β-Cell Replacement Strategies: The Increasing Need for a “β-Cell Dogma”

Type 1 diabetes is an auto-immune disease resulting in the loss of pancreatic β-cells and, consequently, in chronic hyperglycemia. Insulin supplementation allows diabetic patients to control their glycaemia quite efficiently, but treated patients still display an overall shortened life expectancy and an altered quality of life as compared to their healthy counterparts. In this context and due to the ever increasing number of diabetics, establishing alternative therapies has become a crucial research goal. Most current efforts therefore aim at generating fully functional insulin-secreting β-like cells using multiple approaches. In this review, we screened the literature published since 2011 and inventoried the selected markers used to characterize insulin-secreting cells generated by in vitro differentiation of stem/precursor cells or by means of in vivo transdifferentiation. By listing these features, we noted important discrepancies when comparing the different approaches for the initial characterization of insulin-producing cells as true β-cells. Considering the recent advances achieved in this field of research, the necessity to establish strict guidelines has become a subject of crucial importance, especially should one contemplate the next step, which is the transplantation of in vitro or ex vivo generated insulin-secreting cells in type 1 diabetic patients.


INTRODUCTION
Diabetes affects 422 million people worldwide and its increasing prevalence is predicted to reach 552 million patients by 2030 (Whiting et al., 2011;Zhou et al., 2016). The most common feature associated with diabetes is also its principal diagnosis: chronic hyperglycemia. Type 2 diabetes results from a combination of insulin resistance in target organs and defective β-cells (Bergman et al., 2002), while type 1 diabetes is due to the autoimmune-mediated loss of the pancreatic insulinsecreting β-cells, leading to insufficient glucose disposal (WHO, 1999). For both pathologies, the loss of insulin activity causes an imbalance in glucose homeostasis, eventually resulting in multiple cardiovascular complications (Hanefeld et al., 1996;Alwan, 2010;Pascolini and Mariotti, 2012). In the case of type 1 diabetes, the hyperglycemia can be efficiently managed by means of insulin supplementation, but patients still display an overall shorter life expectancy and a relatively altered quality of life (Lind et al., 2014;Morgan et al., 2015). In this context, finding an alternative to daily injections of exogenous insulin has become a crucial research goal. Toward this goal, many current efforts focus on β-cell replacement therapies using different strategies, alongside the development of efficient ways to protect such newly generated cells from the autoimmunology inherent to type 1 diabetes (detailed by Desai and Shea, 2016).
During the last decade, impressive progresses have been made toward the generation of functional insulin-secreting β-like cells (Vieira et al., 2016). Most of the strategies employed initially relied on deciphering the molecular mechanisms underlying βcell (neo) genesis and applying this knowledge to in vitro or in vivo (trans) differentiation: the purpose being to drive progenitor cells (either stem cells or multipotent cells) or differentiated cells toward a β-cell phenotype. To validate the identity of the resulting "β-like" cells, a number of tests have been employed, ranging from marker gene analyses to functional challenges. However, while browsing the recent literature, we noticed important differences between the features examined by various authors. Importantly, our survey indicates that the number of key features assessed to establish whether neo-generated insulinproducing cells are indeed "true" β-cells has not progressed in the last years. These observations clearly establish the need of an "initial β-cell profiling."

Methodology
Our analyses were focused on the following β-cell features: -Glucose Stimulated Insulin Secretion (GSIS) was confirmed when the authors reported at least one insulin and/or C-peptide ELISA measurement increasing upon glucose stimulation, or when an improved response for mice subjected to an intraperitoneal or oral glucose tolerance test was observed. Of note, the sole presence of C-Peptide as a sign of GSIS was not considered. -Gene expression of bone fide β-cell markers was validated when RT-PCR, transcriptomics analyses or immunolabeling was used. -Mice reverting from an established diabetic state (NOD/Akita background, streptozotocin or alloxan treatment) to stable euglycemia due to the presence of neogenerated insulinproducing cells validated the feature "Hyperglycemia Recovery." This could be achieved either by in vivo transdifferentiation or allogenic transplantation of in vitro differentiated cells.

Validation of β-Cell Features
Aiming to summarize the β-like cell features assessed, a survey of the recent literature reporting β-like cell neogenesis was conducted by analyzing all the data provided by the authors in order to deliver an accurate compilation. In the resulting 59 original publications, all the properties used to characterize neogenerated β-like cells were inventoried, ranking them by year of publication and the frequency of their use as a validation tool ( Table 2).

Insulin and β-Cell Function
Expectedly, insulin expression was the only feature commonly displayed by all reported neo-generated β-like cells. Interestingly, the responsiveness of such β-like cells to glucose stimulation was assessed in 88% of the publications analyzed, indicating a satisfying physiological response for most of these newly generated cells. However, the recovery upon induced hyperglycemia was validated in only 46% of the publications listed. In the case of insulin-secreting cells generated in vitro and challenged in vivo, this can most likely be attributed to transplantation-related issues and the need to host immunedeficient mice, in vitro differentiated allogeneic or xenogeneic cells being rejected upon graft in wild-type animals. In the case of in vivo transdifferentiation, on the contrary, the immunological rejection is bypassed by the creation of autologous β-like cells, and consequently the hyperglycemic recovery was assessed in all publications except one.

Transcription Factors
The Pdx1 gene appeared second in ranking, while being a disputed proof of completed β-cell differentiation ( Table 2). Indeed, during the course of pancreas morphogenesis, Pdx1 is first detected in all pancreatic progenitor cells, its expression being subsequently detected in mature β-cells (Ahlgren et al., 1996(Ahlgren et al., , 1998. Pdx1 should therefore not be considered as a mature β-cell marker in approaches aiming at recapitulating pancreas development, as one cannot exclude that an undifferentiated proportion of the cells still expresses this transcription factor. We consequently suggest that its presence should solely be assessed in insulin-secreting cells using double labeling. During pancreas development, an initial expression in pancreatic/endocrine precursors and a subsequent expression in mature β-cells is in fact a feature displayed by numerous transcription factors considered as bona fide β-cell markers. Indeed, HlxB9, Nkx6.1, Pax4, MafA, Nkx2.2, Isl1, NeuroD1, Pax6, Foxa2 are all involved in pancreas organogenesis, their expression being maintained at adult age in β-cells (for the first four) and additional cell subtypes (for the remaining- Ahlgren et al., 1997;Naya et al., 1997;Sander et al., 1997Sander et al., , 2000Sosa-Pineda et al., 1997;Sussel et al., 1998;Li et al., 1999;Edlund, 2002;Henseleit et al., 2005;Zhao et al., 2005). It is thus necessary to validate their expression in insulin-secreting cells either using (q)RT-PCR after FACS sorting or immunohistochemical analyses coupled to insulin detection.

Enzymes and Hormones
In the pancreas, glucokinase is expressed in mature α-and βcells, such enzyme being involved in glucose-sensing (Pierreux et al., 2006). PC1/3 and PC2 correspond to enzymes essential for proinsulin processing and are thus necessary for the normal function of mature β-cells (Marzban et al., 2004;Ugleholdt et al., 2006). Iapp is co-released with insulin by β-cells and acts as a satiation signal (Akesson et al., 2003), while urocortin3 is a hormone secreted by β-cells, acting to induce somatostatin secretion by δ-cells (van der Meulen et al., 2015). Altogether, these proteins are involved in the β-cell metabolism and function, and therefore should mostly be expressed only in mature insulin-secreting cells. Accordingly, they represent markers of the maturation state of differentiated insulinsecreting cells and they should therefore be tested in a more systematic way to ascertain a terminally-differentiated β-like cell phenotype.

Channels
Even though they are not markers of differentiation or maturation per se, potassium channels Kir6.1 and Kir6.2 and ATPbinding cassette channel Sur1 are required for proper insulin secretion (Proks et al., 2002;Kefaloyianni et al., 2013). Coupled to GSIS assessment, the presence of these proteins should be used in order to establish optimal β-like cells response to glucose.

DISCUSSION
Following the outstanding progresses made in the fields of stem-cell differentiation and in vivo trans-differentiation, human applications appear increasingly conceivable. However, one could only contemplate such an exciting clinical outcome after ensuring that the neo-generated insulin-secreting cells are genuine and could therefore fully replace endogenous β-cells. The features displayed in Table 2 rank the common features classically assessed in neo-generated insulin-secreting cells which, taken together, could theoretically constitute an "initial profiling" for β-like cells. Obviously, numerous additional aspects of the β-cell phenotype should be considered when aiming at establishing a standard validation protocol for β-like cells.
Regarding the prerequisites listed in Table 2, as previously discussed, appropriate levels and correct localization of the βcell-specific marker genes undoubtedly should be confirmed by immunohistochemistry using double labeling, especially in the case of developmental transcription factors. Concerning insulin itself, since its release in response to a stimulus is the main property requested from β-like cells, careful examination of its glucose-stimulated secretion and proper storage of insulin are essential. For the latter, the visualization of secretory vesicles by electronic microscopy appears as a valuable tool. Combined with the PC1/3, PC2, and C-peptide expression analyses, proper processing of proinsulin could be convincingly demonstrated.
In addition, the analysis of single insulin-producing cells could provide cues on their ability to behave as endogenous β-cells.
While global proteomic and transcriptomic analysis of neogenerated cells would give detailed information about their state of differentiation, one of the main issues is the heterogeneity of β-cells both in human and rodents (Rutter et al., 2015;Dorrell et al., 2016;Roscioni et al., 2016). A detailed analysis of these aspects, as well as a list of putative routine experiences, are described by James D. Johnson in his elegant review detailing the remaining steps prior to reaching clinical applications (Johnson, 2016). This report provides a thorough analysis of the current state-of-the-art from the point of view of a β-cell biologist, also highlighting the need for standardized protocols validating β-like cells functionality.
In addition to the initial profiling of neo-generated β-like cells, systematic single-cell next generation transcript sequencing (RNA-seq) would be the most decisive validation for β-like cells, providing the complete expression profile of these cells and thus their state of differentiation. Importantly, this transcriptomic phenotyping would not only assess the activation of necessary features, it would ensure the correct repression of non-βcell genes, including the disallowed genes known to interfere with appropriate β-cell functionality (Pullen and Rutter, 2013;Lemaire et al., 2016;Pullen et al., 2017).
A chronological display of the average number of features assessed in neo-generated β-like cells is provided in Figure 1. Importantly, this ranking clearly shows a scattering in the number of validated features, the average values not increasing in time (which would be indicative of an enhanced scrutiny FIGURE 1 | Graphical representation of the number of β-like features validated, per year of publication. The data displayed represent the average number of validated markers ± range, for each year of publication (from the list displayed in Table 2). cells over time). These discrepancies clearly reflect the lack of a canonical list of features to be validated. We thus propose to systematically assess most (if not all) of the features displayed in Table 2 as an initial roadmap toward the establishment of the β-cell identity.

AUTHOR CONTRIBUTIONS
AV conceptualized the study, chose the methodology and wrote the original draft. ND, FA, TN, SN, and SS contributed to the formal analysis and investigation. PC validated the results, supervised the study, edited the draft and reviewed the final version prior to submission.