Pancreatic organoids were first described by Clevers’ laboratory in 2013 (Huch et al., 2013), introducing pancreatic organoid technology as a putative unlimited source of induced-progenitors with the prospect of future differentiation into insulin secreting cells. The adult pancreas does not express the stem marker Lgr5 in homeostatic conditions (Sato et al., 2009). However, mouse pancreatic duct fragments were shown to initiate Lgr5 expression in RSPO1-based cultures followed by organoid formation (Huch et al., 2013). Organoids derived from mouse and human pancreatic tissue expressed ductal markers (SOX9, KRT19, MUC1) and only organoids expressing Sox9 were capable of long-term expansion (Huch et al., 2013; Boj et al., 2015). Together, these observations suggest that ductal cells in vitro display progenitor capacities. In line, organoids derived from EpCAM⁺-TSQ^- cells (i.e., non-endocrine epithelial cells) could be induced to differentiate into ductal as well as endocrine cells upon transplantation with embryonic E13 mouse or E14 rat pancreata under the kidney capsule of non-diabetic immunodeficient mice, although the efficiency of differentiation toward endocrine lineages was ∼5% (Huch et al., 2013). Later, lentivirus-mediated, doxycycline-inducible expression of Neurog3, Pdx1, and MafA in mouse pancreatic ductal organoids has been shown to generate cells that express insulin and resemble β-cells at the transcriptome level (Azzarelli et al., 2017). The combinatorial potential of these three transcriptional factors to induce endocrine differentiation was reported by Melton’s laboratory to induce transdifferentiation of acinar cells to β cells in vivo (Zhou et al., 2008). Finally, efficiency of organoid-derived β-like cell generation can be significantly enhanced by preventing phosphorylation of the Neurog3 protein and further augmented by conditions promoting differentiation (Azzarelli et al., 2017). This suggests that post-translational regulation of key regulators of endocrine differentiation can be exploited to enhance β-cell generation from organoids (Figure 1).

Pancreatic organoids have been subsequently derived from subpopulations of ductal cells defined by already identified progenitor markers. We showed that mouse ductal cells displaying high aldehyde dehydrogenase (ALDH) activity display progenitor features in vitro, in pancreatospheres cultures and upon transplantation into embryonic pancreas (Rovira et al., 2010). These cells demonstrated dramatic expansion in the setting of epithelial injury and pregnancy (Rovira et al., 2010; Ioannou et al., 2013; Socorro et al., 2017). ALDH^hi cells have indeed been identified in human fetal and adult pancreas (Loomans et al., 2018; Oakie et al., 2018). Recently, Koning’s laboratory (Loomans et al., 2018) showed that human ductal pancreatic cells displaying high ALDH activity can efficiently form organoids that can be expanded and maintained overtime. Interestingly, gene expression profiling revealed that ALDH^hi ductal expressing cells are closer to human fetal pancreatic tissue compared with adult pancreatic tissue (endocrine and exocrine). ALDH^hi derived organoids are able to differentiate into insulin expressing cells in vitro and upon being engrafted into the kidney capsule, although with low efficiency (∼1.5% of insulin positive cells) (Loomans et al., 2018) (Figure 1).

Similarly to ALDH, CD133 is a marker whose expression has been described as a cell surface marker in adult progenitor populations (Shmelkov et al., 2005; Li, 2013; Vassalli, 2019). CD133 is also expressed in a fraction of pancreatic ductal cells (Immervoll et al., 2008; Immervoll et al., 2011) and indeed, it has been used as surface marker to isolate ductal cells and assess their progenitor potential in cultures grown as monolayer colonies or spheres and their capacity to differentiate into endocrine cells in vitro (Oshima et al., 2007; Hori et al., 2008). CD133⁺ cells isolated from mouse adult pancreas form organoids (Jin et al., 2013; Jin et al., 2014; Wedeken et al., 2017) but these do not display long-term self-renewal potential and are heterogeneous. However, they displayed endocrine and exocrine differentiation capacity in presence of WNT ligand R-spondin1 (Jin et al., 2013; Jin et al., 2014). In humans, CD133⁺ isolated pancreatic cells form organoids with self-renewal potential (Lee et al., 2013) but their endocrine differentiation was only achieved upon adenoviral induction of the expression of MafA, Neurog3, Pdx1 and Pax4, master regulators of endocrine differentiation. More recently, human pancreatic organoids derived from ductal CD133⁺ cells could be differentiated into insulin⁺ cells (at a frequency of ∼4.6%) just by in vitro treatment with transcribed Neurog3 mRNA and differentiation media (Koblas et al., 2019). Characterization of liver and pancreas organoid-initiating cells in mice showed that these cells are phenotypically and functionally similar and express MIC1-1C3⁺/CD133⁺/CD26^- (Dorrell et al., 2014). Organoids derived from these ductal cells can differentiate into insulin expressing cells (at a frequency of ∼5–22%) following tricistronic adenoviral administration of MafA, Neurog3 and Pdx1. These insulin⁺ cells showed transcriptional identity partially overlapping with murine β cells, but with retained expression of many off-target non-β cell genes (Dorrell et al., 2014). In summary, pancreatic ductal-derived organoids have proven their potential for β cell replacement therapies although still the efficiency of organoids to differentiate into endocrine lineages is limited. A more refined reprogramming/differentiation methodology and a subpopulation of ductal cells that could produce an unlimited source of multipotent cells to generate endocrine cells is still needed (Figure 1).

To note, the media composition of the above studies comprises factors shown to be important in endocrine differentiation during development [e.g., FGFs (FGF10), EGF, RA] (Gittes, 2009; Santosa et al., 2016). These factors might also play a role in the differentiation of adult ductal organoids either potentiating or hindering the process, a field that deserves further investigation.

Finally, protein C receptor (Procr) is a surface protein that has been reported to mark stem cells in several adult tissues (Balazs et al., 2006; Iwasaki et al., 2010; Wang et al., 2015; Yu et al., 2016; Zhou et al., 2016; Fares et al., 2017). Interestingly, a recent comprehensive single cell RNA-seq analysis of murine islets identified a novel Procr⁺ population (Wang et al., 2020). Procr⁺ cells were characterized by a transcriptional signature indicative of epithelial-to-mesenchymal transition, lacked expression of terminally differentiated endocrine and exocrine markers, and were able to form islet-like organoid-like spheroids in vitro. Islet-like derived organoids could be long-expanded in vitro and upon co-culture with endothelial cells differentiated into endocrine cells with a remarkable high efficiency: after 1 month, up to 80% of cells expressed insulin. In vivo, long-term cultured organoids restored glucose homeostasis in streptozotocin-induced T1D mice (Wang et al., 2020). (Figure 1).

Several reasons could explain the limited endocrine differentiation capacity of organoids:

First, organoid maintenance media has been established based on the intestinal organoid media with minimal modifications (Fujii et al., 2019), therefore it is possible that a modified media that best mimics pancreatic progenitors’ requirements will influence positively their expansion, self-renewal and endocrine differentiation capacity. The stem cell/progenitor niche concept, initially proposed by Raymond Schofield in 1978, defines niches as compartments that are conductive for the maintenance of definitive stem cell properties (Schofield, 1978). The niche thus represents a defined anatomical compartment that provides signals to stem cells in the form of secreted and cell surface molecules to control the rate of stem cell proliferation, determine the fate of stem cell daughters and protect stem cells from exhaustion or death. Such a niche has not yet been identified in the adult pancreas. However, we know that through pancreas development the neuronal, mesenchymal and endothelial compartments are key for the proper maintenance and differentiation of embryonic progenitors into exocrine and endocrine lineages (Miralles et al., 1999; Lammert et al., 2001; Gittes, 2009; Borden et al., 2013). Potentially, complex organoid co-cultures with neuronal, mesenchymal and endothelial cells could shed light into the proper settings to mimic a progenitor niche and improve differentiation potential as it has been already proven in pluripotent stem cell cultures (Aghazadeh et al., 2021).

Second, the low regenerative capacity of the adult pancreatic tissue suggests that if a resident progenitor exists, it should be of really low abundance. Thus, not all ductal cells may bear progenitor capacity nor show similar organoid formation ability. Therefore, a better understanding of ductal heterogeneity could highlight ductal subpopulations to be studied in organoid culture and assess their endocrine differentiation capacity (Hendley et al., 2021).

Third, ductal-derived organoids could have a limited endocrine differentiation capacity due to epigenetic features. A better understanding of the differences between embryonic progenitors and organoids may identify an epigenetic brake to be modulated for proper endocrine differentiation.

Fourth, extracellular matrix (ECM) components might be crucial in vitro, as they are in vivo, for the proper morphogenesis and differentiation of the pancreatic lineages. Current organoid systems mostly rely on intrinsic or extrinsic biochemical signals (i.e., growth factors) and cell-cell interactions to control stem cell fate, but there are important elements to consider including the ECM and biophysical signals that are still unknown. This is because current in vitro models rely on animal-derived Matrigel which contains laminin, collagen IV, entactin, heparan sulfate, and growth factors (Rezakhani et al., 2021) in proportions that are difficult to manipulate. A controlled matrix rather than standard Matrigel could be key for organoid-to-β-cell differentiation.

Finally, the above published studies did not analyze the possible presence of endocrine cells in the primary ductal cultures. In this scenario, it cannot be discarded that some of the insulin producing cells observed at the end of the protocol were present in the cultures from the beginning. Further studies where the absence of insulin expressing cells in the original preparation is carefully ratified prior to organoid formation are still required to properly validate and estimate the efficiency of ductal to β cell transdifferentiation.

Human healthy ductal derived organoids have also been used to model pancreatic cancer. Introduction of PDAC driver mutations in normal organoids via CRISPR-Cas9 or overexpression vectors has been used to study PDAC initiation and progression (Lee et al., 2017; Seino et al., 2018). Lee and colleagues (Lee et al., 2017) engineered organoids derived from healthy CD133⁺ ductal cells to express mutant KRAS ^G12V (K), and deleted CDKN2A (C), TP53 (T), and SMAD4 (S) (KCTS organoids). Following orthotopic implantation, the mutant organoids gave rise to early pancreatic lesions (PanINs), but these did not progress to PDAC. Differently, KCTS organoids developed by Seino et al. generated in vivo lesions comprising the full histological transformation from PanINs to invasive PDACs (Seino et al., 2018) (Figure 2A). Instead of healthy pancreata, Seino et al. used ductal cells isolated from “normal-like” regions adjacent to tumor tissue. Additionally, the two studies differed in the methodology to introduce KRAS ^G12V (overexpression vs. knock-in), the culture media used to establish the organoids, and the in vivo implantation model (orthotopic vs. subcutaneous). Whether some or all of these factors influenced the results deserves further investigation and encourages working on the development of standardized models. In the meantime, it is tempting to think that the origin of ductal cells (healthy vs. normal-like, CD133⁺ vs. bulk) may influence organoids’ nature and that microenvironmental cues (culture media, site of implantation) will need to be adjusted in search of more representative models.

The above studies employed organoids derived from ductal cells to study PDAC formation. However, the cell of origin of PDAC and its implication on the progression and biology of tumors are still a matter of debate. Several studies have shown that both acinar and ductal cells can give rise to tumors in genetic mouse models (Guerra et al., 2007; Von Figura et al., 2014; Bailey et al., 2016; Ferreira et al., 2017; Lee et al., 2019; Flowers et al., 2021). Some of these studies have recently highlighted that mouse PDAC tumors arising from either acinar or ductal cells show differences at the transcriptomic and phenotypic levels (Ferreira et al., 2017; Lee et al., 2019; Flowers et al., 2021). The existence of different cells of origin shall, in part, explain the great heterogeneity observed in the PDAC human scenario (Raphael et al., 2017; Collisson et al., 2019; Backx et al., 2021a; Espinet et al., 2021). Thus, models where the impact of mutations can be studied in both acinar and ductal human cells may help us to better understand patient tumor complexity. Unfortunately, acinar cells fail to grow organoids (Seino et al., 2018) even when engineered to express the main PDAC mutations (Lee et al., 2017). In general, acinar cells are refractory to in vitro culture, even if successfully isolated, as they rapidly undergo transformation into ductal-like cells (acinar-to-ductal metaplasia) (Houbracken et al., 2011; Wollny et al., 2016).

An approach to overcome the above limitation consists in inducing the differentiation of human PSCs (hPSCs) into acinus- and ductal-like exocrine organoids (Huang et al., 2015; Hohwieler et al., 2017). To note, hPSCs have been widely studied also as a source to obtain endocrine cells via differentiation, a topic that will not be discussed here as it has been extensively reviewed elsewhere (Melton, 2021). hPSCs exocrine differentiation requires a sequential protocol in which ESCs are first differentiated into pancreatic progenitor-like cells and then into acinus or ductal-like cells (Huang et al., 2015; Hohwieler et al., 2017). Using this approach, Huang and colleagues recently showed that PDAC driver mutations lead indeed to cell-lineage-specific phenotypes (Huang et al., 2021). While KRAS ^G12D in acinus-like organoids gave rise mainly to early PDAC lesions upon orthotopic transplantation, KRAS ^G12D duct-like organoids were less efficiently engrafting and mainly developed IPMN-like lesions (Figure 2B). Differently, if presenting with a GNAS ^R201C mutation neither acinus- nor ductal-like organoids engrafted in vivo (Huang et al., 2021). Another study, by Breunig and colleagues, in which ESCs were differentiated into pancreatic ductal-like organoids (PDLOs) confirmed the tropism of these different mutations, although with different results (Breunig et al., 2021). In this study, KRAS ^G12D in PDLOs resulted in early PDAC lesions upon orthotopic engraftment while GNAS ^R201C PDLOs developed IPMNs-like structures (Figure 2B).

Differentiated hPSCs are a valuable model to study early mutagenic events in human cells. However, it is important to investigate the reasons behind divergent results obtained in, otherwise, similar studies. If possible, unifying protocols and definitions of hPSC-derived exocrine organoids may help conciliate the appearance of contrasting results towards a better understanding of early mutagenesis in human PDAC. Additionally, the existence of acinar and ductal heterogeneity, recently revealed by single cell sequencing technologies (Arda et al., 2016; Muraro et al., 2016; Enge et al., 2017; Martens et al., 2021; Tosti et al., 2021), will need to be considered in future studies. In promising latter single cell analyses of hPSC derived ductal-like organoids, Wiedenmann, Breunig and colleagues have shown the appearance of different types of ductal-like cells that share characteristics with ductal subsets identified in normal pancreas (Huang et al., 2021; Wiedenmann et al., 2021). This suggests that certain relevant ductal heterogeneity can be recapitulated in the hPSC model. Although scRNA-sequencing of hPSC-derived acinus-like organoids also depicted certain heterogeneity (Huang et al., 2021), overall, acinus-like organoids resembled fetal rather than mature pancreatic acini. Thus, the relevance of this observed “acinar heterogeneity” in the adult pancreas context might be questionable. Using adult mouse pancreas studies showed that only a small subset of acinar cells have long term proliferative capacity in vivo (Wollny et al., (2016) . While these results suggest exciting functional differences in the adult acinar compartment, the study of the role of these and other acinar cells in the human and cancer scenario is still hindered by the lack of stable acinar in vitro models. Adaptation of the current culture conditions used to grow acinus-like organoids derived from hPSCs or from adult acini directly might result in improved acinar organoid formation. The big challenge is the fact that acinar cells rapidly lose their acinar identity upon injury or stress to move towards a ductal-like phenotype. The simple isolation of acinar cells from adult tissue might thus induce dedifferentiation of acinar cells. New studies analyzing these early events in vitro might help us to find the proper conditions required by acinar cells to retain their identity during longer experimental times (Backx et al., 2021b).

Two major transcriptomic epithelial subtypes have been described in PDAC after a collection of studies analyzing primary tumor samples: the classical and the basal-like subtypes (Martens et al., 2019). Tumors with a classical epithelial subtype show expression of genes characteristic of pancreatic progenitor cells and tend to be differentiated with better outcome. The basal-like epithelial subtype associates to more dedifferentiated and advanced tumors and correlates with worse outcome and lesser response to treatment (Aung et al., 2018). The emergence of PDAC PDOs offers a practical platform to study the biology of these two epithelial subtypes and to investigate subtype specific vulnerabilities. Additionally, efforts are done to generate transcriptomic signatures according to the response of different PDOs in vitro, that can be used to predict patient responses (Tiriac et al., 2018; Nicolle et al., 2020; Nicolle et al., 2021). The power of these signatures, some of which are currently being tested in clinical trials, will need to be carefully evaluated to find the best predictive markers.

Some studies have adapted the original PDAC-organoid media composition (Boj et al., 2015) to generate PDOs (Huang et al., 2015; Seino et al., 2018; Driehuis et al., 2019; Krieger et al., 2021). Although finding a consensus media to unify and standardize PDO growth across laboratories seems desirable, defining a unique media might be, in this case, delicate. Different tumor organoid subtypes present distinct dependency towards external soluble factors (Seino et al., 2018) likely reflecting existing differences of their tumor microenvironments (Nicolle et al., 2017; Maurer et al., 2019; Grünwald et al., 2021). Additionally, in vivo treatment with cytokines or implantation of organoids in different in vivo niches has shown to change the transcriptome of PDAC cells highlighting the plasticity of these cells under the influence of external factors (Miyabayashi et al., 2020; Tu et al., 2021). In line, the recent study by Raghavan and colleagues described that PDOs acquire culture-specific transcriptomic programs that are not present when the cells are in the tumor (Raghavan et al., 2021). More importantly, different media conditions (rather than 3D versus 2D structural changes) resulted in transcriptomic changes and influenced drug response of tumor cells (Raghavan et al., 2021). While we desire to move towards a standardized model as a tool for personalized medicine, these new observations question if we rather need to move towards fully personalized models first.

To add to this complex scenario, chemotherapeutic treatment has also been described to induce plasticity of PDAC tumor cells with transition from basal to classical subtype or vice versa having been observed (Chan-Seng-Yue et al., 2020; Hwang et al., 2020). This suggests that cancer cells undergo transcriptional subtype switching to adapt to the administered drug. In addition, Peschke, Jakubowsky and colleagues have recently reported a classical to classical case after FOLFIRINOX treatment (Peschke et al., 2022) where PDOs derived from the pre- and the post-treated tumors showed remarkable differences to drug responses in vitro despite sharing genetic drivers and being both classified as classical (Peschke et al., 2022). These results indicate that a transcriptional subtype switch does not occur always, and that drug adaptation can be achieved by other, unknown, mechanisms. Together, the use of current transcriptomic classifications to predict treatment response might need to be revised in the post-treatment scenario. Further knowledge needs to be acquired to understand how the plasticity-emerged subtypes compare to the original subtypes defined in the naïve treatment scenario and how this impacts prediction of treatment response. Whenever possible, performing longitudinal functional assays with PDOs obtained pre- and post-treatment will be a valuable tool for a precision oncology approach.