Early induction of type 1 diabetes

Type 1 diabetes is the second most common chronic childhood illness, after asthma, in Europe and North America. The disease is due to destruction of the insulin secreting (cells of the pancreatic islets of Langerhans [1]. At the clinical onset of type 1 diabetes about 80% of islets contain no (cells and the islets may be infiltrated with mononuclear cells [2]. Mononuclear cell infiltration is not a consistent feature, but includes macrophages and T lymphocytes with lymphocytes predominating and CD8 positive cells being most prevalent [1]. It is thought that the destructive process is mediated directly or indirectly by these immune effector cells. In the years before the clinical onset of type 1 diabetes, immune and metabolic changes can be detected in peripheral blood [3]. Since the nature, intensity, extent and persistence of these immune changes distinguishes individuals who develop diabetes from those who do not, these changes can predict type 1 diabetes [4]. The long prediabetic period and the potential for prediction have led to attempts to prevent clinical disease. But for how long is this prediabetic period and when are the immune changes which lead to diabetes induced? In this issue of Clinical and Experimental Immunology, Mikhail Knip and his colleagues seek evidence that the process may be induced in utero [5].

Antibodies have been identified against a number of antigens in patients with newly diagnosed type 1 diabetes and antibodies against two of these, glutamic acid decarboxylase (GAD) and a protein tyrosine phosphatase (IA-2), are detected in about 90% of such patients, while insulin autoantibodies (IAA) are prevalent in young newly diagnosed diabetic children [3]. These antibodies are probably induced by the interaction of genetic and environmental factors. Type 1 diabetes represents a heterogeneous and polygenic disorder in which genes within the major histocompatibility complex (MHC) HLA class II region on chromosome 6p21 account for about 45% of disease susceptibility [6] while a non HLA loci on chromosome 11p5.5 [7] contributes about 10% toward disease susceptibility [6].

The most powerful evidence that nongenetic factors play a major role in causing type 1 diabetes comes from the study of monozygotic (MZ) and dizygotic (DZ) twins [8]. Differences or discordance between MZ twins must be due to nongenetically determined factors; while greater disease concordance in MZ than DZ twins indicates a genetic effect. Table 1 summarizes the results of concordance rates in population based cross-sectional twin studies for three immune-mediated diseases and is consistent with a major nongenetic effect causing these diseases, in association with limited genetic susceptibility. Although concordance is greater in MZ than DZ twins with autoimmune disease, the majority of MZ twins do not display concordance [8–11]. Recently a large prospective study of MZ diabetic twins found a diabetes concordance rate after up to 40 years review of about 50%, though the concordance rate was only 5% in adult onset cases, confirming that a major nongenetic effect causes type 1 diabetes [12].

Table 1.

Concordance rates (%) in population-based studies of immune-mediated diseases [9–11]

	MZ twins	DZ twins
Rheumatoid arthritis	12	4
Multiple sclerosis	26	2
Type 1 diabetes	13	3

Potential nongenetic factors associated with the critical diabetes-inducing event

Differences between MZ twins could be due to nongermline genetic events such as somatic mutations or environmental factors, though the latter is strongly favoured; for example, diabetes-associated autoantibodies are polyclonal [13]. However, the nature of such environmental factors is unclear, notwithstanding the evidence that they could be either viruses, toxins or early infant dietary factors (e.g. breast feeding vs. early introduction of cow's milk components) [14]. For while both viruses and toxins can cause diabetes in man we cannot be certain that they are responsible for the generality of the disease. Other potential nongenetic disease modifying factors include vaccine administration, psychological stress and climatic influences.

The critical event can operate in early childhood

Whatever the nature of the nongenetic or environmental factors inducing diabetes, they are likely to be ubiquitous and probably occur in a proportion of cases in early childhood. The qualitative age specific pattern of incidence of type 1 insulin-dependent diabetes is similar worldwide: extremely rare before 9 months of age, it has a peak incidence between the ages of 5 and 15 years and declines sharply thereafter, though with a long tail so that adults can present with it in old age [15]. The consistent and striking decline in disease incidence after puberty could be due to either a loss of a genetic or environmental effect; the latter is most likely. Loss of the genetic effect could result from attrition of the genetically susceptible pool or an increasing impact of protective genes, but we know from twin studies that the majority of genetically susceptible individuals do not develop the disease. Furthermore data from population studies show that the impact of HLA protective genes falls with age [6,16–18]. The fall in disease risk is even seen in MZ twins of patients with type 1 diabetes, at least in those with an index twin under 15 years of age at diagnosis in whom the risk of developing diabetes falls sharply one year after the diagnosis of the index twin [12]. Such a decline in disease risk in young twins, suggests that the nongenetic events leading to type 1 diabetes occur within a finite period in early childhood. A disproportionate maternal influence suggests that this event may be very early indeed, perhaps even in utero.

Maternal-related events influence diabetes risk

Maternal-related events influence the development of type 1 diabetes based on the increased risk in children:

• born to diabetic fathers rather than mothers;

• if their diabetic mother was aged less than 8 years at diagnosis compared with mothers diagnosed later;

• first born;

• with increasing maternal enterovirus infection rate;

• with ABO incompatibility with the mother;

• with increasing maternal age at delivery;

• according to the season of delivery;

• with early cessation of breast feeding.

Children of diabetic mothers are less likely to develop type 1 diabetes than children of diabetic fathers and the risk in mothers is less than the expected risk based on their HLA status [19,20]. This low risk is confined to offspring of mothers who had become diabetic after the age of 8 years. It remains unclear whether the reduction in transmission rates is due to genetic or nongenetic factors. The risk of offspring developing diabetes increases with maternal age; and paternal age also has an effect, but it is smaller than the maternal effect [21]. The risk of diabetes is reduced by 15% per child born so that the first born has the highest risk. Blood group incompatibility between a mother and her child may also predispose to type 1 diabetes [22], though the nature and implications of this association remains obscure. Group specific enteroviral IgG and IgM antibodies during pregnancy are higher in mothers of children who later develop type 1 diabetes than in controls [23]. Infants born in autumn and winter may also be at risk compared with those born in other months [24]. Finally, the early infant diet may affect type 1 diabetes onset in later life. Epidemiological studies have established that the risk of developing type 1 diabetes is higher in nonbreast-fed children than breast-fed children and that breast-feeding for more than 3 months protects from type 1 diabetes [14]; possibly because early cessation of breast-feeding is associated with the early introduction of foreign antigens, for example cow's milk proteins [14].

Activation of the diabetes-associated immune response

If the critical event which induces the destructive immune process operates in early childhood then it follows that diabetes-associated immune changes such as the production of serum autoantibodies may also be detected at this early age. Of those schoolchildren who develop islet cell antibodies (ICA), seroconversion to ICA positivity occurred before 5 years of age in 85% [15]; such seroconversion appears to be rare after 10 years of age [15]. At birth the child of a diabetic mother often has islet specific autoantibodies (ICA, IAA and GAD antibodies) but these antibodies can also be found in the maternal serum [25,26]. In the study by Hämäläinen et al. in this issue of Clinical and Experimental Immunology [5], the levels of ICA in cord blood were found to be comparable to maternal levels, and antibody specificities were similar in both the mother and cord blood, presumably reflecting transplacental passage of maternal IgG antibodies [5]. Importantly, diabetes-associated antibodies were not detected in any of the infants of mothers without such antibodies emphasizing the likely maternal origin of cord blood diabetes-associated antibodies. Nor was there any evidence of increased intrauterine antibody production [5]. Passively acquired maternal antibodies disappear after birth as expected, but they can subsequently be replaced by the infant's own antibodies. In one recent study, 3 out of 58 infants of diabetic mothers developed IAA, ICA and GAD by 2 years of age [25]. But at birth none of the offspring had islet autoimmunity other than maternally acquired antibodies. Antibodies detected at 9 months were usually also detected in the sample at two years of age and only then were they associated with a risk of diabetes [25]. The overall cumulative risk of type 1 diabetes in the 1353 offspring of parents with type 1 diabetes was 1·8% at 5 years but 50% in those with more than one diabetes-associated antibody [25]. In another study antibodies were detected in cord blood of 17% (14/81) of children who subsequently developed type 1 diabetes while only 4% of controls had such antibodies suggesting, but not proving, that in some cases induction of autoimmunity can be intrauterine, though this difference might simply reflect the increased genetic susceptibility in the offspring of diabetic mothers [26].

These observations, taken together, suggest that the immune process, associated with the development of type 1 diabetes, is usually activated in early childhood. It is even possible that in some cases the disease process starts in utero; however, the data presented in the current issue strongly suggest that foetal induction of autoimmunity is an exception and not the norm.