Proinsulin‐specific T‐cell responses correlate with estimated c‐peptide and predict partial remission duration in type 1 diabetes

Abstract Objective Type 1 diabetes (T1D) is an autoimmune disorder in which autoreactive T cells destroy insulin‐producing β‐cells. Interventions that preserve β‐cell function represent a fundamental therapeutic goal in T1D and biomarkers that predict and monitor β‐cell function, and changes in islet autoantigenic signatures are needed. As proinsulin and neoantigens derived from proinsulin peptides (hybrid insulin peptides, HIPs) are important T1D autoantigens, we analysed peripheral blood CD4+ T‐cell autoantigen‐specific proliferative responses and their relationship to estimated β‐cell function. Methods We recruited 72 people with and 42 without T1D, including 17 pre‐diabetic islet antibody‐positive and 9 antibody‐negative first‐degree relatives and 16 unrelated healthy controls with T1D‐risk HLA types. We estimated C‐peptide level at 3‐month intervals for 2 years post‐diagnosis and measured CD4+ T‐cell proliferation to proinsulin epitopes and HIPs using an optimised bioassay. Results We show that CD4+ T‐cell proliferation to any islet peptide and to multiple epitopes were significantly more frequent in pre‐diabetic islet antibody‐positive siblings and participants with T1D ≤ 3 months of duration, than in participants with T1D > 3 months or healthy controls. Among participants with T1D and first‐degree relatives, CD4+ T‐cell proliferation occurred most frequently in response to proinsulin33‐63 (full‐length C‐peptide). Proinsulin33‐63‐specific responses were associated with HLA‐DR3‐DQ2 and/or HLA‐DR4/DQ8. In children with T1D, proinsulin33‐63‐specific T‐cell proliferation positively associated with concurrent estimated C‐peptide and predicted survival in honeymoon. Conclusion CD4+ T‐cell proliferative responses to proinsulin‐containing autoantigens are common before and immediately after diagnosis of T1D but decline thereafter. Proinsulin33‐63‐specific CD4+ T‐cell response is a novel marker of estimated residual endogenous β‐cell function and predicts a better 2‐year disease outcome.


INTRODUCTION
Type 1 diabetes (T1D) is a chronic, incurable autoimmune disorder in which insulin-producing β cells are destroyed by islet-infiltrating T cells. 1,2 Before and after clinical onset, pancreatic islets undergo β-cell loss. Despite recent advances in disease management tools, the medical and psychological burden on children and their families is high, and diabetic complications including hypoglycaemia and microvascular disease are common. There has been progress towards understanding the genetic, environmental and immunologic basis for TID 3 ; however, prevention and/or cure of this condition remain elusive. The clear benefits of preserving endogenous insulin secretion include a longer honeymoon period, better metabolic control and fewer complications. 4,5 Moreover, improved metabolic control early in the course of T1D may be protective by producing a 'metabolic memory'. 4,6 Therefore, interventions that preserve β-cell function in T1D represent a fundamental therapeutic goal. Indeed, many patients have detectable β-cell function for many years after diagnosis, 7 and immune intervention may be capable of preserving a significant proportion of β-cells. 8 The rate of decline of β-cell function is heterogeneous, 9 and the factors that preserve βcell function are poorly understood. Furthermore, the response to immune interventions also varies amongst individuals. Robust immune biomarkers to predict and monitor β-cell function are needed, to assess the short-term impact of immune interventions, with the aim of improving longerterm metabolic outcome in T1D. Islet-reactive CD4 + and CD8 + T cells play a central role in the pathogenesis of β-cell destruction. For CD8 + T cells, the balance between β-cell-specific stem cell memory and exhaustion contributes to the rate of disease progression and may be skewed towards exhaustion by immunotherapies such as teplizumab. For CD4 + T cells, the frequency of peripheral blood (PB) T helper 2-like CD25 + CD127 hi T cells was recently associated with improved β-cell function. 10 Favorable responses to agents such as Alefacept are also seen in individuals with a higher frequency of CD25 + CD127 hi T cells, suggesting that CD127 hi cells maintain an anti-inflammatory environment that supports improved β-cell function and response to immunotherapy. 10 Islet-infiltrating CD4 + T cells recognise epitopes derived from the proinsulin C-peptide, and from neoantigens such as hybrid insulin peptides (HIPs) formed in β-cells by the fusion of C-peptide fragments to peptides of chromogranin A, islet amyloid polypeptide or C-peptide itself. 1,11,12 Several lines of evidence indicate that insulin or its precursor proinsulin is a primary autoantigenic target of T cells in T1D. 13,14 In humans, antibodies to insulin are the first marker of pre-diabetes in genetically at-risk infants followed from birth, and precede antibodies specific for other islet antigens. 15 The insulin gene locus is a T1D susceptibility gene, 16,17 and its regulatory VNTR regions play an essential role in insulin-specific self-tolerance. [16][17][18][19] Insulin expression is restricted to β-cells of pancreatic islets, whereas other T1D autoantigens are expressed more widely. Postmortem analyses of pancreatic tissue of T1D show that insulitis is usually detected in islets with insulin-positive β-cells. 20,21 Accordingly, proinsulin peptides have been successfully used in T-cell assays [22][23][24][25] and the 'C19-A3' epitope employed in several early phase immunotherapy trials. 26,27 Proinsulin is also a primary β-cell antigen target for autoreactive T cells in spontaneous autoimmune diabetes in the non-obese diabetic (NOD) mouse model. 28 Studies in transgenic mice suggest that other β-cell antigens may subsequently be recognised as a result of epitope spreading. 29 We hypothesised that circulating CD4 + T cells in patients with recent-onset T1D and AB+ siblings at high risk would respond to specific islet auto-antigenic peptides. Since epitopes of proinsulin and hybrid islet peptides containing proinsulin sequences have been described, 1,11,12,30 we measured autoreactive CD4 + T-cell proliferation towards a panel of proinsulin peptides. We used our optimised fit-for-purpose CFSE-based T-cell proliferation assay, 31 to evaluate the frequency of positive responses to each epitope, and to determine the relationship between the level of proliferation and the clinical features of the patients with T1D. We found that CD4 + T-cell responses to proinsulin  were the most frequent of the islet peptides tested, and that the level of proinsulin 33-63 -specific CD4 + T-cell proliferation was positively associated with β-cell function and predicted survival in remission, suggesting that proinsulin 33-63 -specific CD4 + T-cell proliferation is a novel immune biomarker of predicted β-cell function.

DISCUSSION
Characterisation of islet antigen-specific CD4 + and CD8 + T cells during the prodrome and after onset of T1D is crucial, in order to understand their role in disease pathogenesis, to relate T-cell responses to treatment outcomes, to identify subjects potentially suited to immunotherapy and to characterise T-cell responses to antigen-specific immunotherapies. However, a robust T-cell biomarker that tracks immunological correlates of β-cell function has remained an unrealised goal. Using a previously optimised assay, we show that proinsulin epitope-specific proliferative responses were most frequent towards PI  in children and adults with T1D and in individuals at risk, and that PI 33-63 -specific responses were higher in individuals with HLA class II T1D-associated risk alleles than in individuals without these risk alleles. While high levels of PI epitope-specific PB CD4 + T-cell proliferation were evident in AB− and AB+ at-risk FDR and within 3 months of T1D diagnosis relative to healthy controls carrying HLA risk alleles, they declined in individuals with T1D for longer than 3 months. These data indicate that PI 33-63 -specific CD4 + T-cell proliferation is a marker of active islet autoimmunity, which is demonstrable in FDR at genetic risk of T1D even before the development of islet AB. Furthermore, after T1D diagnosis, PI 33-63 -specific PB CD4 + T-cell proliferation correlates with estimated C-peptide and predicts survival in partial remission. Thus, in children diagnosed with T1D, high PI 33-63 -specific CD4 + T-cell responses identify residual predicted endogenous β-cell mass and predict a better 2year survival in partial remission.
Previous studies also identified antigen-specific CD4 + T cells in PB 25,[32][33][34][35] and in the islets of organ donors with T1D, including multiple islet antigens and HIPs. 11,30 This study extends that work and demonstrates that proinsulin immune reactivity is best captured by assays incorporating the PI 33-63 full-length C-peptide sequence, which contains multiple epitopes, 25,32,36 potentially increasing the pool of antigen-reactive T cells. Shorter epitopes within PI 33-63 are presented by several HLA-DR and HLA-DQ molecules, thus generating T-cell responses from individuals with varied HLA types. 25 Unlike PI-specific immunogenicity, which was common across the groups tested, T-cell autoreactivity to HIP neoepitopes 12 was more heterogeneous. Emerging evidence suggests that the generation of neoepitopes is linked to β-cell stress and activates T cells, amplifying the autoreactive repertoire. However, the presence of HIPs within the islets does not appear sufficient to cause disease, as HIPs have been found in islets of donors without T1D. 37 Heterogeneity in the immune response to HIPs may relate to different levels of β-cell stress, HLA genotype, 38 and the timing and extent of HIP formation in vivo. In NOD mice, PB T cells recognising HIPs were detected with increasing frequency as they progressed towards diabetes. 39 While we did not undertake longitudinal studies, HIP-specific T-cell reactivity correlated with PI-specific reactivity, indicating that epitope spreading may be related to the magnitude of the autoimmune response. In participants with T1D ≤ 3 months, AB-negative and AB-positive FDR, the increased islet-specific CD4 + T-cell autoimmunity presumably indicates ongoing priming of immune cells and epitope spreading to multiple β-cell antigens prior to, or soon after the clinical manifestation of disease. Future longitudinal studies could map changes and potential epitope spreading in autoantigenspecific CD4 + T-cell responses before and after the onset of T1D.
We showed that the level of PI 33-63 -specific Tcell proliferation was associated with concurrent β-cell function and predicted survival in partial remission. Furthermore, proliferation of PI 33-63specific T cells correlated with proliferation to HIPs. If these islet-specific T cells contribute to βcell destruction, this observation is unexpected. The association with concomitant β-cell function may reflect a larger β-cell mass and thus antigen availability to sustain antigen-specific CD4 + T-cell memory, and potential for clonal expansion upon peptide re-exposure in vitro. 40 However, the association of PI 33-63 -specific proliferating T cells with future honeymoon duration suggests they may have regulatory function, with capacity to delay disease progression. Proliferating cells produce IL-2, which is needed to support CD25 + regulatory T cells, including Foxp3 + CD25 hi CD127 lo and CD25 + CD127 hi T cells. The frequency of CD25 + CD127 hi T cells was also recently shown to predict survival in honeymoon. 41 Polymorphisms in multiple genes in the IL-2 pathway can predispose to T1D, [42][43][44] and defects in IL-2 receptor signalling reduce Foxp3 expression in CD25 + regulatory T cells in T1D patients, prompting trials of both low-dose IL-2 and Treg cell therapy. [45][46][47] Autoimmune β-cell damage occurs before the onset of clinically apparent disease, and PI 33-63 CDI may be a useful biomarker of autoimmunity in people at genetic risk for T1D, even in the absence of islet AB. We show that almost all FDR responded to at least one of the proinsulin epitopes. Since many of these individuals are unlikely to develop T1D, this observation also supports the hypothesis that the autoantigenspecific CD4 + T cells may have regulatory function. In a recent analysis of PB cytokine responses to a partially overlapping set of islet epitopes (PI 40-52 , hEL:A-chain and hEGGG:IAPP2) using ELISPOT in FDR, the IFNg/IL-10 ratio in response to certain HIPs correlated with clinical parameters of progression to T1D, including islet AB titres and evidence of glucose intolerance. 48 Given that HIPspecific proliferative responses correlated with PI 33-63 -specific responses and that a PI 33-63 CDI ≥ 3 predicted survival in honeymoon after T1D onset, it will be of interest to determine whether PI  CDI also stratifies the risk of developing T1D, adding new insights into the classification of presymptomatic T1D. 49 Further work is also needed to characterise the phenotype and function of the proliferating PI 33-63 -specific CD4 + T cells, and to map changes in progressors and nonprogressors to T1D, and in individuals with sustained remission in clinical trials of immunotherapy.
Although the presence of islet antibodies is associated with the risk of developing T1D, 15 their predictive role after disease onset is not well understood. An inverse correlation between humoral and cellular islet autoimmunity has been previously described. 50 PI-specific CDI did not correlate with the number of autoantibody specificities in AB-positive siblings. Furthermore, both AB-negative and AB-positive siblings had PIspecific CDI, confirming that the presence of AB was not necessary for islet peptide-specific T-cell proliferative responses.
A key strength of this study is that we quantified PI-specific responses in PB using an optimised, fitfor-purpose assay, validated for intra-assay and inter-assay repeatability. 31 However, our study has limitations. While the at-risk and T1D participant groups comprised adults and children, the healthy control group was significantly older. We limited the pool of peptides tested to those derived from proinsulin, based on previous publications. However, other published epitopes 51,52 within the proinsulin A chain, B chain and other islet antigens such as GAD-65 or ZnT8 could also be informative, particularly as many AB-positive siblings developed antibodies against GAD-65 and ZnT8 (Table 2). Using ELISPOT assays to measure IFN-γ and IL-10 CD4 + T-cell responses to a variety of islet epitopes, a previous study found fewer adults than children with recent-onset T1D made IFN-γ responses. 24 While we found no relationship between CDI and age, we did not assess cytokine production in the current studies. Furthermore, although our cohort had sufficient power to determine the relationships of CDI to clinical features and HLA, it was not large enough to fully explore the immunological heterogeneity of T1D. Validation of these findings in additional cohorts would add value and may reveal interactions between variables.
In summary, CD4 + T-cell proliferative responses to proinsulin-containing autoantigens are frequently detectable prior to and immediately after T1D onset, followed by a decline. Proinsulin 33-63 -specific CD4 + T-cell response is a novel marker of current and future endogenous β-cell function.

Subjects
T1D was defined according to the criteria from the American Diabetes Association. 53 Healthy controls were locally collected participants or Australian Bone Marrow Donor Registry participants carrying either HLA-DQ2 or HLA-DQ8 alleles but with no personal or family history of T1D and without T1D-related antibodies. The absence of AB against insulin, glutamic acid decarboxylase 65 (GAD65) and islet antigen insulinoma-associated protein-2 (IA-2) was confirmed in all healthy controls and ABnegative FDR.

Estimate of β-cell function
In paediatric patients with T1D, we estimated endogenous C-peptide levels according to a previously reported method. 3 The estimated C-peptide model incorporates age, gender, BMI-Z score, HbA1c, time since diagnosis and insulin dose. Based on this model, an estimated C-peptide of 0.4 correlates with a stimulated C-peptide of 0.3 nmol L −1 . We estimated C-peptide levels at 3-month intervals over 2 years. Paediatric patients were defined as being in partial remission (honeymoon) if their estimated Cpeptide measured ≥ 0.4.

Cell preparation, CFSE Staining and T-cell stimulation
Peripheral blood mononuclear cells (PBMC) were freshly isolated and labelled with carboxyfluorescein diacetate succinimidyl ester (CFSE, CellTrace ™ CFSE Cell Proliferation Kit for flow cytometry, Invitrogen, Thermo Fisher Scientific, Waltham, Massachusetts, USA) as described. 31,54 CFSE-stained cells (2 × 10 6 PBMC mL −1 , 200 μL per well) were cultured for 7 days with medium alone (negative control), peptides or with positive controls, tetanus toxoid and/or anti-CD3 to obtain a cell division index (CDI). 54 Three replicates were tested for each experiment. The CDI is the ratio of CD4 + T cells that proliferated in response to antigen, relative to cells that proliferated in the absence of antigen. The CDI was calculated based on a fixed number of 5000 CD4 + CFSE (undivided) cells using the following formula: CDI ¼ number of divided CD4 þ T cells per 5, 000 CD4 þ T cells in CFSE ðundividedÞ from ' with antigen ' group number of divided CD4 þ T cells per 5, 000 CD4 þ T cells in CFSE ðundividedÞ from ' without antigen ' group As the assay was run in triplicate, the mean of unstimulated proliferation was used to calculate the CDI. A CDI of ≥ 3 was determined to represent the threshold for the positive control responses. Data from assays in which the CDI for the positive controls, tetanus toxoid and anti-CD3, did not exceed 3.0 were excluded from the analysis.

Statistical analysis
Pairwise comparison between categorical variables was conducted with the Chi-square test, or the Fisher's exact test if one or more cells in the contingency table had an expected frequency of ≤ 5 using Graphpad Prism (Version 7, San Diego, CA, USA) and R Statistical Software (Version 3.5.3, Foundation for Statistical Computing, Vienna, Austria). The Kruskal-Wallis test was used for multiple comparisons, followed by post-hoc Dunn's test, and P-values were adjusted with the Benjamini-Hochberg method. Comparisons between group data were made using the paired two-tailed t-test. Statistical significance was defined as P < 0.05. The Kaplan-Meier survival analysis was performed with the R package Survival (https://cran.r-projec t.org/web/packages/survival/index.html).
Patients were assigned to two groups high score (≥ median PI 33-63 CDI of 3.15) and low score (< median PI 33-63 CDI of 3.15). A Kaplan-Meier survival using the default setting extracted a P-value for a honeymoon period of up to 30 months.