ABSTRACT
Organoids have become one of the fastest progressing and applied models in biological and medical research, and various organoids have now been developed for most of the organs of the body. Here, we review the methods developed to generate pancreas organoids in vitro from embryonic, fetal and adult cells, as well as pluripotent stem cells. We discuss how these systems have been used to learn new aspects of pancreas development, regeneration and disease, as well as their limitations and potential for future discoveries.
Introduction
The pancreas plays an important role in energy homeostasis. Its exocrine function is ensured by acinar cells, which produce digestive enzymes, and ductal cells, which secrete a protective mucous layer and bicarbonate (Fig. 1A). This basic fluid neutralises the acidic juice of the stomach. The exocrine cells are the most abundant, and they form a tree of hierarchical ducts starting in the duodenum and ending at distally located acini (Fig. 1A). Interspersed in the exocrine pancreas are the islets of Langerhans, which ensure the endocrine function of the organ (Fig. 1A). Endocrine cells here include β cells, which produce insulin to lower blood glucose and α cells, which produce glucagon to increase blood glucose. Other hormones, such as somatostatin and pancreatic polypeptide (PP), are secreted by δ and PP cells, respectively. Although much is known about pancreas development from work in model organisms, predominantly mice (Fig. 1B-D) and zebrafish, in vitro investigations, notably using explants, have played an important role (Cozzitorto and Spagnoli, 2019; Larsen and Grapin-Botton, 2017; Prince et al., 2017). This was recently complemented by cultures of dissociated cells from the embryonic, fetal or adult murine pancreas seeded in 3D in gel-forming extracellular matrices, the so-called organoids. In vitro models have also enabled the recreation of human development starting from pluripotent stem cells (PSCs) or fetal cells, initially with 2D cultures, and more recently in 3D (Gaertner et al., 2019; Petersen et al., 2018).
Comparison of cell types and morphogenesis between the developing pancreas and pancreatic organoids. (A) Schematic of the fully developed pancreas of a mouse embryo. The entire organ is in pink, and its branched ductal network in black. Fluid in the lumen is in green. The tips of the ductal network are magnified in the middle panel, representing the different cell types. The right panel shows a whole-mount immunostaining of the network (mucin; MUC1) and endocrine islets (insulin; INS). (B,C) 3D view of a mouse pancreas at E10.5 (B) and E11.5 (C) with related schematics below. H2B-GFP labels slow-dividing cells after a pulse-chase labelling of PDX1+ cells. (D) 3D view of a pancreas at E12.5 and E14.5. The luminal network and organ outlines are shown below. Images in A courtesy of Lydie Flasse and B-D courtesy of Hjalte List Larsen. (E) Example of an epithelial organoid. Confocal section of a pancreatosphere grown in Matrigel for 4 days from dissociated E10.5 mouse pancreas. Courtesy of Phil Seymour, using the method from Greggio et al. (2013). (F) Example of a complex epithelial organoid. Maximum intensity projection of a 7-day pancreas organoid grown in Matrigel from dissociated cells from E10.5 pancreas. Courtesy of Siham Yennek, using the method from Greggio et al. (2013). (G) Example of a multi-tissue organoid. Maximum intensity projection of a 3D confocal stack of a mouse pancreatoid with epithelium (PDX1) and mesenchyme (vimentin; VIM). Reproduced from Scavuzzo et al. (2017) where it was published under a Creative Commons CC-BY license (http://creativecommons.org/licenses/by/4.0/). (H) Example of a multi-organ organoid showing the apposition of posterior gut endoderm produced from human pluripotent stem cells (CDX2) and anterior endoderm (SOX2, white). Pancreas (PDX1, red) and liver (HEX, green) form at the interface. Reproduced with permission from Koike et al. (2019). Scale bars: 10 µm (A,E); 100 µm (B-D,F-H).
Comparison of cell types and morphogenesis between the developing pancreas and pancreatic organoids. (A) Schematic of the fully developed pancreas of a mouse embryo. The entire organ is in pink, and its branched ductal network in black. Fluid in the lumen is in green. The tips of the ductal network are magnified in the middle panel, representing the different cell types. The right panel shows a whole-mount immunostaining of the network (mucin; MUC1) and endocrine islets (insulin; INS). (B,C) 3D view of a mouse pancreas at E10.5 (B) and E11.5 (C) with related schematics below. H2B-GFP labels slow-dividing cells after a pulse-chase labelling of PDX1+ cells. (D) 3D view of a pancreas at E12.5 and E14.5. The luminal network and organ outlines are shown below. Images in A courtesy of Lydie Flasse and B-D courtesy of Hjalte List Larsen. (E) Example of an epithelial organoid. Confocal section of a pancreatosphere grown in Matrigel for 4 days from dissociated E10.5 mouse pancreas. Courtesy of Phil Seymour, using the method from Greggio et al. (2013). (F) Example of a complex epithelial organoid. Maximum intensity projection of a 7-day pancreas organoid grown in Matrigel from dissociated cells from E10.5 pancreas. Courtesy of Siham Yennek, using the method from Greggio et al. (2013). (G) Example of a multi-tissue organoid. Maximum intensity projection of a 3D confocal stack of a mouse pancreatoid with epithelium (PDX1) and mesenchyme (vimentin; VIM). Reproduced from Scavuzzo et al. (2017) where it was published under a Creative Commons CC-BY license (http://creativecommons.org/licenses/by/4.0/). (H) Example of a multi-organ organoid showing the apposition of posterior gut endoderm produced from human pluripotent stem cells (CDX2) and anterior endoderm (SOX2, white). Pancreas (PDX1, red) and liver (HEX, green) form at the interface. Reproduced with permission from Koike et al. (2019). Scale bars: 10 µm (A,E); 100 µm (B-D,F-H).
Organoids are miniaturised and simplified versions of an organ produced in vitro from stem or progenitor cells, self-organising in 3D and forming realistic micro-anatomy. In some realisations of these avatars, the organ shape is simplified, as seen in spheroids, which form a ball full of cells, or in hollow spheres or cysts (Fig. 1E) (Greggio et al., 2013). Organoids can also exhibit more complex structures (Fig. 1F) (Greggio et al., 2013). For the pancreas, organoids can be classified according to their complexity and the diversity of cell types they contain as epithelial organoids (Fig. 1E,F), multi-tissue organoids (Scavuzzo et al., 2017) (Fig. 1G) and multi-organ organoids (Fig. 1H) (Koike et al., 2019). For each type, the cell of origin defines further subtypes, of which fetal progenitor-derived, adult duct-derived and PSC-derived are the major classes. In addition, there are a few reports of islet-derived or centroacinar-derived organoids that would be worthy of further explorations (Rovira et al., 2010; Wang et al., 2020). Pancreatic organoids have been primarily produced from two species, mouse and human (Fig. 2A), but can also be produced from other species (Montesano et al., 1983; Weegman et al., 2016).
Culture methods used to grow pancreas organoids from tissue samples. (A) Sources of cells used to grow pancreas organoids include mouse and human pancreatic tissues originating from embryonic, fetal or postnatal stages, up to adults. (B) One of the methods uses monolayers of cells that are either dense or at clonal density from which 3D colonies bud. An example is shown in which islet-like structures emerge from a monolayer of pancreatic cells from E16.5 mouse pancreata (Saito et al., 2011). The inset shows expression of insulin (INS) and glucagon (GCG). Reproduced under the terms of the Creative Commons Attribution License. (C) Another method consists of growing cells in suspension. This is illustrated for ALDEFLUOR-positive cells (exocrine) isolated from adult mice (Rovira et al., 2010). The four panels show amylase (AMY; top left), E-cadherin (CDH1; top right), insulin (C-PEP; bottom left) and SOX9 (bottom right). Reproduced with permission from the authors. (D) The most common culture system consists of growing dissociated ductal cells in Matrigel, and sometimes subsets expressing specific surface markers after sorting. Two examples show ductal organoids from adult human pancreata from Loomans et al. (2018) (left; reproduced with permission) or individual mouse ductal cells (Huch et al., 2013) (right; reproduced under the terms of the Creative Commons Attribution License). (E) The Matrigel dome culture system is also permissive to β-cell expansion from PROCR+ cells, as exemplified by insulin staining (Wang et al., 2020). Reproduced with permission. (F) Acinar cells can also be seeded in Matrigel. The three images show a sequence of events at day 1, 5 and 8 after seeding (left to right), showing the amylase+ cells quickly losing acinar markers to become ductal (KRT19) (Wollny et al., 2016). Reproduced with permission. Scale bars: 100 µm (B-D); 20 µm (E,F). Other examples for each method are specified in Table 1.
Culture methods used to grow pancreas organoids from tissue samples. (A) Sources of cells used to grow pancreas organoids include mouse and human pancreatic tissues originating from embryonic, fetal or postnatal stages, up to adults. (B) One of the methods uses monolayers of cells that are either dense or at clonal density from which 3D colonies bud. An example is shown in which islet-like structures emerge from a monolayer of pancreatic cells from E16.5 mouse pancreata (Saito et al., 2011). The inset shows expression of insulin (INS) and glucagon (GCG). Reproduced under the terms of the Creative Commons Attribution License. (C) Another method consists of growing cells in suspension. This is illustrated for ALDEFLUOR-positive cells (exocrine) isolated from adult mice (Rovira et al., 2010). The four panels show amylase (AMY; top left), E-cadherin (CDH1; top right), insulin (C-PEP; bottom left) and SOX9 (bottom right). Reproduced with permission from the authors. (D) The most common culture system consists of growing dissociated ductal cells in Matrigel, and sometimes subsets expressing specific surface markers after sorting. Two examples show ductal organoids from adult human pancreata from Loomans et al. (2018) (left; reproduced with permission) or individual mouse ductal cells (Huch et al., 2013) (right; reproduced under the terms of the Creative Commons Attribution License). (E) The Matrigel dome culture system is also permissive to β-cell expansion from PROCR+ cells, as exemplified by insulin staining (Wang et al., 2020). Reproduced with permission. (F) Acinar cells can also be seeded in Matrigel. The three images show a sequence of events at day 1, 5 and 8 after seeding (left to right), showing the amylase+ cells quickly losing acinar markers to become ductal (KRT19) (Wollny et al., 2016). Reproduced with permission. Scale bars: 100 µm (B-D); 20 µm (E,F). Other examples for each method are specified in Table 1.
Development of organoids from fetal-derived pancreata
Emergence of pancreas organoid models: from explants to spheres
In vitro models, such as 2D culture of primary cells and embryonic explants, were the precursors of organoids. Explant cultures from rodent embryonic and fetal pancreata were performed in different conditions, including on a culture plate (Burke et al., 2010; Percival and Slack, 1999), at the air-liquid interface (Chen, 1954; Duvillié et al., 2006) or in collagen (Miralles et al., 1998; Montesano et al., 1983). In vitro cultures extended from rodents to the human fetal pancreas in the 1980s, usually making use of tissue fragments (Maitland et al., 1980; Sandler et al., 1987; Tuch et al., 1985).
It was only about 20 years later that attempts were made to start with single cells from embryonic pancreata, including specific fluorescence-activated cell sorting (FACS)-isolated populations, which could initially be cultured only for a few days (Sugiyama et al., 2007). The development of cultures using a matrix produced by Engelbreth–Holm–Swarm mouse sarcoma, commonly commercialised and referred to in publications as Matrigel, enabled great progress. This matrix had been used for 3D culture of mammary spheres (Petersen et al., 1992), invasion assays and as a 2D culture matrix. Embedding cells in 3D in this gel-like matrix was an important landmark that made possible the culture of early [embryonic day (E) 10.5-E11.5] (Fig. 1B,C) fetal pancreatic progenitors dissociated to single cells. In this matrix, they formed hollow spheres (pancreatospheres) composed of epithelial cells surrounding a central lumen either in the presence (Sugiyama et al., 2013) or absence (Greggio et al., 2013) of mesenchyme (Fig. 1E, Table 1). These pancreatospheres could be expanded and passaged, although the endocrine differentiation capacity decreased with time (Sugiyama et al., 2013).
Although Sugiyama et al. (2013) showed that spheres mostly formed from single cells, this event was found to be rare by Greggio et al. (2013), who showed a cooperative effect whereby multiple cells coalesced and formed spheres with a greater efficiency than single cells. This discrepancy may be due to the composition of the culture medium, the presence of mesenchyme, or the cell density used (Table 1). Notably, SRY-box transcription factor 9 (Sox9)-expressing progenitor cells were successfully seeded, whereas, in contrast, cells expressing neurogenin 3 (Neurog3), a marker of endocrine progenitors, could not generate colonies in these conditions (Sugiyama et al., 2013). Although the culture medium may have been inadequate to support replication of endocrine-committed cells, this agreed with their reported low proliferation (Kim et al., 2015).
EpCAM-sorted progenitors from later stages (E12/E13) (Fig. 1D), embedded in Matrigel, also formed pancreatic spheres that could be expanded for long periods through regular passaging (Table 1) (Bonfanti et al., 2015). They formed duct-like structures in the presence of epidermal growth factor (EGF) and differentiated more efficiently toward endocrine cells in its absence (Bonfanti et al., 2015). Endocrine cell formation was also studied by live imaging after dissociating cells at E14.5 (Fig. 1D) and growing them in suspension supplemented with 2% Matrigel, instead of the high percentages used in prior studies (Table 1) (Bakhti et al., 2019).
Evolution towards morphogenesis: improving in vivo relevance
Although these sphere cultures are morphologically dissimilar to the branched architecture of the pancreas (compare Fig. 1E and 1A,D), they appear to recapitulate pancreas progenitors lining the fetal ducts. This is corroborated by the progenitor markers they express, notably pancreatic and duodenal homeobox 1 (PDX1) and SOX9, although their similarity to the pancreatic cells found in vivo has not been systematically assessed by single-cell sequencing. Moreover, the ability of the E10.5 cells to form endocrine and acinar cells suggests that they are progenitors, that some of them are multipotent and that the culture condition enables not only their differentiation but also morphogenesis (Fig. 1F) (Greggio et al., 2013). In contrast, the spheres reported by Sugiyama et al. did not contain acinar cells (Sugiyama et al., 2013), possibly because their culture medium did not support them or the multipotent progenitors initially expected to still be present at E11.5 were not (Larsen et al., 2017). The spheres at later stages could also form endocrine cells, although acinar cell presence was not assessed. These progenitors do not form elongated ducts in culture but instead form spheres with polarised cells forming a central lumen (Fig. 1E). This is likely the result of the lack of a symmetry-breaking event intrinsic to the epithelium or provided by the surrounding matrix.
Interestingly, when the E10.5 progenitors were cultured in Matrigel in a more complex medium than the sphere medium, they generated larger organoids with a network of ducts, as well as branches harbouring acinar cells at the tips and a few endocrine cells (Fig. 1F, Table 1) (Greggio et al., 2013, 2014). Sugiyama et al. (2013) also reported organoids that appeared similar, although only superficially studied. Media comparisons between the two methods suggest that canonical WNT pathway activation in the absence of retinoic acid enables the formation of these complex organoids. In these early studies, benchmarking of cultured cells to in vivo fetal pancreas cells was limited to about 20 markers assessed by immunocytochemistry or single-cell PCR. A better assessment of the similarity to cells at different stages of development and notably of their maturity would be important.
Mouse organoid culture without Matrigel
Scaffold-free cultures of embryonic and fetal pancreatic cells provide alternative in vitro models of pancreas development. In the absence of matrix, E10.5 progenitors in the complex organoid medium developed by Greggio et al. (2013) could be re-aggregated in round-bottom, 96-well plates with their mesenchymal cells and formed dense aggregates called pancreatoids (Fig. 1G, Table 1) (Scavuzzo et al., 2018b, 2017). Pancreatoids express markers of different pancreatic lineages, including progenitors and acinar cells, as well as about 22% of endocrine-like cells, which tend to cluster away from other cells (Scavuzzo et al., 2018b, 2017). These experiments suggest that the presence of the native mesenchyme increases endocrine differentiation and decreases acinar cell differentiation. 2D cultures of E14.5 or E16.5 pancreatic cells on gelatin proliferated as a monolayer that formed spherical aggregates made of all endocrine cell types by day 12 in culture (Fig. 2B) (Saito et al., 2011). Their function was ascertained by in vivo transplantation; however, this differentiation capacity was lost at E18.5 and thereafter.
Organoid models from human fetuses
Pancreatic organoids can also be derived from human fetuses originating from pregnancy termination (mostly 7-11 weeks post-conception). For example, Bonfanti et al. (2015) showed that small fragments dissociated from the human pancreas could, like the mouse counterparts, form hollow spheres that express pancreas progenitor markers (Table 1). The expansion could be continued for 5 months and required the presence of EGF. As in mouse, removal of EGF from the culture medium increased endocrine differentiation. Using a different culture medium, Goncalves et al. (2021) also observed the formation of a mixture of hollow spheres and dense spheroids (Table 1). The spheroids contained ducts with narrow lumens, but very few endocrine cells were observed. They compared the cultured cells with the cells in vivo by single-cell sequencing and observed that many cultured cells drifted from their original phenotype by co-expressing ductal and pancreatic progenitor markers. This suggested that the culture conditions promote ductal differentiation, but these organoids could be passaged, indicating the maintenance of pancreas progenitors (Goncalves et al., 2021). Starting from minced human pancreata from 9 to 22 weeks post-conception, Loomans et al. (2018) could grow organoids harbouring compact but seemingly branched structures, as well as swollen areas (cysts), expressing progenitor/ductal markers (Fig. 2D, Table 1). Global transcriptome analyses were performed on these organoids after passaging, and they showed greater similarity to the fetal human pancreas than to islets or adult pancreata.
Pancreas organoids derived from adult tissues
The 2D primary culture of adult pancreatic cells was reported for the three main adult cell populations (acini, ducts and islets; Fig. 2A) and provided a valuable tool, but the cells lost functionality after a few days to a week (Dodge et al., 2009; Marciniak et al., 2013; Weinberg et al., 2007). The growth of colony-initiating cells in 3D cultures had already been observed in the 1990s (Kerr-Conte et al., 1996). Ductal contaminants of human islet preparations were indeed shown to grow as cysts in rat tail collagen and more efficiently in Matrigel in the presence of serum. The cells retained a ductal identity, notably the expression of KRT1 and CA19-9. Although some endocrine cells survived in culture as independent islets, no endocrine cells were observed in the ductal epithelium before 7 days in culture, suggesting that few endocrine cells formed during culture. Nevertheless, this method did not become widely adopted in the field. A decade later, it was shown that either dissociation of islets (and contaminants) or manual handpicking of ducts led to the formation of spherical non-adherent PDX1+ pancreatic progenitor colonies in suspension, though at very low frequency (Fig. 2C) (Seaberg et al., 2004; Smukler et al., 2011). Cells expressing neuronal markers were found around these colonies, which contained a few endocrine cells. This suggested the presence of cells that were able to proliferate, but their potency and in vivo counterpart remained an enigma. Another early work adopting clonal density on low-attachment, 96-well plates revealed that a population co-expressing aldehyde dehydrogenase 1 (ALDH1) and E-cadherin (cadherin 1), potentially centro-acinar/terminal duct cells, could seed spheroids in suspension (Fig. 2D, Table 1) (Rovira et al., 2010). The resulting spheroids could be passaged and were composed of acinar and endocrine cells, as well as a small SOX9+ population. It was hypothesised that these were exocrine or centroacinar cells, although the endocrine production from such cells was surprising as in vivo lineage tracing has previously shown an early segregation of the acinar lineage (Gaertner et al., 2019; Kopp et al., 2011; Solar et al., 2009; Zhou et al., 2007). However, it is possible that in vitro cultures uncovered a potency the cells have but do not reveal in vivo.
The initial success of Matrigel-based intestinal organoid cultures, initiated from crypt-resident intestinal adult stem cells (Sato et al., 2009), led the scientific community to try similar media on other adult organs. This included fragmented or dissociated single cells, including those that were not typically recognised to be maintained by adult stem cells under homeostasis. Ductal cells isolated from mouse pancreas ducts were clonally propagated in Matrigel in a medium similar to that used for intestinal stem cells, with the addition of FGF10 (Fig. 2D), and the addition of noggin and nicotinamide was essential for expansion beyond 2 months (Table 1) (Huch et al., 2013). The authors showed the emergence of a cell population expressing LGR5, a marker that is not expressed in the pancreas unless it is injured by partial duct ligation. Following aggregation with embryonic pancreatic cells, some endocrine differentiation potential of the propagated ductal cells was revealed after transplantation in vivo under the kidney capsule (Boj et al., 2015; Huch et al., 2013). The method could be adapted to grow ductal organoids from digested human pancreas samples with serum-free medium optimisation by adding TGFβ inhibitor, forskolin and prostaglandin E2 (Table 1). Although expansion was limited to 2 months, in contrast to the unlimited expansion of murine ductal spheres (Boj et al., 2015; Driehuis et al., 2020), longer expansion was subsequently obtained by medium refinement (Georgakopoulos et al., 2020). This protocol was later adapted to good manufacturing practice conditions for the large-scale production of pancreas organoids (Dossena et al., 2020). However, even with these optimisations, the organoids stayed in a simple sphere form.
Sorted cell populations of the adult pancreas, notably CD44+/CD24+ (mouse) ductal cells (Rezanejad et al., 2018), ALDH-high cells (mouse and human) (Loomans et al., 2018; Mameishvili et al., 2019; Rovira et al., 2010), CD133 (PROM1)+ (human) (Lee et al., 2013) and CA19-9+/CD51 (ITGAV)+ (human) (Rezanejad et al., 2018) cells were found to expand when seeded in 3D (Fig. 2D). Sorted CD133+ mouse ductal cells could also grow in a matrix containing methylcellulose and 5% Matrigel (Jin et al., 2013, 2014) (Table 1). Further analysis of the heterogeneity of these cell populations is required, however.
Taken together, these studies show that multiple media and culture environments, with or without matrices, may re-induce a progenitor state from adult cells, although benchmarking to embryonic stages would be necessary (Loomans et al., 2018; Mameishvili et al., 2019; Rezanejad et al., 2019, 2018). The differentiation capacity of adult cells reported was, however, variable. Most media did not promote spontaneous endocrine or acinar differentiation, with some exceptions (Jin et al., 2013, 2014). Some endocrine cell differentiation was triggered by differentiation media (Jin et al., 2013, 2014; Loomans et al., 2018; Rezanejad et al., 2019, 2018) or upon transduction with viruses expressing the transcription factors PDX1, MAFA, NEUROG3 and/or PAX6 (Azzarelli et al., 2018; Lee et al., 2013).
Although expansion of ductal cells was the most prevalently reported, two reports suggested that the expansion of endocrine and acinar cells was possible. A promising recent report suggested that cells expressing protein C receptor (PROCR) could be isolated from adult islets and seeded to form islet organoids comprising multiple endocrine cell types (Fig. 2E, Table 1) (Wang et al., 2020). Although there is a paucity of surface markers to isolate acinar cells, Wollny et al. sorted acinar cells based on their large size (Wollny et al., 2016). They observed that, although single acinar cells could not form clones in vitro, doublets could form organoids with 5% efficiency when the starting cells were not multi-nucleated (Fig. 2F). The authors proposed that acinar cells were heterogeneous and identified a new marker (stathmin 1, STMN1) that correlated with a proliferative status and was upregulated by caerulein-induced injuries. However, acinar identity was not maintained in culture. The cells started to co-express amylase and ductal marker KRT19 within a day of seeding and totally lost amylase by day 8.
Organoids derived from pluripotent stem cells
Alternatively, organoids could be generated from PSCs, including human embryonic stem cells (ESCs) and induced PSCs (iPSCs) as well as, in rare cases, mouse ESCs. Although the field started by making embryoid bodies in suspension and relying on the spontaneous differentiation of endoderm, 2D protocols gained momentum owing to the exquisite control and uniform stepwise cell differentiation (Fig. 3A) (D'Amour et al., 2005, 2006; Kroon et al., 2008). The differentiation protocols continued to evolve with a manufacturing process mind-set, notably increasing efficiency, functional maturity and quality control of the final product with the aim of β-cell production for a cell therapy of diabetes. With time, the 2D protocols were further developed into three alternative 3D culture methods: (1) air-liquid interface after initial production of pancreas progenitors in 2D (Rezania et al., 2014) (Fig. 3B), (2) embedding in Matrigel after pancreas progenitor production in 2D (Fig. 3C) (Goncalves et al., 2021; Hohwieler et al., 2017; Huang et al., 2015) or in suspension (Fig. 3F) (Huang et al., 2021), as well as (3) suspension cultures from the onset (Nair et al., 2019; Nostro et al., 2011; Pagliuca et al., 2014; Russ et al., 2015; Velazco-Cruz et al., 2019) (Fig. 3D) or from later steps (Fig. 3E) (Balboa et al., 2022; Jiang et al., 2007; Zhu et al., 2016).
Organoids from pluripotent stem cells. (A-C) Protocols with several steps, usually four, enable the production of pancreas progenitors in 2D. After pancreas progenitor production (A), differentiation into endocrine cells can be conducted at the air-liquid interface (B) or after Matrigel embedding (C). Differentiation of pancreas progenitors into acinar or ductal cells can be further promoted. (D) Other differentiation protocols were adapted to 3D suspension culture by first assembling aggregates of hPSCs. (E) It is also possible to assemble 2D pancreas progenitor into aggregates in suspension and differentiate them further into endocrine cells. (F) Pancreatic progenitors differentiated in suspension can be also embedded in Matrigel and further differentiated. Created with BioRender.com.
Organoids from pluripotent stem cells. (A-C) Protocols with several steps, usually four, enable the production of pancreas progenitors in 2D. After pancreas progenitor production (A), differentiation into endocrine cells can be conducted at the air-liquid interface (B) or after Matrigel embedding (C). Differentiation of pancreas progenitors into acinar or ductal cells can be further promoted. (D) Other differentiation protocols were adapted to 3D suspension culture by first assembling aggregates of hPSCs. (E) It is also possible to assemble 2D pancreas progenitor into aggregates in suspension and differentiate them further into endocrine cells. (F) Pancreatic progenitors differentiated in suspension can be also embedded in Matrigel and further differentiated. Created with BioRender.com.
While most investigators focused on the production of endocrine cells and developed protocols to generate β cells as efficiently as possible, others focused on the development of other pancreatic cell types. In 2015, Huang et al. (2015) seeded human pancreas progenitors produced with a 2D protocol (Nostro et al., 2015) in Matrigel (Fig. 3C). Although the cells appeared to expand mostly as pancreatic progenitors, which were benchmarked to embryonic human pancreata, the differentiation medium enabled an increase of ductal and acinar markers, though at much lower level than when transplanted in vivo in the mammary fat pad. The transplantation also resulted in a loss of NK6 homeobox 1 (NKX6-1), suggestive of the disappearance of pancreas progenitors (Huang et al., 2015). No endocrine differentiation was reported after either in vitro differentiation or transplantation. In 2021, we also observed that in a simpler medium, pancreas progenitors produced with multiple protocols in 2D (Ameri et al., 2017; Rezania et al., 2014), embedded in Matrigel, could be propagated for at least a year from multiple PSC lines (Goncalves et al., 2021) (Fig. 3C, Table 1). The benchmarking to fetal pancreas showed that the organoids consisted mostly of pancreas progenitors and some cells on the path to ductal fate, but no acinar cells. Although only 0.1% of cells found in expansion conditions were endocrine cells, progenitor cells could differentiate into endocrine cells upon application of endocrine differentiation media (Rezania et al., 2014).
It was found that after generating pancreas progenitors in 2D with similar methods and seeding them in 3D, some media can prioritise either the maintenance and expansion of progenitors (Goncalves et al., 2021) (Fig. 3C) or the differentiation of acinar and ductal cells (Hohwieler et al., 2017) (Table 1). In the latter study, different 3D protocols were used either in suspension media (Fig. 3E) or in Matrigel culture (Fig. 3C). Previously defined media (Greggio et al., 2013) and new ones were assessed and, although it is difficult to match each condition tested to a specific outcome, acinar cells and ductal cells could be obtained in some conditions, based on the presence of acinar enzymes, as well as SOX9 staining and carbonic anhydrase activity for ducts (Hohwieler et al., 2017). Here, a probable NKX6-1-expressing progenitor subpopulation was retained in culture. As no benchmarking to adult or fetal tissue was provided in the study, it is difficult to compare the acinar and ductal differentiation to that of in vivo counterparts.
In 2021, Breunig et al. screened about 30 compounds to identify those that could enhance ductal markers (Fig. 3C) (Breunig et al., 2021a,b). They found a positive effect of ZnSO4 and keratinocyte growth factor (KGF) on the levels of KRT19. Additionally, moderate levels of the putative Notch activator and WNT inhibitor MSC2530818 enhanced cystic fibrosis transmembrane conductance regulator (CFTR) expression. The transcriptome of the cells obtained in an optimised ductal differentiation medium was compared with that of pancreas spheres produced from adult ducts, and they showed that the induced cells gradually acquired a transcript profile similar to that of adult ductal organoids. Comparisons with adult ductal transcriptomes (Baron et al., 2016; Qadir et al., 2020) revealed the similarity to ductal cells and a decrease of the progenitor marker NKX6-1 (Sander et al., 2000). A comparison with pancreatic progenitors from embryos and with the previous acinar/ductal differentiation protocol of the same authors (Hohwieler et al., 2017) would have been desirable, as well as some insight about heterogeneity. Some differences were found with duct-derived organoids, notably lower levels of CFTR, KRT8, KRT18, KRT19, ANXA2 and HNF1B in human PSC (hPSC)-derived ductal organoids and higher levels of CA4 and NKX6-1. This may be indicative of these hPSC-derived ductal organoids being closer to the subpopulation of ductal cells expressing MUC1 rather than the CFTR+ ductal cells identified by Baron et al. (2016). Further maturation was observed upon transplantation in the pancreas. The ductal differentiation conditions were subsequently adapted to a microwell culture chamber without Matrigel after optimisation of cell number and microwell size (Table 1) (Wiedenmann et al., 2021). Single-cell transcriptome analysis revealed the genes marking the progressive maturation of ductal cells over 31 days, very few progenitors and endocrine cells by the end of the differentiation and two types of ductal cells enriched either in CA2 and MUC13 or in CFTR. The different cell types appeared to be mixed in single organoids.
In another study, more acinar-like organoids were promoted by activation of the canonical WNT, FGF and cortisol pathways, and inhibition of the hedgehog, Notch, BMP and ALK5 (TGFβR1) pathways (Fig. 3F, Table 1) (Huang et al., 2021). The acinar spheres exhibited a columnar epithelial shape, whereas ductal spheres they differentiated in the same study exhibited a squamous shape. A comparison with adult acinar cells would have been useful, but the small number of secretory vesicles observed by electron microscopy, the moderate increase of acinar markers and the absence of late markers, such as amylase, suggest that the cells had just initiated acinar differentiation. Comparing the cell states obtained in all these studies and their relation to fetal and adult cell types would be valuable (Breunig et al., 2021a,b; Goncalves et al., 2021; Hohwieler et al., 2017; Huang et al., 2021, 2015; Wiedenmann et al., 2021). Moreover, clarifying the heterogeneity reported among ductal and acinar cells in a consistent way (Baron et al., 2016; Hendley et al., 2021; Qadir et al., 2020; Tosti et al., 2021) and defining the differences between progenitors and ductal cells would be useful.
All the organoids presented above remain rather simple, forming essentially pancreatic epithelia. In addition, some attempts have been made to co-culture the epithelium with relevant mesenchyme (Fig. 1G) (Ghezelayagh et al., 2022). Co-culture without contact, using microfluidic chips, could be an alternative or complementary strategy, as was achieved recently with immortalised stellate cells isolated from a patient with pancreatitis (Wiedenmann et al., 2021). They enabled only short-term culture, and stable co-culture has proven challenging for many organoid systems. Other cell types relevant to pancreatic function would be interesting to co-culture, notably endothelial cells and their assembly into blood vessels, as well as neurons. Assembling endothelial cells has so far been limited to human umbilical vein endothelial cells, which are only distantly related to the endothelial cells found in the adult pancreas (Candiello et al., 2018).
Mouse ESCs and iPSCs could also be coaxed to form 3D islet-like structures comprising bona fide endocrine cells (Saito et al., 2011; Wang and Ye, 2009), but most studies focused on human cells. Compared with adult pancreas-derived organoids, the field of PSC-derived organoids is still largely in a phase of model development and characterisation. This, however, provides an interesting possibility of studying human development, which is discussed below.
Applications of pancreas organoids to the study of development and regeneration
After a phase of model development, research is entering a phase in which pancreas organoids are used as one of the available methods to study mouse development and as a crucial method enabling us to delve mechanistically into human development.
Many organoids can be initiated from single cells. These clonal assays are very convenient to investigate the potency of individual cells, their proliferation potential and whether they can give rise to single or multiple cell types (Fig. 4A) (Loomans et al., 2018; Sugiyama et al., 2013). Although this can also be done in vivo by clonal labelling (Larsen et al., 2017), it is easier to implement in vitro, although it is not perfectly equivalent. Indeed, the in vitro assay examines what the cells can do, which is highly influenced by the culture medium and growth conditions, whereas in vivo clonal labelling reveals what they actually do.
Pancreas organoids as a developmental biology discovery tool. (A) Potency and clonality assays enable exploration of the cell types and structures a cell can give rise to, within the limits of the culture medium used. (B) By seeding different numbers of cells, one can explore the minimal cell number that enables a specific self-organisation. (C) Organoids enable the study of signalling between cell types, notably epithelial-mesenchymal interactions or signalling from the medium or the extracellular matrix. (D) The mechanical role of the matrix can be further explored. (E) Morphogenesis modules, such as compaction, polarity acquisition or lumen formation, can be explored with organoids. Mechanobiology can be explored using physical perturbations. (F) Genetic modifications can be implemented (as symbolised by the red jagged arrow), including chemically or optically induced mutations, interfering with gene function or recapitulating disease alleles. In all these assays, the ease of applying perturbations and performing live imaging are two assets of organoids.
Pancreas organoids as a developmental biology discovery tool. (A) Potency and clonality assays enable exploration of the cell types and structures a cell can give rise to, within the limits of the culture medium used. (B) By seeding different numbers of cells, one can explore the minimal cell number that enables a specific self-organisation. (C) Organoids enable the study of signalling between cell types, notably epithelial-mesenchymal interactions or signalling from the medium or the extracellular matrix. (D) The mechanical role of the matrix can be further explored. (E) Morphogenesis modules, such as compaction, polarity acquisition or lumen formation, can be explored with organoids. Mechanobiology can be explored using physical perturbations. (F) Genetic modifications can be implemented (as symbolised by the red jagged arrow), including chemically or optically induced mutations, interfering with gene function or recapitulating disease alleles. In all these assays, the ease of applying perturbations and performing live imaging are two assets of organoids.
Organoids are also capable of uncovering the scale of interactions during organ formation (Fig. 4B). For example, whereas the pancreas starts from a primordium of several hundred cells (Larsen et al., 2017), we observed that organoids could be initiated from groups of four cells with a probability of about 10%, which reached 100% with more than 12 cells (Greggio et al., 2013). This suggests that only a few cells are required for morphogenesis reminiscent of the pancreas, including the formation of acinar and endocrine cells. The pancreas can thus be viewed as a composite of small autonomous units that can assemble or scale up into a superstructure. The connection of all ducts in organoids shows that all units can interact. However, the structural evolution of the organ likely requires interactions with neighbouring organs, which can be studied using multi-organ organoids (Fig. 1H).
At a larger scale, scientists have started to assemble different organoids, collectively referred to as assembloids (Kanton and Paşca, 2022). In this spirit, Koike et al. (2019, 2021) have differentiated hPSCs into anterior [low CHIR99021 (a WNT pathway activator) and noggin] and posterior (high CHIR99021) gut spheroids. When they apposed the two spheroids, they fused, and pancreas- and liver-like tissues formed at the boundary between the foregut and midgut (Fig. 1H). Long-term culture up to 32 days revealed the formation of hepato-biliary-pancreatic structures connected by ducts. Although some repression of domain-specific factors between domains have been discovered in mouse (Willet et al., 2014) and chick (Grapin-Botton et al., 2001) models, the signalling factors that are likely to be involved remain to be identified.
Organoids are useful models to study how extracellular signals affect development as they can easily be included in the culture medium or blocked by inhibitors (Fig. 4C). For example, organoids from five different endodermal organs were compared to identify common WNT targets (Dugnani et al., 2018) and organ-specific ones (Boonekamp et al., 2021). Mesenchymal cells in vivo produce many signals essential for morphogenesis (Landsman et al., 2011). Although appropriate cell types and structures, notably the branching patterns, could form in the absence of mesenchyme (Greggio et al., 2013), essential mesenchymal factors such as FGF10 were included in the medium. The complexity of Matrigel precludes a more systematic assessment of specific mesenchymal and extracellular matrix factors involved in pancreas development. The development of functionalised synthetic hydrogels would enable us to study the role of specific extracellular matrix molecules and mesenchymal factors (Fig. 4D).
Pancreas organoids have also been used as a cell biology tool (Fig. 4E). Compaction and polarisation acquisition have been studied using organoids, showing that they polarise with their basal side out when cultured in Matrigel, and their apical side lining a lumen (Bakhti et al., 2019; Greggio et al., 2013). The start of the apical membrane initiation site was seen at the two-cell stage (Bakhti et al., 2019). More investigations are expected to study the mechanical control of pancreas development, notably lumen formation and remodelling (Lee et al., 2022). Imaging and image analysis will be crucial for such studies (Hof et al., 2021; Keshara et al., 2022). The culture of pancreatic cells in 3D often leads to the formation of large spherical lumens (spheres), but we also reported the formation of a network of interconnected narrow lumens reminiscent of what is observed in vivo (Fig. 1F). Although this appears to be linked to the presence of acinar cells, more work will be necessary to understand how different types of lumen form and the potential role of hydrodynamics in the process. However, the organoids did not recapitulate the outlet to the duodenum, which enables fluid release from the ducts, possibly explaining why they did not recapitulate the maturation of a hyperconnected ductal network to a hierarchical tree (Dahl-Jensen et al., 2018). The development of synthetic hydrogels would enable us to control the stiffness of the environment and its effect on morphogenesis, notably the formation of branched patterns (Fig. 4D). However, no such gel has so far been as effective as Matrigel in promoting growth (Greggio et al., 2013).
Organoids can also be used to test the function of specific genes with high throughput. An early example exploited the sphere system to introduce short hairpin RNAs using lentiviruses and screen for genes controlling pancreas development. Notably, the role of PR/SET domain 16 (PRDM16) in the differentiation of endocrine cells, most prominently α cells, was uncovered (Fig. 4F) (Sugiyama et al., 2013). Using the pancreatoid models that re-aggregate E10.5 progenitors, Scavuzzo et al. characterised the role of HNF1A, a transcription factor with an unknown role in the pancreas, complementing in vivo investigations (Scavuzzo et al., 2018a). Gain- and loss-of-function experiments in pancreatoids showed that HNF1A functions to reduce endocrine fate by limiting Notch activation through blocking Delta-like 1 endocytosis (Scavuzzo et al., 2018a). More recently, in a study demonstrating the role of RE1 silencing transcription factor (REST) in mouse pancreas development, human adult duct-derived pancreas organoids complemented an in vivo investigation in mice to show that REST inhibition was not sufficient to induce endocrine cells in the adult ducts (Rovira et al., 2021).
Developmental origins of disease: where organoids can help
Organoids are not only useful for understanding development, organ homeostasis and regeneration, but also for studying human disease. In the pancreas, the most prevalent are diabetes and pancreatic cancers, and other diseases, such as pancreatitis or cystic fibrosis are also relatively common (Uc and Fishman, 2017). Together with the growing information on genetic variants that predispose to disease, organoids are powerful instruments for functional investigations. Although there are questions regarding whether the organoids engineered from PSCs represent adult cell types, functional investigations of development appear possible. From a developmental viewpoint, some forms of diabetes begin before birth, such as neonatal diabetes, and it is also the case for some exocrine disorders, such as cystic fibrosis (Durie, 1996). Neonatal diabetes (Jennings et al., 2020) and maturity-onset diabetes of the young (MODY) are rare monogenic diseases resulting from mutations in loci of genes involved in β-cell development, identity and function (Balboa et al., 2021; Burgos et al., 2021). 2D culture systems where hPSCs produce pancreatic progenitors and endocrine cells have already been used to study monogenic forms of diabetes, but the use of 3D culture is just emerging. One strategy consists of differentiating patient-derived iPSCs and comparing them to CRISPR/Cas9 genetically corrected autologous cells (Braverman-Gross and Benvenisty, 2021). This method, combined with 3D culture, was used to show that human mutations in the insulin locus cause neonatal diabetes, in part, by impaired β-cell differentiation (Ma et al., 2018), as well as endoplasmic reticulum stress causing reduced β-cell proliferation (Balboa et al., 2018). Another example revealed that patient-relevant mutations in the transcription factor STAT3 directly induce NEUROG3 expression and thereby cause premature β-cell differentiation (Saarimaki-Vire et al., 2017). As an alternative approach, one can genetically modify hPSCs to recapitulate patient mutations or knockouts of the disease gene. This method was employed to study the role of MAFB in endocrine subtype specification (Russell et al., 2020). Additionally, PDX1 (Wang et al., 2019; Zhu et al., 2016), NEUROG3 (Zhang et al., 2019; Zhu et al., 2016) and RFX6 (Zhu et al., 2016) have also been studied using this approach, revealing their role in the production of hormonal cells or the formation of glucose-responsive β cells. So far, only 3D suspension cultures or cultures at the air-liquid interface have been used in disease modelling. Organoids in extracellular matrix hydrogels would represent a model of choice to provide a holistic view of the diabetes forms, associated exocrine phenotypes, such as cystic ducts, and their relations to β-cell defects leading to diabetes.
Pancreatic organoids can also be used to model cystic fibrosis, a disease which can manifest prenatally by ductal malformations and obstructions in human and pig (Durie, 1996; Rogers et al., 2008). Hohwieler et al. used their organoid system to show that a patient mutation prevented swelling triggered by forskolin and IBMX, two activators of wild-type CFTR (Table 1) (Hohwieler et al., 2017). They also identified how the CFTR mutation modified the transcriptome of pancreatic cells and could be used to screen for compounds that could rescue the swelling phenotype. Moreover, they developed the transfection of chemically modified wild-type CFTR and could also rescue cystic phenotypes. Additionally, iPSCs from CFTR patients exhibited dysfunctional chloride excretion (Shik Mun et al., 2019; Simsek et al., 2016). In a microfluidic system, secondary consequences on insulin secretion from islets were also observed (Shik Mun et al., 2019).
In the future, pancreas organoids would be potentially useful to study the effects of maternal health and environment on the developing fetal pancreas. Based on rodent models and human studies, it is known that fetuses exposed to nutrient excess or deficiency are prone to develop metabolic disorders, such as diabetes, later in life (Bruce, 2014; Remacle et al., 2007). Environmental toxins may also contribute to the vulnerability of the offspring to pancreatic diseases, as shown for endocrine disruptors such as Bisphenol A and Δ9-THC (Alonso-Magdalena et al., 2015; Gillies et al., 2020). By easing manipulations and enabling observations, organoids may help clarify the mechanisms by which developmental conditions lead to increased risk of offspring metabolic disease. A study by Beyaz et al. performed on intestinal organoids is a valuable example that addresses the consequences of a high-fat diet on stem cell function (Beyaz et al., 2016). Applying similar strategies to pancreatic organoids mimicking the fetal developmental stages could provide a valuable source of information on epigenetic and metabolic consequences of pancreatic progenitor exposure to altered diets and toxins.
Limitations of organoids for recapitulating pancreatic development
Although pancreas organoids have enabled some discoveries that would not have been possible without these tools, the field is still very focused on developing models. To be able to answer additional biological questions, several improvements are still required, especially to control the differentiation of pancreatic cell types, notably acinar cells. These cells are present in only a few models (Table 1). A more systematic assessment of how they are produced would be important as there is scarce information on the development of this lineage even in vivo (Larsen and Grapin-Botton, 2017). Being able to include these cells, trigger full maturation and maintain them in culture would open avenues to understand the early steps of cancer, notably acinar-to-ductal metaplasia (Huang et al., 2021). The development of the ductal lineage is also an enigma. It is currently viewed as the default state a progenitor will end up in if it does not make acinar or endocrine cells, but the steps that lead progenitors to ductal cells and their diversity deserve more attention. For endocrine cells, promoting further maturation would enable us to study postnatal development and improve modelling of adult-onset diabetes. Benchmarking the nature of the cells generated in organoids and their maturity will be important to tackle these questions and understand the developmental stages modelled as well as the limitations of the systems (Childs et al., 2022). As more and more cell subtypes (with some stability) and states are discovered even in vivo, it is difficult to evaluate the significance of transcriptome differences. This is also associated with technical challenges. Indeed, when previously published reference single-cell transcriptomes are used, batch-correction algorithms are usually applied, which minimise not only the technical noise between studies but also biological differences between organoid cells and their in vivo references. Another difficulty will be to parse which differences are meaningful and which ones are not.
Some cell types are currently not included in pancreas organoids, such as endothelial and pericytes, which could be useful to incorporate to study, for example, how endocrine cells release hormones in the blood. Lymphatics are also missing, as well as neurons, immune cells and glial cells. As endocrine hormone release has an important neuronal control, including them would enable us to improve our understanding of neuronal-epithelial interactions and how they are established (Burris and Hebrok, 2007). An additional example is adding endothelial cells, which could be used to study their signals to pancreatic epithelial cells and would not necessarily require their arrangement into blood vessels and their perfusion. However, if the goal is to enable better perfusion of organoids for more growth, blood vessel connectivity needs to be included as well as perfusion by appropriate media.
Another challenge is to improve morphogenesis to get closer to that of the organ. Many pancreatic 3D models are spheroids, instead of elongated tubes. Although organoids with an architecture close to the embryonic pancreas, including a network of ducts, branches, acinar cells at the tips and endocrine cells, could be produced from E10.5 pancreata, this ability decayed at E11.5 and only spheres were produced thereafter (Greggio et al., 2013; Scavuzzo et al., 2017). In human organoids, a network or a branched architecture was reported mostly when starting from small biopsies rather than fully dissociated cells, which tend to form spheres (Goncalves et al., 2021; Loomans et al., 2018). All the cultures described in this Review rely on self-organisation constrained by the developmental potential of the seeded cells. How specific external biochemical and structural components trigger the programme often remains elusive. Moreover, as soon as several cells are associated, through either aggregation or division, they produce signals that may influence their neighbours. Understanding symmetry-breaking events that produce heterogeneity and feedback loops, enabling self-organisation, will require tools measuring cell-state dynamics at the system scale and theoretical modelling (Lewis et al., 2021; Tan et al., 2022). An alternative avenue is to employ engineered 3D moulds (Gjorevski et al., 2022) or cell and matrix bioprinting strategies (Hagenbuchner et al., 2021) to constrain the cells into organ-relevant shapes, for example tubes or a tree. These methods would also enhance the reproducibility and homogeneity, but have the disadvantage of constraining the organoid to a final shape, ignoring its possible intermediate shapes. They could be combined with large-scale screening formats enabling high-throughput perturbations or drug screens (Lukonin et al., 2021).
Although matrices replacing Matrigel would be desirable to reduce the environment complexity, batch-to-batch variation and cost (Kozlowski et al., 2021), so far organoids in defined matrices remain small (Greggio et al., 2013). Further optimisation and establishment of defined matrices that provide a scaffold with proper stiffness and a specific extracellular signalling niche will enhance the control over the system, improve reproducibility and reduce variability.
Looking ahead, although the development of many systems should enable more biological questions to be addressed, notably to understand human pancreas development and diseases, there is a risk that a plethora of models limits our ability to compare studies. Development in vivo is canalised, and model organisms have the benefit of providing a reproducible reference. Removing the constraints may enable us to explore what the cells can do at the expense of being sure that this is what they actually do. This is a risk to take and a challenge to solve as we know so little about human development and organoids provide a fantastic opportunity to further our knowledge.
Acknowledgements
We thank past and current Grapin-Botton lab members for generating collective knowledge leading to this Review and particularly Lydie Flasse, Hjalte List Larsen, Siham Yennek and Phil Seymour for unpublished images.
Footnotes
Funding
A.G.-B. is funded by the Max-Planck-Gesellschaft, the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) (project number 288034826–IRTG 2251; as well as the cluster Physics of Life under Germany's Excellence Strategy EXC-2068–390729961) and a Human Frontier Science Program grant (RGP0050/2018).
References
Competing interests
The authors declare no competing or financial interests.