Type 1 diabetes mellitus (T1D) is the result of a slow progressive multistep autoimmune destruction of pancreatic insulin-producing cells (1–3). Early studies in European and derived populations have shown that susceptibility to T1D is strongly associated with HLA-DR3, -DQ2 and HLA-DR4, -DQ8 haplotypes, while protection is associated with HLA-DR2, -DQ6 haplotypes (4, 5). Subsequent studies have shown that within these haplotypes certain HLA-DQB1, and -DRB1 alleles and residues confer susceptibility, while others provide resistance to the disease (6–11). For instance, several studies have shown that differences of a few amino acid residues in HLA-DRB1*04 alleles are per se sufficient to modify the risk of developing T1D conferred by the high-risk HLA-DQB1*03:02 allele (12–15). Indeed, the spectrum of HLA-DQB1 and -DRB1 association, which constitutes the major component of T1D risk, is more complex than initially outlined and can be primarily grouped into very high risk, intermediate and very low risk haplotypes which, in turn, are the result of the structure of the peptide-binding pockets (15, 16).

In recent years, the intricacy and the molecular nature of immunogenic T cell epitopes of pancreatic autoantigens have been further elucidated (17). It is generally believed that T cell responses against autoantigens may become less evident with the progressive destruction of the islets of Langerhans (18, 19). Moreover, the spreading of T cell reactivity to other not primarily involved regions, of preproinsulin (PPI) or glutamate decarboxylase 65 (GAD65) autoantigens or to other pancreatic proteins/neoantigens generated from post-translational modifications may be an important confounding factor complicating the identification of pathogenically relevant T cells (19–22). Secondly, autoreactive immune responses are not completely disease-specific, since T cell reactivity against autoantigenic proteins is detected also in control subjects who carry disease-associated HLA-DR and -DQ molecules (18, 23, 24).

A useful tool to identify immunodominant T cell epitopes from pancreatic autoantigens is the use of humanized HLA transgenic mice (25). Indeed, numerous PPI and GAD65 T cell epitopes have been identified in triple HLA, human CD4 (hCD4) and IA knock-out transgenic mice (25–29) and confirmed in T1D patients (30–32). The majority of these studies have been conducted using triple HLA-DRB1*04:01, hCD4, IA knock-out mice (25–29), in HLA-DQ8 transgenic mice (33), whilst literature regarding studies on patients carrying different HLA-DR4 subtypes, such as the HLA-DRB1*04:05-DQA1*03:01-DQB1*03:02 haplotype is lacking. Interestingly, this is the most frequent HLA-DR4 haplotype found in Sardinian T1D patients (13, 16, 34).

Thus, we tested in children with recent-onset T1D and HLA-matched healthy controls from Sardinia a set of PPI and GAD65 peptides derived from previous studies in both HLA-DR4 and -DQ8 transgenic mice and humans (26–28, 33, 35).

Here, we report the capacity of PPI and GAD65 derived peptides obtained from humans and from HLA-DR4 and -DQ8 transgenic mice (21, 25–33, 35) to induce spontaneous T cell proliferation responses in PBMCs from recent-onset T1D pediatric patients.

In terms of global reactivity, our results indicate a higher response against peptides of the PPI protein and less reactivity against GAD65 peptides.

Analysis of proliferative responses to PPI showed that reactivity is mainly directed against PPI_1-18, and PPI_7-19 peptides of the leader sequence. These 2 peptides are included in the immunodominant peptide PPI_1-24 described as HLA-DQ8-restricted in HLA-DQ8 transgenic mice (33). However, we found responses against these 2 peptides in both HLA-DR4, -DQ8 and -DR3, -DQ2 positive T1D patients compared to HLA-matched controls ( Tables S1, S2 ; Figure 1A ).

Since we found significant responses for both peptides, we hypothesize that the sequence responsible for reactivity in T1D patients is included in the PPI_7-19 shared sequence LLPLLALLALWG. Prevalent reactivity of T1D PBMCs against the leader sequence may indicate that precursors of insulin, normally confined to the ER of the β-cells, are involved in targeting T cell autoreactivity to the islets. Interestingly, the same region of PPI is recognized by CD8+ T cells from recent-onset T1D patients (38, 39), which may suggest that the leader sequence of PPI is targeted by both CD4+ and CD8+ autoreactive T cells. In addition to HLA-DQ8, the leader sequence of PPI (LALLALWGPDPAAAFV) may be presented also by HLA-DRB1*04:01 as previously shown in HLA-DRB1*04:01 transgenic mice (27). Therefore, in analogy to other autoantigens such as GAD65 (26, 28), the leader sequence of PPI may show overlapping peptide sequences that are presented by both HLA-DR and -DQ molecules.

Also for PPI_31–49 peptide, including the B chain peptide PPI_B9-23, reported immunodominant in both humans and NOD mice (40–42), significant proliferative responses were observed ( Table S1 ; Figure 1A ).

Interestingly, CD8+ T cells in long-standing patients have been shown to recognize the B-chain peptide PPI_33–42 (38), that is also targeted by both human and NOD mice CD4+ T cells (41, 42).

Among the 16 GAD65 peptides tested, only GAD65_271-285 and GAD65_431-450 peptides induced significant proliferative responses in T1D patients. These patients carried both HLA-DR4 and HLA-DR3 haplotypes ( Tables S3, S4 ; Figure 2 ). These findings are in line with previous data showing that GAD65_271-285 was found immunodominant in HLA-DRB1*04:01 transgenic mice and humans (26), whilst a peptide similar to GAD65_431-450 (GAD65_431-445) is recognized by HLA-DQ8 molecules in HLA-DQ8 transgenic mice (28). Interestingly, GAD65_431-450 has been also reported as an inducer of IL-13 in CD4+ T cell lines derived from HLA-DR3, -DQ2 homozygous individuals (43).

Noteworthy, the GAD65 region 245-450 that includes both GAD65_271-285 and GAD65_431-450 has been identified as the main target of the earlier anti-GAD65 response in pre-diabetic, healthy high-risk subjects and early onset T1D patients (44–46). Finally, in genetically at-risk T1D patients, autoantibodies against GAD65 are found mainly in HLA-DR3, -DQ2 and less commonly in HLA-DR4, -DQ8 patients (19, 47).

Our findings demonstrate the crucial importance of validating in T1D patients the prediction of immunogenic epitopes. Indeed, even epitopes that are immunodominant in HLA class II transgenic mice may be not immunogenic in humans. As an example, GAD65_206-220 peptide is an immunogenic/immunodominant T cell epitope in HLA-DQ8 transgenic mice (28, 48) and in NOD mice after immunization with murine GAD65 (49) and, as such, could be considered a good pathogenic candidate for T1D in mice and humans. Instead, T cell receptor transgenic mice specific for GAD65_206-220 have been demonstrated to be protected against T1D development (50). In accordance, our data show that this peptide elicits proliferative responses not only in T1D patients but also in controls ( Tables S3, S4 ; Figure 2 ), making a pathogenic role highly unlikely also in humans.

Our study has several limitations that we tried to mitigate. The large number of potential T cell epitopes of pancreatic autoantigenic proteins that need to be tested challenges the relatively small amount of blood that can be drawn from children. To circumvent this limitation, we restricted the number of epitopes by prior analysis of the T cell responses in transgenic mice carrying human HLA haplotypes. Another limitation is the T cell spreading in long standing T1D patients (19–21), which is why we focused on early onset T1D pediatric patients recruited within three to four days after diagnosis. Discrepancies in the interpretation of T cell responses from PBMCs, the usage of diverse experimental methodologies among laboratories and the variability in the frequencies of HLA-DR4 subtypes in different T1D populations, may lead to different results. In this work we tried to reduce these variables taking advantage of the genetically homogenous Sardinian population, less prone to both genetic and clinical confounds that are present in more cosmopolitan collections (13, 51). Finally, a further possible limitation of this work was the significant age difference between T1D patients and controls. However, the 18 controls were healthy blood donors who were selected for a negative history of autoimmune diseases, thus making spontaneous not specific responses less likely than what could have been derived from age matched random pediatric controls attending the outpatient clinic for different disorders.

In summary, we identified for both PPI and GAD65 specific proliferative responses addressed to the PPI leader sequence and to GAD65_271-285, GAD65_431-450 in T1D patients. We suggest that these could be primary epitopes involved in the early phase of the clinical onset of T1D. Finally, these data hint towards the possibility that the leader sequence of PPI is targeted by both CD8+ and CD4+ T cells, with HLA class I and class II presenting slightly different overlapping peptide epitopes and thus inducing a stronger pathogenic immune response involved in the β-cell destruction.

These results may have implications in the design of immunogenic PPI and GAD65 peptides for peptide-based immunotherapy. Finally, this work could pave the way to test these epitopes in early stage T1D patients (stage 1 or 2) who have islet autoantibodies but no overt clinical diabetes requiring treatment with exogenous insulin.

The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding author.

The studies involving human participants were reviewed and approved by Comitato Etico Indipendente AOUCA. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.

RJ and MC conceptualized, designed and led the research. DB, ES, SM contributed to data interpretation and drafting the manuscript. RR provided HLA typing. MMan, MMar and CC collected and processed PBMC. CR and MR provided T1D patients. FC critically revised the manuscript. MC is the guarantor of this study. All authors contributed to the article and approved the submitted version.