Found in Translation: Novel Insights Into Type 1 Diabetes and β-Cell Biology

Although essential features in the natural history of type 1 diabetes are generally understood (1), many key questions remain. In particular, despite decades of research, the precise events that cause immune tolerance mechanisms to fail, triggering autoimmunity, remain unclear (2). One theory that has attracted considerable recent attention is that β-cell “stress” can trigger or potentiate autoimmunity by generating neoantigens (3–5), for example, by posttranslational modifications, hybrid peptide formation, or protein splicing (6). In this issue, Thomaidou et al. (7) significantly expand our knowledge of two other potential sources of “neo-antigens,” namely, the products of translation from noncanonical start sites, and alternatively spliced transcripts. Using a combination of ribosomal profiling (8) and long read RNA sequencing, they show that β-cells can use at least 15,014 unique translation start sites within transcripts from the 5,529 genes whose expression they detected. Ribosome profiling was initially developed by Ingolia, Weissman, and colleagues (8) and is based on next-generation sequencing of the ∼30 base pairs fragments of mRNA that are protected from nuclease digestion by bound ribosomes. Combining this approach with pharmacological treatment to accumulate ribosomes at the sites of initiation allows accurate inference of translational reading frames. This powerful technique is now widely used to study many aspects of translation but had not previously been applied to human β-cells.

Surprisingly, only 17% of the start sites defined by Thomaidou et al. mapped to the reference for that transcript, thus revealing a wide range of previously unsuspected open reading frames (ORFs) and expanding the “translatome” to at least 22,670 distinct proteins. Unique ORFs included the products of upstream initiation (including some that would terminate before the annotated coding sequence and others that would extend it), translation in alternative reading frames, and downstream initiation, including some ORFs originating from annotated 3′ untranslated regions (7). Cytokine treatment resulted in initiation at 4,427 unique sites. As expected, many of these were from genes in pathways related to an inflammatory response, but importantly, some novel ORFs were in secretory granule proteins such as CHGA and so might be sources of “neoantigens,” though this remains to be confirmed.

A second surprising finding was that 20% of the start sites did not map to the reference transcriptome. While some of the discrepancies might possibly have arisen from sequencing or bioinformatics errors, long read sequencing revealed 6,892 new transcripts, suggesting that the β-cell transcriptome is also much more diverse than previously appreciated. Most (74%) of the new transcripts spanned a splicing junction, while others extended the annotated 3' untranslated region and so might represent novel polyadenylation sites. Importantly, 1,745 of the initiation sites that Thomaidou et al. (7) detected were located on the novel transcripts, including 499 that were specifically detected upon cytokine treatment, and so might also give rise to neoantigens. Preproinsulin is a major target of autoreactive T cells (9), and it is therefore of interest that Thomaidou et al. identified five novel transcripts from the INS gene. Four of these were of relatively low abundance, but the fifth was a novel 3' extension that comprised 3% of the total INS transcripts and would allow synthesis of a longer version of the defective ribosomal product that the authors have previously shown contains at least one major epitope recognized by diabetogenic T cells (10). Bioinformatics analyses indicate that the longer variant may contain additional T-cell epitopes, further highlighting the potential importance of the INS-IGF2 locus as a source of neoantigens (11).

Beyond autoimmunity, the study by Thomaidou et al. may also provide new insight into basic β-cell biology. A considerable fraction (7%) of the novel start sites they found mapped to noncoding RNA, and particularly to long noncoding RNA, thereby revealing previously unsuspected polypeptides within them (7). This finding is consistent with many other related studies. Indeed, with the increasing use of ribosome profiling it has become apparent that there are a multitude of short ORF-encoded peptides and small proteins in eukaryotic cells whose existence and true significance are only gradually being revealed (12,13). Such “micropeptides” may play important cellular roles, for example, in regulating translation (14) and modulating the function of larger interaction partners (15). At present the functional significance of micropeptides in β-cell biology remains largely unknown. However, Thomaidou et al. have now cast new light on the “dark proteome” of the β-cell, which is a critical step toward gaining a complete understanding of its function in health and disease.

Although providing important new information, the study by Thomaidou et al. had several limitations. First, for technical reasons, it focused exclusively on EndoC-βH1 cells (16). This line is generally accepted as the best available model for primary human β-cells and shows many functional similarities to isolated human islets. However, recent analyses have highlighted some key differences between EndoC-βH1 cells and islets (17,18), which likely reflect its fetal/embryonic origin and/or its transformed state. Indeed, ORFs for several key β-cell proteins such as IAPP and IGRP were not detected (7), indicating that the expanded proteome still remains incomplete. Thus, as acknowledged by the authors, this model exhibits an incomplete phenotype compared with bona fide β-cells but, nonetheless, represents a logical first step toward defining their translatome. Using primary tissue will be a critical next step, but the lower cell numbers and cellular heterogeneity of cadaveric islets present experimental challenges that will need to be overcome first. Pluripotent stem cell differentiation into β-cells is an alternative model that could provide large numbers of β-cells to further probe the translatome. However, meta-analysis of the translatomes of all available human β-cell models will likely provide the most thorough insights.

A second limitation was the lack of independent validation for the majority of the newly described ORFs. While comprehensive validation was beyond the scope of this study, it leaves open the possibility that some ORFs may represent bioinformatics artifacts. Thus, it will be important for future follow-up studies to be conducted to validate the newly identified polypeptides and determine their roles.