Introduction

The past two decades has seen remarkable advances in understanding the genetics and pathophysiology of spontaneous animal models of immune mediated diabetes (Type 1A) including structural characterization of class II MHC presentation of insulin peptide B:9-23¹ as well as the chromagranin peptide WE-14 (previously elusive target of the BDC2.5 T cell receptor²), the creation of new animal models ³ with either genetic manipulation of autoantigens^{4, 5} or immunologic function (e.g., toll receptor activation; peptide immunization, mutations of regulatory pathways such as AIRE⁶ or foxP3⁷), or a combination of genetic and environmental manipulation ^8-18. “Humanized” mouse strains are being created and studied and promise additional insights relevant to type 1A diabetes ^19-21. In addition at the Jackson laboratory a repository of type 1 diabetes related mouse models has been created which is an important resource for the long-term preservation of the multiple strains that have been created and a means to facilitate sharing of strains by different research groups.

Spontaneous type 1 diabetes-susceptible models include the non-obese diabetic (NOD) mouse, the BioBreeding Diabetes-Prone (BB-DP) rat, the Komeda Diabetes-Prone (KDP) sub-line of the Long-Evans Tokushima Lean rat Lew.1.WR1 and the Lew.1AR1/Ztm rat. Multiple experimentally-induced models of type 1 diabetes are available including: 1) T cell receptor (TCR) transgenic (Tg) and retrogenic mice with the T cell receptors of naturally occurring diabetogenic clones 2) Neo-antigen (Ag) expression under the control of the rat insulin promoter (RIP) to establish neo-self antigen pancreatic expression that can be the target of autoimmunity, and 3) RIP-driven expression of costimulatory molecules on beta cells. Mice with knockouts of putative islet autoantigens have allowed direct testing of the pathogenic significance of specific target molecules. Strains of mice with mutations of genes associated with type 1 diabetes in man (FoxP3 and AIRE) are being studied (including an autosomal dominant “human” AIRE mutation⁶). Such strains usually do not develop diabetes, but rather have novel autoimmune phenotypes ^22-24.

A major conclusion from these models is that type 1 diabetes is not the result of a single pathway but can be the result of numerous distinct mechanisms ²⁵. Nonetheless, some generalities do exist including the polygenic nature of the disease, the involvement of T cells in the destruction of pancreatic beta cells, and the incomplete penetrance of the disease implying that environmental or “stochastic” factors influence disease susceptibility. These models provide useful tools for studying the autoimmune process as well as testing potential treatments for the prevention of type 1 diabetes. There has been some pessimism²⁶, appropriately expressed ¹⁰, that therapies effective in animal models such as the NOD mouse may not be directly relevant to man. No doubt the usual animal models are inbred (thus not diallelic at any locus) and finely “balanced” in terms of progression to diabetes while raised in pathogen-free environments. Thus, it is likely that many more therapies will be effective in the animal models compared to man (Entelos has developed a computer model of NOD mice and catalogued therapeutics²⁷). Nevertheless, the initial experimental success of anti-CD3 monoclonal antibody therapy in man to delay loss of insulin secretion is a direct result of studies in the NOD mouse ²⁸, and large trials such as the Diabetes Prevention Trial (DPT) have provided preliminary data that oral insulin (also first studied in the NOD mouse) in subjects with elevated insulin autoantibodies may delay disease progression to diabetes²⁹. We suspect that informed optimism relative to the utility of animal models is in order²⁶ , and that the more robust the therapy in the animal model (e.g., ability to abrogate insulitis, ability to prevent diabetes in models engineered to be more highly penetrant [e.g., insulin 2 gene knockout NOD mice ³⁰) , reversal of hyperglycemia, and the closer the human studies mimic the animal studies (e.g., dose and timing), the more relevant the study will be to human type 1 diabetes.

The Nonobese Diabetic Mouse

There are now more than 6,000 articles listed in PubMed on the NOD mouse making any comprehensive review of the model a difficult task with the certainty of failure to cite all important contributions and a certainty that not all the different pathways being pursued will be covered. Given the ability to electronically access primary articles through the internet with services such as PubMed and Google my purpose in this review on such a large topic is to provide an overall organization, a particular viewpoint (with caveats), and to attempt to highlight many important “relevant” observations. The general hypothesis my laboratory group is pursuing is that NOD mice 1. have relatively mild defects of immune tolerance (particularly in comparison to foxP3 or AIRE mutant mice) and thus the polygenic nature of the disease and the small influence of most loci; 2. that the disease results because of an immune response that once directed at a specific beta cell peptide is essentially normal, though the targeting leads to disease and there is spreading of autoimmunity, with implication that all mechanisms available to the normal immune system for destruction are operative; 3. that there is a primary autoantigenic epitope presented by the class II I-A^g7 molecule and for the NOD mouse it is the insulin B:9-23 peptide presented in “register 3” of I-A^g7 ³¹ and recognized primarily by a conserved alpha chain (TRAV5D-4) of the T cell receptor³² and 4. The killing of islet beta cells is asynchronous with different islets destroyed over time until enough destruction has occurred for hyperglycemia to develop. These “biases” are stated, as to some extent they influence “commissions” and “omissions” of this review that many investigators will hopefully correct in the future.

Figure 1. Derivation of the spontaneous animal model of type 1A diabetes, the NOD mouse.

The NOD mouse is the most-studied animal model of type 1A diabetes^33-37^;^{26, 27, 38-51}. The NOD inbred mouse line was established at the Shionogi Research Laboratories in Japan through sib breeding of a hyperglycemic female mouse from the CTS sub line ⁵² (Figure 1). The line that actually became the NOD mouse strain was originally a control line for the modestly hyperglycemic line that became the NON strain (Non-obese Non-diabetic). NOD mice show islet infiltration by lymphocytes around 5-7 weeks of age followed by the spontaneous development of overt diabetes in approximately 70% of the females and 40% of the males by 30 weeks of age ^{53, 54}. Similar to humans, NOD mice usually (but not always) express anti-insulin autoantibodies in their serum prior to hyperglycemia ^{55, 56};⁵⁷ (Figure 2). Of note NOR mice which do not usually develop diabetes follow the same age dependent pattern of expression of insulin autoantibodies. NOR mice are a complex congenic inbred strain with approximately 80% NOD genome and following depletion of CD4+CD25+ T cells NOR splenocytes can mediate diabetes in NOD-scid mice⁵⁸.

Figure 2. Insulin autoantibodies and glucose of individual NOD mice followed from 4 weeks of age.

Although all NOD mice develop insulitis, this is not always followed by diabetes. Diabetes incidence varies among established NOD colonies, although in all cases, unlike in humans, gender influences disease incidence with higher disease frequencies in female mice than males ⁵⁴. Type 1 diabetes incidence is also affected by environmental conditions with higher disease frequencies in sterile conditions ⁵⁹ and lower disease penetrance with infectious agents ^{60, 61}. The disease can be prevented in numerous ways (>134) ²⁶ including immunological or genetic manipulations of NOD mice ⁶².

Diabetes susceptibility in the NOD mouse results from the interaction of multiple genes⁶³, with the strongest predisposing effect deriving from genes within the major histocompatibility complex (MHC) ^64-66. Multiple studies suggest that the effect of the MHC ⁶⁴ is due to the combination of the unique sequence of the class II I-A^g7 allele with its beta chain lacking aspartic acid at position 57 and proline at position 56 ⁶⁷ (the I-Aalpha chain sequence of NOD is identical to BALB/c), a lack of expression of I-E ⁶⁸ due to a common mouse mutation (also present in C57BL/6 mice) and specific class I alleles ⁶⁹, with additional polymorphisms of other genes within the MHC(e.g. idd16^{70, 71}. In general it has been assumed that the lack of aspartic acid at position 57 of the I-A^g7 beta chain creating a basic I-A^g7 pocket 9 would favor the binding of amino acids with negatively charged side chains into pocket 9. The recognition of these peptides by the CD4 T cells then creates class II MHC mediated diabetes susceptibility. Kappler and coworkers have recently discovered that for the insulin peptide B:9-23, just the opposite occurs. A given peptide can bind to I-A^g7 in different registers with side chains of different amino acids of the peptide binding in pockets 1, 4, 6 and 9 depending on where the peptide docks along the linear I-A^g7groove (register). T cell receptors then recognize the peptide MHC complex in a specific register. For each specific register that a single peptide binds in, the amino acid chains facing the T cell receptor are different. To go from one register to another, the peptide must rotate for specific side chains to bind in pockets 1,4,6 and 9. By covalently coupling the B:9-23 peptide in the groove of I-A^g7 with the arginine (peptide amino acid 22) in pocket 9 (unfavored basic amino acid in a basic pocket) it was found that all of the studied anti-B:9-23 autoreactive T cells reacted only when the peptide was in this low affinity register¹. This has led to the hypothesis that such unusual recognition may allow anti-B:9-23 T cells to avoid thymic deletion in that the concentrations of insulin in the thymus are very low. T cells should be able to recognize the peptide in a low affinity register in the islets where the concentrations of insulin and presumably the B:9-23 peptide presented by I-A^g7 are huge. Of note Unanue and coworkers have provided data that the B:9-23 peptide is likely created within islets, potentially within insulin secretory granules with subsequent uptake by antigen presenting cells⁷².

A great deal of effort has been involved in mapping the diabetes-associated alleles in both NOD mice and humans. In both cases, numerous loci have been identified (>50), although to date, few predisposing genes are identified for the NOD mouse (MHC genes ⁶⁴, potentially CTLA-4 ⁷³, IL2^{63, 74} and beta-2 microglobulin important exceptions ^{66, 75, 76}). Genetic analysis has involved mapping of the mouse genome usually with crossing of loci from nonautoimmune-prone strains onto the NOD background as well as analysis of recombinant and related strains. Loci (identified as Iddn in the mouse and IDDMn in the human) with significant LOD scores for association have been analyzed for candidate genes. Multiple congenic strains containing Idd locus (n) from disease-resistant strains have been generated and analyzed for disease incidence, insulitis, and immunological phenotypes (e.g., insulin autoantibodies ⁷⁷) associated with diabetes in the NOD mouse ^{76,
78, 79}. A general finding is that several identified regions contain multiple-disease associated loci which alone rarely confer protection but in multitudes of two or more can nearly completely block disease ⁷⁶. Positional cloning of numerous Idd loci is underway with potential for success of identifying “causative” polymorphisms likely to be dependent upon the number of genes within a given locus. For example a 1.52Mb region of locus Idd5.2 contains 45 genes, while the locus Idd5.1 (2.1-Mb region) contains four genes, including CTLA-4 ⁶³, with evidence for influence of CTLA-4 polymorphisms influencing disease in man and mouse.

A list of many of the multiple identified Idd loci, their positions on mouse chromosomes, identified phenotypes, and candidate genes is presented in Table 3.1 and a more comprehensive list is provided by Serreze and coworkers ³⁷. These loci can either confer susceptibility or resistance to disease development and even some alleles of the NOD mouse confer resistance rather than susceptibility. Given the identification of these loci, it is possible to rapidly introduce loci from other strains onto the NOD background to create what has been termed speed congenics. Usually within five generations of backcrossing, essentially all NOD diabetogenic loci can be fixed with the locus of interest introduced ⁸⁰. Though there are reports of some success, a major disadvantage of the NOD mouse is the lack of embryonic stem cell lines to directly create gene knockouts on the NOD background, a lack that is partially alleviated with the availability of “mixed” NOD strain embryonic cell lines ⁸¹.

Table 3.1 Loci of the NOD mouse (idd)

Locus	Chromosome	NOD Allele	Phenotype protective allele	Disease Protection	Genes/ (Candidates)	Reference:
idd1	17	Susceptible	No Diabetes But some insulitis	100%	MHC class I and II I-A^g7	Hattori 1986 Todd 1988 Ikegami 2004 Serreze
idd2	9	Susceptible				Wicker 1995 McAleer 1995
idd3	3	Susceptible	Moderate IAA, Insulitis	69%	(IL2, IL21,Fgf2,Cetn4)	Podolin 1997 Hill 2000 Lyons 2000 Ikegami 2003 Rabinow 2008
idd4	11	Susceptible	age of onset			Wicker 1995 Devidi-.. 2007
idd5	1	Susceptible	Low IAA, Insulitis	45%		Colucci 1997 Hill 2000
idd5.1	1 (2.1 Mb)	Susceptible	CTLA-4 Ligand independent splice	26%	(CTLA-4, Icos, Als2cr19, and Nrp2 (neuropilin))	Hill 2000 Wicker 2004
idd5.2	1 (1.52 Mb)	Susceptible		0%	(Nramp1, 45 gene interval)	Hill 2000 Wicker 2004
idd6	6	SusceptiblexB10 ResistantxNON				Wicker 1995 McAleer 1995
idd7	7	Resistant	Low TCR, CD8			Gonzlez 1997 McAleer 1995
idd8	14	Resistant				Wicker 1995
idd9	4	Susceptible	High insulitis and IAA	90%	(Vav3, Cd30,Tnfr2,Cd137)	Lyons 2000
idd9.1	4	Susceptible			(Jak1, Lck)	Lyons 2000
idd9.2	4	Susceptible			(Tnfr2)	Lyons 2000 Siegmund 2000
idd9.3	4	Susceptible		20%		Lyons 2000
idd10	3	Susceptible		36%	Fcgr1 ruled out, CD101	Podolin 1998 Ikegami 2003
idd11	4	Susceptible	Marginal zone B cells	62%		Brodnicki 2000
idd12	14	Susceptible		74%		Wicker 1995
idd13	2	Susceptible	Decrease insulitis	100%	β-2 microglobulin	Serreze 1998
idd13a	2	Susceptible		38%		Serreze 1998
idd13b	2	Susceptible		25%		Serreze 1998
idd14	13	Susceptible				McAleer 1995
idd15	5	Susceptible			(Xmv65)	McAleer 1995
idd16	17	Susceptible		52%	H-2^k	Ikegami 1995
idd17	3	Susceptible				Podolin 1997
idd18	3	Susceptible		9%		Podolin 1997
Combined Congenics
10/18	3	Susceptible	Low IAA, Insulitis	38%	(Cfsm, Cd53, Kcna3, Rap 1a)	Podolin 1998 Robles 2003
3/10/18