The pool of tRNA-derived fragments is altered in NOD mouse islet cells during insulitis

NOD mice constitute the best characterised animal model of type 1 diabetes [2, 24,25,26]. At 4 weeks of age, the islets of NOD mice display a normal morphology and function but they are then progressively invaded by immune cells, resulting in beta cell dysfunction and death and, starting from weeks 12–14, in the appearance of diabetes [27]. To assess the potential contribution of tRFs to the initial phases of type 1 diabetes, we isolated the RNA from islets of NOD mice and treated the samples to remove the most common modifications [28]. We then compared, by small RNA-seq, the level of tRFs in the islets of 4 weeks old NOD mice (control) and of 8 weeks old NOD mice, which are still normoglycaemic but display insulitis [ESM Fig. 1] [29]) (Fig. 1). This led to the identification of ~20,000 tRFs, 158 of which were upregulated and 178 downregulated in prediabetic mice (p<0.05, fold change ≥2) (Fig. 1a, b and ESM Table 3; GEO accession no. GSE242568). Among the upregulated tRFs, 76 were generated from mitochondrial tRNAs (mt-tRFs, indicated by squares in Fig. 1b), while only ten downregulated tRFs originated from mitochondrial tRNAs (indicated by red squares in Fig. 1a). The modulation of several tRFs was confirmed by qPCR (Fig. 1c). The PCR primers were specifically designed for the sequences indicated in ESM Table 2 and ESM Fig. 2. Small RNA-seq of human islet samples (GEO accession no. GSE256343) permitted identification of close homologues of these tRFs (ESM Table 2).

Fig. 1

Identification of tRFs displaying changes in their level in the islets of prediabetic NOD mice. (a) tRFs significantly downregulated in the islets of 8-week-old mice compared with the islets of 4-week-old mice. (b) tRFs significantly upregulated in islets of 8-week-old mice compared with the islets of 4-week-old mice. Squares, tRFs derived from mitochondrially encoded tRNAs; circles, cytoplasmic tRFs. Colour codes indicate fragments originating from the same isoacceptor tRNA. (c) Real-time PCR confirmation of the upregulation of selected tRFs in islets of 8-week-old mice. The data were normalised to the level of the small ncRNA miR-184–5p, which displays no changes under prediabetic conditions. The level of the fragments in 4-week-old mice has been set to 1 and the data shown are the means ± SD. n=3–7. *p<0.05, **p<0.01 (unpaired t test). FC, fold change

Proinflammatory cytokines modify the level of some tRNA fragments

Part of the changes in the tRF profile of prediabetic NOD mice may be triggered by exposure of islet cells to proinflammatory factors released by the invading immune cells. To determine the impact of proinflammatory cytokines, islet cells were treated with IL-1β, TNF-α and IFN-γ for 24 h, prior to tRF profiling. Prolonged exposure to these cytokines affected the level of 4974 tRFs (p<0.05, fold change ≥2), 179 of which displayed an FDR of <0.1. Among these, 50 were upregulated and 129 downregulated (ESM Table 4; GEO accession no. GSE242568).

As expected, part of the changes elicited by the cytokines overlapped with those observed in the islets of prediabetic NOD mice. Indeed, the level of mt-Met-CAT, mt-Ser-GCT-5′ and His-GTG-i were increased both in prediabetic NOD islet cells (Fig. 1c and ESM Table 3) and in response to cytokines (Fig. 2). However, at the time point investigated, many of the changes observed in prediabetic NOD mice were not reproduced by exposure to proinflammatory cytokines (Fig. 2). Treatment of human islets (see Human islet checklist in the ESM) with proinflammatory cytokines did not significantly affect the level of the measured tRFs (ESM Fig. 3).

Fig. 2

Identification of tRFs displaying changes in their levels in cytokine-treated islet cells. C57BL/6NRj mouse islet cells incubated with or without a mix of cytokines (IL-1β, IFN-γ and TNF-α) for 24 h before RNA collection. Real-time PCR validation of tRFs differentially expressed in cytokine-treated islet cells (Cyt) compared with control (CTRL). The data were normalised to the level of miR-7a and miR-375, which displays no changes in response to proinflammatory cytokines. The data are shown as fold changes vs CTRL and are presented as mean ± SD. n=3–6. *p<0.05 (paired t test). FC, fold change

IFN-α, a cytokine secreted by microbially infected cells, has been proposed to initiate innate immune responses and to contribute to type 1 diabetes development [30]. Thus, we analysed the impact of IFN-α on the selected mouse islet tRFs and found that only the level of mt-Met-CAT was significantly affected by this cytokine (ESM Fig. 4).

tRNA fragments can be transferred from T lymphocytes to beta cells via extracellular vesicles

tRFs can be produced endogenously but can also be released and transferred in EVs to other cells [31, 32]. Since T cells are known to release EVs [33], we next investigated if the increased levels observed in prediabetic NOD islets could result from a transfer from T cell EVs to beta cells. EVs produced by activated CD4+/CD25− T lymphocytes of 8-week-old female NOD mice displayed a mean diameter of ~150 nm and characteristic exosomal properties (ESM Fig. 5 and [4]). Small RNA profiling of these EV preparations (GEO accession no. GSE242568) revealed that, consistent with previous reports [34], tRFs represent 80% of the small non-coding RNA fraction (ESM Fig. 6a). Computational analysis identified 3518 unique tRF sequences in NOD T cell EVs (mean transcripts per million ≥0.5). Of note, most of the fragments present in the EVs originated from five isodecoder tRNAs (ESM Fig. 6b). Interestingly, many tRFs carried by EVs were upregulated in the islets of prediabetic NOD mice, pointing to a possible delivery of EV tRFs to beta cells (Fig. 3a and ESM Table 5).

Fig. 3

T cell EVs contain tRFs that are transferred to islet cells resulting in changes in the tRF pool. (a) Venn diagram showing the number of tRFs present in T cell EVs displaying changes in the islets of prediabetic NOD mice and in EV-treated islet cells. (b) Volcano plot indicating the changes in tRF level in islet cells after 24 h incubation with NOD mouse T cell EVs. Fragments above the horizontal dashed line display significant changes (padj<0.05). Vertical dashed lines indicate a fold change of ±2. (c) Significantly upregulated tRFs in EV-treated mouse islet cells. Selected tRFs that originate from the same isodecoder tRNA are labelled with the same colour. (d) Analysis by qPCR of the level of selected tRFs in islet cells after 24 h and 48 h incubation with EVs released by NOD mouse T cells. The data were normalised to the level of miR-7a and miR-375, which display no changes under these experimental conditions; values are shown as fold change vs CTRL and are presented as mean ± SD. n=3 or 6. *p<0.05, **p<0.01 (ratio paired t test). FC, fold change

To verify this hypothesis, islet cells were incubated in the presence of EVs released by NOD mouse T lymphocytes. RNA-seq analysis at the end of the treatment (GEO accession no. GSE242568) identified 57 tRFs that were increased in beta cells (padj<0.05) upon exposure to EVs (Fig. 3a, b). Almost half of the increased tRFs originated from the fragmentation of three isodecoder tRNAs: Phe-GAA; Glu-CTC; and His-GTG (Fig. 3c and ESM Fig. 6c). The increase in several tRFs was confirmed by qPCR analysis of islet cells incubated for 24 h or 48 h with lymphocyte EVs (Fig. 3c). Similar changes were also observed in human islet cells incubated in the presence of EVs released by human CD4+ T cells (ESM Fig. 7) [4].

To investigate whether tRFs can be transferred from immune cells to beta cells, we developed an RNA-tagging approach. The tagging technique involved the incubation of donor cells with 5′-EU, a nucleotide derivative incorporated in cellular RNAs [35,36,37]. The tagged RNA released in EVs can be recovered after biotinylation of the EU residues and purification on streptavidin-coated beads (ESM Fig. 8a). To verify the efficiency of RNA tagging, Jurkat T cells and murine CD4+/CD25− T lymphocytes were treated with EU for 24 h and 48 h, and the incorporation of the tagged nucleotide was confirmed by dot blot (ESM Fig. 8b). Quantitative PCR analysis of the RNA recovered on streptavidin beads revealed an efficient tagging of tRFs (ESM Fig. 8c). Moreover, the presence of tagged tRFs was also confirmed inside the EVs released by EU-treated T cells (ESM Fig. 8d).

Next, islet cells were cultured in the presence of EVs produced by NOD mouse T lymphocytes previously incubated with (EU-EVs) or without (CTRL-EVs) EU. RNA was then extracted from islet cells, biotinylated and pulled down with streptavidin beads. An EU-tagged spike-in oligonucleotide containing the sequence of C. elegans miR-238 (ESM Fig. 8e) was used as internal control for the biotinylation and pull-down steps. The analysis of the pulled-down RNA revealed that islet cells treated with EU-EVs contained several tRFs produced by NOD mouse T lymphocytes (Fig. 4a). The tRF mt-Gln-TTG was used as negative control, since it was expressed in islet cells but not in T cell EVs (ESM Fig. 8f).

Fig. 4

tRFs of CD4+/CD25− T cells transferred in vitro and in vivo to pancreatic beta cells. (a) Real-time PCR of islet cell tRFs pulled down on streptavidin beads upon incubation with EVs released by EU-treated T cells (EU-EVs) or untreated control T cells (CTRL-EVs). An EU-tagged C. elegans miR-238 mimic was spiked in cell extracts as internal control. *p<0.05, **p<0.01 (ratio paired t test). (b) EU-tagged RNAs from FAC-sorted beta cells of NOD.SCID mice injected with EU-tagged T cells (EU) or with saline solution (CTRL) were purified on streptavidin beads and analysed by qPCR. An EU-tagged oligonucleotide containing the sequence of C. elegans miR-238 was spiked in the samples and was used as internal control to normalise the data. Data are presented as mean ± SD. n=3–5. *p<0.05 (ratio paired t test)

Some tRNA fragments are transferred from T lymphocytes to beta cells during the early phases of type 1 diabetes

These in vitro findings were confirmed in vivo, using CD4+/CD25− T lymphocytes from NOD BDC2.5 mice that, upon inoculation into NOD.SCID mice, promote the appearance of diabetes within less than 10 days [26, 38,39,40,41] (ESM Fig. 9a). CD4+/CD25− T cells from NOD BDC2.5 mice were isolated and incubated with EU for 48 h. The cells were then inoculated into NOD.SCID mice by i.v. injection. We verified by dot blot that RNA molecules were still tagged 3 days after the removal of EU (ESM Fig. 9b) and that the injected T lymphocytes contained EU-tagged RNA (ESM Fig. 9c). Control mice were injected either with saline solution (SHAM) (CTRL), or with BDC2.5 mouse T cells not treated with EU (CTRL2). Two days after injection (the time taken for T cells to invade the islets [38]), animals were killed and beta cells sorted by FACS. At the time of islet collection, the animals were normoglycaemic (ESM Fig. 9d), suggesting that they were still in the initial phases of type 1 diabetes [26, 38,39,40,41]. The RNA isolated from purified beta cells was biotinylated and pulled down with streptavidin beads prior to RNA-seq profiling (ESM Fig. 9a). An EU-containing spike-in oligonucleotide was used as internal control for the biotinylation and the pull-down steps. As expected, the same amount of spike-in oligonucleotide was recovered in EU and CTRL samples (Fig. 4b). Bioinformatics analysis led to the identification of 141 unique tRF sequences with more reads in beta cells from NOD.SCID mice injected with EU-tagged T cells than in beta cells injected with saline solution (CTRL1) (ESM Table 5; GEO accession no. GSE242568). Similar results were obtained by comparing the RNAs pulled down from beta cells of NOD.SCID mice injected with EU-tagged T cells with those pulled down from beta cells of NOD.SCID mice injected with untagged T cells (CTRL2) (ESM Fig. 9e). The sequencing results for several tRFs were confirmed by qPCR (Fig. 4b). Interestingly, these tRFs were present in T cell EVs and were upregulated in islet cells of 8-week-old NOD mice (Fig. 1c and ESM Table 4) but not in islet cells of NOD.SCID mice (ESM Fig. 10), which do not develop diabetes. This confirms the hypothesis that the increased levels of these tRFs in islet cells under prediabetic conditions is caused by the transfer from the invading T cells rather than by the exposure of beta cells to proinflammatory cytokines.

Impact of transferred tRNA fragments on beta cell function

Five tRF candidates were selected for functional studies based on the following criteria: (1) their increase in NOD islet cells during insulitis; (2) their increase in beta cells after exposure to T cell EVs; and (3) their transfer from T cells to beta cells in vitro and/or in vivo, in a model of autoimmune diabetes.

Exposure to NOD T cell EVs induces beta cell death [4]. We investigated whether this effect was at least in part mediated by an increase in tRFs. For this purpose, the levels of candidate tRFs were increased by transfecting mouse islet cells with oligonucleotide mimics (ESM Fig. 11a) and by assessing beta cell death. Propidium iodide / Hoechst staining and immunofluorescence analysis with anti-insulin and anti-cleaved caspase-3 antibodies showed a rise in beta cell apoptosis when the levels of Phe-GAA-3ʹH, Ser-GCT-3′ or Gly-GCC-5ʹH were increased (Fig. 5a and ESM Fig. 11b). In contrast, overexpression of Lys-CTT-5′ and Phe-GAA-5′ did not affect beta cell survival (Fig. 5a). The effect of Gly-GCC-5’H on beta cell apoptosis could be partially explained by the inhibition of the anti-apoptotic protein Bcl-XL. Indeed, the overexpression of Gly-GCC-5ʹH triggered a 40% decrease in Bcl2l1 mRNA levels (Fig. 5b). This effect was specific, since overexpression of Ser-GCT-3′ had no effect on the expression of this anti-apoptotic gene (Fig. 5c).

Fig. 5

Modulation of the levels of selected tRFs affects beta cell apoptosis. (a) Mouse islet cells were transfected with different tRF mimics (as indicated) or with a scrambled control sequence (CTRL) for 48 h. Part of the cells were incubated for 24 h with a mix of proinflammatory cytokines (IL-1β, IFN-γ and TNF-α) (Cyt) prior to staining using antibodies against insulin and cleaved caspase-3 (CASP-3). Between 600 and 1000 cells per condition were counted and the percentage of cleaved caspase-3 positive beta cells was calculated. *p<0.05, **p<0.01 (one-way ANOVA), n=4 (Dunnett Correction). (b) MIN6 cells were transfected with an oligonucleotide containing the sequence of Gly-GCC-5′H (Gly-GCC) or with a scrambled sequence (CTRL) for 48 h and were subsequently incubated with or without IL-1β for 24 h. RNA was collected and Bcl2l1 expression was measured by qPCR. *p<0.05, **p<0.01 (one-way ANOVA), n=4 (Dunnett correction). (c) MIN6 cells were transfected with an oligonucleotide containing the sequence of Ser-GCT or with a scrambled sequence (CTRL). Two days later, RNA was collected and Bcl2l1 expression was measured by qPCR. (d, e) Mouse pancreatic islet cells were transfected with a control oligonucleotide inhibitor (CTRL) or with a tRF inhibitor as indicated. After 24 h, the cells were treated with T cell EVs and incubated for another 48 h. Islet cell death was assessed by scoring the cells displaying pycnotic nuclei upon Hoechst/propidium iodide staining (around 5000 cells were counted per condition). **p<0.01 (paired t test), n=5 (d); *p<0.05 (one-way ANOVA, Dunnett post hoc test), n=4 (e). (f) MIN6 cells were transfected with Ser-GCT-3′ inhibitor or a negative control (CTRL). After 24 h, the cells were incubated with or without T cell EVs for another 24 h. RNA was collected and Ccl2 expression was measured by qPCR. *p<0.05 (one-way ANOVA, Šidák correction), n=3. (a) Data are presented as median, with 25th and 75th percentile, and whiskers showing minimum and maximum. (b–f) Data are presented as mean ± SD

As expected, incubation of mouse islet cells with EVs released by NOD mouse T cells led to an increase in cell death (Fig. 5d). While inhibition of Ser-GCT-3′, Gly-GCC-5ʹH or Phe-GAA-3ʹH did not affect apoptosis under basal conditions (ESM Fig. 11d), the introduction of Gly-GCC-5ʹH antisense oligonucleotides into islet cells 24 h before exposure to T cell EVs prevented the increase in tRF levels (ESM Fig. 11c) and reduced cell death induced by EVs (Fig. 5e). Inhibition of Ser-GCT-3′ and Phe-GAA-3′H (ESM Fig. 11c) in the receiving beta cells was not sufficient to prevent apoptosis triggered by EVs (Fig. 5e). However, blockade of Ser-GCT-3′ was found to prevent the rise in Ccl2 expression levels induced by EV (Fig. 5f), suggesting the possible involvement of this tRF in the recruitment of immune cells.

To gain insight into the role of the tRFs transferred in beta cells, we performed bulk RNA-seq analysis in mouse islet cells overexpressing Lys-CTT-5′, Phe-GAA-5′ or Ser-GCT-3′ (GEO accession no. GSE242568). After 48 h overexpression of Lys-CTT-5′, 167 genes were upregulated (fold change >1.5, p<0.05) and 151 were downregulated (fold change >−1.5, p<0.05) (Fig. 6a and ESM Table 7). Gene ontology (GO) analysis of the downregulated genes revealed an enrichment for transcripts associated with immune functions (Fig. 7a), including those regulating immune effector processes and leucocyte-mediated immunity. In contrast, the upregulated genes were mainly related to hormone metabolic processes (Fig. 7b).

Fig. 6

Overexpression of Lys-CTT-5′, Phe-GAA-5′ or Ser-GCT-3′ affects islet cell gene expression. Volcano plots of the differentially expressed transcripts in Lys-CTT-5′ (a), Phe-GAA-5′ (b) or Ser-GCT-3′ (c) overexpressing islet cells compared with cells transfected with a scrambled oligonucleotide. Upregulated transcripts are shown in green and downregulated transcripts are shown in orange. FC, fold change

Fig. 7

GO analysis of the transcripts affected by the overexpression of Lys-CTT-5′, Phe-GAA-5′ or Ser-GCT-3′. (a, b) Top enriched GO terms for downregulated (a) and upregulated (b) genes in mouse islet cells overexpressing Lys-CTT-5′. (c, d) Top enriched GO terms for downregulated (c) and upregulated genes (d) in mouse islet cells overexpressing Phe-GAA-5′. (e) Top enriched GO terms for upregulated genes in mouse islet cells overexpressing Ser-GCT-3′. The size of the circles is proportional to the ratio of the observed vs the expected overlaps. p values correspond to hypergeometric tests (colour-coded)

Interestingly, GO analysis revealed that the upregulation of Phe-GAA-5′ also affects genes involved in the activation of the immune system. Indeed, many of the 334 genes differentially expressed in Phe-GAA-5′ transfected cells (Fig. 7c, d and ESM Table 8) are associated with inflammation and immune functions, such as the regulation of TNF production.

The analysis of mouse islet cells overexpressing Ser-GCT-3′ led to the identification of 152 upregulated and 158 downregulated genes (fold change >1.5, p<0.05) (Fig. 7e and ESM Table 9). GO analysis of the upregulated transcripts highlighted an enrichment for genes regulating Jun N-terminal kinase activity. Since the activation of this kinase is known to induce beta cell death [42, 43], this may contribute to the apoptosis observed after Ser-GCT-3′ overexpression.

Taken together, these observations indicate that the transfer of tRFs from CD4+/CD25− T lymphocytes to beta cells during insulitis modify the transcriptomic landscape of the insulin-secreting cells, making them more engaged in the regulation of the immune cells and rendering them more susceptible to the autoimmune reaction.