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

Mapping of the oxidised amino acid hotspots in the oxPTM-INS

For the epitope mapping, we used multiple size-exclusion chromatography (ÄKTA), ELISA and LC-MS/MS experiments for Sigma insulin and Humulin R insulin. We first confirmed that the reactivity pattern of type 1 diabetes samples to Humulin R oxPTM-INS was similar to Sigma oxPTM-INS (ESM Fig. 2). Size-exclusion chromatography fractions of oxPTM-INS corresponding to small insulin fragments resulting from oxPTM were collected and analysed by ELISA. Fractions that showed reactivity by ELISA were dried and analysed by LC-MS/MS (ESM Fig. 3, ESM Table 2). We have previously reported that amino acids His5, Cys7, Tyr16, Phe24 and Tyr26 in the beta-chain are oxidised hotspots [18]. In the current study, additional new oxidised amino acid modification hotspots were discovered: His10, Leu17, Cys19 and Phe25 of the beta-chain and Cys6, Cys7, Cys11, Tyr14 and Cys20 of the alpha-chain. Oxidation of Cys6 in the alpha-chain was also seen in the Nt-INS (Fig. 1).

Fig. 1figure 1

Mapping the oxidised amino acid hotspots in oxPTM-INS. The analyses showed that the main hotspots in oxPTM-INS involved His5, Cys7, His10, Tyr16, Leu17, Cys19, Phe24, Phe25 and Tyr26 in the insulin beta-chain. Additional hotspots were identified in the alpha-chain: Cys6, Cys7, Cys11, Tyr14 and Cys20. Red boxes indicate the newly discovered oxidation hotspots and blue boxes indicate previously described amino acid hotspots

LC-MS/MS experimental data mapped neoepitopes to six potential oxPTM-INSPs that span both insulin alpha- and beta-chains. Candidate INSPs included SLYQLENYCN (A:12–21, INSP-3) from the alpha-chain and an additional five peptides from the beta-chain: YLVCGERGFF (B:16–25, INSP-1), LVEALYLVCGER (B:11–22, INSP-2), LVEALYLVCGERGFFYTPKT (B:11–30, INSP-4), FVNQHLC (B:1–7, INSP-5), ERGFFYTPKT (B:21–30, INSP-6). We also included another version of INSP-6 with the addition of a C-terminal arginine (R), as this sequence was seen in several MS profiles and R is the amino acid in the junction with the proinsulin C-peptide (Table 2).

Table 2 Six oxidative insulin neoantigenic peptidesAntibody reactivity of type 1 diabetes serum against the candidate oxPTM-INSPs

Identified peptide candidates were synthesised and exposed to either HOCl or ●OH to generate oxPTM-INSPs that were first assessed by time of flight–electrospray ionisation positive mode–total ion chromatogram (TOF-MS/ES+TIC) to confirm modification (ESM Fig. 4). Antibody response against Nt-INSPs and oxPTM-INSPs was evaluated by ELISA using sera from Study cohort 1. Serum antibody binding experiments revealed the highest number of type 1 diabetes binders for ●OH-modified oxPTM-INSP-3 (86% binders, mean absorbance=0.667±0.044; cut-off defined as mean binding of healthy control to Nt-INSP-3 plus 3×SEM), HOCl-modified oxPTM-INSP-4 (66% binders, mean absorbance=0.563±0.053; cut-off defined as mean binding of healthy control to Nt-INSP-4 plus 3×SEM) and ●OH-modified oxPTM-INSP-6 (83% binders, mean absorbance=0.461±0.013; cut-off defined as mean binding of healthy control to Nt-INSP-6 plus 3×SEM) (Fig. 2a–c, ESM Table 3). No significant reactivity was observed for INSP-1, INSP-2 and INSP-5 (data not shown). For oxPTM-INSP-3, we observed high background binding of healthy control samples (p>0.05; Fig. 2a, ESM Table 3). For oxPTM-INSP-4 and oxPTM-INSP-6, binding of type 1 diabetes serum was significantly stronger compared with control samples (p=0.0204, p=0.0176 and p=0.0005 for native, ●OH and HOCl oxPTM-INSP-4; and p<0.0001, p<0.0001 and p=0.0187 for native, ●OH and HOCl oxPTM-INSP-6; Fig. 2b,c, ESM Table 3). INSP-6 showed the highest specificity and sensitivity, with AUCs of 0.879, 0.875 and 0.740 for native, ●OH-modified and HOCl-modified INSP-6 (ESM Table 3, ESM Fig. 5).

Fig. 2figure 2

Antibody binding reactivity to neoantigenic INSPs (oxPTM-INSPs) in type 1 diabetes. (ac) After oxidation of INSPs by either ●OH or HOCl, reactivity of type 1 diabetes serum samples against each of the native or oxPTM-INSP candidates was tested by ELISA. (a) oxPTM-INSP-3 (SLYQLENYCN) showed higher reactivity compared with Nt-INSP-3 (p<0.001). The highest reactivity for oxPTM-INSP was observed for ●OH-modified oxPTM-INSP-3. High background binding in healthy control samples was observed for oxPTM-INSP-3. (b) oxPTM-INSP-4 (LVEALYLVCGERGFFYTPKT) and (c) oxPTM-INSP-6 (ERGFFYTPKT) displayed significantly greater reactivity in type 1 diabetes samples compared with healthy control samples (p=0.0204, p=0.0176 and p=0.0005 for native, ●OH and HOCl oxPTM-INSP-4, respectively; and p<0.0001, p<0.0001 and p=0.0187 for native, ●OH and HOCl oxPTM-INSP-6, respectively). (d, e) In silico-oxidised oxPTM-INSP-3 derivatives of SLYQLENYCN included SL-DOPA-QLENY-Cysteate-N and SL-DOPA-QLEN-DOPA-Cysteate-N (d). In silico-oxidised oxPTM-INSP-6 derivatives of ERGFFYTPKT included ERGYYYTPKT, ERGYY-DOPA-TPKT and ERGYYYTPKTR (e). (d) Binding to SL-DOPA-QLENY-Cysteate-N and SL-DOPA-QLEN-DOPA-Cysteate-N was similar to binding to SLYQLENYCN (p=NS). (e) In type 1 diabetes patients, binding to ERGYYYTPKT or ERGYYYTPKTR was higher than binding to native ERGFFYTPKT (p≤0.008). Similarly, a significant increase in binding to ERGYY-DOPA-TPKT was observed compared with the native ERGFFYTPKT (p=0.008). Multiple comparisons were adjusted for using the Holm–Sidak test; *p<0.05; **p<0.01; ***p<0.001. Cut-off points of positivity (binders) in the antibody ELISA for each peptide were defined by the mean absorbance of healthy control samples to the corresponding Nt-INSP plus 3×SEM. HC, healthy control; T1D, type 1 diabetes

Designing in silico oxPTM-INSPs

We designed in silico multiple oxPTM-INSP derivatives corresponding to one or more aminoacidic modification. For INSP-3 (SLYQLENYCN) we synthesised the following oxPTM-INSP-3 derivatives: SL-DOPA-QLENY-Cysteate-N where tyrosine (Y) was converted to DOPA only in one position and cysteine (C) to Cysteate. An additional oxPTM-INSP-3 was synthesised where both Y residues were converted to DOPA: SL-DOPA-QLEN-DOPA-Cysteate-N. To make the in silico oxPTM-INSP-6 of ERGFFYTPKT, phenylalanine (F) was converted to Y, and Y to DOPA. We thus synthesised two oxPTM-INSP-6 versions of ERGFFYTPKT: ERGYYYTPKT and ERGYY-DOPA-TPKT. We also included another oxPTM-INSP-6 version with a C-terminal arginine (R), ERGYYYTPKTR, as this sequence was seen in several MS profiles and R is the amino acid in the junction with the proinsulin C-peptide (ESM Table 2). Sequences of native peptides and their corresponding in silico oxPTM peptides were confirmed by UPLC-qTOF/MSe (ESM Results, Peptide sequence confirmation by UPLC-qTOF/MSe, ESM Fig. 6). Structural changes induced by oxPTM were then studied by circular dichroism analysis (ESM Results, Structural changes in the oxPTM-INSPs compared with native peptides, ESM Fig. 7).

Antibody reactivity of type 1 diabetes serum against in silico-modified oxPTM-INSPs

In type 1 diabetes patients (Study cohort 1), we observed a non-significant increase in binding to SL-DOPA-QLENY-Cysteate-N and SL-DOPA-QLEN-DOPA-Cysteate-N (54% and 57% binders, respectively) compared with SLYQLENYCN (49% binders, p>0.05); binding of type 1 diabetes samples was, however, significantly more frequent compared with healthy control samples (9%, 17% and 30% of control samples bound to SLYQLENYCN, SL-DOPA-QLENY-Cysteate-N and SL-DOPA-QLEN-DOPA-Cysteate-N, with p=0.0006, p=0.0112 and p=0.0029 vs type 1 diabetes, respectively) (Fig. 2d, ESM Table 3). There was no increase in specificity/sensitivity of binding to oxPTM-INSP-3 derivatives compared with the Nt-INSP-3, with AUCs of 0.670, 0.707 and 0.664, for SLYQLENYCN, SL-DOPA-QLENY-Cysteate-N and SL-DOPA-QLEN-DOPA-Cysteate-N, respectively (ESM Table 3, ESM Fig. 5).

We observed a significantly increased binding of type 1 diabetes samples to both ERGYYYTPKT and ERGYYYTPKTR, with 100% and 88% binders, respectively, compared with 25% and 48% binders in control samples, respectively (p≤0.004). In type 1 diabetes patients, binding to oxPTM-INSP-6 derivative ERGYYYTPKT or ERGYYYTPKTR was significantly higher compared with the native ERGFFYTPKT (p≤0.008). Similarly, a significant increase in binding to ERGYY-DOPA-TPKT was observed, compared with the native ERGFFYTPKT (p=0.008; Fig. 2e). We did not observe a significant difference in specificity/sensitivity of Nt-INSP-6 vs in silico-modified oxPTM-INSP-6 derivatives, with AUCs of 0.8686, 0.8542 and 0.8340, respectively (ESM Table 3, ESM Fig. 5).

A competitive displacement assay was performed to evaluate serum binding specificities to oxPTM-INSPs by pre-incubating sera with Nt- or oxPTM-INSPs. Interestingly, oxPTM-INSP-3 and oxPTM-INSP-6, but not Nt-INSP-3 or Nt-INSP-6, were able to inhibit the binding of type 1 diabetes samples to oxPTM-INS (p<0.001), but not to Nt-INS (Fig. 3a–c,g–i). Competition with combined oxPTM-INSP-3 and oxPTM-INSP-6 did not increase blocking to oxPTM-INS binding compared with a single peptide (data not shown). Nt-INSP-4, however, displayed a comparable inhibition compared with oxPTM-INSP-4 (Fig. 3d–f).

Fig. 3figure 3

Serum binding specificity to neoantigenic INSPs (oxPTM-INSPs). The figure shows the residual antibody binding to Nt-INS or oxPTM-INS with and without preincubation of type 1 diabetes serum samples with Nt- or oxPTM-INSP-3 (a, b, c), Nt- or oxPTM-INSP-4 (d, e, f), or Nt- or oxPTM-INSP-6 (g, h, i). Preincubation of type 1 diabetes serum samples with oxPTM-INSP-3 and oxPTM-INSP-6, but not with unmodified native peptides, strongly inhibited binding to oxPTM-INS (p<0.001), indicating the presence of antigen-binding sites specific to these oxPTM-INSPs. Nt-INSP-4 displayed a comparable inhibition to oxPTM-INSP-4. Percentage residual binding to Nt-INS (b, e, h) and oxPTM-INS antibodies (c, f, i) is shown for each type 1 diabetes sample tested. Antibodies to ●OH-INS are used as an example for oxPTM-INS. Each line in the figure panels represents the percentage binding of a serum sample from a single donor with type 1 diabetes to either Nt-INS (b, e, h) or oxPTM-INS (c, f, i) after preincubation with Nt-INSP-3 or oxPTM-INSP-3 (b, c); Nt-INSP-4 or oxPTM-INSP-4 (e, f); or Nt-INSP-6 or oxPTM-INSP-6 (h, i), relative to binding to Nt-INS or oxPTM-INS without peptide competitors (100%). Multiplicity was adjusted for using the Holm–Sidak test; *p<0.05; ***p<0.001

T cell stimulation with oxPTM-INSPs

To evaluate the immune cell response against the oxPTM-INSPs, we performed CD4+ and CD8+ T cell proliferation experiments using freshly isolated PBMCs (Study cohort 2, Table 1). Response was calculated as SI over unstimulated T cells.

We found that Nt-INSP-4 (LVEALYLVCGERGFFYTPKT) induced the strongest stimulation in type 1 diabetes compared with control samples for both CD4+ (mean SI: 119.8±51.69 vs 6.89±3.4, p<0.001; Fig. 4a) and CD8+ T cells (mean SI: 405.8±325.5 vs 5.948±3.125, p=0.049; Fig. 4c). Of note, as highlighted by the heatmaps in Fig. 4b,d, heterogeneous responses also to other peptides were evident across different individuals with type 1 diabetes, with some preferentially responding to various derivatives of oxPTM-INSP-3 (SL-DOPA-QLENY-Cysteate-N, SL-DOPA-QLEN-DOPA-Cysteate-N) and oxPTM-INSP-6 (ERGYYYTPKT, ERGYY-DOPA-TPKT, ERGYYYTPKTR). To better assess specificity of T cell stimulation in type 1 diabetes compared with control participants, we analysed response according to different SI cut-offs. When using an SI>3, we found a larger number of individuals with type 1 diabetes with a CD4+ response to oxPTM-INSP-6 derivatives compared with control participants (66.7% vs 27.3%; p=0.039), while response to Nt-INSP-4 and oxPTM-INSP-3 was similar between type 1 diabetic and control participants (Nt-INSP-4: 66.7% vs 45.5%; oxPTM-INSP-3: 22.2% vs 9.1%) (ESM Table 4). When comparing response to oxPTM-INSPs and Nt-INSPs among type 1 diabetes patients, we found that CD4+ response to oxPTM-INSP-6 was more frequent compared with Nt-INSP-6 (66.7% vs 27.8%; p=0.045) (Fig. 4a,b, ESM Table 4). CD8+ T cell responses to the tested peptides were also common in individuals with type 1 diabetes, who responded with similar frequency to oxPTM-INSP-6 and Nt-INSP-4 (72.2% of patients showed SI>1 for both); such response was higher in type 1 diabetic compared with control participants for oxPTM-INSP-6 (72.2% vs 27.3%; p=0.02), but not for Nt-INSP-4 (72.2% vs 63.6%; p=NS) (ESM Table 5). Higher SI cut-offs did not reveal significant differences between groups (ESM Table 5).

Fig. 4figure 4

oxPTM-INSPs stimulate T cell and autoantibody responses in type 1 diabetic individuals. The figure shows the CD4+ (a, b, red), CD8+ (c, d, blue) and IgG autoantibody (e, f, green) responses against Nt-INSPs and oxPTM-INSPs. Nt-INSP-4 (LVEALYLVCGERGFFYTPKT) and different oxidative derivatives of oxPTM-INSP-6 (ERGYYYTPKTR) are the main targets. Heatmaps show the degree of response heterogeneity within type 1 diabetic individuals, with some individuals preferentially responding to different oxPTM-INSP formats derived from the same peptide sequence. (b, d, f) Heatmaps of reactivity for each individual with type 1 diabetes tested against the various Nt-INS-Ps and oxPTM-INSPs. Multiplicity-adjusted p values: *p<0.05; **p<0.01; ***p<0.001

Correlation analysis showed association between T cell responses to the oxPTM-INS-6 derivative ERGYYYTPKTR (but not Nt-INSP-6) and Nt-INSP-4, for both CD4+ (r=0.59, p=0.12; Fig. 5a) and CD8+ (r=0.83, p=0.002; Fig. 5b). The CD4+ T cell response to Nt-INSP-4 was also strongly correlated with the CD8+ T cell response to oxPTM-INSP derivatives ERGYYYTPKTR and SL-DOPA-QLENY-Cysteate-N (r≥0.83, p≤0.002), but not their native counterparts (Fig. 5c), suggesting an overlap in CD4+ and CD8+ T cell responses involving Nt-INSP-4, oxPTM-INSP-3 and oxPTM-INSP-6.

Fig. 5figure 5

Correlation matrix of CD4+ (a) and CD8+ (b) responses to oxPTM-INSPs. T cell stimulation with Nt-INSP-4 (LVEALYLVCGERGFFYTPKT) strongly correlated with stimulation with oxPTM-INSP-6 (ERGYYYTPKTR) containing two aminoacidic oxidations (conversion of Phe24 and Phe25 to Tyr), for CD8+ (r=0.83, p=0.002) responses. There was no correlation between stimulation by Nt-INSP-4 and stimulation by Nt-INSP-6 (ERGFFYTPKTR) containing the native aminoacidic sequence (a, b). CD4+ stimulatory response to Nt-INSP-4 correlated with CD8+ responses to Nt-INSP-4 (r=0.89; p=0.0017), oxPTM-INSP-6 (r=0.83; p=0.002) and oxPTM-INSP-3 (SL-DOPA-QLENY-Cysteate-N) (r=0.86; p=0.0018) (c). All analyses are corrected for multiple comparisons, with statistically significant correlations highlighted by green squares

We next utilised surface staining for CD45 receptor type C (CD45RA; the long isoform of CD45 that is expressed on naive T cells) and C-C motif chemokine receptor 7 (CCR7) on CD154+CD69+ T cells to classify epitope-specific T cells as naive (CD45RA+CCR7+), central memory (TCM, CD45RA−CCR7+), effector memory (TEM, CD45RA−CCR7−) or effector memory cells re-expressing CD45RA (TEMRA, CD45RA+CCR7−). Across five representative individuals with established type 1 diabetes (Cohort 3, ESM Table 1), we detected TCM, TEM and TEMRA, with naive cells also present. Nt-INS- and oxPTM-INS-specific T cells had a higher percentage of naive cells than the influenza control (44.7% and 41.1%, respectively) but appreciable percentages of TCM and TEM were also present, suggesting that there is an existing pool of memory T cells that recognised these INSPs in individuals with type 1 diabetes (ESM Figs 8, 9).

Correlation between T cell stimulation and antibody response

Participants evaluated for T cell stimulation were also tested for antibody reactivity to either oxPTM-INSPs or oxidised intact insulin (oxPTM-INS) to assess correlations between humoral and cellular responses (Fig. 4e,f). In Study cohort 2, antibody reactivity to oxPTM-INSP-6 was the highest, as observed in Study cohort 1, with 11/18 (61.1%) binding to at least one oxPTM-INSP-6 derivative (p<0.001 oxPTM-INSP-6 vs Nt-INSP-6). Detailed analysis of autoantibody response in this cohort is described in the ESM (ESM Results, Antibody binding to oxPTM-INSPs in Study cohort 2).

We then analysed the extent of correlation between CD4+, CD8+ and IgG antibody responses. CD4+ and CD8+ responses to oxPTM-INSP-3 overlapped in 9/18 (50.0%), but only 1/18 patients (5.5%) showed concordant antibody reactivity (Fig. 6a). The CD4+ T cell response to Nt-INSP-4 frequently overlapped with CD8+ (13/18 [72.2%]), and to a lesser extent with antibodies (7/18 [38.8%]). Overall, 4/18 (22.2%) patients had a concordant CD4+, CD8+ and antibody response to Nt-INSP-4 (Fig. 6b). CD4+ response to oxPTM-INSP-6 was linked to both CD8+ and/or antibodies: 12/18 (66.7%) patients had concordant CD4+ and CD8+ responses, while 9/18 (50%) patients had concordant CD4+ and antibody responses. Overall, 8/18 (44.4%) patients showed an immune response involving simultaneously CD4+, CD8+ and antibodies (Fig. 6c). CD4+ T cell stimulation with Nt-INSP-4, oxPTM-INSP-6 and oxPTM-INSP-3 was associated with antibody reactivity to oxPTM-INS in 8/18 (44.4%), 8/18 (44.4%) and 7/18 (38.9%) participants with type 1 diabetes. Concordant autoimmune response to oxPTM-INSPs involving simultaneously CD4+ and CD8+ T cells and autoantibodies to oxPTM-INS was seen in 5/18 (27.8%), 6/18 (33.3%) and 4/18 (22.2%) participants with type 1 diabetes for Nt-INSP-4, oxPTM-INSP-6 and oxPTM-INSP-3, respectively (Fig. 6d–f), suggesting that CD4+ T cell response to these peptides is required to generate CD8+ and/or antibody responses to oxPTM-INS.

Fig. 6figure 6

Venn diagrams of CD4+, CD8+ and antibody responses to oxPTM-INSPs. (a, b, c) Overlap between IgG autoantibody and T cell (CD4+ and CD8+) responses specific to oxPTM-INSP-3 (a), Nt-INSP-4 (b) and oxPTM-INSP-6 (c). (d, e, f) Overlap between T cell responses to the three INSPs and IgG antibody responses to oxPTM-INS modified by ●OH (●OH-INS). Overall, 5.5%, 22.2% and 44.4% of patients showed concordant responses involving, simultaneously, CD4+, CD8+ and IgG towards oxPTM-INSP-3, Nt-INSP-4 and oxPTM-INSP-6, respectively. There is concordance between reactivity to oxPTM-INS and CD4+ and CD8+ reactivity for all three tested peptides (7/17, 8/18 and 8/18 for oxPTM-INSP-3, Nt-INSP-4 and oxPTM-INSP-6, respectively). An SI>1 was used for definition of a positive T cell response

We next performed hierarchical cluster analysis (Euclidean distance, Ward’s method) of patients, and of peptides. Hierarchical cluster analysis and PCA revealed association between responses to different oxPTM-INSPs and identify clustering of type 1 diabetes vs healthy control samples. We observed association between Nt-INSP-4 and oxPTM-INS-P6, ERGYYYTPKTR for CD4+ and CD8+. For IgG response, ERGYY-DOPA-TPKT is associated with insulin modified by ●OH (●OH-INS). We also observed clustering of response of type 1 diabetes samples using PCA of all responses, CD4+, CD8+ and IgG. We observed a cluster of 11 type 1 diabetes samples with principle component 1 (PC1)>0, while the rest clustered with healthy control samples with PC1<0 (ESM Fig. 10).

留言 (0)

沒有登入
gif