Insulin autoantibodies (IAA) were first convincingly demonstrated by Palmer and co-workers in 1983. These autoantibodies appear prior to insulin treatment and are present in approximately 70% of children and adolescents at the diagnosis of type 1 diabetes. The levels of IAA show a strong inverse correlation with age, being found in more than 90% of children under the age of 5 years at diagnosis. The early appearance of IAA makes them particularly useful for diabetes prediction in young children. Induction of insulin antibodies to exogenous insulin means that IAA measurement is no longer informative once insulin therapy has been given for more than 2 weeks.
Pro-insulin; a “primary” target of autoantibodies in type 1 diabetes
Of the major islet antigens, proinsulin and glutamate decarboxylase (GAD) are currently the best candidates as primary autoantigens; i.e. those that could play a role in the initiation of the autoimmune response that leads to beta cell destruction and type 1 diabetes. Proinsulin is favoured by many, since autoantibodies to insulin are often the first to be detected in young children at increased genetic risk of type 1 diabetes. Proinsulin, unlike GAD is expressed almost exclusively in beta cells, which is consonant with the specific targeting of beta cells by T-cells infiltrating the pancreatic islets. The very high frequency of insulin autoantibodies (IAA) found in young children at diabetes onset shows that loss of tolerance to proinsulin is common in those who progress rapidly . Insulin autoantibodies are also found in the non-obese diabetic (NOD) mouse model of spontaneous autoimmune diabetes , which further supports a role for proinsulin as a primary autoantigen.
Figure 1. Proportion of 424 children and young adults with newly-diagnosed type 1 diabetes from the UK found positive for autoantibodies to zinc transporter 8 (ZnT8A), islet antigen-2 (IA-2A), glutamate decarboxylase (GADA) and insulin (IAA) according to age. The prevalence of IAA at diagnosis falls with age.Birth cohort studies have shown that IAA can be detected from 6 months of age in children at increased genetic risk of type 1 diabetes. Recent analysis of the BABYDIAB and DIPP birth cohorts has shown that the incidence of IAA peaks between one and two years of age . In DIPP, peak IAA levels were reached 6 months after IAA first appeared. The prevalence and levels of IAA in patients at diagnosis of type 1 diabetes show an inverse correlation with age . In first-degree relatives of patients with type 1 diabetes, IAA may be found in more than 90% of children below 5 years of age, but in only half of young adults aged 15 to 21.
IAA prevalence and levels are similar between boys and girls across all ages, although the peak incidence in BABYDIAB was at 9 months in boys and 2 years in girls . A male excess of IAA at diagnosis in young adults has also been reported, but not confirmed by other studies.
IAA prevalence and levels are strongly associated with HLA DRB1*04 in both patients with type 1 diabetes and their relatives . This effect in relatives has also been ascribed to HLA DQA1 alleles, with DQA1-*0101, *0102, *0103, *0201, or *0301 associated with higher IAA levels in comparison to DQA1*0401, *0501, *0601. An association of IAA at diagnosis with the diabetes susceptibility class I allele of INS VNTR has been reported in Swedish patients , but not in other populations. Two further studies suggest that INS VNTR may influence IAA; IAA development was limited to young offspring of patients with type 1 diabetes carrying the class I allele of INS VNTR, while IAA were more frequent in ICA positive children in DIPP who were homozygous for the class I allele.
Use in prediction
The best models for diabetes prediction are currently based on combined measurement of islet autoantibodies . Insulin autoantibodies are a critical component of this mix when seeking to identify young children at increased risk of type 1 diabetes, but their lower prevalence at diagnosis in older individuals means that they are less useful for prediction in adolescents and adults. The early appearance of IAA make them indispensible for following the development of islet autoimmunity from birth in natural history and primary intervention studies. Eligibility for recruitment to secondary intervention trials is also often based on IAA measurement. IAA are of limited use for characterising the disease once insulin therapy has been initiated however, as they may be masked by the development of antibodies to exogenous insulin. Disease sensitivity and specificity of IAA are dependent on several factors, including the characteristics of the population, selection of assay threshold and assay performance. Common thresholds used for defining IAA positivity usually fall between the 97.5th and 99th percentile of control populations such as healthy schoolchildren or adult blood donors. Assay performance may vary widely, but recent international workshops have shown improved measurement by many laboratories. In an effort to improve the reliability of prediction models additional antibody characteristics such as antibody level, affinity and epitope specificity have been considered .
The intensity of the IAA response, as reflected in the antibody level, alters the risk of progression to diabetes. Achenbach et al. found that IAA positive first-degree relatives of patients with type 1 diabetes with levels in the upper quartile had a 77% risk of progression within 10 years compared with 37% for those in the lower 3 quartiles . Young age at seroconversion and persistent positivity for IAA were found to be significant risk factors for diabetes progression in siblings of patients with type 1 diabetes, while age at seroconversion and mean IAA levels predict age at onset in children at increased genetic risk.
Relatives and schoolchildren with moderate-high affinity IAA (>108-109 l/mol) have been found to be at increased risk of developing multiple islet autoantibodies and type 1 diabetes . High affinity IAA were associated with the HLA DRB104/DQB10302 haplotype, which may explain why IAA affinity did not discriminate risk of progression in IAA positive children with increased HLA-conferred genetic susceptibility. Lower affinity IAA were found to have a more restricted IgG subclass distribution and some had IgM dominated responses not found in individuals with high affinity IAA. Using a single insulin concentration as a surrogate measurement of IAA affinity, was also effective at discriminating progression to multiple islet autoantibodies and diabetes in relatives of patients with type 1 diabetes.
Early IAA responses are dominated by IgG1. Other IAA IgG subclasses are usually of lower titre, but their presence in addition to IgG1 has been associated with higher risk or more rapid progression in IAA positive relatives and children with increased genetic susceptibility. The number of IgG subclasses detected is also associated with IAA level. The presence of IgG4 IAA however, indicative of a Th2 phenotype, was not associated with protection from diabetes even when these responses were dominant.
Evidence that insulin autoantibodies were present in serum of patients with prior to exogenous insulin treatment was first provided in 1963, but they were first demonstrated convincingly in patients with type 1 diabetes by Palmer et al. in 1986  using a sensitive radiobinding assay (RBA). Enzyme-linked immunosorbent assays (ELISAs) were subsequently also used for IAA measurement, until international workshops in the early 1990s showed that IAA measured using RBAs were more closely associated with type 1 diabetes . The lower specificity of ELISAs may be because they detect lower affinity IAA, while reduced sensitivity could be explained by hindered binding to some of the major IAA epitopes when insulin is coated on the plastic of the ELISA plates. Except in very young children, the proportion of labelled insulin bound by IAA in RBAs is generally very low (<10 %) when compared with that bound by autoantibodies to GAD or islet antigen-2 (up to 75%); reliable and sensitive measurement is therefore very challenging. Early RBAs used polyethylene glycol (PEG) to precipitate immunocomplexes and A-14 mono-iodinated human insulin label . A number of modifications have been used in an effort to improve performance, including pre-extraction to remove endogenous insulin, competitive displacement with excess insulin to demonstrate specificity, use of large serum volumes (up to 700 μl) and incubation of serum and label for up to 7 days to achieve binding equilibrium. The use of protein A or G sepharose to precipitate immunocomplexes allowed the minimum assay volume required to be reduced to 20 μl, without loss of performance, largely because non-specific binding was reduced to less than 0.1% . Variations of this IAA “microassay” include washing by filtration or centrifugation, addition of salt and/or bovine serum albumin to the buffers, and incubation times of between 1 and 5 days. Efforts to develop alternative assay formats for IAA measurement have so far been unsuccessful, but recently an electrochemiluminescence (ECL) assay for proinsulin autoantibodies has shown promise of improved sensitivity and specificity compared with the IAA microassay .
Insulin autoantibodies are conformational, with no binding of isolated insulin A- and B-chains. The precise epitopes are poorly defined, although the major diabetes associated epitopes are dependent on intact disulphide bonds and on residues located in regions 8-13 of the A-chain and 1-3 of the B-chain . IAA which are less closely associated with diabetes and of lower affinity are often directed to the carboxy-terminal region of the B-chain, an area also targeted by antibodies to exogenous insulin. Early experiments suggested that use of A-14 monoiodinated insulin was superior to other insulin labels for detection of IAA, although binding to A-19 125I-insulin has been found to correlate with IAA affinity. There is little evidence to suggest that other regions of pre-proinsulin distinct from the insulin molecule are recognised by diabetes relevant antibodies, but these regions can be targeted by T-cells. Closer association of proinsulin autoantibodies with type 1 diabetes has been found in some studies and individuals with antibodies to both insulin and proinsulin are at increased risk of progression . This increased risk may be because binding of low affinity IAA to epitopes dependent on the carboxy-terminal region of the insulin B-chain is inhibited by the presence of the intact C-peptide linkage domain. Proinsulin autoantibodies measured by a sensitive and specific ECL assay were also found to be more closely associated with diabetes development than IAA measured using RBAs, although it is not clear whether this improved performance derives from the use of proinsulin or the altered assay format .
Significance of Insulin Antibodies
The observation that insulin autoantibodies appear so early in the prodrome to type 1 diabetes in children suggested that recognition of IAA-associated epitopes might be the initiating step in the autoimmune cascade resulting in diabetes. This formed an important part in the rationale of intervention studies using injected, oral or intranasal insulin, but results from these trials have been disappointing.
^ Ziegler AG et al. Age-related islet autoantibody incidence in offspring of patients with type 1 diabetes. Diabetologia 2012 55(7):1937-43
^ Parikka V et al. Early seroconversion and rapidly increasing autoantibody concentrations predict prepubertal manifestation of type 1 diabetes in children at genetic risk. Diabetologia 2012 55(7):1926-36
^ Ziegler AG et al. Radioassay determination of insulin autoantibodies in NOD mice. Correlation with increased risk of progression to overt diabetes. Diabetes 1989 38(3):358-63
^ Vardi P et al. Concentration of insulin autoantibodies at onset of type I diabetes. Inverse log-linear correlation with age. Diabetes Care 1988 11(9):736-9
^ Ziegler R et al. 1991. Specific Association of HLA-DR4 with Increased Prevalence and Level of Insulin Autoantibodies in 1st-Degree Relatives of Patients with Type-I Diabetes. Diabetes 40: 709-14
^ Graham J et al. Diabetes Incidence in Sweden Study Group; Swedish Childhood Diabetes Study Group. Genetic effects on age-dependent onset and islet cell autoantibody markers in type 1 diabetes. Diabetes 2002 51(5):1346-55.
^ Achenbach P et al. Stratification of type 1 diabetes risk on the basis of islet autoantibody characteristics. Diabetes. 2004 53(2):384-92. Erratum in: Diabetes. 2004 53(4):1175-6.
^ Achenbach P et al. 2004. Mature high-affinity immune responses to (pro)insulin anticipate the autoimmune cascade that leads to type 1 diabetes. J Clin Invest 114: 589-97
^ Palmer JP et al. Insulin antibodies in insulin-dependent diabetics before insulin treatment. Science. 1983 23;222(4630):1337-9
^ Greenbaum CJ et al. Insulin autoantibodies measured by radioimmunoassay methodology are more related to insulin dependent diabetes mellitus than those measured by enzyme-linked immunosorbent assay: results of the fourth international workshop on the standardization of insulin autoantibody measurement. J.Clin.Endocrinol.Metab. 1992; 74: 1040-1044
^ Williams AJK et al. A novel micro-assay for insulin autoantibodies. J.Autoimmun.1997; 10: 473-478
^ Yu L et al. Distinguishing persistent insulin autoantibodies with differential risk: nonradioactive bivalent proinsulin/insulin autoantibody assay. Diabetes 2012 61(1):179-86