Genetic susceptibility is considered a valuable clue to the molecular mechanisms of T1D. In humans, over 50 gene variants have been identified as carrying a risk for T1D, the most important being HLA genes contributing to 40-50% of the lifetime risk of T1D (13). HLA class-II genes provide the highest susceptibility. They are involved in the initiation of the T cell response to β-cells autoantigens in T1D. Among HLA class-II genes, HLA-DQ8 molecule carries the highest risk and is present in over 40% of at-risk and pediatric T1D patients (14), while HLA-DQ6 molecule is protective with a relative risk of 0.2 (15). HLA-DR molecules carry an independent risk with the particular case of HLA-DR*04:01 (14), and to a lesser extent DR*04:05 and 04:02 (16). HLA class-I molecules provide a lower risk and are associated to the progression of the disease. The strongest class-I susceptibility is conferred by HLA-B*39:06, HLA-B *57:01, HLA- B*18:01, HLA-A*02:01, HLA-A*24:02 and HLA-C*05:01 (15, 16). Class-I and class-II MHC genes directly control the peripheral T cell repertoire and the spectrum of antigen epitopes that are presented to T cells. Thus, introducing HLA transgenes in the mouse allows to characterize HLA-restricted autoantigen peptides at play in human T1D (11). “Humanized mouse” models have been developed by introducing HLA class-I or class-II transgenes into different mouse strains with or without invalidating corresponding murine H2 class-I or class-II genes (11). The expression of HLA class-I transgenes and their interaction with murine CD8⁺ T cells has been obtained using constructs encoding a human β2-microglobulin (β2m) covalently linked to HLA alpha₁ and alpha₂ and H2 cytosolic and transmembrane alpha₃ chain domains (11, 17, 18). H2 class-I genes have been invalidated by deleting either the murine β2m or the MHC class-I locus. On the NOD genetic background, mice expressing HLA-A2.1 or HLA-B39 transgenes developed accelerated T1D (19–22) while the expression of HLA-DQ6 decreased the incidence of spontaneous diabetes and insulitis (23, 24). HLA-DQ8 or HLA-DR4 transgenic NOD mice depleted for IAg⁷ were resistant to diabetes probably because T cells shift toward a tolerogenic regulatory profile (25). HLA transgenic mouse models have not been limited to the NOD background. HLA transgenes have been introduced in non-diabetes prone strains, mainly C57/BL6 mice. The expression of T1D susceptibility HLA class-II genes in these mice is not sufficient to induce diabetes. However, immunizing these transgenic mice with β-cells autoantigens allowed homing of T cells to the pancreas and the development of insulitis. The expression of the human costimulatory B7.1 molecule under the control of the rat insulin gene promoter as a transgene in these mice led to the development of diabetes. HLA-DQ8/RIP-B7.1 transgenic mice showed the highest incidence, whereas the genotype HLA-DR*04:01/HLA-DQ8 attenuated this effect and HLA-DQ6/RIP-B7.1 mice were protected from diabetes (24, 26). Table 1 shows the different HLA transgenic models that have been reported.

HLA transgenic mice allowed the identification of key players in T1D development. Adoptive transfer of T cells from HLA-DQ8 transgenic mice immunized with GAD65 and having evidence of insulitis, induced insulitis in recipients (32). Challenging HLA-DR4/RIP-B7.1 mice with murine proinsulin-2 peptides accelerated T1D development (46). The level of thymic insulin 2 gene expression determined the timing and the incidence of T1D in an HLA-B*39:06 transgenic mouse, a similar effect of that of the invalidation of the insulin 2 gene in NOD mice (47) or of the insulin variable number of tandem repeats (VNTR)in humans (22).

Besides pointing to the role of HLA in T1D development, HLA transgenic models have allowed identifying β-cell peptides recognized by T cells using either T cell hybridomas, or T cell assays or class-I or class-II peptide-MHC tetramers (29, 34, 38, 48). The main autoantigens that are recognized by T cells in T1D patients are insulin and its precursor preproinsulin (PPI) hereafter described as insulin, GAD65, ZnT8, IA2 and islet-glucose-6-phosphatase catalytic subunit-related protein (IGRP) (49, 50). Differences in β-cell peptides have been seen depending on whether a murine or a human autoantigen was expressed (34). Therefore, HLA transgenic mice that express human GAD65 or human PPI have been developed (41, 44).

HLA transgenic NOD models has been a unique model to advance our understanding of T1D. However, the genomic inflammatory responses in humans and mice do not overlap, possibly explaining the failure of translating therapies from mice to humans (60). Immunodeficient mice engrafted with human immune cells and tissues provide NOD mice with a humanized functional immune system to overcome this problem. Immunodeficient NOD (NSG) mice were obtained by deleting the IL2-receptor γ C gene, although not the SCID-Prkdc gene, from NOD-SCID mice (61). These mice are engrafted by a human immune system and/or by human islets (8). In NSG mice, the lack of B, T, and NK cells and the poor lymph node organization and development support the engraftment with human cells and tissues. The human immune system engraftment could originate from human peripheral blood monocytes or from human stem cells isolated from the umbilical cord, from fetal liver or mobilized to the periphery through G-CSF. It can also be obtained by transplanting the human fetal liver and autologous thymus fragments under the renal capsule while injecting the autologous human HSC intravenously (8, 12) ( Figure 1 ). This leads to murine models harboring a functional human immune system. The proper technological approach for engraftment and the proper mouse model are chosen depending on study objectives, i.e study of autoreactive or alloreactive T cells, or HLA-restricted epitopes, or induction of autoimmune diabetes. However, the scope of these models is limited by the murine component of many immune determinants: cytokines, murine major histocompatibility complex (H2), homing molecules, poorly developed lymph nodes and in case of diabetes, the cutoff level of a normal glycemia (62, 63).

To optimize these humanized models, mice can be manipulated to induce diabetes or to express human genes such as IL3, MG-CSF, SCF, thrombopoietin, SIRP alpha and HLA class-I or class-II depending on the study outcome (62, 64). Table 2 shows the hyperglycemic and the HLA transgenic NSG mouse models.

These models are valuable to decipher the pathophysiology of T1D. They serve to study T1D triggering factors. Fifty percent of NSG mice transplanted with human islets and infected with cocksackie virus developed hyperglycemia (78). Also, these models serve in studying human β-cells proliferation in vivo. NSG strains have been genetically modified to develop hyperglycemia either spontaneously or chemically (79). These hyperglycemic models could be transplanted with human islets or human stem cells derived from β-cells or progenitor cells to revert the hyperglycemia (66). Engraftment of hyperglycemic NRG Akita mice with human islet cells increased the β-cell proliferation by 6 folds as compared to normoglycemic NRG Akita mice (80).

Additionally, such models allow to identify key players in T1D and mechanisms behind β-cells destruction. Destruction of pancreatic islets and infiltrates with human CD4⁺ T cells was observed in humanized NSG mice after injection of irradiated monocytes from diabetic NOD mice (81). The adoptive transfer of T cells transduced to express human autoantigen-specific TCR allows to isolate a larger number of human diabetogenic T cells and dissect the role of islet autoantigens in T1D. A mouse model can be engrafted by human or murine stem cells transduced to express autoantigen-specific TCRs to create retrogenic humanized models. The retrogenic mouse model allows to study human autoreactive T cells phenotype and function and to study thymic selection (82). The retrogenic mouse can express multiple autoreactive or control TCRs to better mimic the physiological setting. In a TCR-transgenic humanized mouse model, thymocytes expressing TCRs specific to the HLA-DQ8 restricted peptide hPPI 33-47 (insulin B9-23) were negatively selected in an HLA-DQ8 positive human immune system more efficiently than on an HLA-DQ8 negative immune system (83). In another model, HLA-DR4 retrogenic mouse expressing monoclonal or polyclonal TCRs reactive to native or deamidated GAD-65_115-127 peptides showed that post-translational modifications epitopes do not support T reg development (76).

The diagnosis of autoimmunity in full-blown T1D is based on the detection of autoantibodies (1). However, in prediabetic individuals, while positivity for three or four different antigenic specificities is highly predictive, positivity for one or two autoantibodies has a low predictive value, highlighting the importance of developing new assays for early and accurate diagnosis (84). Providing the key role of T cells in driving the autoimmune process against β-cells, T cell assays need to be developed. T cell responses to epitopes recognized in the context of HLA class-I and class-II transgenic mice will be useful in helping to develop these assays for diagnosing and immune monitoring in patients under immunotherapy.

Identifying epitopes recognized by T cells in T1D pave the way to developing antigen-specific immunotherapies which are likely to carry a high benefit/risk ratio (85). Among recent examples, injecting NOD-β2m^null HHD mice with a nanoparticle-peptide complex (PSB coupled to HLA-A2 restricted ZnT8 or IGRP epitopes) induced immune tolerance and prevented diabetes by decreasing the numbers of autoreactive CD8⁺ T cells (86). Altered peptide ligand for insulin B1(5-14) induced antigen-specific anergy in a similar model (87). Vaccinating NSG-HLA-DQ8 transgenic mice with insulin mimotopes stimulated Foxp3⁺ Tregs in vivo (88).

These models can allow discovering and testing new-targeted therapies. The study of teplizumab in HLA-A2/NSG mice allowed the identification of CCR6⁺ Treg cells secreting IL-10 which could be considered as a therapeutic target (89). HLA-DQ8/hGAD65 transgenic mice have been used to test a targeted therapy using GC7 molecule which inhibits the eucaryotic translation initiation factor A-1 (eIF5A) activating enzyme. In this model, the onset of T1D was delayed and the function of β-cells improved (90).

In other preclinical models, humanized mice engrafted with human immune precursors have been used to evaluate the translational potential of promising therapies. Currently, pancreatic islet transplantation can restore normoglycemia in patients with long-onset T1D. However, it faces the shortage in human donors and the risk of graft rejection. Manipulation of the hematopoietic stem-cells or PBMC engrafted NSG strain has generated mice in which chemically or spontaneously induced diabetes was reversible by islet engraftment (61, 91). This allowed the identification of new potential therapeutic targets and the study of the mechanisms of islet graft rejection and the means to prevent this rejection (92). Combining human immune system and islet engraftment in these models allow the optimization of protocols for inducing remission in T1D through islet engraftment and suppression of graft rejection. Treatment with IL-2 and rapamycin suppressed effector T cells and stimulated regulatory (CD4⁺FOXP3⁺) T cells reducing human islet allograft rejection in NSG mice transfused with human spleen mononuclear cells (93). Combination therapy with ethylcarbodiimide, rituximab and rapamycin limited the rejection of xenogeneic porcine islets in humanized mice (94).

As another approach, costimulation blockade has been shown to prevent the rejection of allogeneic pancreatic endoderm by human PBMCs in a humanized model in vivo (95). Co-transplantation of human bone marrow-derived mesenchymal stem cells (hBMSCs) could prevent immune rejection and improve human islet transplantation in a humanized NSG mouse (96). Another NSG mouse model was created by transferring genetically modified human embryonic stem cells that lacked CIITA and expressed HLA-A2 as the only HLA class-I molecule. The differentiation of these cells into β-cells then the engraftment with human PBMCs allowed to study the immune response and the islet rejection (97). Genetically modified β-cells engraftment is another promising therapy to prevent T1D recurrence post engraftment; human β-cells engineered to express Herpesvirus encoded immune-evasion proteins prevent islet destruction in NSG mice by degrading MHC class-I molecules and inhibiting granzyme B activity (98). Beyond allograft rejection, NSG mice can be used to study xenogeneic GVHD reactions. An option has been developed that replaced human islets by genetically modified porcine islet. Engraftment of neonatal porcine islet-like cell clusters overexpressing CTLA-4 Ig analogue in diabetic Hu-HSC-NSG mice reverted diabetes without a xenogeneic GVHD reaction (99). Despite these advantages, the translation of treatment to humans is not straightforward. The dosing, frequency, and route of administration of immunotherapies are still to be refined.

T1D involves an auto-immune destruction of the β-cells. Therefore, a therapeutic approach aiming at modulating the immune response represents an attractive means of treatment approach. So far, therapies have met with varying clinical success despite efficiency in murine preclinical models. At best, the response to short-term treatments such as anti-CD3 antibodies had time limited effect. Pre-clinical models expressing the human targeted molecules might fill this gap and allow the optimization of therapeutic protocols. Immunomodulatory treatments have been attempted. Humanized murine models expressing human CD3ϵ and CD20 were developed to study the therapeutic potential of combined protocols in restoring tolerance in T1D (100). Treating VH125.hCD20/NOD mice with anti-human CD20 delayed diabetes development by reducing the effect of costimulatory molecules on B cells, by decreasing the INFγ production and by limiting T cell activation in the islets. Combining a histone deacetylase inhibitor with low-dose CD3 antibodies abrogated local inflammation, improved pancreatic β-cell survival and metabolic function, and led to long-lasting diabetes remission (101).

β-cell antigen-based therapy is another attractive approach, as it precludes the long-term side effects of immune modulating therapy as being antigen-specific. To study the capacity of dendritic cells to induce an antigen-specific immune tolerance, a humanized mouse model expressing human CD205 on a NOD background was produced. CD205, as the endocytic receptor of antibodies coupled to islet antigens on myeloid dendritic cells, allows antigen processing and presentation by MHC class-I and II and modulate antigen-specific T cell responses (102).