The longitudinal loss of islet autoantibody responses from diagnosis of type 1 diabetes occurs progressively over follow-up and is determined by low autoantibody titres, early-onset, and genetic variants

Abstract

The clinical usefulness of post-diagnosis islet autoantibody levels is unclear and factors that drive autoantibody persistence are poorly defined in type 1 diabetes (T1D). Our aim was to characterise the longitudinal loss of islet autoantibody responses after diagnosis in a large, prospectively sampled UK cohort. Participants with T1D [n = 577] providing a diagnosis sample [range -1.0 to 2.0 years] and at least one post-diagnosis sample (<32.0 years) were tested for autoantibodies to glutamate decarboxylase 65 (GADA), islet antigen-2 (IA-2A), and zinc transporter 8 (ZnT8A). Select HLA and non-HLA SNPs were considered. Non-genetic and genetic factors were assessed by multivariable logistic regression models for autoantibody positivity at initial sampling and autoantibody loss at final sampling. For GADA, IA-2A, and ZnT8A, 70.8%, 76.8%, and 40.1%, respectively, remained positive at the final sampling. Non-genetic predictors of autoantibody loss were low baseline autoantibody titres (P < 0.0001), longer diabetes duration (P < 0.0001), and age-at-onset under 8 years (P < 0.01--0.05). Adjusting for non-genetic covariates, GADA loss was associated with low-risk HLA class II genotypes (P = 0.005), and SNPs associated with autoimmunity RELA/11q13 (P = 0.017), LPP/3q28 (P = 0.004), and negatively with IFIH1/2q24 (P = 0.018). IA-2A loss was not associated with genetic factors independent of other covariates, while ZnT8A loss was associated with the presence of HLA A*24 (P = 0.019) and weakly negatively with RELA/11q13 (P = 0.049). The largest longitudinal study of islet autoantibody responses from diagnosis of T1D shows that autoantibody loss is heterogeneous and influenced by low titres at onset, longer duration, earlier age-at-onset, and genetic variants. These data may inform clinical trials where post-diagnosis participants are recruited.

Keywords: autoantibodies; autoantibody loss; autoantibody specificity; autoimmunity; islet autoimmunity; type 1 diabetes.

Graphical Abstract

Figure 1:

Prevalence of autoantibody positivity at type 1 diabetes onset and longitudinal follow-up. Autoantibody positivity profiles for GADA, IA-2A, and ZnT8A in 577 subjects sampled at onset of T1D (<2 years) that also had at least one follow-up serum sample (range 2.0–32.0 years). (a) Positivity for GADA, IA-2A, and ZnT8A. (b) The proportion of subjects positive for three and two autoantibodies decreased from onset (P < 0.0001) and the proportion of subjects positive for 1 autoantibody or 0 autoantibodies increased from onset (P < 0.0001). (c) Proportions of positivity for each autoantibody over longitudinal follow-up were compared to proportions at onset (all P_corr < 0.0001) and to the previous category of longitudinal follow-up (shown below data points); Red: GADA; Blue: IA-2A; Black: ZnT8A at significance with Bonferroni correction *P_corr < 0.05, **P_corr < 0.01, ***P_corr < 0.001 and ****P_corr < 0.0001.

Figure 2:

Non-genetic associations of autoantibody loss for GADA, IA-2A, and ZnT8A. OR: odds ratio; 95% CI: confidence interval; OR of 1 denotes the reference category of variable; ORs over 1 favours autoantibody loss; red dots and text denote alpha significance (<0.05). ORs, CIs, and P values were calculated from multivariable logistic regression models for autoantibody loss at final sampling time (years) adjusted for all non-genetic covariates [gender, age at onset, quartile of baseline autoantibody titre at onset, and quartile of disease duration from onset (years) at final sampling]. This multivariable non-genetic model was used as a baseline model for further analysis.

Figure 3:

Longitudinal GADA, IA-2A, and ZnT8A titres according to the quartile of titre present at type 1 diabetes onset. The prevalence of longitudinal autoantibody positivity in all autoantibody responses was higher in subjects with high quartiles of baseline autoantibody titre (P < 0.0001–0.05). Independent of baseline titre, longitudinal autoantibody titres sequentially decreased over follow-up in most subjects [GADA: 403/487 (82.8%); IA-2A: 423/452 (93.6%); ZnT8A: 389/395 (98.5%)].

Figure 4:

Genetic associations of autoantibody positivity around type 1 diabetes onset and autoantibody loss after onset. OR: odds ratio; 95% CI: confidence interval; OR of 1 denotes the reference category of variable; ORs over 1 favours autoantibody positivity at initial sampling (a) or autoantibody loss at final sampling (b); red dots and text denote alpha significance (<0.05). ORs, CIs and P values were calculated from multivariable logistic regression models adjusted for all non-genetic covariates [gender, age-at-onset, sampling time (months/years), and autoantibody titre at onset (final sampling only)]. All genetic covariates were considered independently in all models.