It is possible to derive T-cell lines and to produce T-cell clones specific to vari-ous β-cell autoantigens (insulin, glutamic acid decarboxylase, or GAD) in healthy individuals. Diabetes does not develop, though such β-cell-specific T cells are likely to have access to β cells. Similarly, double-RIP LCMV transgenic mice (in which large amounts of LCMV glycopro-tein are expressed in β cells and a large proportion of CD8 T cells expressed an LCMV glycoprotein-specific TCR) do not become diabetic unless they are infected with LCMV. This state of ignorance implies that development of diabetes requires that autoreactive T cells be activated to become pathogenic.
T cells differentiate in the thymus where they undergo two successive stages of selection: the first positive, giving rise to the emergence of the TCR repertoire, and the second negative, eliminating high-affinity self-reactive TCR-carrying T cells. In any event, negative selection is not absolute, because autoreactive T cells are found in the periphery. They can then be stimulated by the corresponding autoantigen when it is adequately presented to them in the context of MHC molecules. The hypothesis has recently been put forward that IDDM could result, at least in part, from defective intrathymic negative selection due to either underexpression of intrathymic β-cell autoantigens (perhaps under the control of the AIRE gene) or to partial expression of these autoantigens in the thymus due to alternative splicing. Also, NOD mouse thymocytes show a defect in thymocyte apoptosis, which could reduce the efficacy of negative selection.
There is strong evidence that the diabe-togenic autoimmune response is driven by β-cell autoantigens. (Diabetogenic T cells are rapidly exhausted in the absence of β cells.) The nature of the IDDM tar-get autoantigens remains elusive. Several candidates have been proposed, namely, insulin (or proinsulin), GAD, IA.2, a tyro-sine phosphatase, and IGRP. All four molecules have an exclusive or a prefer-ential β-cell distribution. They induce the production of T cells or autoantibodies in NOD mice or diabetic patients. Insulin was recently proposed as the primary autoantigen.
Strong experimental evidence in the NOD mouse indicates that regulatory T cells slow down disease progression. Diabetes onset is accelerated by thymectomy performed at weaning (three weeks of age) or by high doses of cyclophosphamide, which selec-tively destroys regulatory T cells. Disease transfer by diabetogenic T cells derived from overtly diabetic mice into predia-betic syngeneic recipients only operates when the recipient is immunoincompetent (neonate, NOD-SCID). The protective role of such regulatory T cells is demonstrated by the capacity of purified CD4+ T cells to inhibit diabetes transfer when they are co-injected with T cells from diabetic mice into NOD-SCID or irradiated NOD recipients. This model has allowed the phenotyping of the regulatory T cells, which are either CD4+, CD25+, or CD4+, CD25−, CD62L+ (CD62L is l-selectin). It has also allowed the demonstration of their dependency on TGF-β (but not on IL-4 or IL-10).
The question remains to determine whether regulatory T-cell function declines at the time of diabetes onset or whether it is overridden by a burst of effector cells. Another possibility is that effector cells resist regulation. There is experimental evidence in favor of each of these three hypotheses, which are not mutually exclusive.