Antibodies against insulin measured by electrochemiluminescence predicts insulitis severity and disease onset in non-obese diabetic mice and can distinguish human type 1 diabetes status

Abstract

Background: The detection of insulin autoantibodies (IAA) aids in the prediction of autoimmune diabetes development. However, the long-standing, gold standard 125I-insulin radiobinding assay (RBA) has low reproducibility between laboratories, long sample processing times and requires the use of newly synthesized radiolabeled insulin for each set of assays. Therefore, a rapid, non-radioactive, and reproducible assay is highly desirable.

Methods: We have developed electrochemiluminescence (ECL)-based assays that fulfill these criteria in the measurement of IAA and anti-insulin antibodies (IA) in non-obese diabetic (NOD) mice and in type 1 diabetic individuals, respectively. Using the murine IAA ECL assay, we examined the correlation between IAA, histopathological insulitis, and blood glucose in a cohort of female NOD mice from 4 up to 36 weeks of age. We developed a human IA ECL assay that we compared to conventional RBA and validated using samples from 34 diabetic and 59 non-diabetic individuals in three independent laboratories.

Results: Our ECL assays were rapid and sensitive with a broad dynamic range and low background. In the NOD mouse model, IAA levels measured by ECL were positively correlated with insulitis severity, and the values measured at 8-10 weeks of age were predictive of diabetes onset. Using human serum and plasma samples, our IA ECL assay yielded reproducible and accurate results with an average sensitivity of 84% at 95% specificity with no statistically significant difference between laboratories.

Conclusions: These novel, non-radioactive ECL-based assays should facilitate reliable and fast detection of antibodies to insulin and its precursors sera and plasma in a standardized manner between laboratories in both research and clinical settings. Our next step is to evaluate the human IA assay in the detection of IAA in prediabetic subjects or those at risk of type 1 diabetes and to develop similar assays for other autoantibodies that together are predictive for the diagnosis of this common disorder, in order to improve prediction and facilitate future therapeutic trials.

Trial registration: ClinicalTrials.gov NCT00896610.

Figures

Figure 1

Protocol for murine IAA ECL assay. A) Overview of the murine IAA ECL procedure. B) IAA in serum is precipitated by biotinylated-insulin conjugated to streptavidin-coated magnetic beads, while a ruthenium (Ru)-tagged secondary antibody detects the bound IgG autoantibodies. 2° = secondary.

Figure 2

Murine IAA ECL assay has a broad dynamic range and negligible background from non-diabetic strains. A) The chemistry behind ECL technology, demonstrating the regenerating electrochemiluminescent reaction. A magnet below the electrode attracts the magnetic beads, bringing the Ru-tagged immune complex near the electrode. The Ru is then oxidized by the current and reacts with the free radical form of tripropylamine (TPA) to produce a photon. Superscripted asterisk indicates the higher energy state of the chemical and superscripted dot indicates a radical species. Bipyridine = bpy. B) The murine IAA ECL was tested using an anti-insulin mouse (Ms) monoclonal antibody or an isotype control antibody at the same concentrations. C) The murine IAA ECL assay was also performed on serum from 8-week-old NOD mice (n = 10) and non-diabetic control strains (C57BL/6 and BALB/c, n = 10 each). ECL S/N is the average signal from triplicate wells divided by the "noise" (background signal) from triplicate wells with beads alone.

Figure 3

IAA levels peak at age 8-10 weeks for NOD that become diabetic by 20 weeks. IAA levels were followed in NOD mice from 4 weeks up to 36 weeks. A) IAA values from mice that developed diabetes by 20 weeks are plotted, with each line depicting the data from a single mouse. The IAA levels from mice that had an IAA value below the positive predictive threshold at 8.5 weeks (described in Table 1) are represented by dotted lines. The 95% confidence interval of IAA values from mice not diabetic by 20 weeks (shaded region, n = 22-29) is plotted for comparison. B) IAA values were normalized to the highest IAA value for each mouse (n = 18-25). The average (avg) IAA values at the indicated ages are graphed. Data after the age of 18 weeks are calculated from an avg of less than 10 mice due to euthanasia of diabetic mice. C) Percent of diabetes-free mice from 4 weeks up to 20 weeks are plotted (n = 54). The number of mice that became diabetic at each time point is indicated by bars. IAA value is calculated as the average signal from triplicate wells divided by the background signal from triplicate wells with negative control sera.

Figure 4

Insulitis index correlates with murine IAA ECL values. NOD mice were followed from 4 weeks up to 20 weeks of age. Serum samples were collected weekly and IAA levels were measured by our murine IAA ECL assay. Mice were euthanized at various ages and pancreatic tissue was extracted and stained with H&E. A) Insulitis was scored as depicted in the H&E stained tissue sections. The formula for calculating insulitis index is indicated below. B) Insulitis index was correlated with the IAA values at the age the pancreas was removed for histology (final measurement). C) Insulitis index was correlated with the maximum IAA values measured for each mouse during the study. Data were plotted for NOD pancreas tissues obtained from mice at 5-20 weeks and 5-10 weeks of age. IAA value is calculated as the average signal from triplicate wells divided by the background signal from triplicate wells with beads alone. The Spearman's correlation coefficient (rs) and the corresponding p-value for each age group are displayed on the graphs. The trends showing highest correlation (rs values) when restricting the analysis to younger mice is graphed (bottom panels), where the x-axis is the upper age limit (in weeks) of the mice included in the analysis.

Figure 5

ROC analysis of sensitivity-to-specificity ratios for IAA values in 7-10 week old mice. Area under the curve (AUC) is shown for each age analyzed. Receiver operating characteristic (ROC) curves were generated using the IAA values at the indicated ages for the NOD mice that did or did not develop diabetes by 20 weeks of age.

Figure 6

IAA levels measured at 8-10 weeks of age can distinguish mice progressing to disease. Positive predictive IAA thresholds for each age were calculated as in Table 1 and are indicated by the dotted lines. IAA values measured at 7, 8-8.5, and 10 weeks of age are shown for each mouse that did (Db+) or did not (Db-) become diabetic by 20 weeks. IAA value is calculated as the average signal from triplicate wells divided by the background signal from triplicate wells with negative control sera.

Figure 7

Using the murine IAA ECL format, IA negative and IA positive individuals cannot be distinguished. Serum from 23 individuals (IA-, n = 14; IA+, n = 9) measured for IA by RBA were tested for IA using the murine IAA ECL format, but with an anti-human IgG ruthenium-labeled secondary. IA value is calculated as the average signal from triplicate wells divided by the background signal from triplicate wells with beads alone. NS = non-significant.

Figure 8

Schematic diagram of ECL-based assay for human IA. A) Non-specifically binding material is removed from 1:10 diluted serum or plasma by incubation with streptavidin coated paramagnetic beads (SA-beads), followed by magnetic removal to yield a clarified sample. B) Biotinylated-insulin is incubated with the clarified sample and binds to IA in solution. C) Bound IA-biotinylated insulin complexes are precipitated with SA-beads. D) Ruthenium-labeled secondary goat anti-human IgG antibody binds to IA-bead complexes and is detected on the M1MR Analyzer.

Figure 9

Evaluation of specificity and stability of the ECL-based assay for human IA. A) Samples from five IA+ and one IA- donors as determined by RBA were incubated with an increasing percentage of non-biotinylated insulin prior to addition of biotinylated insulin. B) Example dataset from the competition assay for one IA+ and one IA- donor. IA value for each donor was determined from the mean signal from triplicate wells with insulin divided by the mean signal from triplicate background wells without insulin. C) Curve generated by ROC analysis of the ability of the ECL assay to distinguish between samples of RBA determined IA status. Analysis was performed using Prism software. D) Samples from IA+ and IA- donors were used to evaluate the reproducibility of the ECL assay across multiple batches of reagents. TAG-IgG = ruthenium labeled goat anti-human IgG secondary antibody. HAD = Human Assay Diluent.

Figure 10

Performance, dynamic range, and reproducibility of the ECL-based IA assay across three independent test sites. Plasma samples from 59 non-diabetic individuals (30 males, 29 females, Mean Age: 28.61 ± 8 yr) and serum samples from 34 long-standing diabetic individuals (11 males, 21 females, Mean Age: 38.23 ± 15.73) were evaluated using our human IA ECL-assay at three separate laboratories. A) IA values from all samples were analyzed by receiver operating characteristic (ROC) analysis for the ability to distinguish between non-diabetic and diabetic donors. B) Plot of mean values obtained across the three test sites for each sample. Samples could be distinguished by IA values spanning a 3-log range (0.77-943.3). A cutoff IA value of 3.0 (dashed line) determined by ROC analysis of these samples. C) Plot of ascending intra-laboratory rank vs ascending mean rank for the 93 samples across the three test sites. Samples were ranked 1 through 93 in order of increasing IA value as determined by our IA ECL assay. Black circles = Lab 1 (Wellstat), black squares = Lab 2 (NIAID), and black triangles = Lab 3 (DIL). All statistical analyses were performed using Prism software. *AUC, Area under the curve;† 95% CI, 95% confidence interval; AS95 = sensitivity at 95% specificity; Lab1-Wellstat = Wellstat Diagnostics in Gaithersburg, MD, USA. Lab2-NIAID = Laboratory of Immunology, National Institute of Allergy and Infectious Disease in Bethesda, MD, USA; and Lab3-DIL = Diabetes and Inflammation Laboratory, University of Cambridge in Cambridge, UK.