inhibition disrupted the consolidated structure of PDX1+ cells (d), and a schematic summary of phenotypes (e). Scale bar, 150 μm. Created with BioRender.com released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license in (e). f, g Representative images (f) of Stage 5-7 clusters that were treated with DMSO or EphB3/4 inhibitors at budding stage. Quantification by histogram and Gaussian fitting (g) showing a left shift of INS-GFP intensities in Stage 7 clusters that had been treated with EphB3/4 inhibitors. n = 50 clusters for each from 4–5 independent differentiations. Scale bar, 150 μm. h Static GSIS assays showing insulin secretion from Stage 7 clusters that were treated with DMSO or EphB3/4 inhibitors. n = 4 independent differentiations, unpaired two-tailed t-test (*p < 0.05, **p < 0.01) for comparisons between DMSO control and EphB3/4 inhibition at indicated secretagogue, one-way ANOVA with Tukey test (with different letters) for multiple comparisons between secretagogues within same treatment group. i Representative images of Stage 5 clusters (that were treated with DMSO or EphB3/4 inhibitors) stained for endocrine precursor marker NGN3. Data were normalized to PDX1. n = 4 independent differentiations, unpaired two-tailed t-test (**p < 0.01) for comparisons between DMSO control and EphB3/4 inhibition. Scale bar, 100 μm. Representative flow plots (j) and quantification (k) of Stage 5 entire clusters and Stage 7 isolated buds (that were treated with DMSO or EphB3/4 inhibitors) examined for various markers. n = 4 independent differentiations, ns not significant, *p < 0.05, **p < 0.01, one-way ANOVA with Dunnett test for multiple comparisons to DMSO control clusters. Budding differentiation was induced by 100 ng/mL GDF8 plus 1–1.5 μM CHIR99021 for Stage 1 Day 1 using the HUES4 PDXeG line (b–d) and Mel1 INS^GFP/W line (f–k). Source data are provided as a Source Data file.

Full size image

Next, we assessed how disrupting PDX1+ cell budding morphogenesis impacts subsequent endocrine cell development in our stem cell differentiation systems. As the mechanism of action for EphB inhibitors is reversible, we transiently added EphB3/4 inhibitors during Stage 5 and removed inhibition at later stages (Fig. 6a). This EphB3/4 co-inhibition impaired islet bud formation and INS+ cell induction in Stage 6-7 cultures (Fig. 6f). Quantification showed that INS-GFP fluorescent intensities (mean values by Gaussian fitting: 36 ± 2 versus 50 ± 3 of control) and insulin secretion of Stage 7 islet buds that had been treated with EphB3/4 inhibitors were lower than those of vehicle control islet buds (Fig. 6g, h), indicating that the establishment of PDX1+ bud niche and endocrine aggregates is essential for beta cell formation and function. The transient inhibition of EphB signaling during Stage 5 did not impact the proportion of PDX1+ cells, but its effects on interrupting PDX1+ cell clustering were associated with reductions in pro-endocrine cells (NGN3+ cells: 6.7% ± 0.8% in DMSO versus 1.7% ± 0.4% in EphB3/4 co-inhibition; NEUROD1+ cells: 25.2% ± 1.1% in DMSO versus 11.9% ± 0.7% in EphB3/4 co-inhibition) and islet cell types (INS+ cells: 43% ± 2.6% in DMSO versus 19.5% ± 3.7% in EphB3/4 co-inhibition; GCG+ cells: 18.7% ± 1.2% in DMSO versus 8.1% ± 1.6% in EphB3/4 co-inhibition) (Fig. 6i–k). Altogether, these data demonstrate the involvement of EphB3/4 signaling in the PDX1+ cell budding morphogenesis and highlight the importance of the local niche in early pancreatic patterning as well as islet development.

• SEO Powered Content & PR Distribution. Get Amplified Today.
• PlatoData.Network Vertical Generative Ai. Empower Yourself. Access Here.
• PlatoAiStream. Web3 Intelligence. Knowledge Amplified. Access Here.
• PlatoESG. Carbon, CleanTech, Energy, Environment, Solar, Waste Management. Access Here.
• PlatoHealth. Biotech and Clinical Trials Intelligence. Access Here.
• Source: https://www.nature.com/articles/s41467-024-50109-2