Islet organoid as a promising model for diabetes

Studies on diabetes have long been hampered by a lack of authentic disease models that, ideally, should be unlimited and able to recapitulate the abnormalities involved in the development, structure, and function of human pancreatic islets under pathological conditions. Stem cell-based islet organoids faithfully recapitulate islet development in vitro and provide large amounts of three-dimensional functional islet biomimetic materials with a morphological structure and cellular composition similar to those of native islets. Thus, islet organoids hold great promise for modeling islet development and function, deciphering the mechanisms underlying the onset of diabetes, providing an in vitro human organ model for infection of viruses such as SARS-CoV-2, and contributing to drug screening and autologous islet transplantation. However, the currently established islet organoids are generally immature compared with native islets, and further efforts should be made to improve the heterogeneity and functionality of islet organoids, making it an authentic and informative disease model for diabetes. Here, we review the advances and challenges in the generation of islet organoids, focusing on human pluripotent stem cell-derived islet organoids, and the potential applications of islet organoids as disease models and regenerative therapies for diabetes.


INTRODUCTION
Diabetes mellitus, characterized by hyperglycemia, is a prevalent metabolic disease affecting more than 400 million people worldwide that can lead to various severe complications and premature death (Saeedi et al., 2019). Type 1 diabetes (T1D) is an autoimmune disease that results in the dysfunction of β cells. Type 2 diabetes (T2D), the most common form in adults, is characterized by peripheral insulin resistance and relatively insufficient insulin secretion (Ashcroft and Rorsman, 2012). In addition, an increasing number of rare monogenic diabetes, including maturity-onset diabetes of the young (MODY) and neonatal diabetes (ND), which result from mutations in a single gene involved in pancreatic β cell development or function, have been identified (Ashcroft and Rorsman, 2012;Bishay and Greenfield, 2016). Unfortunately, both the pathogenesis of diabetes and effective therapeutic regimens remain largely unclear.
Organoid is a cutting-edge technology providing new models for developmental biology and disease research (Lancaster and Knoblich, 2014;Fatehullah et al., 2016;Dutta et al., 2017;Schutgens and Clevers, 2020). Organoids are three-dimensional (3D) cultures derived from stem cells, either pluripotent stem cells (PSCs) or organ-restricted adult stem cells (ASCs) and can mimic specific organ functions (Lancaster and Knoblich, 2014). Compared with animal models, organoids can be of human origin and thus can obviate the need to extrapolate discoveries from model animals to humans; compared with organ models derived directly from humans, organoids are more readily available and can be personalized. In clinical applications, organoids can serve as personalized high-throughput drug screening platforms with which to evaluate potential therapeutic efficacies, and as new sources for organ transplants due to their functionality (Fatehullah et al., 2016;Dutta et al., 2017;Schutgens and Clevers, 2020). However, current organoid systems have many problems, including limited cell compositions, insufficient maturity, incomplete function, and poor vascularization.
Because islets are closely involved in the occurrence and development of diabetes, islet organoids have attracted increasing attention. Pancreatic β cells are considered the most essential components of islets; thus, previous studies paid more attention to the generation of β cells from stem cells, termed stem cell-derived β (sc-β) cells . Recently, rapid increasing progress has been made in sc-β cell generation and islet organoid model development.
Here, we review the advances and challenges in establishing islet organoids and discuss their potential application.

STATE-OF-THE-ART STRATEGIES TO ESTABLISH ISLET ORGANOIDS
The mystery of human pancreas development has been partially uncovered. Some key signaling pathways, including the BMP, NOTCH, and canonical and noncanonical WNT pathways, and a few regulatory genes, including NKX2.2, NEUROD1, and ISL1, have been characterized (reviewed in (Jennings et al., 2015;Nair and Hebrok, 2015;Petersen et al., 2018;Yu and Xu, 2020)). Based on current knowledge about the mechanism of human islet development, a series of studies have developed various recipes combining panels of growth factors and small molecules to sequentially manipulate the signaling pathways to guide the commitment of hPSCs to a β cell fate (D'Amour et al., 2006;Kroon et al., 2008;Rezania et al., 2012). However, in these previous studies, the immaturity of sc-β cells, as demonstrated by the poor glucose responsiveness, polyhormonal signatures, and preferential commitment of the polyhormonal cells to α instead of β cells, has been reported (Kelly et al., 2011;Basford et al., 2012;Rezania et al., 2013;Bruin et al., 2014;Hrvatin et al., 2014). Currently, functional maturation relies mostly on in vivo transplantation (Rezania et al., 2013Augsornworawat et al., 2020), but the underlying mechanisms remain unknown. Considerable efforts have been made to derive functional islet organoids in vitro.

Optimization of recipes to obtain functional β-like cells
Recently, several groups have reported refined recipes to generate sc-β cells or accomplish islet regeneration in vitro. The design of these recipes relied mainly on current knowledge of pancreas development to in vitro mimic the natural microenvironment corresponding to specific developmental stages; therefore, most protocols share certain induction pathways, although disparities exist (summarized in Fig. 1A).
Pagliuca et al. systematically tested >150 combinations of >70 compounds to formulate a 6-step protocol, which generated ∼33% sc-β cells resembling primary β cells at the molecular, ultrastructural, and functional levels . These sc-β cells expressed certain canonical β cell marker genes, including PDX1 and ZNT8, and possessed both developing and mature crystallized insulin granules, where normal insulin processing occurs . Functionally, these cells repeatedly increased the intracellular Ca 2+ and secreted insulin upon sequential glucose changes in vitro and rapidly restored euglycemia in a diabetic mouse model after transplantation, closely resembling the native islets . As determined by multiomics analysis, three other major cell types in addition to sc-β cells existed among the final induction products, namely, α-like cells, an unexpected population of enterochromaffin cells, and SOX9 + pancreatic progenitors, which tend to generate exocrine cells upon further induction (Veres et al., 2019). Multiomics analysis further delineated the process of in vitro islet specification and identified two sequential lineage bifurcations, which lay at the initiation points of endocrine cell and β cell formation (Sharon et al., 2019;Veres et al., 2019;Alvarez-Dominguez et al., 2020). Each of the two bifurcations led to a dramatic decrease in induction efficiency and an increase in induction product heterogeneity (Sharon et al., 2019;Veres et al., 2019).
The immaturity of the sc-β cells was attributed partially to the premature induction of early precursors due to the premature expression of NGN3 in the early stages (Johansson et al., 2007). Rezania et al. introduced vitamin C during induction of the pancreatic endoderm to suppress precocious NGN3 expression and the downstream targets, NEU-ROD1 and NKX2.2 . Vitamin C also  Figure 1. State-of-the-art strategies to establish islet organoids. (A) Signaling pathways typically manipulated to induce differentiation of sc-β cells. Generally, hPSCs are sequentially differentiated to definitive endoderm (DE), pancreatic progenitors (PP), endocrine precursors (EP) and endocrine cells (EC), as shown by markers of each stage. The pathways in black are commonly manipulated in the most widely used protocols Rezania et al., 2014;Russ et al., 2015;Nair et al., 2019), and the pathways in red are specifically reported to facilitate endocrine specification (Ghazizadeh et al., 2017;Rosado-Olivieri et al., 2019;Sharon et al., 2019;Velazco-Cruz et al., 2019;Helman et al., 2020;Hogrebe et al., 2020). (B) Strategies currently leveraged to establish 3D islet organoids. Strategies in black are used to establish 3D islet organoids, relying on either suspension or scaffold culture. 3D culture starting from different time points has been reported. Strategies in red are applied to improve the maturity of 3D islet organoids, focusing mainly on improving vascularization and recovering metabolic defects of the immature islet organoids. (C) Strategies for ASC-derived islet organoids. Islet progenitors can be isolated from the adult pancreas and then form islet organoids with endothelial cells or other cells.
New insights into islet organoid REVIEW regulates extracellular matrix (ECM) production and increases cell confluency (Choi et al., 2008). With addition of vitamin C, a population containing approximately 50% insulin + /NKX6.1 + and KCl-responding cells appeared after β cell induction; however, these cells failed to respond to high glucose concentrations, indicating their immaturity and the requirement for additional maturation steps . Further screening of compounds capable of inducing MAFA, a marker of mature β cells, formulated a 7-step protocol. With this updated protocol, the majority of endocrine cells in the final products were β-like cells, ∼5%-10% of which rapidly increased the cytosolic Ca 2+ concentration upon glucose challenge. Although significant glucose stimulated insulin secretion (GSIS) was not observed, the β-like cells slowly accumulated insulin upon glucose responsiveness . The role of vitamin C in suppressing NGN3 expression is somewhat cell-specific. Russ et al. showed that NGN3 transcripts were not reduced in other cell lines upon vitamin C treatment (Russ et al., 2015). They omitted BMP inhibitor treatment during the induction of pancreatic progenitors to prevent premature endocrine commitment, instead treating progenitors with the BMP inhibitor noggin to induce endocrine precursors, allowing spontaneous differentiation into ∼20% monohormonal glucose-responding β-like cells in basal medium (Russ et al., 2015). However, omission of noggin during the early pancreatic progenitor step was recently found to induce the expression of CDX2, a marker of intestinal development, at the expense of losing NKX6.1 expression and the ability for further differentiation into endocrine cells (Wesolowska-Andersen et al., 2020). However, different cell lines were employed in these studies; therefore, the action of noggin could be a cell-specific effect, as also observed with other factors, including TPB or PdBU for PKC activation, and dorsomorphin or noggin for BMP inhibition (Kunisada et al., 2012;Rezania et al., 2012;Pagliuca et al., 2014;Nostro et al., 2015). Most recently, Hogrebe et al. developed a two-dimensional (2D) protocol using latrunculin A to depolymerize the cytoskeleton during endocrine induction and strongly improve the in vitro and in vivo function of sc-β cells .
In addition to adjusting the combinations of growth factors and small molecules, decreasing the exposure time to the progenitor-inducing cocktail yielded fewer polyhormonal cells and more NKX6.1 + multipotent progenitors (Nostro et al., 2015). Intriguingly, increased induction efficiency was observed by halving the cell density after the endodermal stage and extending the progenitor-inducing time (Memon et al., 2018). These results indicate that the efficiency of sc-β cell differentiation from hPSCs is mysteriously correlated with cell density, induction recipe, and cell lines.

Strategies to establish 3D islet organoids
In addition to elaborate chemical recipes, the role of 3D microenvironments for in vitro islet development has been extensively investigated. Islets are densely packed 3D architectures with specific cell distribution patterns in which cell-cell interactions have significant roles in regulating cell fate specification and islet function . These roles were ignored in early attempts to establish β-like cells in vitro, which partially accounted for the immaturity of the derived β-like cells due to diminished growth factor gradients. Spontaneous clustering was observed when hPSC-derived endodermal cells were replaced with a quartered density (Memon et al., 2018) or when monolayer-cultured hPSCs were induced toward the posterior foregut on Matrigel-coated dishes (Broda et al., 2019). The former generated clusters containing unusual PDX1 − /NKX6.1 + progenitors, and the latter have not been used for islet organoids.

Suspension-based 3D culture
The self-organization capacity of hPSCs has been widely employed to aggregate dissociated hPSCs into 3D embryonic bodies (EBs) with variable diameters dependent on the culture system (Fig. 1B). Pagliuca et al. utilized a stir plate to apply a constant rotation to hPSCs cultured in spinner flasks to generate ∼100-200-μm cell clusters. Morphologically and functionally, these clusters resembled native human islets; however, few non-β endocrine cells were detected compared with those in human islets . Dissociation and reclustering of islet organoids in the final stage have been reported to further improve functionality (Velazco-Cruz et al., 2019).
Microfluidic devices offer a more sophisticated suspension-based strategy. Microfluidic perfusable systems continuously deliver induction signals or nutrients while removing metabolic wastes. Furthermore, perfusion enables the application of mechanical cues, such as physiologically relevant flow-induced shear stress. Compared with the products obtained under static culture conditions, perfusable system-generated islet organoids have increased cell viability, elevated expression of mature endocrine cell markers, and significantly improved islet function (Tao et al., 2019).
Although 3D architecture plays a pivotal role in maturing islet organoids, the appropriate stage at which to apply 3D culture remains debatable. Compared with 3D suspension culture, 2D planar culture prevented pancreatic progenitors from prematurely expressing NGN3 but inhibited their further differentiation toward the endocrine lineage, indicating the significance of timing in 3D culture initiation . Initiation of 3D culture at different stages has been reported. Inspired by the aggregated distribution of PDX1 + / NKX6.1 + pancreatic progenitors in monolayer culture (Toyoda et al., 2015;Mamidi et al., 2018), starting suspension culture from progenitors increased the proportion of PDX1 + / NKX6.1 + cells and the expression levels of PDX1 (Toyoda et al., 2015). The transfer of planar pancreatic endodermal cells to an air-liquid interface allowed apical-basal polarity and generated ∼1-2-mm clusters with enhanced NGN3, INS, GCG and SST expression . Direct evidence showing the impact of the 3D structure on islet organoid maturation was provided by comparing 2D sc-β cells and their clustered counterparts. A static suspension culture of dissociated sc-β cells generated ∼50-150-μm endocrine cell clusters (ECCs) with significantly augmented coexpression of PDX1/NKX6.1 and PDX1/GLUT1; reduced expression of MAFB, a marker of endocrine progenitors; and improved functional maturation (Kim et al., 2016). The ECCs contained all endocrine cell types except α cells; however, cell heterogeneity has not been promoted by clustering, as shown by decrease of PPY and GCG expression (Kim et al., 2016). Nair et al. further examined the metabolic mechanisms underlying clustering-facilitated β cell maturation and revealed the role of clustering-promoted mitochondrial oxidative respiration in the metabolic maturation of β cells (Nair et al., 2019).

Scaffold-based 3D culture
Instead of relying on self-organization to establish 3D structures, other strategies incorporated ECM components as scaffolds to promote 3D structure formation and cellmatrix interactions (Fig. 1B). Dissociated mouse embryonic pancreatic progenitors can survive in 2D culture, but only in Matrigel organoids can be formed, highlighting the role of cell-matrix interactions in cell fate specification (Greggio et al., 2013). 3D islet organoids built within a scaffold combining collagen and Matrigel (C-M scaffold) yielded α, β, γ and PP cells . The β-like cells expressed elevated levels of INS, GLUT2 and MAFA and harbored 12-fold more insulin granules than their 2D cultured counterparts. In addition to facilitating late-stage maturation of endocrine cells, the C-M scaffold also provides a favorable microenvironment for endocrine commitment in the early developmental processes, as manifested by the reduced expression of HNF6 (restricted in exocrine cells in the native pancreas) in 3D pancreatic progenitors compared with planar counterparts . Candiello et al. devised a novel hydrogel-based platform called "Amikagel" by polymerizing amikacin hydrate and polyethyleneglycol diglycidyl ether (PEGDE), which facilitated the self-aggregation of hESC-derived pancreatic progenitor cells and their coaggregation with supporting endothelial cells (Candiello et al., 2018). Amikagel-produced 3D cell aggregates had increased populations of cells coexpressing NKX6.1/PDX1, increased expression of insulin/C-peptide, and enhanced glucose responsiveness (Candiello et al., 2018). Islet organoids cultured in a microporous scaffold showed improved control over islet organoid size and cell-cell interactions. These ∼250-425-µm islet organoids had more mature marker expression and performed better in GSIS than their counterparts in suspension culture (Youngblood et al., 2019).
Using a decellularized pancreas as a scaffold is another elaborate and effective strategy to provide an intact biomimetic ECM microenvironment for supporting the differentiation and maturation of sc-β cells. This strategy includes the perfusion of decellularization reagents to completely remove cells residing in the pancreas followed by perfusion of a cell suspension to recellularize the scaffolds. The scaffold retains biochemical signals with ECM proteins and can mimic biophysical signals, such as shear stress, during recellularization. Mouse iPSC-derived β-like cells have been perfused into decellularized scaffolds and matured into functional β-like cells (Wan et al., 2017). Decellularized ECM hydrogels offer another more flexible microenvironment, in which the ECM composition can be manipulated and the embedding of cells is more convenient than recellularization. Native islets cultured in decellularized pancreatic ECM exhibited improved glucose responsiveness and enhanced viability of both islets and islet-resident endothelial cells (Jiang et al., 2019).
Proteomic analysis of decellularized rat pancreas revealed that collagen V was a crucial component in the islet niche (Bi et al., 2020). Suspension cultures of hPSCs in collagen V-augmented Matrigel-coated dishes generated more mature islet cells (Bi et al., 2020). Notably, in addition to biochemical signals guiding cell fate determination, the ECM components surrounding the developing cells are also in continuous flux to regulate cell differentiation to different stages. Enrichment of certain ECM components, such as laminin, promoted bipotent pancreatic progenitors to commit to endocrine cell specification, while exposure to other ECM components led to duct cell differentiation (Mamidi et al., 2018), implying that stage-specific scaffolds may facilitate endocrine differentiation in vitro and improve the induction efficiency.

Innovative strategies facilitating functional maturation of islet organoids
In addition to optimization of the induction protocol and establishment of 3D structures, instructive strategies derived from the current understanding of the islet developmental process have facilitated the maturation of sc-β cells.

Vascularized islet organoids
In vivo maturation of islet organoids or hPSC-derived pancreatic progenitors typically generates islet organoids that are morphologically and functionally similar to native islets (Rezania et al., 2013Augsornworawat et al., 2020). One of the key disparities between the features of islet organoids derived from in vitro environments and those derived from in vivo environments is their vascular networks, which are abundant in native islets and transplanted islet organoids but scarce in in vitro-derived products (Ranjan et al., 2009;Eberhard et al., 2010;Sneddon et al., 2018). Islets are considered one of the most highly vascularized organs due to their endocrine features. Therefore, vascularization is vital for islet development and endocrine function, and lack of vascular networks reduces not only the New insights into islet organoid REVIEW fidelity of islet organoids but also their viability during in vitro culture because of the insufficient nutrient and oxygen supply to deeply buried cells. Many strategies have attempted in vitro vascularization by coculturing endothelial cells with endocrine cells. Taniguchi's group reported that coculturing of cell lines, native tissue fragments, and iPSC spheroids with human umbilical vein endothelial cells (HUVECs) and mesenchymal stem cells (MSCs) in Matrigel enabled the formation of vascularized islet organoids (Fig. 1B) (Takahashi et al., 2018a;Takahashi et al., 2018b). Gene expression patterns of the vascularized islet organoids reflected native islets better than those of non-vascularized islets (Takahashi et al., 2018a). Furthermore, a vascular-inductive bed was established by functionalizing the bioactive polyethylene glycol (PEG) hydrogel with vascular endothelial growth factor A (VEGFA), which is widely believed to be secreted by early β cells during embryonic development to recruit endothelial cells and promote the formation and expansion of the vasculature (Phelps et al., 2013). This protocol facilitated intraislet vasculature formation and significantly improved the functional performance of islets embedded in this bed and delivered to the small bowel mesentery in a diabetic mouse model (Phelps et al., 2013).

Other strategies
Core circadian clock activators have been predicted to facilitate maturation of the epigenomic dynamics of islet organoids (Alvarez-Dominguez et al., 2020). Daily circadian rhythms in islet organoids tuned, but did not elicit, glucose responsiveness by increasing the glucose threshold of GSIS (Alvarez-Dominguez et al., 2020). Defective GSIS but normal KCl responsiveness of β-like cells have been demonstrated in various studies. A possible reason for this impaired glucose sensing and GSIS is reduced anaplerotic cycling stemming from reduced GAPDH and PGK activity . The use of cell-permeable intermediate metabolites downstream of these two enzymes successfully improved islet function .
Gene manipulation has also been considered for the maturation of islet organoids. Inducible knockout of LIN28B after definitive endoderm formation significantly improved GSIS of the derived islet organoids, but the expression of mature islet markers was not significantly upregulated (Zhou et al., 2020). Overexpression of ERRγ, an important regulator of mitochondrial oxidative ATP biosynthesis, increased the glucose responsiveness of β-like cells in vitro and in vivo (Yoshihara et al., 2016).
Noncanonical WNT signaling was reported to participate in pancreas specification and islet maturation (Rodriguez-Seguel et al., 2013;Yoshihara et al., 2020). Correspondingly, treatment of immature islet organoids with WNT4, an activator of noncanonical WNT signaling, significantly increased mitochondrial content and oxidative metabolism and improved GSIS in vitro, without affecting cell fate determination (Yoshihara et al., 2020). Enrichment of the CD177 + cell subpopulations from induced definitive endoderm, in which WNT/PCP signaling was specifically activated, generated a higher percentage of MAFA + mature sc-β cells with elevated glucose responsiveness (Mahaddalkar et al., 2020).

Islet organoids derived from ASCs and other sources
ASCs residing in adult tissues have the capacity for selfrenewal and differentiation under certain conditions, and they are considered another potential source for organoids (Clevers, 2016). In contrast to PSC-derived organoids, ASCderived organoids have the identity of a given organ (Fig. 1C) (Habener, 2004) and are able to avoid tumorigenicity in potential clinical applications. There is a longstanding debate regarding whether the adult pancreas contains stem or progenitor cells. Several particular cell groups have been proposed as pancreatic progenitor or progenitorlike cells, which have the ability to differentiate into insulinsecreting cells in vitro, including Nestin-positive cells isolated from mouse pancreas (Wei et al., 2012), cells with high aldehyde dehydrogenase enzyme activity sorted from human fetal and adult pancreas (Loomans et al., 2018;Oakie et al., 2018), and P2RY1-and ALK3-positive cells sorted from human pancreatic ducts (Qadir et al., 2018;Qadir et al., 2020). Recently, a new population of Procr + endocrine progenitors was identified from adult mouse islets . Procr + progenitors can generate functional endocrine cells in vitro and in vivo, and islet organoids derived from Procr + cells are capable of reversing hyperglycemia in a mouse model of streptozotocin-induced T1D . Additionally, non-β endocrine cells, pancreatic exocrine cells, and even nonpancreatic cells such as intestinal enteroendocrine cells have been demonstrated to transdifferentiate into insulin-secreting cells by reprogramming in some cases (Thorel et al., 2010;Chen et al., 2014;Lemper et al., 2015). However, genetic lineage-tracing studies indicated that the exocrine compartment was unlikely to contribute to the formation of endocrine cells in vivo (Aguayo-Mazzucato and Bonner-Weir, 2018; Dominguez-Bendala et al., 2019). The current dominant view is that the replenishment of adult pancreatic β cells relies principally on their self-duplication instead of potential pancreatic stem cell differentiation (Dor et al., 2004).

APPLICATIONS OF ISLET ORGANOIDS
As the developmental and functional properties of islet organoids have become more representative of physiological human islets, islet organoids can function as surrogates for inaccessible human islets, especially those from diabetic individuals. The combination of gene editing technologies, Xiaofei Zhang et al. such as CRISPR/Cas9 and refined iPSC technology, allows the manipulation of islet organoids to investigate the developmental, functional, and pathological mechanisms of normal and diabetic human islets and to develop personalized and informative drug screening platforms (Fig. 2).

Models for studying islet development
Although the mechanisms governing islet development are largely conserved between human and murine models, human-specific developmental events and structural or Figure 2. Promising applications of islet organoids in disease modeling and personalized medicine. Personalized islet organoids can be derived from either iPSCs reprogrammed from fibroblasts/peripheral blood mononuclear cells (PBMCs) or ASCs sorted from pancreatic fragments. Islet organoids derived from hESCs or iPSCs reprogrammed from healthy individuals are able to serve as a novel model system for human islet development and function research. Patient-derived iPSCs/ASCs can be used to generate functional islet organoids with or without mutation correction (Millman et al., 2016;Manzar et al., 2017;Saarimaki-Vire et al., 2017;Balboa et al., 2018;Ma et al., 2018;Maxwell et al., 2020), providing unlimited sources of islets for autologous transplantation. The abnormal islet organoids derived from patients or gene-edited hESCs faithfully recapitulate developmental and functional defects under pathological conditions, offering novel valuable platforms for the study of pathological mechanisms and personalized drug screening.
New insights into islet organoid REVIEW functional characteristics arise remarkable attention (Nair and Hebrok, 2015). Islet organoids derived from hPSCs are a promising system for investigating pivotal niche signals and biochemical or biophysical mechanisms governing islet development and developmental defects that result in congenital diabetes. In turn, this information can be further used to improve the design of differentiation strategies to enhance the maturation of islet organoids. Using sc-β cells, Sharon et al. found that WNT inhibition was a major regulatory mechanism in endocrine differentiation, capable of adjusting the induction output of β-like cells (Sharon et al., 2019). Adjusting the balance between proliferation and differentiation of pancreatic progenitors could expand the progenitor population to generate increased numbers of endocrine cells and force the differentiation of the pancreatic progenitors to eliminate undesired progenitors. However, effective strategies to manipulate this balance have not yet been developed. A possible solution is inhibition of YAP to manipulate the Hippo signaling pathway, whose activation facilitates pancreatic progenitor proliferation and whose inhibition favors endocrine specification (Rosado-Olivieri et al., 2019). Overexpression of YAP in β-like cells of derived islet organoids revealed a subpopulation of β cells with activated LIF pathways and increased proliferative capacity (Rosado-Olivieri et al., 2020), indicating the ability of these cells to reflect subtle characteristics of human islets.
Combined with gene editing technologies such as CRISPR/Cas9, organoids have the capacity to accomplish the incomplete gene regulatory network of islet development and pathogenesis. CRISPR screening in sc-β cells identified that the vitamin D receptor (VDR) is required for β cell differentiation and maintenance in vitro, and subsequent studies revealed an unrecognized role of VDR in β cell survival under inflammatory stress (Wei et al., 2018). Combining a novel gene editing platform termed iCRISPR and hPSC-directed differentiation, Zhu et al. confirmed the roles of conserved transcription factors in mice and humans, revealed the novel regulator RFX6 in the specification of pancreatic progenitors, and reported an intriguing haploinsufficiency requirement of PDX1 in endocrine differentiation (Zhu et al., 2016). Moreover, NGN3 −/− hPSC derivatives showed negligible functional endocrine specification in several independent studies (McGrath et al., 2015;Zhu et al., 2016), and some patients carrying NGN3 mutations retained at least partial endocrine function. Using NGN3 −/− hPSC derivatives, Zhang et al. found that NGN3 mutations affected protein instability and DNA binding capacity during pancreas development (Zhang et al., 2019). Another transcription factor, MAFB, is expressed in adult human β-cells but absent from the murine counterparts. Using CRISPR/Cas9 gene editing coupled with endocrine cell differentiation strategies, Russell et al. generated MAFB knockout hPSCs and revealed that MAFB is essential for human β-cell identity during endocrine cell development (Russell et al., 2020). Applying state-of-the-art strategies to establish 3D islet organoids may further systematically elucidate the developmental and functional mechanisms involving β cells and their regulatory interactions with non-β endocrine and nonendocrine cells.

Models for disease research
Current model systems limit the research and exploration of diabetes and other disease mechanisms; for example, cell lines respond poorly to glucose stimulation and cannot model interactions between β cells and other cell types, which may play roles in islet function and the onset of diabetes. Islet organoids provide a novel system in which to study diabetes. β cells in islets are destroyed during disease progression in T1D and thus were previously difficult to be studied. Currently, iPSCs derived from patients with T1D are used to generate new models demonstrating that cells respond to different forms of β cell stress in vitro (Millman et al., 2016). Analysis of three T2D-related genes, CDKAL1, KCNQ1, and KCNJ11, which were identified by GWAS, in gene-edited hESCs showed normal endocrine specification but impaired insulin secretion in vitro and in vivo (Zeng et al., 2016). These phenotypes mimicked symptoms in patients carrying mutations in these genes, whereas animal models showed irrelevant phenotypes (Asahara et al., 2015). Wolfram syndrome, an autosomal recessive disease caused by mutations in WFS1, is characterized by juvenile-onset diabetes (Rohayem et al., 2011). β-like cells differentiated from iPSCs derived from Wolfram syndrome patients showed decreased insulin content and increased activity of molecules related to endoplasmic reticulum (ER) stress, indicating that β cell failure is caused by WFS1 deficiency (Shang et al., 2014;Maxwell et al., 2020). Congenital hyperinsulinism (CHI), a rare genetic disorder related to mutations in sulfonylurea receptor 1 (SUR1), encoded by ABCC8, is characterized by excessive insulin secretion and hypoglycemia (Nessa et al., 2015). ABCC8-deficient sc-β cells recapitulated the disease phenotypes of CHI in vitro, indicating that these cells could be an attractive model for further elucidating SUR1 function . We have summarized the disease models related to islet organoids and sc-β cells in Table 1.

Platforms for drug screening
Islet organoids are better platforms than other models for drug screening in diabetes. Via a chemical drug library, T5224 was discovered to be able to rescue β cell dysfunction in CDKAL1 −/− sc-β cells by inhibiting the FOS/JUN pathway, providing a potential treatment for T2D (Zeng et al., 2016). Induction of GLIS3 −/− hESCs showed impaired β cell differentiation and increased apoptosis. Moreover, a highcontent chemical screening based on GLIS3 −/− hESCs derivatives identified a drug candidate, galunisertib, that could be used for treatment of GLIS3-associated diabetes .

REVIEW
With advances in iPSC technology and personalized medicine, patient-derived β cells and islet organoids can serve as informative disease models recapitulating the pathogenesis and phenotypes of specific patients and can be used to assess patient-specific drug reactions upon screening. In a proof-of-concept experiment to evaluate possible applications for drug screening, sc-β cells from T1D patient-derived hiPSCs treated with three antidiabetic compounds, tolbutamide, liraglutide and LY2608204, showed increased insulin release (Millman et al., 2016). Additionally, 4PBA successfully restored the glucose responsiveness of islet organoids derived from patients with Wolfram syndrome (Shang et al., 2014). However, the limited functionality and reproducibility of established islet organoids require validation in other systems.

Models for studying infection with SARS-CoV-2 and other viruses
During the current worldwide COVID-19 pandemic, diabetes has been reported to be a high-risk factor for both SARS-CoV-2 infection and severe symptoms in infected patients (Muniyappa and Gubbi, 2020). Significantly increased mortality and comorbidities have been observed in patients with COVID-19 and pre-existing T1D and T2D, and these outcomes can be ameliorated with glycemic control . On the other hand, COVID-19 can also induce or worsen hyperglycemia through unclear mechanisms (Bornstein et al., 2020). Hence, islet organoids have potential as a promising human model at the whole-organ level for revealing the mechanisms underlying putative SARS-CoV-2 infection and destruction of pancreatic islets, as well as for assessing therapeutic drugs for both treatment of COVID-19 and glycemic control in the COVID-19 context (Fig. 3). As a proof-of-concept experiment, islet organoids and several other organoids were recently used to model SARS-CoV-2 infection in various human organs Lamers et al., 2020;Suzuki et al., 2020;Yang et al., 2020). hESC-derived sc-β and sc-α cells expressed the coronavirus spike protein receptor ACE2 and were permissive to SARS-CoV-2 infection with robust viral replication in vitro and in vivo, consistent with the results in human primary β and α cells, suggesting the authenticity of islet organoids in modeling virus infection . SARS-CoV-2-infected sc-β and sc-α cells showed increased insulin resistance and chemokine expression and a decrease in endocrine functional pathways due to increased apoptosis, resulting in insulin deficiency and even diabetes onset in some COVID-19 patients . However, the expression of ACE2 is relatively low in endocrine compartments compared with exocrine compartments. Evidence has shown that pancreatic ducts and microvasculature are more likely targets in SARS-CoV-2 infection (Coate et al., 2020;Kusmartseva et al., 2020). Therefore, whether islet organoids are an ideal SARS-CoV-2 infection model remains to be observed and determined. As a safe and convenient system, organ-on-a-chip technology can identify the crosstalk between different SARS-CoV-2-infected organs and elucidate the mechanisms underlying the elevated mortality in diabetic COVID-19 patients.
In addition to SARS-CoV-2, several other viruses are involved in the onset of T1D (Op de Beeck and Eizirik, 2016), and the islet organoid system has shown the capacity to model both virus infection and endocrine cell dysfunction, offering cellular and molecular insights into human-relevant pathogenesis.

Potential organ sources for tissue replacement and personalized therapy
Islet organoids have the potential to overcome the shortage of cadaveric islet donors by offering an unlimited number of functional islets on demand for regenerative therapy. The combination of iPSC and CRISPR technologies makes it possible to derive patient-personalized islet organoids with or without gene correction (Shang et al., 2014;Millman et al., 2016;Saarimaki-Vire et al., 2017;Balboa et al., 2018), paving the way for autologous transplantation and minimizing immune rejection. Compatibility of these methods with novel technology such as PD-L1 generated immune-evasive islet organoids which allows them to offer a new therapeutic strategy for autoimmune diabetes, like T1D (Yoshihara et al., 2020). Recently, phase I/II clinical trials for subcutaneous implantation of hESC-derived pancreatic endodermal or progenitor cells, which will further mature into functional endocrine cells in vivo, into T1D patients were approved  New insights into islet organoid REVIEW (Viacyte, 2014(Viacyte, , 2016(Viacyte, , 2017a(Viacyte, , b, 2019. Different kinds of encapsulation devices have been applied to contain and protect the implanted cells (Cooper-Jones and Ford, 2016). The PEC-Encap (VC-01) prevents immune cell accession from direct contact with the implanted cells, but allows vital nutrients and proteins to travel between the inside cells and the outside blood vessels (Agulnick et al., 2015). The PEC-Direct (VC-02) allows blood vessels to enter the device and directly interact with the implanted cells, while immune cells are also capable of entering the device (Cooper-Jones and Ford, 2016).

REMAINING CHALLENGES AND PROSPECTIVE SOLUTIONS
Although improvements have allowed 3D hormone-producing islet organoids to partially reproduce islet morphology and function, certain limitations need to be addressed.

Immaturity
Generally, established islet organoids do not fully recapitulate the behavior of native islets at the molecular and functional levels; for example, they have poor insulin secretion ability and lack dynamic insulin release. To improve the performance of islet organoids, one of the key fundamental efforts is characterization of the discrepancies between derived islet organoids and native islets, as well as updating induction strategies. Multiomics analysis approaches, especially single-cell RNA sequencing (scRNA-seq), have widely been used and shown to be informative for revealing molecular defects hindering the maturation and application of islet organoids. The results showed that most derived islet organoids rarely express certain mature β cell transcription factors, including MAFA, UCN3 and SIX3, differently express mature β cell functional genes, including PCSK1,PCSK2, CPE (Suckale and Solimena, 2010;Wang et al., 2011;Li et al., 2018), and the ratio of major endocrine cells is much different from that of native islets (Bakhti et al., 2019;Veres et al., 2019;Augsornworawat et al., 2020). Real-time imaging of the mouse embryonic pancreas demonstrated how molecular interactions and organ architecture influence the cell fate specification of pancreatic progenitors to commit to endocrine differentiation (Nyeng et al., 2019). Similar studies on islet organoids or native human islets will substantially contribute to the current understanding of islet development.

Heterogeneity
Currently established islet organoids have lost at least part of the heterogeneity compared with their in vivo counterparts. Although most islet organoids contain all or several of the five major islet cell types, their proportion of β cells is much higher than that of native human islets. Human islets contain more non-β endocrine cells, which disperse more widely and without a prominent core, than rodent islets (Cabrera et al., 2006;Arrojo e Drigo et al., 2015;Roscioni et al., 2016). In addition, β cells are highly heterogeneous and comprise several subpopulations differing in electrical activity, glucose responsiveness, and proliferative activity (Meda et al., 1984;de Vargas et al., 1997;Van de Casteele et al., 2013;Bader et al., 2016;Roscioni et al., 2016). The mechanism and biological relevance of such heterogeneity in humans are not fully understood, and high-fidelity islet organoids may help elucidate this issue. Nonendocrine supporting cells, including endothelial cells and neurons, are also barely observed in current islet organoids. These supporting cells are essential components that form vascular and innervation networks, which facilitate islet development by integrating islets into the circulatory and nervous systems (Ahren, 2000;Ranjan et al., 2009;Borden et al., 2013). Incorporation of supporting cells in islet organoids may not only improve the maturity and functionality of the derived islet organoids but also contribute to identifying the interactions between different cell types in human islets. Vascularized islet organoids have been attempted by several elaborate strategies (summarized above); however, innervated islet organoids still need to be pursued in the future. Islet organoids have certain undesired heterogeneity. ScRNA-seq revealed that most derived islet organoids comprise large numbers of proliferating progenitors or precursors and irrelevant or uncharacterized cell types (Veres et al., 2019;Weng et al., 2020), which hinders their in vitro application and in vivo transplantation. Elimination of such undesired heterogeneity requires either improvement of the induction efficiency or cell isolation to remove the unwanted populations and enrich the endocrine populations. CD49a has been identified as a surface marker for sc-β cells and used to purify β-like cells (Veres et al., 2019). GP2 has been determined to be a surface marker for PDX1 + /NKX6.1 + pancreatic progenitors. This protein has been employed to enrich PDX1 + /NKX6.1 + cells, eliminate ∼20%-40% of undifferentiated hPSCs, early derivatives or irrelevant cell types, and generate a higher percentage of βand α-like cells (Ameri et al., 2017;Cogger et al., 2017). Other innovative strategies to remove the undesired products have been introduced to avoid teratoma formation after transplantation and to indirectly elevate induction efficiency. Insertion of the suicide gene iC9 in the SOX2 locus enabled induction of apoptosis of undifferentiated hPSCs upon treatment of the final-stage products with the iC9 inducer AP1903, while the differentiated hPSC derivatives were unaffected (Wu et al., 2019).

Batch-to-batch variations
Batch-to-batch variations in the efficiency of differentiation, organoid quality (e.g., size and shape), and organoid viability or functionality are significant. This stochasticity reflects the absence of pivotal niche signals in the culture cocktail; thus, a complete and detailed understanding of the islet REVIEW Xiaofei Zhang et al. development process is required to optimize growth factor cocktail recipes. Additionally, cell-specific subtle modifications of induction strategies are needed, especially for patient-derived iPSCs. iPSCs derived from patients with T1D lost resistance to differentiation after transient demethylation treatment (Manzar et al., 2017). The demethylation-treated sc-β cells displayed no significant disparities compared with sc-β cells derived from healthy individuals, highlighting the necessity of cell-specific induction strategies (Manzar et al., 2017).

Challenges in the clinical applications of islet organoids
Islet organoids have shown potential as an organ source for the treatment of diabetes. However, several issues related to islet organoid transplantation must be carefully addressed.

Tumorigenicity
Although iPSCs show promise for personalized therapy, their issues with clinical safety remain to be addressed. The incorporation of stem cell derivatives in the human body raises concerns with respect to tumors (Zhu and Fan, 2018), especially teratoma formation, which is further exacerbated by the presence of uncharacterized proliferative cells in established organoids. The propensity for teratoma formation from iPSC derivatives has been reported to be affected by many factors, including iPSC tissue origin and strategies used for reprogramming, differentiation or transplantation (Miura et al., 2009). For example, c-Myc, an indispensable transcription factor in the reprogramming cocktail, is an oncogene (Okita et al., 2007). If it is omitted during the reprogramming, iPSCs would be generated with low efficiency (Nakagawa et al., 2010). Thus, prudent selection of induction strategies and rigorous examination before transplantation are indispensable.

Low survival and immunogenicity
Low survival and limited duration of euglycemia after transplantation have hindered transplantation of both donor islets to patients and islet organoids to mice. Low vascularization and immune rejection have accounted for significant islet cell loss. To overcome these issues, scientists have proposed elaborate encapsulation devices to shelter organoids from the immune system. Embedding a catheter into subcutaneous tissues at the transplant site in host mice before islet implantation can generate a vascularized space (Pepper et al., 2015). In addition, a recent study proposed using CRISPR/Cas9 to reduce the immunogenicity of stem cells by eliminating the expression of human leukocyte antigen (HLA) (Han et al., 2019).

Undefined ingredients and contamination
Most established islet organoids rely on Matrigel. However, Matrigel is not an approved material for clinical applications due to the presence of undefined ingredients. Some groups also use mouse embryonic fibroblasts (MEFs) as feeders for hPSC passaging or fetal bovine serum (FBS) during differentiation, thus resulting in xenogeneic contamination. The influence of xenogeneic contamination and undefined ingredients on the cellular microenvironment and organoid morphology or function remains unclear. To render the induced products amenable to transplantation, organoids need to be fully purified by enzymatic digestion to remove scaffolds Rossen et al., 2020). Another strategy is culture of the organoids in medical grade scaffolds, either derived from decellularized porcine tissues (Giobbe et al., 2019) or synthesized with defined components. Various kinds of hydrogels, including PEG-based hydrogels functionalized with laminin (Greggio et al., 2013), PLG scaffolds containing exendin-4 microparticles (Kasputis et al., 2018), and electrospun PLLA/PVA scaffolds (Ojaghi et al., 2019), have been developed to facilitate islet or β cell specification and maturation, substantially accelerating the clinical applications of islet organoids.

SUMMARY AND PERSPECTIVES
Islet organoids enable scientists to transform the understanding and treatment of diabetes. An ideal islet organoid should recapitulate the features of islet development and function of mature islets under physiological and pathological conditions. To improve the performance of current islet organoids, a more detailed understanding of islet development is necessary, and interdisciplinary methods, including organ-on-a-chip, 3D scaffold or bioprinting, are welcomed to construct more appropriate physiological niches for islet organoid development. As a highly expected application, islet organoids are considered an unlimited organ resource for transplantation and the treatment of diabetes, although several issues and safety concerns regarding organoid transplantation remain to be addressed. Organ transplantation techniques dedicated to reducing host immune rejection are also evolving (O'Sullivan et al., 2011;Pathak et al., 2019). The development of islet organoids will be a milestone in understanding and treating diabetes.

COMPLIANCE WITH ETHICS GUIDELINES
All the authors of this paper declare that they have no conflict of interest. This article does not contain any studies with human or animal subjects performed by the any of the authors.

OPEN ACCESS
This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creativecommons.org/licenses/by/4.0/.