Type 1 Diabetes: Cellular, Molecular & Clinical Immunology
Theoretical Essay B - Pathogenesis of Type I Diabetes
George S. Eisenbarth
Updated 3/05 Click to download Powerpoint slide set
Theoretical essays provide an opportunity for "going"
out on a limb, but hopefully not so far on the "limb" as to hang oneself.
With recent information I will go further than I ventured in the first edition
of this essay.
There is now compelling evidence in rats with the RT1-U MHC haplotype, and mice
with the H-2d MHC haplotype and individuals with DR3/4, DQ2/8 haplotypes that
developing diabetes is "EASY" (1-3).
Given the right ("wrong") major histocompatibility genotype, induction
of insulin autoantibodies, insulitis and even diabetes is a relatively common
phenomenon. At the time of the first edition it was known that for BB rats and
the related strain of DR (Diabetes Resistant BB rats that lack the lymphopenia
gene) diabetes was dependent upon RT-U. Of interest was the infection of the
DR strain with Kilham rat virus induced diabetes. Stimulation of immunity with
poly-IC (poly-inosinic, cytidylic acid: a mimic of viral RNA that activates
Toll receptors and induces interferon-alpha) was also sufficient to induce diabetes.
It now appears that Kilham rat viral infection is similar to poly-IC induction
and the virus does not infect islets. Many strains of "normal" rats
that share RT1-U develop either insulitis or diabetes following poly-IC administration
(1).
In a similar manner, mice homozygous for H-2d (the MHC of the Balb/c mouse)
with islet B7-1 expression (RIP-B7-1: Rat Insulin Promoter Transgene inducing
beta cell expression of the T stimulatory molecule B7-1), following the administration
of poly-IC develop diabetes (4). In these mice
there is a predominant CD8 T cell islet infiltrate. Even in the absence of B7
in islets, administration of the insulin peptide B:9-23 to normal Balb/c mice
induces insulin autoantibodies (3), and if combined
with poly-IC, insulitis. The insulin autoantibodies are not absorbed by the
peptide but react with intact insulin. Thus normal Balb/c mice have autoreactive
B-lymphocytes reacting with intact insulin, and more important, administration
of a specific peptide of insulin (other insulin peptides do not induce insulin
autoantibodies) induces autoantibodies and insulitis. The "simple"
combination of human DQ8 and the islet B7-1 transgene also leads to diabetes
in mice (5).
These observations lead to a series of hypotheses whose central component is
autoimmunity directed against insulin with the concept that anti-islet autoimmunity
(presumably anti-insulin) is "hard-wired" into the immune system.
The "hard-wiring" is predominantly dependent upon MHC alleles (This
is reminiscent of Irun Cohen’s immunologic "homunculus"). One
could substitute other islet antigens and other peptides for insulin for most
of the discussion that follows, but as has been reviewed we favor insulin as
a primary autoantigen (6).
A large number of observations concerning the pathogenesis of type 1 diabetes
that should guide pathogenic hypotheses are now available (7;8).
Beginning at the end of the process that leads to type 1A diabetes, it is clear
that patients with type 1 diabetes and animals such as the nonobese diabetic
(NOD) mouse produce autoantibodies and T cells reactive with a relatively large
series of autoantigens including insulin, glutamic acid decarboxylase (GAD)
and ICA512 (IA-2), etc. What is more striking is that T cell clones derived
from NOD mice recognize multiple proteins and that different CD4 clones recognizing
different antigens are able to transfer diabetes. This statement can be derived
by combining studies of Haskins and coworkers on the initial BDC diabetogenic
clones (e.g. BDC 2.5), Wegmann’s studies of insulin and in particular insulin
peptide B:9-23 reactive clones, as well as the clones of Wong, Serreze, and
Santamaria. Both CD4 and CD8 clones reacting with different peptides and molecules
can produce beta cell destruction and diabetes. For the NOD mouse the B cell
repertoire at least as represented by high affinity autoantibodies is more limited
and at the recent Snow Mountain Ranch workshop, only specific insulin autoantibodies
were demonstrated (9). Given so many different
specificities for T cells in the NOD and for autoantibodies in man, the simplest
hypothesis is that many clones arise subsequent to beta cell destruction and
are "antigen" driven. It is also likely that at least a subset of
clones are not secondary to islet destruction and may be the rate-limiting factor
for development of type 1 diabetes. The clones isolated to date follow the classic
rules of T cell recognition (e.g. trimolecular complex of major histocompatibility
molecule, peptide and T cell receptor for recognition). There may however be
some unique aspects to recognition of "pathogenic" self peptides.
For example we have found a marked restriction of T cell receptor alpha chain
usage for clones recognizing insulin peptide B:9-23 with no junctional conservation
(10). Overall, diabetogenic clones recognize peptides
presented by classic MHC molecules, and can be deleted within the thymus by
their cognate peptide.
Type 1A diabetes develops only in the setting of genetic susceptibility, but
the genetic determinants vary. The major determinants are alleles of the major
histocompatibility complex and in particular class II HLA alleles. It is however
very likely that class I alleles will (as for a usual immune response) present
peptides to cytotoxic T lymphocytes that will participate in islet destruction.
There is some evidence that HLA-A2 and A24 enhances diabetes risk. In addition
to HLA alleles increasing risk, a series of HLA alleles provide dominant protection.
This is well illustrated by the NOD mouse, which have a "natural"
mutation such that I-E is not expressed. Transgenic introduction of I-E prevents
diabetes. For man DQA1*0102,DQB1*0602; DQA1*0201, DQB1*0303, and DRB1*1401 all
appear to provide dominant protection (11). Approximately
20% of patients from the general population have DQB1*0602, while less than
1% of children developing type 1 diabetes have this allele. DQB1*0602 does not
prevent all autoimmunity and for instance is part of the high-risk haplotype
for multiple sclerosis. Thus it is unlikely that DQB1*0602 has a global effect
on autoimmunity, but rather its influence is likely to be specific for anti-islet
T cell clones. Thus MHC alleles both protect and "cause" type 1 diabetes.
One can speculate that these alleles, by the islet peptides that they bind and
the type of immune response they engender (e.g., Th1 versus Th2, T regulatory
cells, etc) either increase or decrease the risk of type 1 diabetes.
In man, in addition to MHC alleles, polymorphisms of regulatory sequences and
in particular a 3’ variable nucleotide tandem repeat, of the insulin gene
influence disease risk. Insulin is the only known beta cell specific islet autoantigen
(e.g. GAD is expressed in non-beta cells of human islets). The insulin VNTR
polymorphism associated with protection from diabetes is also associated with
enhanced insulin messenger RNA production within the thymus. A significant number
of cells both within the thymus and other lymphoid organs produce proinsulin
(12). Given this observation and the functional
data from Hanahan’s transgenics concerning Peripheral Antigen Expressing
Cells (13), it is hypothesized that insulin within
the thymus alters diabetes risk.
In animal models and man there are multiple genes outside of the MHC and insulin
locus that influence disease. Several "autoimmune genes" have been
identified, including mutations in AIRE (Autoimmune Regulator for the APS-I
syndrome see Chapter 8) and the Scurfy gene
of the XPID syndrome. Both of these syndromes are very rare with multiple autoimmune
manifestations. A major effort is underway to define other loci and genes influencing
diabetes risk, with two complementary hypotheses:
1. Type 1 diabetes is polygenic with multiple small influences of polymorphisms
of many genes determining disease in each individual or
2. Type 1 diabetes is oligogenic but heterogeneous with a few major genes determining
risk in individual families, but different genes for different families.
Both hypotheses may be true for different families, but until actual genes are
better defined we will lack essential information. It is however very clear
that the MHC locus on chromosome 6 provides the bulk of genetic susceptibility.
It is a striking observation that approximately 50% of siblings of patients
with type 1 diabetes with DR3/4, DQ2/8 genotype (and also HLA identical to their
sibling with diabetes, thus sharing all the same class I and other MHC locus
polymorphisms) develop islet autoantibodies prior to age 3, with the great majority
progressing to diabetes by age 5 (3). Such a potent
influence of this MHC genotype leaves little room for influence of non-MHC alleles
in such children with the highest risk HLA genotype.
The above observations in man and mouse suggest that given specific HLA haplotypes
the immune system is poised to destroy islet beta cells. Such destructive potential
is probably a function of the islets possessing one or more peptides that T
cells are either likely to target or unable to appropriately regulate an immune
response directed at those peptides. For example, it appears that recognition
of the B:9-23 peptide of insulin is dependent primarily upon the presence of
a specific T cell alpha chain (AV13.3) (10). If
recognition is not dependent upon the complex of TCR A and B chains and specific
junctional sequences, one would expect a huge number of T cells able to recognize
this peptide. In terms of protection from diabetes one can imagine MHC alleles
that delete or "deviate" AV13.3 T cell receptors and thereby provide
dominant suppression of disease. These self-reactive T lymphocytes are present
in normal mice and presumably normal man, and in mice/rats can be activated
with non-specific stimuli such as poly-IC. If this viral RNA mimic is able to
induce disease in rodents, it is likely multiple relatively innocuous (and multiple
different) viral infections may in a similar manner activate disease in man.
This may explain the difficulty in isolating specific environmental factors
responsible for triggering anti-islet autoimmunity.
Hypotheses are only useful if they can be tested. To begin to test the hypothesis
that a single peptide of insulin may be central to disease, we are developing
NOD mice lacking the two insulin genes, but with a transgene with an altered
insulin sequence, namely an alanine at position 16 of the insulin B chain. We
hypothesize that such an alteration will abrogate the development of diabetes
if the specific sequence of the B:9-23 insulin peptide is essential. Of note
this specific sequence has been modified to produce an altered peptide ligand
of the B:9-23 molecule that is currently in clinical trials.
The accompanying figure illustrates a view of pathogenesis:
Figure 1.
MHC alleles of normal individuals allow the selection (positive and lack of negative) of T cells with receptors that recognize islet autoantigens. Given a number of potentially non-specific stimuli (e.g. poly-IC), such T cells are expanded and activated. At the same time regulatory T cells recognizing the same peptide (e.g. insulin peptide B:9-23) are selected, expanded and activated. Finally the balance of pathogenic and regulatory T cells, both potentially recognizing the same peptide, determines whether islet destruction begins. Once islet destruction begins many islet determinants are presented and pathogenic T lymphocytes recognizing a multitude of antigens produce enough beta cell destruction, over time (regulation is still active over time) to lead to diabetes. This hypothesis suggests that the prevention of type 1 diabetes is likely to be best accomplished in an antigen specific manner, potentially with vaccination with the very peptides central to pathogenesis. This is a hypothesis we are pursuing.
Figure 2.
Reference List - links to PubMed available in Reference List.
Click to download Powerpoint slide set - Updated 6/05
For comments, corrections or to contribute teaching slides, please contact Dr. Eisenbarth at: george.eisenbarth@ucdenver.edu