Autoantibody and T cell responses to oxidative post-translationally modified insulin neoantigenic peptides in type 1 diabetes

Aims/hypothesis Antibodies specific to oxidative post-translational modifications (oxPTM) of insulin (oxPTM-INS) are present in most individuals with type 1 diabetes, even before the clinical onset. However, the antigenic determinants of such response are still unknown. In this study, we investigated the antibody response to oxPTM-INS neoepitope peptides (oxPTM-INSPs) and evaluated their ability to stimulate humoral and T cell responses in type 1 diabetes. We also assessed the concordance between antibody and T cell responses to the oxPTM-INS neoantigenic peptides. Methods oxPTM-INS was generated by exposing insulin to various reactive oxidants. The insulin fragments resulting from oxPTM were fractionated by size-exclusion chromatography further to ELISA and LC-MS/MS analysis to identify the oxidised peptide neoepitopes. Immunogenic peptide candidates were produced and then modified in house or designed to incorporate in silico-oxidised amino acids during synthesis. Autoantibodies to the oxPTM-INSPs were tested by ELISA using sera from 63 participants with new-onset type 1 diabetes and 30 control participants. An additional 18 fresh blood samples from participants with recently diagnosed type 1 diabetes, five with established disease, and from 11 control participants were used to evaluate, in parallel, CD4+ and CD8+ T cell activation by oxPTM-INSPs. Results We observed antibody and T cell responses to three out of six LC-MS/MS-identified insulin peptide candidates: A:12–21 (SLYQLENYCN, native insulin peptide 3 [Nt-INSP-3]), B:11–30 (LVEALYLVCGERGFFYTPKT, Nt-INSP-4) and B:21–30 (ERGFFYTPKT, Nt-INSP-6). For Nt-INSP-4 and Nt-INSP-6, serum antibody binding was stronger in type 1 diabetes compared with healthy control participants (p≤0.02), with oxidised forms of ERGFFYTPKT, oxPTM-INSP-6 conferring the highest antibody binding (83% binders to peptide modified in house by hydroxyl radical [●OH] and >88% to in silico-oxidised peptide; p≤0.001 vs control participants). Nt-INSP-4 induced the strongest T cell stimulation in type 1 diabetes compared with control participants for both CD4+ (p<0.001) and CD8+ (p=0.049). CD4+ response to oxPTM-INSP-6 was also commoner in type 1 diabetes than in control participants (66.7% vs 27.3%; p=0.039). Among individuals with type 1 diabetes, the CD4+ response to oxPTM-INSP-6 was more frequent than to Nt-INSP-6 (66.7% vs 27.8%; p=0.045). Overall, 44.4% of patients showed a concordant autoimmune response to oxPTM-INSP involving simultaneously CD4+ and CD8+ T cells and autoantibodies. Conclusions/interpretation Our findings support the concept that oxidative stress, and neoantigenic epitopes of insulin, may be involved in the immunopathogenesis of type 1 diabetes. Graphical abstract Supplementary Information The online version contains peer-reviewed but unedited supplementary material available at 10.1007/s00125-022-05812-4.

Masslynx v4.2 (Waters, Milford, USA), whereas data were processed using the UNIFI Scientific Information System v1.8 software (Waters, Milford, USA). Three accumulations were averaged for each spectrum. Data analysis was carried out using the BeStSel server [21].

Assay for detection of peptide specific T cells. An activation induced marker (AIM) to label and characterize native (Nt-INS) and modified insulin (oxPTM-INS) responsive T cells and influenza
specific T cells (as a positive control) was performed as previously described [22]. Briefly, PBMC from subjects with type 1 diabetes were plated at a density of 10 million cells per ml in 24 well plates and pulsed with solvent only (Mock) or specific peptides of interest (oxPTM-INSP, Nt-INSP, or an influenza peptide) in individual wells for 10-14 hours in the presence of an anti-CD40 blocking antibody (Miltenyi Biotec). Activated T cells were labeled with anti-CD154 APC (Miltenyi Biotec) followed by anti-APC beads (Miltenyi Biotec) and magnetically enriched using columns (Miltenyi Biotec), reserving a 1% fraction of the non-enriched cells to determine the total number of T cells in the sample. The enriched and reserved cells were labeled with anti-CD4 BUV395 (BD Biosciences), anti-CD69 PE-Cy7 (BioLegend), anti-CD45RA AF700 (BD Biosciences), anti-CCR7 APC-Cy7 (BioLegend), plus anti-CD14 PerCP-Cy5.5 (BD Biosciences) anti-CD19 PerCP-Cy5.5 (BD Biosciences) and Viaprobe (BD Biosciences) as cell exclusion markers acquired using an BD LSR II cytometer and analyzed using FlowJo software (Tree Star).

ESM results
Peptide sequence confirmation by UPLC-qTOF/MS e . ESM Fig. 6 shows the fragmentation spectra of the three versions of synthetically oxidative modified peptides 3 (oxPTM-INSP-3) and 6 (oxPTM-INSP-6). Native and oxPTM-INSP-3 correspond to common sequence backbone of SLYQLENYCN, Detailed CID spectra with 67%-83% matched first generation primary ions within a filtered upper limit of error of measurement of 30 ppm of peptide y and b assigned fragment ions positively identified against their respective putative sequences and expected modifications was collected (not shown).
Structural changes in the oxPTM-INSPs compared to native peptides. Within the context of intact insulin, residues within native peptide 3 form a helix (ESM Fig. 7a). However, the free peptide is significantly unstructured (52%) but with sheet forming propensity (27%) and minor helical conformations (6%). After oxPTM modification, the SL-DOPA-QLENY-cysteate-N peptide became more structured with increased helicity (16%). However, the SL-DOPA-QLEN-DOPA-cysteate-N peptide was more disordered (60%) with complete loss of helical structure. Residues corresponding to native peptide 6 are extended within intact insulin and the free peptide (ERGFFYTPKT) showed both coil (54%) and sheet (29%) conformations. After oxPTM modification ERGYYYTPKT, ERGYY-DOPA-TPKT and ERGYYYTPKTR became significantly more sheet-like in structure (10 to 14% increase), while addition of arginine at the C-terminus in peptide Ac-ERGFFYTPKTR had no effect.

ESM Table 5 -CD8 + T-cell response in subjects with type 1 diabetes and controls according to different stimulatory index (SI) cut-offs.
Type 1   to optimized for Humulin R ® insulin that was formulated differently by the manufacturer. The terms "2x " and "5x " denotes the scaled-up of the above treatment conditions. Humulin R ® oxPTM displays similar reactivity as for the Sigma insulin.
ESM Figure 3. The size exclusion chromatography ÄKTA. Superdex 30 column suitable for the detection of low molecular weights (100-7000 kDa) was used for separating the various insulin fragments that were obtained after the oxidative posttranslational modification (oxPTM) of either Sigma insulin or Humulin R ® insulin. Sigma and Humulin R ® native insulin (Nt-INS) display a single peak at around 10 ml. Following oxidation, oxPTM-INS different small (< 6kDa) insulin fragments were observed. Sigma insulin exposed to HOCl showed a major peak at around 10 ml with additional peak at lower molecular weight. Exposing Sigma insulin to • OH resulted in a broad peak of small molecular fragments < 6 kDa at elution volumes between 10 ml and 25 ml, and an additional peak at 25 ml. Humulin R ® insulin, subjected to oxidation with HOCl displayed a major peak at 10 ml with an additional shoulder of a smaller molecular weight. For • OH modification of Humulin R ® insulin, the major native peak corresponding to native insulin disappeared but instead we observed a set of multiple peaks towards the end of the chromatograph corresponding to small molecular weights < 6 kDa. c. Type 1 diabetes (T1D) serum samples but not healthy controls (HC) bound to the small molecular weight fractions that resulted from either HOCl or • OH modification. Plotting the same data as individual data points for each subject preserved the same tendencies, but reveals some heterogeneity in the relative proportion of Naïve, TCM, and TEM cells. TPKT is associated with • OH -Insulin. We also observed clustering of response of type 1 diabetes samples using PCA analysis of all responses, CD4, CD8 and IgG (d). We observed cluster of 11 type one diabetes samples with PC1>0 while the rest clustered with healthy control samples with PC1<0.