1. IntroductionType 1 diabetes is a worldwide autoimmune disease characterized by chronic hyperglycemia due to the destruction of insulin-producing pancreatic beta cells. Beta cell death occurs by cell-to-cell contact with autoreactive T lymphocytes [1,2,3] and by contact-independent mechanisms where inflammatory cytokines play a key role [4,5]. The cytotoxic effect of cytokines on whole pancreatic islets [6,7,8] and on beta cells [9,10,11,12,13] is well-known.Among these cytokines, IFNγ has been detected in pancreatic lymphocytes from human diabetic patients [14]. Its mRNA is expressed by the immune cells infiltrating the islets, and correlates with destructive insulitis in rodent models of diabetes [15,16]. An IFNγ signature has been observed both in human pancreatic endocrine cells from diabetic patients [17] and in islets from the non-obese diabetic (NOD) mouse [18], a model of T1D [19,20]. In mice, antibodies against IFNγ decrease the incidence of cyclophosphamide-induced diabetes [21] and IFNγ-deficient mice are resistant to virus-induced diabetes [22]. Conversely, transgenic mice with beta cell specific IFNγ expression become diabetes prone [23]. IFNγ participates to diabetes progression by increasing beta cell visibility to immune cells through the upregulation of HLA class I molecules [24,25,26] and presentation of beta cell-derived peptides [27,28]. Another key feature of IFNγ is its capacity to induce at the surface of beta cells the expression of programmed death-ligand 1 (PD-L1) [29,30], an immune checkpoint protein [31]. PD-L1 may slow-down beta cell destruction and thus T1D development since its deficiency increases the incidence of diabetes both in NOD mice and in a model of diabetes induced by splenocytes transfer [32]. Additionally it has been suggested that pancreatic expression of PD-L1 delays beta cell destruction [33].While beta cells are at the centre of diabetes research, they are not the sole endocrine cell type in the pancreas. Indeed, beta cells cluster with glucagon-secreting alpha cells and somatostatin-secreting delta cells to form the endocrine units of the pancreas: the islets of Langerhans. Importantly, alpha and delta cells also show dysregulation during diabetes. For example, impaired glucagon secretion among diabetic patients has been identified for decades [34,35]. This result is supported by recent experiments performed on islets isolated from diabetic patients showing both a decreased expression of key alpha cells-enriched transcription factors, and a loss of glucose-regulated glucagon secretion [36,37]. A limited number of studies also suggest an alteration of delta cells function in type 1 diabetes [38,39]. Yet, information on alpha and delta cell functions and fates in T1D remains scarce. Interestingly, the different endocrine cell types within islet are tightly interacting [40,41,42], making conceivable that beta cell dysfunction or loss imbalances the neighboring alpha and delta cell function. Another hypothesis is that within islets, alpha and delta cells directly suffer from the diabetogenic milieu. Here, we asked whether IFNγ that impacts beta cells, also signals on alpha and delta cells.To address this question, we treated pancreatic mouse islet with IFNγ, FACS-sorted alpha, beta and delta cells using our recently developed FACS-based approach [43] and performed RNA-seq analysis. We show that alpha, beta and delta cells are surprisingly similar in their response to IFNγ. We also observed two populations of beta cells based on their response to IFNγ, reflecting a spatially restricted effect of IFNγ linked to its gradual diffusion across the islet. 4. DiscussionCytokines such as IFNγ are important mediators of inflammation during T1D progression [21,22,58,59]. As an example, an IFNγ signature is observed in islets from NOD mice as early as 6 weeks of age [18]. Transcriptomic changes induced by cytokines were also deeply studied in primary beta cells or beta cell lines [59,60,61,62,63]. However, our knowledge of IFNγ responses in other islet cell populations (glucagon-producing alpha cells and somatostatin-secreting delta cells) is more limited. In the present study, using our recently described FACS-based strategy to sort the major endocrine populations from mouse islets [43], we investigated the effects of IFNγ on mouse alpha and delta cells, and compare to beta cells. Our work pointed out two major findings: (i) alpha, beta and delta cells responses to IFNγ are similar and (ii) the diffusion of IFNγ within an islet is impeded, thus creating a gradient of concentration across the islet.The current study has been made possible by the development in the laboratory of a FACS-based strategy to purify live alpha, beta and delta cell populations [43]. This approach opens new avenues for pancreatic islets study. As an example, RNA-seq on bulk alpha, beta and delta populations solves both the requirement of complex reporter mice [40,64,65] and the eventual pitfalls of single-cell RNA sequencing [66,67], the main technic used for alpha and delta cells transcriptomic analyses [68,69,70]. The present protocol was set up on untreated mouse islets and we now demonstrate that this method is also robust on islets treated with IFNγ. As described here, this strategy allows the sorting of cells on multiple parameters. Here, we sorted beta cells based on PD-L1 expression following IFNγ treatment. PD-L1 is a stress response factor known to be induced by cytokines such as IFNγ [71], oncogenes such as epidermal growth factor [72] and hypoxia inducible-factor-1 [73]. Sorting was followed by reaggregation into pseudo-islets for further analyses. Such protocol permits to address the question of cell subtype stability, a current topic of major interest [56].We observed here that IFNγ transcriptomic signature of sorted mouse primary beta cells was similar to previous observations on human beta cells [74], demonstrating the strength of our strategy. Confident with our protocol, we also characterized alpha and delta cells responses to IFNγ. Remarkably, transcriptomic responses to IFNγ were similar between alpha, beta and delta cells, pointing out that beta cells are not the only source of inflammatory response genes in the islet. In this context, gene products activated by IFNγ in alpha and delta cells might impact diabetes development. This might be the case for Cxcl10 that encodes a chemokine expressed by beta cells, that attracts T lymphocytes and thus participates to diabetes onset and progression [75,76]. We showed here that alpha and delta cells also increase Cxcl10 expression upon IFNγ exposure. Recently, alpha cells expressing the CXCL10 protein were reported both in diabetic mouse model and in human islets from diabetic patients [77]. Our data suggest that this induction of CXCL10 relies on IFNγ, and add the fact that delta cells also contribute to CXCL10 production within islets. Thus, alpha and delta cell may favor, along with beta cells, infiltration of the pancreatic islets by immune cells. PD-L1 also represents another example of proteins upregulated by IFNγ in alpha and delta cells that might impact diabetes development. It has been shown that inflamed islets cells may limit immune cells activation through PD-L1/PD1 interaction [32,33]. With our experimental model, we confirmed that beta cells express high levels of PD-L1 in inflammatory condition as previously observed [29,49,78]. Interestingly, we demonstrate here that alpha and delta cells also increased Cd274 mRNA, leading to an increase of PD-L1 surface expression confirmed by flow cytometry. Whether PD-L1 expressed by alpha and delta can slow down the immune attack remains an open question.It is well documented that beside inducing gene expression, a combination of cytokines—including IFNγ—precipitates beta cell death [10,79]. A limited number of studies performed on alpha cell lines suggest that cytokines can also induce alpha cell apoptosis, but to a lesser extent than beta cells. [80]. It was recently hypothesized that the alpha cell resistance to cytokine-induced apoptosis relies on their high expression of poly-ADP Ribose Polymerase PARP-14, that modulates the expression of apoptosis-related genes [81]. In our dataset, using IFNγ, we indeed observed an increased expression of Parp14 in alpha cells exposed to IFNγ, but similar to beta and delta cells. So, our results do not support an alpha cell resistance to IFNγ based on PARP-14.Another major point derived from our study concerns differences in individual beta cell sensitivity to IFNγ within islets. Heterogeneity among beta cells has been identified for several decades. Data published in the 80th and 90th and reviewed in a landmark publication [82] indicated functional beta cell heterogeneity. As an example, glucose stimulates proinsulin biosynthesis in a dose-dependent manner by recruiting additional beta cells, that might suggest differences in beta cell sensitivity to glucose [83]. Moreover, in vivo, in rats, beta cells showed a regional heterogeneity in terms of their ability to respond to stimulation by insulin secretagogues [51]. In the present study, by measuring cell surface levels of PD-L1 following low-dose IFNγ stimulation, we observed two beta cell populations: PD-L1high cells were IFNγ-responsive while the PD-L1low cells were not, without difference in expression of surface IFNGR. PD-L1high and PD-L1low populations were previously described in other cell types such as a fibroblast cells line, but also murine lung, liver and kidney where they were linked to senescence [84] Heterogeneity in PD-L1 expression was also previously described in NOD mice where a population of beta cells that resists auto-immune attack and express high level of PD-L1 was described [49]. Interestingly, in that model, cells expressing higher levels of PD-L1 seemed to be less differentiated and expressed lower levels of beta cell markers such as Ins1, Ins2, Mafa, Pdx1, Nkx6.1, Slc2a2 while re-expressing markers of pancreatic endocrine progenitors such as Ngn3 [49]. However, in our model, we did not observe any difference in the expression of such genes, indicating similar differentiation status between PD-L1high and PD-L1low beta cells.We also assessed whether PD-L1high and PD-L1low beta cells represent stable populations fixed in a specific state or whether cells might navigate from one state to the other. Recent data based on sorting of beta cell populations [50,53] or single cell transcriptomic analyses [68,69,70,85,86] suggested that beta cells are fixed in transcriptional and functional states, giving rise to stable and specific beta cell populations entitled Beta-1, -2, but this remains a matter of discussion. As examples, cell tracking [87] and pseudo-time trajectory [88] support the notion that cells swap between different states. Our data indicate that PD-L1high and PD-L1low beta cells represent transient populations. Indeed, PD-L1low beta cells were able to induce PD-L1 surface expression in response to a second pulse of IFNγ. Moreover, IFNγ stimulation of dispersed islet cells resulted in all beta cells being PD-L1high, further suggesting that the location of a beta cell within an islet might dictate its sensitivity to IFNγ. Interestingly, tissular diffusion of IFNγ was recently addressed with melanoma cells as the penetrance of the cytokine was dependent on the density of cytokine-consuming cells in the tissue, creating a gradient of concentration [57,89]. Our data derived from pancreatic islets treated in vitro with IFNγ and from pancreases of NOD mice suggest that a similar gradient mechanism might contribute to our observations, pancreatic islet being a cell-dense structure. There, IFNγ would be captured by the first layer of cells, protecting the cells located inside the islet. Another eventual actor of this process might be the basal membrane surrounding the islet, which has been shown to protect islets from immune invasion [90]. The basal membrane might limit the diffusion of soluble molecules such as cytokines, as suggested by the observation that transplanted islets show better resistance to cytokines when components of extracellular matrix are added [91,92].

Taken together, our data, suggest that we should re-consider the importance of endocrine non-beta cells in T1D. Indeed, alpha and delta cells are similarly affected by inflammation and respond with induction of genes that may influence the recruitment and activation of autoreactive T cells, and thus disease progression, such as Cxcl10 and Cd274. Our work also proposes a new insight into the propagation of inflammation across an islet, where an islet would shield itself against inflammatory stimuli such as IFNγ entry. Experiments focusing on islet inflammation should take this notion into account. In this study, experiments were performed using mouse islets. Future studied will aim at determining whether similar results can be obtained with either pseudo-islets made of homogeneous human beta cell lines or with human islets. It will also be interesting to test whether other cytokines (Il-1β, TNFα) show similar gradient-like action.