Mice used in this study included ROSA-YFP/R26R-EYFP (Jackson labs) (Srinivas et al., 2001), Hemizygous (Pdx1-Cre)6Tuv/J (abbreviated as Pdx1-Cre; Jackson labs) (Hingorani et al., 2003), and Pdx1-Cre; RosaYFP.

Single mouse and hESC-derived pancreatic colonies were fixed in 4% paraformaldehyde (PFA, Sigma) for 10–11 min at room temperature after being plated down. They then were incubated for 1 h in blocking solution containing 1X PBS, 5% normal goat serum (NGS, Jackson Immunoresearch) and 0.3% Triton™ X-100 (Sigma) to permeabilize the cells. Primary antibodies were incubated overnight at 4°C in 5% NGS in 1X PBS, and secondary antibodies were incubated for 1 h at room temperature in 5% NGS in 1X PBS. Nuclei were stained with Hoechst (Sigma, 33,258, 10 μg/mL or 1:1,000) for 10 min at room temperature. Images were taken using an inverted fluorescence AxioVision Zeiss UV microscope with an AxioCam MRm camera and AxioVision v4.6 imaging software (Zeiss) or Olympus Fluoview FV1000 confocal laser scanning microscope. Primary antibodies include anti-insulin, mouse monoclonal (1:1,000, #I2018, Sigma); anti-insulin, guinea pig polyclonal (1:200, #7842, Abcam), anti-glucagon, mouse monoclonal IgG₁ (1:2,000, #G2654, Sigma), anti-GFP, chicken polyclonal IgY (1:1,000, #GFP-1020, Aves), anti-C-Peptide, rat monoclonal (concentrated version, 1:1,000, #GN-1D4, Developmental Studies Hybridoma Bank).

hESC to beta cell cytokine differentiation was performed until the pancreatic progenitor stage, according to Korytnikov and Nostro (2016). All H1 hESC-derived DE, PF, PE, and PP cells were provided directly by the Nostro lab. HES3/ins-GFP reporter hESC-derived DE, PF, PE and PP cells (Micallef et al., 2012) were either provided directly by the Nostro lab or isolated independently according to their protocol (Korytnikov and Nostro, 2016). The ins-GFP hESC cell line was approved for use by the Murdoch Children’s Research Institute (ABN 21 006 566 972) of the Royal Children’s Hospital in Australia and The Governing Council of the University of Toronto.

When we independently differentiated the HES3/ins-GFP cells, the original vial was provided by Dr. Christina Nostro, expanded upon and frozen down into multiple vials for later use. To start the differentiation, a frozen vial was thawed across about 3 wells of a 6-well plate using hES media. The hES media components included 200 mL of DMEM/F12, 50 mL of knockout serum replacement multispecies, 3.2 mL of sodium bicarbonate 7.5% solution, 2.5 mL of Glutamine, 2.5 mL of MEM non-essential amino acids solution, 2.5 mL of penicillin/streptomycin, 0.5 mL sodium pyruvate solution, 0.250 mL of 2-mercaptoethanol, and 7 ng/mL FGF2 (which was added fresh right before feeding cells each time). These materials are listed in Korytnikov and Nostro (2016). They were expanded upon until at least two full 6-well plates were covered with a monolayer of pluripotent hESCs. Their pluripotency was checked with immunostaining for pluripotency markers such as October 4 (data not shown). Incubation with TryplE for 3 min was always used when cells to lift cells from the plate for passaging/expansion and freezing. For freezing, freezing media containing 40% ES qualified (Wisent) FBS, 50% hES media, and 10% DMSO were used.

On Day 0 (when hESCs are ready to start the differentiation process), 1% L-Glutamine, 3 μL/mL of diluted monothioglyrcerol (MTG), 100 ng/mL Activin A, and 2 μM CHIR99021, all diluted in RPMI were used to feed cells. On Day 1, 1% L-Glutamine, 3 μL/mL of diluted MTG, 100 ng/mL Activin A, 5 ng/mL FGF2, and 10uL/mL ascorbic acid, all diluted in RPMI were used to feed cells. By Day 3, cells have reached the DE stage. On Day 3, 1% B27, 3uL/mL of diluted MTG, 1% L-Glutamine, 50 ng/mL FGF10, and 0.75 μM Dorsomorphin, all diluted in RPMI media were used to feed cells to bring cells to D6 (PF stage). On D6, 1% L-Glutamine, 10 μL/mL ascorbic acid, 1% B27, 50 ng/mL Noggin, 50 ng/mL FGF10, 0.25μM SANT1, and 2 μM retinoic acid (RA), all diluted in DMEM HG media were used to refeed cells until D8 at which point cells transition to the PE stage. On D8, 1% L-Glutamine, 10 μL/mL ascorbic acid, 1% B27, 50 ng/mL Noggin, 100 ng/mL hEGF, and 10 mM Nicotinamide, all diluted in DMEM HG were used to refeed cells until Day 13, at which point cells transition the PP stage.

Ascorbic acid was reconstituted in water at 10 mM stock concentration and stored at −20°C. Nicotinamide was reconstituted in IMDM media and aliquoted into ready-to-use vials at a stock concentration of 1,637.7 mM and stored at −20°C. RA was stored in aliquots at −80°C. Feeding of cells that involved RA was done in the dark. Noggin, L-Glutamine, B27, FGF2, and hEGF were stored at −20°C. All other components were stored at 0°C.

At each stage of differentiation, DE, PF, PE, and PP cells underwent a clonal sphere-forming assay, which is a highly reproducible assay (Reynolds and Weiss, 1992; Coles-Takabe et al., 2008; Pastrana et al., 2011). Stem cell frequency is calculated based on the number of clonal spheres generated from a given tissue sample. Briefly, cells were lifted and dissociated with TryplE Express (Gibco). Incubation times with TryplE Express were as follows: DE: 3 min, PF: 3 min, PE: 7 ½ mins, PP: 3 min), and centrifuged at 350 g for 5 min in 1X PBS and 10 μM Rock Inhibitor (RI). Cells were resuspended in defined serum-free medium (SFM) (Tropepe et al., 1999) containing 10 ng/mL basic fibroblast growth factor (FGF2) (Sigma), 20 ng/mL EGF (Sigma), 2 mg/mL heparin (Sigma), 1 × B27 Supplement (Gibco-BRL) and 10 μM Rock Inhibitor (RI). Cells were counted using Trypan Blue exclusion and hemocytometer and immediately replated at a low clonal density in SFM/FGF2/EGF/heparin/1xB27/RI either at clonal density (Coles-Takabe et al., 2008) in 24-well ultra-low attachment plates (3473/07-200-600, Fisher/Sigma). Specifically, DE- and PP-derived single cells were plated at 10 cells/μL because sphere production from these stages was minimal and plating at 10 cells/μL density maintained clonality where no fusion of spheres occurred but sphere count was not too sparse as to be difficult to count and obtain a sufficient number of spheres. PF- and PP-derived single cells were plated at 1 cell/μL because too many spheres are produced when they are plated at 10 cells/μL, which results in sphere fusion and a loss of clonality. Reducing the plating density to 1 cell/μL was needed to ensure clonality for PF- and PE-derived spheres in the clonal sphere-forming assay.

Cells then were incubated for 1 week without moving to reduce the incidence of non-clonal colonies (Coles-Takabe et al., 2008). To assess differentiation potential, single colonies were removed from the aforementioned mitogen-containing media and transferred to wells coated with either Geltrex or Laminin 411, in SFM containing Stage 3 media and grown for an additional 1 or 14 days before fixation and analysis. To ensure an accurate assay of the sphere progeny, each well of a plate (Nunc) contained only one clonal DE-, PF-, PE-, or PP-derived colony.

The morning of the first appearance of vaginal plugs was regarded as 0.5 days post conception (dpc). Pdx1-Cre; RosaYFP E11 was dissected under a dissection microscope (light microscope), where the location of pancreatic bud was located using the Greggio et al. (2014) video, where they showed a dissection of E10.5 pancreatic bud. In brief, the embryonic dissection was performed in PBS, and the digestive tract was identified and removed. From the successful isolation of an intact digestive tract, the pancreatic bud was identifiable and was isolated using forceps.

Next, the dorsal pancreatic bud that was isolated from E11 tissue was triturated in TryplE 10-20X. The resulting single cells from E11were centrifuged in 1X PBS or SFM/FGF2/EGF/heparin/1xB27 (both had produced similar results) and resuspended in SFM/FGF2/EGF/heparin/1xB27. Following trituration and centrifugation, cells were counted with Trypan Exclusion and plated at a cell density of 10 cells/µL to ensure clonality in SFM/FGF2/EGF/heparin/1xB27 in ultra-low adhesion plates. The plates then were incubated for 1 week without moving. This part constitutes the sphere-forming assay, which is a highly reproducible assay (Reynolds and Weiss, 1992; Coles-Takabe et al., 2008; Pastrana et al., 2011). Stem cell frequency is calculated based on the number of spheres generated from a given tissue sample.

Data are presented as means ± standard errors of the means (SEM). Statistical analyses were performed using GraphPad Prism 9 (GraphPad Software Inc.). Student’s t-tests (two-tailed) were performed for statistical analysis between two groups. A one-way ANOVA or two-way ANOVA with a Holm-Sidak post hoc test (pairwise or versus control comparison) was used when three or more groups were compared. Sample sizes (N) and p-values are provided in the figure legends. Statistical significance was set at p < 0.05.

To determine the embryonic timepoint in which pancreatic stem cells emerge using the hESC to beta cell model, hESCs were differentiated towards the DE, PF, PE, and PP stages as previously described (Korytnikov and Nostro, 2016). All H1-derived DE, PF, PE, and PP cells were obtained fresh directly from the Nostro lab, where they frequently ensure the identity of the cells through flow cytometry at these four initial stages. Although we obtained some HES3/ins-GFP-derived DE, PF, PE, and PP cells directly from the Nostro lab as well, this was only at the beginning of our project for a proof of concept. We were able to later independently differentiate the HES3/ins-GFP line to continue with the majority of our experiments, and we were able to perform flow cytometry to qualify our own differentiation. Through our own differentiation HES3/ins-GFP cells transitioned to the DE stage with about 93% of cells co-expressing specific DE markers CXCR4 and CD117; the PE stage, where about 96% of cells were PDX1 + NKX6.1-; and to the PF stage where about 95% of cells were PDX1 + NKX6.1+ (Supplementary Figure S1). This shows we were able to obtain relatively pure populations of the desired stages.

The cells, once they reached each of the first four stages, were dispersed and passed through a 40 μL cell strainer before plating at a low cell density in low cell-adhesion 24-well plates. Specifically, DE- and PP-derived cells were plated at 10 cells/μL because sphere production from these stages was minimal and plating at 10 cells/μL density maintained clonality. However, PF- and PE-derived cells were plated at a much lower density of 1 cell/μL because too many spheres are produced when they are plated at 10 cells/μL or even at 5 cells/μL, which results in sphere fusion and a loss of clonality. Reducing the plating density to 1 cell/μL ensured clonality when cells from the PF and PP stages were plated in the clonal sphere-forming assay. Coles-Takabe et al. (2008) were able to show that by plating at low density (greater than single cells per well), spheres can grow clonally. After sphere growth, stem cell frequency from the original hESC monolayer culture at each of the four stages of differentiation was determined based on the number of spheres generated within the sphere assay.

We used DE, PF, PE, and PP cells from the H1 and HES3/ins-GFP lines (the latter will be referred to as ins-GFP from now on), for the initial primary sphere assay to compare sphere number results between them. The H1 and ins-GFP cell lines require the same cytokines, with the same exposure times, during hESC differentiation (Nostro et al., 2015; Nostro et al., 2011). Therefore, drastically different outcomes in sphere generation between the 2 cell lines would not be expected. The frequency of hESC-derived PF and PE-generated spheres from the ins-GFP cell line was about 1.6 and 2.6-fold higher than that of H1, respectively (Figure 1D). However, the overall pattern of sphere formation between the different stages was similar between H1 and ins-GFP lines (Figures 1A, B, D). The varying frequencies of spheres generated across the stages indicate that each stage carries varying proportions of proliferating sphere-forming cells.

The ins-GFP reporter line was used to quickly visualize whether any spheres generated were ins-GFP+ and, therefore, part of the pancreatic lineage. About 50% of all spheres generated from the PE stage and about 10% of all spheres generated from the PP stage were ins-GFP+ (Figure 1C). Since insulin is a marker of adult pancreatic precursor cells (Smukler et al., 2011), insulin + cells within the sphere isolated at these early developmental time points are predicted to originate from an insulin + pancreatic stem cell or progenitor. There is more than a four-fold reduction in the percentage of ins-GFP + sphere formation between the PE and PP stages, as well as a significant reduction in the frequency of the total number of spheres (Figures 1A–C). We suggest that the difference in environmental factors between the two stages affects the survival and/or proliferation of both ins-GFP + spheres and ins-GFP- sphere-forming cells. To start characterizing whether these ins-GFP + cells in these primary spheres had stem properties, we assessed their ability to self-renew by bulk passaging the ins-GFP+ and ins-GFP- PE primary spheres. Upon secondary sphere formation, we saw that 77% of secondary spheres were ins-GFP+, suggesting that most of the insulin + cells (within the primary sphere) are precursors that are sphere-initiating cells and, therefore, have stem-like qualities (Supplementary Figures S2A, B). The other 20% of secondary spheres that do not contain ins-GFP + cells were either derived from the ins-GFP- primary spheres, which in turn were derived from a non-pancreatic lineage, or they were from ins-GFP + primary spheres and the ins-GFP + stem or precursor cell lost its insulin expression during secondary sphere growth.

Similar-sized single ins-GFP + PE spheres were incubated with PE stage media on either Geltrex or laminin for 14 days. We observed that the survival of cells within clonal spheres was significantly greater when plated on laminin compared to when plated on Geltrex (Supplementary Figure S3A). The cells also remained closer together on laminin than on Geltrex, but the results were not significant (Supplementary Figure S3B). Future experiments assessing the multipotential nature of PE-generated ins-GFP + spheres should focus on using laminin as the extracellular matrix.

Polyhormonal cells emerge starting at E8.5 in the developing mouse pancreas and during the PE stage of hESC differentiation (Korytnikov and Nostro, 2016; Nostro et al., 2011; Nostro et al., 2015; Gittes et al., 1996). We wanted to determine if the ins-GFP + cells within spheres generated from the PE stage are polyhormonal for glucagon. Due to the enhanced survival (Supplementary Figure S3A) and compactness (Supplementary Figure S3B; although the difference in compactness was not significant) of the cells in our initial observation with laminin versus Geltrex and due to other studies showing that the prevention of cell migration/spreading by laminin (versus Geltrex) in the plate promote pancreatic identity (Mamidi et al., 2018), the PE generated spheres plated on laminin were further analyzed. Therefore, in this case, ins-GFP + spheres were stuck down on the laminin extracellular matrix in the presence of PE stage media for 1 or 14 days before fixation and immunostained for C-peptide, glucagon and GFP. After 1 day, 100% of the total number of ins-GFP+/C-peptide + cells in the spheres analyzed were co-labeled with glucagon (Figure 1E). This indicates that all insulin + cells within PE-generated spheres are polyhormonal with glucagon, and that therefore, they are immature and may be primed to differentiate into the alpha cell lineage, according to Nostro et al. (2015). All C-peptide + cells also expressed ins-GFP, however, some cells expressed GFP weakly and other cells expressed GFP strongly (Figures 1F–J).

Following a 14-day incubation in Stage 3 media, the C-peptide + cells in ins-GFP + spheres either no longer expressed glucagon or expressed significantly less glucagon (Figures 1E–O). Although no lineage tracing was done in this experiment, all C-peptide + cells in hESC-derived PE spheres were initially polyhormonal for glucagon. No single labelled insulin + or single labelled glucagon + cells were seen when spheres were fixed and stained 1-day post-fixation (Figure 1E). When hESC PE generated ins-GFP + spheres were plated on laminin and re-exposed to stage 3 media for a prolonged period of 14 days, C-peptide + cells were still seen, and many of these C-peptide + cells no longer had glucagon expression overlapping them, and those that had still glucagon expression overlapping them had faint levels of this glucagon expression (compared to what had been previously co-expressed in the initial sphere 1-day post-fixation) (Figure 1E).

Although the above results are not conclusive due to the absence of a lineage tracing experiment to trace the descendants of the polyhormonal cells, they suggest, indirectly, that these cells, given enough time in PE Stage media, may show a trend of changing their identity towards monohormonal insulin + progenitors. However, after 14 days, C-peptide also was less strong in expression as well, indicating that optimizing the maintenance of beta cell identity may need further studies.

To determine whether clonal sphere-producing pancreatic precursors first emerged at the PE stage in mice as they did in the hESC to beta cell model, we dissected the developing pancreas at E8.5-9, and single cells were cultured in a sphere-forming assay. After sphere growth, stem cell frequency in the original tissue sample was determined based on the number of spheres generated within the sphere assay. To ensure that the spheres obtained are from pancreatic PDX1 lineage origin, we utilized the early pancreas/duodenum specification marker PDX1 and the Pdx1-Cre; RosaYFP lineage tracing mouse. Homogenously labelled PDX1-YFP+ spheres from mouse E8.5-10 could not be isolated. However, homogenously labelled PDX1-YFP+ spheres were isolated at E11, which corresponds to the pancreatic progenitor developmental stage. Single similarly sized PDX1-YFP+ spheres isolated from the E11 PP stage were plated into a sphere differentiation assay with 1% serum and were immunostained for GFP and either insulin or glucagon. About 16% of progenitors within the sphere differentiated into glucagon + cells, and a separate 8.5% of the progenitors differentiated into insulin + cells (Figure 2B). At the PP stage, progenitors are known to be able to differentiate into three lineages: endocrine, exocrine, and duct cells (Gu et al., 2002). However, due to the limited number of spheres obtained, we did not test for the presence of exocrine and duct cells (via immunocytochemistry), and therefore, it remains unknown whether the precursors isolated at this stage are truly multipotential. Since previously it was shown that the adult mouse PMP could differentiate into various endocrine and amylase cells (Seaberg et al., 2004; Smukler et al., 2011; Arntfield and van der Kooy, 2013), it might be the case for E11 precursors as well, but this is not conclusive. Nonetheless, E11 cells having the ability to differentiate into monohormonal glucagon and insulin + cells indicates that the isolated E11 spheres are from pancreatic origin and that there is a trend toward possible multipotentiality of its progenitors. In the differentiation assay following sphere growth for E11 spheres, there may still be proliferation going on during the differentiation for E11 (Figure 2B), and if the cells are fully differentiating, we would want to stain for the proliferation marker ki67 to make sure proliferation is not continuing. However, this study was unable to conduct this experiment due to limited spheres.

Figures 2C–J shows clearly labelled differentiated glucagon + cells and insulin + cells derived from E11 mouse pancreatic anlage, revealing that pancreatic stem cells do derive from E11 mouse pancreas. Staining for glucagon and insulin was done separately for different spheres due to the availability of reliable antibodies. Therefore, we could not determine whether or not there was any insulin+/glucagon + co-expression in differentiated spheres. Since the mice used in this experiment were Pdx1-Cre; Rosa YFP (where YFP indicates that the cell was derived from the PDX1 pancreatic lineage regardless of whether or not that cell is currently producing PDX1), all cells that were fully differentiated or were in the process of differentiation should also express YFP. The insulin + cells from the E11 differentiation assay were not YFP+ (Figures 2C–F). This may indicate that the PDX1 lineage marker turned off in these cells during differentiation or sphere growth or that YFP expression had become too faint to see once these insulin + cells underwent differentiation. Some of the glucagon + cells from the sphere differentiation assay were not labelled with YFP and some were labelled with YFP strongly (Figures 2G–J). Again, for the glucagon+/YFP- cells, either the lineage marker had turned off during glucagon differentiation, or YFP became too faint to see once differentiation occurred. It has been previously shown that the Pdx1-Cre; Rosa YFP mouse strain can undergo transgene silencing (Arntfield and van der Kooy, 2013), which could occur at any stage in vivo or in vitro. It is undetermined if the differentiated insulin and glucagon cells were actively expressing PDX1 in these cells.

It is interesting to note that there are also YFP + cells with processes (Figures 2C–F). Although this indicates that the PDX1 lineage gives rise to neurons, the differentiated insulin + cells are distinct cells without processes. This aligns with previous studies from our lab (Smukler et al., 2011; Arntfield and van der Kooy, 2013; Brokhman et al., 2019), which reported that adult mouse PMPs (from the PDX1 lineage) also give rise to both beta cells and a small number of separate neurons that are also distinct and separate from the neural crest lineage.

Some glucagon + cells appear to overlap with cells that may have processes (Figures 2G–J). It could be that these glucagon + cells (that do not have processes themselves) are overlapping but not double-labeled with the neurons. Another explanation could be that since these YFP + processes from glucagon + cells are significantly fainter and less robust than the processes coming from the non-insulin cells (presumably neurons) in C-F, the faintly labelled YFP processes coming from glucagon + cells in G-J may indicate an artifact. Due to the limited number of spheres, the passaging capability was not tested.

Homogenous PDX1-YFP+ spheres also were isolated at E16. Passaging of E16 primary pancreatic spheres resulted in an ∼11-fold increase in the number of secondary spheres compared to the number of primary spheres (Figure 2A). This suggests that the original sphere-forming cells of the primary spheres have self-renewal stem cell properties and, further, that this pancreatic stem cell (which was the origin of the clonal sphere) must have engaged in multiple rounds of symmetric division, expanding the stem cell population, and thereby expanding the number of secondary spheres we see in vitro. There were fewer tertiary spheres than secondary spheres after secondary sphere passaging (Figure 2A). This suggests that after the earlier pancreatic stem cell symmetric expansion, the stem cells underwent some proliferative exhaustion, whereby proliferation reduced, switching from symmetric division to asymmetric division during secondary sphere growth, reducing the number of tertiary spheres seen in vitro (Figure 2A). Although hESC-derived PE spheres observed limited self-renewal, most secondary spheres were insulin+ (Supplementary Figure S2), suggesting that the starting sphere-forming cell is insulin+ (or less likely, that insulin + cells turned on after clonal proliferation commenced).