Chapter 4
The Role of T-Cells in Beta Cell Damage in NOD Mice and Humans

(Updated 10/07/2011)

Tomasz Sosinowski, Edwin Liu, George Eisenbarth, and Howard W. Davidson

Type 1A diabetes (T1D) is a chronic disorder that results from the immune-mediated destruction of the insulin-producing ß-cells of the pancreatic islets¹. In its initial phase, which is clinically silent, T lymphocytes and other inflammatory cells invade the islets and eventually destroy them. The disease then becomes clinically evident with the pathological consequences (hyperglycemia, ketosis and long-term complications) resulting from the inability to maintain glucose and lipid homeostasis.

Type 1 diabetes is a T-cell mediated disease

The first indication that T1D is an autoimmune disease came from the results of a comprehensive histological examination of pancreata from diabetic patients who had died shortly after diagnosis. This showed that most of the subjects had significant lymphocytic infiltration of their islets concordant with loss of ß-cell mass ². Subsequently, islet-cell antibodies (ICAs) and anti-pancreatic cell-mediated immunity were detected in recently diagnosed T1D patients ^3-5, suggesting that the lymphocytes accumulated as a result of attraction by antigens derived from pancreatic ß-cells⁶. Consistent with this hypothesis, insulitis was only seen in islets containing ß-cells. With the advent of monoclonal antibodies capable of identifying distinct lymphocyte sub-populations more detailed immunohistochemical examinations of islet infiltrates became possible. One of the earliest of such studies showed a predominance of CD8⁺ T-cells in the islets of a deceased 12-year old girl with newly diagnosed T1D, which, together with the observed up-regulation of MHC class I molecules by islet cells, implicated cytotoxic T-cells (CTLs) in ß-cell destruction ⁷. Additional studies of pancreas from patients with type 1 diabetes have confirmed preponderance of CD8 T-cells and the presence of B-lymphocytes related to extent of ß-cell destruction ⁸. The JDRF nPOD (Network for Pancreatic Organ Donors with Diabetes) program now allows viewing of pancreatic histology of cadaveric donors directly online (http://www.jdrfnpod.org/). Similarly, the strong association of T1D with particular MHC II haplotypes (see chapter 7) suggested a critical role for CD4⁺ T-cells in the disease process (reviewed by ⁹. Additional circumstantial evidence supporting a crucial role for T-cells and MHC-restricted self antigen recognition in diabetogenesis came from the reversal and recurrence of diabetes following twin to twin pancreatic isografts ^{10, 11}, and the inadvertent transfer of disease between HLA-identical siblings by bone marrow transplantation ¹².

The mere presence of T-cells in infiltrates, though highly suggestive, does not by itself establish a direct role for these cells in the development of T1D. However, the histological findings were subsequently followed by reports of T-cell reactivity to ß-cell proteins, providing further support for the hypothesis. Thus, Roep and colleagues established CD4⁺ T-cell lines and clones restricted to HLA-DR from the peripheral blood of new-onset diabetics after stimulation in vitro with rat insulinoma cells ¹³. Of the eight clones examined, five appeared to recognize insulinoma membrane components, one of which was a 38kD protein later termed IMOGEN38 ^14-18. Surprisingly, after expression cloning IMOGEN38 was shown to be a broadly distributed mitochondrial protein (a probable subunit of the mitochondrial ribosome), and studies with the human ortholog suggested that the response was likely xenogeneic (JC Hutton personal communication). Nonetheless, the identification of several islet cell molecular targets allowed subsequent studies to be conducted using defined autoantigens, rather than crude fractions, and have suggested that the peripheral blood of diabetic subjects and their at-risk relatives contain elevated numbers of T-cells able to recognize epitopes from ß-cell proteins ¹⁹. Such studies suggest that T-cells provide a legitimate therapeutic target for intervention, and results from clinical trials of new-onset diabetic patients with humanized anti-CD3 monoclonal antibodies delayed the deterioration of circulating C-peptide levels normally seen in the year following diagnosis in 9 of 12 subjects ^20-22. Attempts to build upon these partial successes and improve the therapeutic regimen are currently in progress.

Animal models of T1D

A. Spontaneous models

Since the target organs of T1D (islets and draining pancreatic lymph nodes) are inaccessible in human subjects, the study of T1D has been greatly facilitated by the availability of animal models such as the Biobreeding-diabetes prone (BB-DP) rat ²³ and the nonobese diabetic (NOD) mouse ²⁴, which spontaneously develop diseases that mimic many features of human T1D. In particular, the NOD mouse has been the subject of extensive studies for over 20 years ^25-27, and has provided key experimental evidence supporting several strategies to treat the human disease, some of which are showing promise in initial clinical trials ^{28, 29}. Numerous abnormalities have been reported in the immune systems of NOD mice, including defects in antigen presenting cells (APCs) ³⁰ and hyporesponsiveness of T lymphocytes ^{31, 32}, which together may compromise both central ³³ and peripheral ³⁴ tolerance to pancreatic ß-cells. Similar abnormalities have been reported in human subjects (for example ^35-37, and it has been proposed that T-cell hyporesponsiveness may be a general feature conferring susceptibility to inflammatory autoimmune disorders ³⁸. However, it must be noted that the immune systems of humans and mice show several key differences (reviewed by ³⁹, which must be kept in mind when extrapolating between these species ^{28, 40-42}. Indeed, there are notable differences between T1D in NOD mice and humans, not the least being that NOD mice only develop disease if kept in specific pathogen free conditions. Moreover, in these animals disease is associated with pronounced cellular infiltrates that surround the individual islets (Figure 1) that begin to form at least 7 weeks prior the onset of overt

disease ^{43, 44}. In contrast, pancreata obtained at post mortem from diabetic subjects who died shortly after clinical manifestation of T1D typically show a much less florid infiltration (e.g. ^{7, 45}). In that NOD mice are inbred they can only be considered to be the equivalent of a single genotype "case study" of T1D and a genotype that is homozygous at all loci.

Although the precise sequence of events that lead to T1D in NOD mice remain uncertain, the central role of T-cells in diabetogenesis in these animals is incontrovertible. Thus, treatment of newly diabetic animals with anti-CD3 antibodies, which suppress immune responses by transient T-cell depletion and modulation of T-cell Receptor (TCR) signaling, induces long-term remission ^{46, 47}. Moreover, diabetes can be transferred to immuno-compromised hosts by mixed T-cell populations, or in some instances, individual T-cell clones ^48-51.  For example, Haskins and colleagues isolated 8 CD4⁺ T-cell clones recognizing islet antigens (including chromagranin and islet amyloid polypeptide) ^{52, 53} presented by the NOD MHC class II molecule, I-A^g7, at least 5 of which are capable of inducing disease after adoptive transfer ^{54, 55}. Transgenic mice expressing the T-cell receptor (TCR) of one of the diabetogenic clones, BDC2.5 (target peptide derived from chromogranin), have been generated and bred onto the NOD background ⁵⁶. Interestingly, evaluation of the lymphoid compartment in NOD/BDC2.5 animals showed no sign of negative selection in the thymus, with normal peripheral T-cell reactivity. This was also true of transgenic C57BL/6-H-2^g7 (B6^g7/BDC2.5) animals. However, autoimmune pathogenesis was highly dependent upon the genetic background of the animal ⁵⁷. In both NOD and B6^g7 transgenics there was no manifestation of disease in the first 2 weeks of life, with pancreatic islets completely free of infiltration. Insulitis appeared abruptly at 18 days, and subsequently progressed to eventually involve essentially all islets. However, whilst the B6^g7/BDC2.5 animals suffered a highly aggressive insulitis and the majority rapidly progressed to overt disease, paradoxically, NOD/BDC2.5 animals were significantly protected from spontaneous disease, and exhibited a more benign "respectful" insulitis. Nevertheless, young NOD/BDC2.5 are significantly more sensitive to cyclophosphamide induced diabetes than their non-transgenic relatives ⁵⁸, and NOD/scid/BDC2.5 mice rapidly progress to overt T1D and typically die of diabetic complications on or before 33 days of age ⁵⁹. The precise mechanisms by which the un-manipulated NOD/BDC2.5 animals restrain their insulitis are currently uncertain, but may involve populations of T-cells expressing alternative a T cell receptor chains due to incomplete allelic exclusion at this locus ^{56, 60}.

In addition to TCR transgenic animals, targeted gene disruption and retrogenic ⁶¹ technologies have also been used to study other features of T1D in NOD mice. For example, mice lacking expression of ß₂-microglobulin (ß₂M), that consequently do not express functional MHC class I molecules, do not develop insulitis ^62-65, implicating CD8⁺ T-cells in the initiation of disease. Interestingly, transgenic restoration of MHC class I expression in NOD-ß₂M^-/- mice using different ß₂M alleles provided contrasting results. Thus, animals reconstituted with the endogenous ß₂M^a allele developed T1D, whilst those given the ß₂M^b allele (which only differs at a single amino acid residue) did not ⁶⁶. At present the precise mechanism of protection is uncertain, although it may reflect subtle differences in the peptide-binding properties of the resultant MHC class I molecules ⁶⁷.

Targeted gene disruption has also provided evidence for a key role for proinsulin in diabetogenesis in NOD mice. In contrast to humans, mice have 2 non-allelic insulin genes (insulin 1 on chromosome 19, and insulin 2 on chromosome 7), and recently NOD mice with disruptions in either one have been created ^{68, 69}.  Surprisingly, the mice show contrasting phenotypes; insulin 1^-/- mice are markedly protected from diabetes (but do develop anti-insulin autoantibodies), while insulin 2^-/- animals show accelerated disease. In contrast, animals whose sole preproinsulin is a mutant form of insulin 2 in which the immunodominant B:9-23 epitope is disrupted are completely protected from diabetes ⁷⁰, although they do develop sialitis confirming the organ-specificity of the effect. Such results are at variance with those from related studies in which NOD mice lacking the autoantigens GAD65, IA-2, phogrin (IA-2ß) and IGRP ⁷¹ were produced and where disease occurrence was not significantly altered ^72-74. This suggests a central role for (pro) insulin in the disease process, although the precise mechanisms of acceleration, or protection from disease, remain to be determined. The potential central role of insulin and insulin peptide B:9-23 is supported by studies where both insulin genes are eliminated and an insulin mutated at position B16 (Y to A). These mice do not develop diabetes ^{70, 75}. Krishnamurthy and coworkers have found that eliminating response to proinsulin eliminates the prominent CD8 T-cell response to IGRP ^{76, 77}. Vignali and coworkers have introduced a new methodology to study T-cell receptor targeting of islet ß cells, namely the creation of retrogenic mice ^{78, 79}. The creation of retrogenic mice utilizes retroviruses to introduce T-cell receptors into bone marrow cells that are then transplanted into immunodeficient mice. This technology greatly accelerates studies of such T-cell receptors compared to creating transgenic mice. Within 8 weeks, the pathogenicity of T-cell receptors can be assessed. Studying a series of 17 retrogenes, those targeting GAD failed to induce diabetes. Insulin peptide B:9-23 reactive TCR caused delayed diabetes and IA-2/phogrin TCR caused insulitis and TCRs targeting chromogranin (e.g. BDC 2.5) caused diabetes as did TCR’s targeting unknown molecules. (Diabetes induction TCR: BDC-10.1 > BDC-2.5 > NY 4.1 > BDC 6.9). Of note, a series of TCRs targeting GAD65 caused fatal encephalitis independent of the induction of anti-GAD autoantibodies⁶¹. No insulitis was observed. Presumably this was related to the lack of GAD in mouse islets.

B. Experimental autoimmune diabetes (EAD)

As only a limited number of animal models exhibiting spontaneous disease are currently available, systems to induce EAD in non-autoimmune prone mice have also been developed. Typically these are based upon the transgenic ß-cell expression of heterologous proteins under control of the rat insulin promotor (RIP). Such models have provided key insights into the establishment and breaking of tolerance to ß-cell antigens. Initial studies showed that some lines of RIP-Tag C57BL/6 mice, which express the SV40 large T antigen (Tag) in their ß-cells, were intolerant of the transgene, developing spontaneous autoimmunity ⁸⁰. Tolerance, or autoimmunity, correlated with the presence or absence of embryonic expression of the transgene, with animals that did not express Tag until adulthood developing disease. Moreover, tolerant animals also expressed the transgene in the thymus ⁸¹, suggesting that, at least for some proteins, central tolerance can be established to apparently organ-specific antigens (reviewed in ⁸². However, central tolerance could be broken if the precursor frequency of peripheral autoreactive T-cells was too high. Thus, tolerant RIP-Tag mice crossed with transgenic mice showing low expression (~10% of peripheral T-cells) of the TCR of a Tag specific CTL were protected from spontaneous disease ⁸³. In contrast, offspring of parents showing high expression (~90%) of the transgenic TCR were intolerant. The importance of central tolerance was also demonstrated in a model where CBA (H-2^k) mice expressing the K^b MHC class I molecule under control of the RIP were crossed with transgenic mice expressing the TCR of an H-2^k-restricted CTL clone recognizing K^b. In this case the progeny rejected skin grafts from H-2^b animals, but were protected from T1D. However, crossing the double transgenic mice with RIP-IL-2 animals caused rapid onset disease ⁸⁴. Neonatal replacement of the thymuses of the double transgenic animals with tissue from non-transgenic mice permitted high avidity K^b-specific T-cells to reach the periphery, which allowed disease to be triggered in animals primed either with allogeneic skin grafts or the injection of irradiated splenocytes from H-2^b donors ⁸⁵. Some animals required multiple primings, suggesting that the duration of the stimulus, as well as the avidity of the peripheral autoreactive T-cells, profoundly influences the course of disease.

The lack of spontaneous disease in RIP-K^b x K^b-TCR mice after thymus replacement, despite the presence of peripheral high affinity K^b-specific T-cells, indicates that protective mechanisms exist in the periphery which can act to negate a lack of central tolerance. Two potential mechanisms have been proposed, namely immune ignorance and active tolerance ⁸⁶, and evidence that both are involved in protection from T1D has come from RIP-based transgenic models. For example, two strains of mice secreting either high (RIP-Ova^hi) or low (RIP-Ova^lo) levels of ovalbumin from their ß-cells were exposed to 5, 6-carboxy-succinimidyl-fluorescein-ester (CSFE) labeled OT-1 T-cells specific for ovalbumin. Analysis of the pancreatic draining lymph nodes 3 days after transfer revealed that the OT-1 cells had proliferated in the RIP-Ova^hi but not the RIP-Ova^lo mice ⁸⁷, indicating that the level of antigen acquired by antigen presenting cells (APCs) in the RIP-Ova^lo animals was insufficient to trigger naive T-cells, and consequently that the immune system was ignorant of the presence of the neo-antigen. Both strains developed T1D if exposed to pre-activated OT-1 CTLs, confirming that the RIP-Ova^lo animals did indeed exhibit ß-cell expression of the transgene. However, despite proliferation in the draining lymph nodes, adoptive transfer of even 10⁷ naive OT-1 cells to RIP-Ova^hi mice did not induce disease ⁸⁸, suggesting that effector CTL generation was inefficient under these conditions. Nonetheless, tolerance to ß-cell expression of ovalbumin could be broken if the efficiency of presentation was enhanced. Thus, studies with RIP-mOva mice, which express a membrane-anchored chimeric fusion protein derived from ovalbumin and the transferrin receptor ⁸⁹, demonstrated that cell-associated ovalbumin was cross-presented to CD8⁺ T-cells ~50,000-fold more efficiently than the secreted form ⁹⁰and adoptive transfer of 5 x 10⁶ naive OT-1 T-cells into non-irradiated RIP-mOva mice rapidly induced disease.

To analyze the fate of OT-1 cells following transfer into RIP-mOva mice without inducing disease these animals were crossed with bm1 mice (that have a mutation in the K^b molecule which prevents presentation to OT-1 T-cells ⁹¹. The resulting RIP-mOva.bm1 mice were lethally irradiated and reconstituted with bone marrow from wild-type C57BL/6 animals so that OT-1 cells could interact with hematopoietic, but not peripheral, cells. Analysis of draining lymph nodes 3 days after adoptive transfer of CSFE-labeled OT-1 T-cells confirmed that they had proliferated, but after 8 weeks transgenic T-cells comprised only ~1% of CD8⁺ T-cells recovered from the spleens and lymph nodes of B6®RIP-mOva.bm1 mice as compared to ~7.4% in B6®bm1 animals, suggesting that activation had led to peripheral deletion