Dissecting immune heterogeneity in Type 1 Diabetes using longitudinal sampling

Type 1 diabetes (T1D) exhibits substantial heterogeneity in disease course between subjects, with clinical variability in rate of progression before and after diagnosis. Autoantibodies are a reliable biomarker of disease progression but do not fully explain the mechanisms underlying clinical heterogeneity. Other immune markers are needed to effectively stratify subjects.

To assess the between-subject and within-subject variance of T1D immune markers, we performed a longitudinal observational study of subjects with and without T1D (n=60), with 9 clinical, immunologic, and metabolic visits over 1 year. Immune assays tested were: islet autoantibodies, antigen-specific T-cell frequencies, cytokine production by ELISPOT, and frequency of T-cell, natural killer and monocyte subsets by flow cytometry. We determined the intraclass correlation coefficient (ICC) for individual analytes within all assays. ICC prioritizes analytes that are highly stable within an individual over time but also vary between individuals. As expected, autoantibodies had the highest ICC over a year. High ICCs were also seen in the majority of T-cell subsets measured by flow cytometry, suggesting they could be prioritized for future studies. For example, the ICC of CD8 effector memory T cells measured over 1 year was 0.88, and 0.89 for CXCR3+ CD4+ T cells (max ICC = 1). Conversely, ELISPOT analytes generally exhibited poor ICC in the context of this study.

To characterize the multivariate relationship among the immune markers and clinical metrics, we used partial least squares regression (PLS). Using PLS, we can characterize the multivariate relationship between immune markers and endogenous insulin secretion, and identify immune markers such as CD8 EM that are most associated with insulin secretion after diagnosis. Identifying biomarkers associated with metabolic parameters, and understanding the longitudinal variability and between-subject heterogeneity of those markers, is expected to be useful in clinical study design and prediction of disease progression, risk, and therapeutic response in T1D.