Discovery and Validation of Serum Protein Changes in Type 1 Diabetes Patients Using High Throughput Two Dimensional Liquid Chromatography-Mass Spectrometry and Immunoassays

Abstract

Type 1 diabetes (T1D) is expected to cause significant changes in the serum proteome; however, few studies have systematically assessed the proteomic profile change associated with the disease. In this study, a semiquantitative spectral counting-based two dimensional liquid chromatography mass spectrometry platform was used to analyze serum samples from T1D patients and controls. In this discovery phase, significant differences were found for 21 serum proteins implicated in inflammation, oxidation, metabolic regulation, and autoimmunity. To assess the validity of these findings, six candidate proteins including adiponectin, insulin-like growth factor binding protein 2, serum amyloid protein A, C-reactive protein, myeloperoxidase, and transforming growth factor beta induced were selected for subsequent immune assays for 1139 T1D patients and 848 controls. A series of statistical analyses using cases and controls matched for age, sex, and genetic risk confirmed that T1D patients have significantly higher serum levels for four of the six proteins: adiponectin (odds ratio (OR) ‫؍‬ 1.95, p ‫؍‬ 10 ؊27 ), insulin-like growth factor binding protein 2 (OR ‫؍‬ 2.02, p < 10 ؊20 ), C-reactive protein (OR ‫؍‬ 1.13, p ‫؍‬ 0.007), serum amyloid protein A (OR ‫؍‬ 1.51, p < 10 ؊16 ); whereas the serum levels were significantly lower in patients than controls for the two other proteins: transforming growth factor beta induced (OR ‫؍‬ 0.74, p < 10 ؊5 ) and myeloperoxidase (OR ‫؍‬ 0.51, p < 10 ؊41 ). Compared with subjects in the bottom quartile, subjects in the top quartile for adiponectin (OR ‫؍‬ 6.29, p < 10 ؊37 ), insulin-like growth factor binding protein 2 (OR ‫؍‬ 7.95, p < 10 ؊46 ), C-reactive protein (OR ‫؍‬ 1.38, p ‫؍‬ 0.025), serum amyloid protein A (OR ‫؍‬ 3.36, p < 10 ؊16 ) had the highest risk of T1D, whereas subjects in the top quartile of transforming growth factor beta induced (OR ‫؍‬ 0.41, p < 10 ؊11 ) and myeloperoxidase (OR ‫؍‬ 0.10, p < 10 ؊43

Peptide serum markers in islet autoantibody-positive children

“…Consistent with previous studies in children with overt type 1 diabetes, we found lower levels of APOA4 [19], APOC4 [19], CF3 [17,18], CF4 [17][18][19], clusterin [16,17], kininogen [17] and transthyretin [17] in children with islet autoantibodies. We also found lower levels of CF9 in autoantibodypositive children, while others [19] have reported slightly increased levels.…”

“…2). Eight of those proteins (represented by 14 peptides) overlapped with findings from previous studies [16][17][18][19] (Table 1). This included four of the 14 peptides that were tested in our study because they had been identified in a previous study [17] (Table 1).…”

“…As expected, children in the highrisk group were younger; however, they did not differ from the other risk groups in islet autoantibody status or HLA genotype (ESM Table 3). Positive effect sizes represent higher abundance and negative effect sizes represent lower abundance in islet autoantibody-positive vs autoantibody-negative children c Significant differences were reported by Moulder et al [19] d Significant differences were reported by Zhi et al [18] e Significant differences were reported by Zhang et al [17] f Peptides originating from Zhang et al [17], but not selected by dCV in the selection phase of this study g Significant differences were reported by Metz et al [16] h Peptide levels were not significantly different in the Zhang et al [17] validation cohorts ALB, albumin; BTD, biotinidase; C, complement component; CLU, clusterin; CPN1, carboxypeptidase N; EFEMP, epidermal growth factor-containing fibulin-like extracellular matrix protein 1; FN1, fibronectin 1; HPX, haemopexin; ITIH, inter-α-trypsin inhibitor heavy chain; KNG1, kininogen 1; PROZ, vitamin K-dependent protein Z; QSOX1, quiescin Q6 sulfhydryl oxidase 1; SERPINF2, α2-antiplasmin; TTR, transthyretin…”

“…We also identified a proteomic signature that, together with age, was able to discriminate fast and slow progression to clinical diabetes in islet autoantibodypositive children. Previous studies have used LC-MS/MS-based proteomics approaches and applied extensive prefractionation techniques on pooled samples [17,18], followed by applying selected candidate proteins using ELISA [18], LC-SRM-MS [18] or other methods [19]. We designed our study in two phases, capitalising on the high analytical depth of a shotgun proteomics approach for selecting interesting peptides followed by an application using sensitive targeted proteomics specifically developed for the subset of potentially relevant peptides.…”

“…In type 1 diabetes, previous proteomic biomarker screening studies have compared patients with type 1 diabetes to autoantibody-negative control participants [16][17][18] and identified protein signatures correlated with clinical disease. A recent longitudinal study in Finland compared islet autoantibody-positive children with autoantibody-negative children, and identified a protein signature that distinguished between healthy children and those with autoimmunity [19].…”