figure b

Introduction

Type 1 diabetes is an autoimmune disease that manifests in adaptive immune responses targeting the insulin-producing beta cells of the endocrine pancreas. However, alterations in innate immune responses also drive autoimmune pathogenesis, with involvement in both the initial break in tolerance and the later failure of regulation [1,2,3,4]. Many known type 1 diabetes genetic risk alleles encode genes expressed by antigen presenting cells or other innate immune cells [5]. Understanding how innate immune responses are altered in type 1 diabetes and other autoimmune diseases is important for elucidating pathogenic pathways that could be targeted with new treatments. Within the spectrum of innate immune responses, inflammatory responses can generally be categorised into type 1 IFN (IFN-1)- or IL-1 dominant [6,7,8,9,10]. IL-1 and IFN-1 responses can counter-regulate each other and are needed for responses to different types of pathogens: IFN-1 increases antigen presentation and is primarily involved in viral responsiveness, whereas IL-1 stimulates secretion of inflammatory cytokine, such as IL-6, and drives responses to bacterial infection [6, 11]. Both types of responses have been postulated to play a role in the pathogenesis of type 1 diabetes and other autoimmune diseases [12, 13].

Autoimmune diabetes-susceptible NOD mice display alterations in both IFN-1 and IL-1 responses. Higher IFN-1 responses are observed in young (2–4-week-old) NOD mice and blocking the IFN-α receptor (IFNAR) at this age inhibits diabetes [14, 15], suggesting that higher IFN-1 responsiveness may contributes to the initial loss of self-tolerance. However, we have shown that myeloid cells from older prediabetic NOD mice display impaired expression of IFN-1-induced genes but higher IL-1-associated gene expression [16].

Previous studies have not provided a clear picture of how innate immune responses are altered in human type 1 diabetes. Viral infections, and the associated IFN response, are a potential environmental trigger for type 1 diabetes in genetically susceptible individuals [17], and individuals at risk for developing type 1 diabetes have shown higher expression of IFN-1 response genes (IFN-stimulated genes [ISGs]) in peripheral blood mononuclear cells (PBMCs) without ex vivo stimulation. This IFN signature has been shown to correlate with the development of islet autoantibodies, but wanes before the development of overt disease [18, 19]. Furthermore, pancreatic islets from individuals with type 1 diabetes may display IFN hyper-responsiveness [20,21,22]. Increased IL-1 has also been associated with type 1 diabetes [23]. For example, monocytes from individuals with type 1 diabetes can make more IL-1 and IL-6 without stimulation [24]. Increased IL-1-associated responses have been observed at the time of diagnosis; these increases were extinguished with IL-1 receptor antagonist treatment [25, 26]. Although blocking IL-1 signals leads to clinical improvement in other autoimmune diseases, such as rheumatoid arthritis, and in type 2 diabetes [27, 28], two type 1 diabetes trials that blocking IL-1 signals with either anakinra or canakinumab failed to alter insulin secretion in individuals with new-onset type 1 diabetes [29]. Together, these studies show that, in different contexts, type 1 diabetes has been associated with innate immune responses dominated by both IFN-1 and IL-1. Less is known about innate immunity in individuals with established type 1 diabetes.

One potential limitation of many previous studies on innate immune changes associated with human autoimmunity is the lack of relevant innate stimulation that can increase the response range. Although some prior studies have assessed cytokine responses after innate stimulation of either PBMCs or purified cell populations, a mix of either increased IL-1- or IFN-1-associated responses were associated with type 1 diabetes [23, 30]. However, the process of cellular isolation can alter phenotypes and may eliminate the inter-cell crosstalk that occurs in vivo. In contrast, a whole blood assay (TruCulture) was used by the Milieu Intérieur project to standardise human immune stimulation directly, ex vivo; this has facilitated measurements of individual response variation, and categorisation of stimulus-specific immune responses [31, 32]. This assay has lower technical variation than PBMC-based assays and includes innate immune cells not present in PBMCs [33]. In this study, we compared monocyte gene expression after CpG stimulation in NOD and control mice, and measured gene expression and cytokine responses after ex vivo innate immune stimulation in 17 individuals with established type 1 diabetes and 17 age-matched control participants. A better understanding of innate immune responses in human type 1 diabetes and related mouse models could help identify the innate pathways that could be targeted therapeutically for successful disease modification.

Methods

Mouse CpG stimulation and monocyte gene expression analysis

Female 8–10-week-old NOD and control B6.g7 mice (Jackson Laboratories, Ellsworth, ME, USA), 3–4 mice/group, were injected intravenously with 170 μl PBS + 30 μl 1,2-Dioleoyloxy-3-trimethylammonium propane (DOTAP; Roche, Indianapolis, IN, USA), with and without 5 μg CpG-A (2216; Invivogen, San Diego, CA, USA). After 6 h, spleen cells were harvested and Ly6C+ monocytes were sorted. Extracted RNA was sequenced and analysed, as previously described [16, 34, 35]. Details and other analysis of the unstimulated mouse samples used in this study were previously published [16]. The experiments were not replicated, with each mouse acting as a biological replicate.

Human Study design

Samples from 17 individuals with type 1 diabetes and 17 healthy control participants matched for age ±1 year were collected, concurrently, over 6 months, usually as age-matched pairs that were collected within 1 week of each other (Table 1 and electronic supplementary material [ESM] Table 1). Individuals with type 1 diabetes had history of clinical diagnosis and presence of type 1 diabetes-associated autoantibodies, were diagnosed 3 months to 8 years prior to blood sample collection, and had good glucose control, as represented by HbA1c <63.9 mmol/mol (8%), although HbA1c at time of collection was not an exclusion criterion. Healthy control participants had no first-degree relatives with type 1 diabetes. Groups were not sex-matched but gene expression changes were not associated with sex (ESM Fig. 1). All participants were enrolled in the Benaroya Research Institute (BRI) Immune Mediated Disease Registry. All participants fasted overnight prior to the blood draw. HbA1c was measured under Clinical Laboratory Improvement Amendments (CLIA)-approved protocols at the University of Washington’s Northwest Lipid Metabolism and Diabetes Research Laboratories. Each sample was only measured once; repeat measurements were not taken from the same sample.

Table 1 Characteristics of human participants

TruCulture tube collection and quantification of cytokine and gene expression

Peripheral blood was drawn directly into TruCulture tubes (Myriad RBM, Austin, TX, USA) for stimulation following manufacturer’s instructions, and as previously described [32]. Tubes were pre-loaded with one of the following immune stimuli: Staphylococcal enterotoxin B (SEB; 0.4 μg/ml), IL-1β (25 ng/ml or 1760 IU/ml), lipopolysaccharide (LPS; 1 ng/ml), IFN-β (1000 IU/ml), polyinosinic-polycytidylic acid (polyIC; 20 μg/ml), with stimulus concentrations previously determined [31], or a no-stimulus null control. After blood draw and incubation, the supernatant and cells were separated for downstream assays of secreted proteins and mRNA, respectively.

Cytokines were quantified in TruCulture tube supernatants using the OptiMAP Luminex panel (Myriad RBM), following manufacturer protocols. IFN-α and IFN-β were quantified using Simoa assays (Quanterix, Billerica, MA, USA), as described previously [36, 37]. The limit of detection (LoD) was 0.6 fg/ml for IFN-α and 0.45 pg/ml for IFN-β.

Gene expression was quantified in stimulated whole blood samples using the nCounter Human Immunology kit v2 (NanoString, Seattle, WA, USA). Total mRNA was extracted from TruCulture cell pellets stabilised in Trizol LS (Sigma-Aldrich, St Louis, MO, USA), using the NucleoSpin 96 RNA tissue kit protocol (Macherey-Nagel, Düren, Germany), as previously described [32]. For each sample, 100 ng of total RNA was analysed.

Gene expression analysis

Data processing and normalisation

Normalisation of Nanostring nCounter data was performed as described previously [32], using nSolver (NanoString; v. 2.5) for analysis of quality and negative and positive controls. Differences in RNA input were normalised using six housekeeping genes (EEF1G, RPL19, PPIA, TBP, G6PD, POLR2A), selected using the geNorm method [38]. We estimated a lower LoD of 4 counts [39], and retained 520 genes with expression above this LoD in ≥1/3 of samples from any stimulation condition in both participant groups. Multi-dimensional scaling (MDS) of gene counts revealed batch differences associated with the sample quantification method; ComBat [40] in the sva R package [41] was used to remove this batch effect prior to downstream analyses.

Differential expression analysis

Differential gene expression was determined using limma-voom [42], accounting for non-independence of samples from the same participant using limma’s duplicateCorrelation. Between-group comparisons of the stimulation effect used the interaction term between disease status and stimulus. Multiple comparisons were adjusted using the Benjamini–Hochberg false discovery rate (FDR). Heatmaps were generated using ComplexHeatmap [43].

Gene set and signature score analyses

We determined differential expression of pre-defined gene sets using ROAST [44]. For human samples, we used gene sets from two sources; first, we used hallmark IFN response gene sets from the Molecular Signatures Database (MSigDB) [45]. For more refined stimulus-specific responses, we used genes that have increased expression after TruCulture tube stimulation with IFN-α, IFN-β, IFN-γ or IL-1β [32] and have an ortholog to mouse ISGs. For mouse samples, we used genes that respond to IFN-α stimulation in vivo in many different cell types [10] and genes specifically induced by IFN-β [32]. ESM Table 2 shows details of gene sets. Signature scores for each gene set are the sample-level means of the gene-level z scores of expression or log-fold change.

Statistics, visualisations and code

Samples were not randomised and the experimenters were not blinded. All statistical analyses were performed using R v. 3.5.2 [46], with visualisations using R base graphics and ggplot2 [47]. t tests were two sided and data are expressed as means (SEM), unless otherwise indicated. Specific tests and significance thresholds are listed in the text; where not noted, tests were two-tailed and p < 0.05 was considered significant.

Study approval

Animals were housed in specific pathogen–free conditions and handled according to guidelines provided by the Animal Care and Use Committee of the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) at the National Institutes of Health (NIH). All human samples were prospectively obtained under written informed consent. All study protocols were approved by the BRI Institutional Review Board (IRB).

Results

Higher CpG-stimulated IFN-1 responses in monocytes from diabetes-susceptible NOD mice

To determine if the acute IFN-1 response is altered in NOD mice, we analysed gene expression in monocytes from NOD and B6.g7 mice 6 h after in vivo stimulation with CpG. Gene set analysis showed that the expression of IFN-1-induced genes was lower in unstimulated monocytes from NOD vs diabetes-resistant B6.g7 mice, as we have shown previously [16], but these same IFN-1-stimulated genes were more strongly upregulated in NOD monocytes after stimulation with CpG (Fig. 1a, b). A more restricted set of IFN-response genes that are orthologs to human IFN-response genes also showed this pattern of lower unstimulated but higher stimulated expression in NOD mice (Fig. 1c). Therefore, monocytes from NOD mice with chronic autoimmunity display lower baseline levels and higher induction of IFN-1 response genes with CpG stimulation, as compared with those from control B6.g7 mice.

Fig. 1
figure 1

Lower baseline and higher stimulated IFN-1-induced gene expression in NOD monocytes compared with control mice. NOD and control B6.g7 mice were injected with either PBS or CpG and, 6 h later, monocytes were sorted from isolated spleen cells for RNA sequencing (n = 3). (a, b) Gene set enrichment analysis of IFN-α-induced genes, with decreased expression in NOD mice after null stimulation (a; p = 0.016) and increased expression after CpG stimulation (b; p = 0.0013). Genes are rank-ordered by the t-statistic comparing NOD with control mice; genes that are more highly expressed in NOD are on the positive side of the t-statistic scale. Position on the x-axis indicates rank. Significance values derived using limma ROAST. (c) Heatmap showing expression of genes in the Milieu Intérieur-derived IFN-β signature in sorted monocytes from diabetes-susceptible NOD or control B6.g7 mice after either control (PBS) or CpG stimulation

Innate immune stimuli induce changes in gene expression and upregulate inflammatory cytokine levels in human whole blood samples

TruCulture ex vivo whole blood stimulation system was used to measure immune response differences between individuals with type 1 diabetes and healthy control participants. Five stimuli were chosen: SEB to induce T cell responses; an IFN-1 response induced directly via IFNAR with IFN-β; an IFN-1 response induced indirectly via pattern recognition receptors, Toll-like receptor 3 (TLR3) and retinoic acid-inducible gene I (RIG-I), using polyIC; LPS to induce both type 1 and type 2 IFN and IL-1 responses; IL-1β [31]. Blood was incubated with the indicated stimuli, and cell pellets and supernatants collected to measure gene expression and cytokine protein levels, respectively, as previously described (Fig. 2a) [31, 32]. As expected, these innate stimuli induced cytokine expression in samples from both type 1 diabetic and healthy control participants, but no significant differences were observed between the type 1 diabetes and healthy control group in the analytes examined, after adjusting for multiple tests (ESM Fig. 2). Five-hundred and twenty genes from the Nanostring human immunology panel with measurable expression in at least one condition were included. Gene expression induced by distinct stimuli showed stimulus-specific patterns (Fig. 2b) and clustered separately from each other by principal component analysis (PCA; ESM Fig. 3). As previously described, stimulation with IL-1β elicited fewer gene expression changes as compared with other stimuli, and samples stimulated with polyIC and IFN-β elicited similar gene expression changes as compared with other stimuli, as captured by PC2 (Fig. 2 and ESM Fig. 3) [32]. Therefore, the expected upregulation of cytokine and gene expression with stimulation was observed.

Fig. 2
figure 2

Innate and adaptive stimuli induce gene expression changes in samples from both type 1 diabetes (T1D) and healthy control (HC) participants. (a) Experimental setup for collection of ex vivo-stimulated whole blood samples. (b) Heatmap of stimuli-induced genes, including all participants and all stimuli. For each stimulation, the ten most significantly upregulated genes with a fold change (FC) ≥2 are shown

Lower IFN-1 response gene expression in unstimulated samples

Without specific immune stimulation (null stimulation), gene expression in the type 1 diabetes and healthy control group were similar, with only three genes showing significant differential expression (Fig. 3a, ESM Table 3). Specific modules of gene expression that can separate responses to key cytokines, including IFN-γ, IFN-β and IL-1β, were previously defined by the Milieu Intérieur Consortium, who used the approach in healthy donors [32]. The genes in these modules are the ones that are most distinct from induction by other cytokine stimuli. To better understand baseline, unstimulated human ISG expression, we applied these cytokine-specific gene expression modules to the TruCulture null samples. Using this method, we found that, without specific stimulation, the IFN-1 gene score was significantly lower in type 1 diabetes (adjusted p = 0.036), but this was not the case with the type 2 IFN or IL-1 gene scores (adjusted p values: 0.93 and 0.12, respectively), as compared with healthy control participants without stimulation (Fig. 3b–d). Individual Milieu Intérieur-derived IFN-1 signature genes are shown in the heatmap in Fig. 3e. Although the effect size without stimulation is small, this suggests that, as observed in NOD mice (Fig. 1a, c) [16], individuals with type 1 diabetes may have reduced tonic IFN-1 signalling.

Fig. 3
figure 3

Unstimulated samples show lower type 1 diabetes(T1D)-associated IFN-β-driven gene expression. (a) Volcano plot showing differential gene expression between the T1D and healthy control (HC) groups in the null stimulation tube. Horizontal dashed line indicates adjusted (adj) p = 0.05; vertical dashed line indicates log fold change equals 0. (bd) Gene signature scores in null-stimulated samples for Milieu Intérieur-derived IFN-β- (b), IFN-γ- (c) and IL-1β- (d) specific gene sets. For IFN-β, p = 0.036; for IFN-γ, p = 0.93; for IL-1β, p = 0.12. (e) Heatmap of Milieu Intérieur-derived IFN-β genes in the null-stimulated samples. *p < 0.05

SEB-driven and IFN-γ-specific gene expression does not differ in type 1 diabetes

Adaptive immune responses, including IFN-γ-producing T helper 1 (Th1) cells, have been shown to be critical for type 1 diabetes pathogenesis [48, 49]. Thus, we examined gene expression changes after exposure to the adaptive stimulus SEB and analysed IFN-γ-driven changes in gene expression. Although SEB stimulation induced robust gene expression changes (Fig. 2b), no significant differences in gene expression were observed between the type 1 diabetes and healthy control group (Fig. 4a). The Milieu Intérieur-derived IFN-γ-induced gene expression module was applied to the two most relevant stimuli, namely SEB and LPS, but no significant differences were observed between type 1 diabetes and healthy control participants in this IFN-γ signature (Fig. 4b). Similarly, at the protein level, IFN-γ was not differentially induced in type 1 diabetes vs healthy control participants (Fig. 4c). Other cytokines strongly induced by SEB stimulation (IL-17, granulocyte-macrophage colony-stimulating factor [GM-CSF] and IL-2) were not differentially upregulated in type 1 diabetes as compared with healthy control participants (ESM Fig. 2). Therefore, no significant type 1 diabetes-associated differences were observed in the adaptive immune responses when induced by SEB or IFN-γ.

Fig. 4
figure 4

SEB and IFN-γ responses are similar in type 1 diabetes (T1D) and healthy control (HC) participants. (a) Volcano plot comparing SEB-induced gene expression in T1D vs HC. Horizontal dashed line indicates adjusted (adj) p = 0.05; vertical dashed line indicates log fold change equals 0. (b) Milieu Intérieur-derived IFN-γ-specific gene signature scores for samples from T1D or HC after stimulation with SEB or LPS. For SEB, p = 0.19; for LPS, p = 0.26. (c) IFN-γ protein levels quantified from supernatants of T1D or HC samples stimulated with SEB or LPS. For SEB, p = 0.63; for LPS, p = 0.99. Values shown on log10 scale

IL-1 induced gene expression is not significantly altered in type 1 diabetes

IL-1 is a key innate immune signal that may contribute to type 1 diabetes pathogenesis [16, 26, 50]. Surprisingly, IL-1β stimulation resulted in only two genes being differentially induced when comparing type 1 diabetes and the healthy control group (Fig. 5a), though IL-1β stimulation overall induced smaller changes in gene expression as compared with other stimuli (Fig. 2b) [32].

Fig. 5
figure 5

IL-1-associated responses are similar in type 1 diabetes (T1D) and healthy control (HC) participants. (a, b) Volcano plots showing genes with significantly differential induction between T1D and HC samples after stimulation with IL-1β (a) or LPS (b). Horizontal dashed lines indicate adjusted (adj) p = 0.05; vertical dashed lines indicate log fold change equals 0. (c) Milieu Intérieur-derived IL-1β-specific gene signature scores for samples from T1D or HC participants after stimulation with IL-1β or LPS. For IL-1β, p = 0.076; for LPS, p = 0.21

LPS also induces IL-1-specific gene expression, and type 1 diabetes-associated changes in LPS-stimulated responses were observed (LPS vs null), with ten genes being more significantly induced by LPS in the type 1 diabetes vs healthy control group, and two genes (NCF4 and S100A9) being less induced (Fig. 5b, ESM Table 3). However, LPS also can induce IFN-1 responses. For example, LPS-stimulated complement factor B (CFB) is higher in individuals with type 1 diabetes, but stimulation with IFN-1 has been previously shown to induce higher CFB expression than stimulation with IL-1 or IFN-γ [51]. To further understand IL-1-driven changes in gene expression and separate them from other innate immune responses, the Milieu Intérieur-derived IL-1-specific gene module was applied to both IL-1 and LPS stimuli. LPS stimulation gave the highest level of IL-1-induced gene expression (ESM Table 3) but no significant differences in the IL-1 signature were observed between the type 1 diabetes and healthy control group (Fig. 5c). Together these data show that the IL-1 response, induced either by LPS or IL-1 itself, is not significantly altered in type 1 diabetes vs healthy participants. In addition, some of the LPS-induced changes in gene expression in type 1 diabetes may be due to the effects of increased IFN-1 signals.

Preferential induction of IFN-1 response genes in type 1 diabetes as compared with healthy control participants

Next, we compared differences between individuals with type 1 diabetes and healthy control participants following stimulation with polyIC and IFN-β, both strong inducers of the IFN-1 response. We found 21 genes that were more highly upregulated in individuals with type 1 diabetes as compared with control participants after IFN-β stimulation (Fig. 6a), and 14 genes after polyIC stimulation (Fig. 6b). No genes showed significantly higher induction in healthy control participants after either stimulus. Upon subsetting the analysis to genes in the Milieu Intérieur-derived IFN-β module to assess IFN-1-dependent gene expression, we found a significant increase in the IFN-1-driven gene signature in type 1 diabetes, as compared with healthy control participants, with both IFN-β and polyIC stimulation (Fig. 6c). A heatmap of the genes in the Milieu Intérieur-derived IFN-β module after IFN-β stimulation shows higher expression of IFN-1-induced genes in individuals with type 1 diabetes (Fig. 6d, ESM Fig. 4a). A heatmap showing the pattern of key signature scores in healthy control and type 1 diabetic individuals from this cohort is presented in ESM Fig. 4b. Enrichment of genes differentially induced in type 1 diabetes after IFN-β or polyIC stimulation was then tested using the broader MSigDB hallmark set of IFN-1 response genes and a similar pattern was observed, with upregulation of a significant proportion of the IFN-response genes in individuals with type 1 diabetes, as compared with healthy control participants (Fig. 6e).

Fig. 6
figure 6

Higher IFN-1-induced responses in type 1 diabetes (T1D) samples stimulated with IFN-β or polyIC. (a, b) Volcano plots showing genes with significantly differential induction between T1D and healthy control (HC) samples after stimulation with IFN-β (a) or polyIC (b). Horizontal dashed lines indicate adjusted (adj) p = 0.05; vertical dashed lines indicate log fold change equals 0. (c) Milieu Intérieur-derived IFN-β-specific gene signature scores for samples from T1D or HC particpants after stimulation with IFN-β and polyIC. For IFN-β, p = 0.0022; for polyIC, p = 0.0056. (d) Heatmap of IFN-β-induced expression of Milieu Intérieur-derived IFN-β-response genes in T1D and HC samples, with T1D and HC clustered separately. (e) Proportion of hallmark IFN-⍺ response genes preferentially induced in T1D vs HC. (f, g) IFN-α (f) and IFN-β (g) protein levels in supernatants from T1D or HC samples after polyIC stimulation, as measured by Quanterix Simoa assay. For IFN-α, p = 0.20; for IFN-β, p = 0.70. Values shown on a log10 scale. **p < 0.01 vs HC

Differential cellular composition could contribute to the cytokine levels or gene expression changes observed in whole blood. We compared major cell populations from cell blood counts (CBCs) between type 1 diabetes and healthy control participants. In our cohort, monocyte levels were higher in type 1 diabetes than in the healthy control group (ESM Fig. 5a); however, levels of monocytes did not significantly correlate with IFN responses (r2 = 0.025, p = 0.39), and the differences in IFN response gene expression were not explained by cell levels (ESM Fig. 5b, c). Further, neither demographic nor clinical variables were associated with stimulation-induced IFN gene expression, either in individual genes (ESM Fig. 1) or gene signature scores (ESM Fig. 6).

Specific genes from the Milieu Intérieur-derived IFN-β module may be relevant for type 1 diabetes pathogenesis. For example, IFIH1, an ISG linked to type 1 diabetes, shows higher expression after IFN-β, polyIC and LPS stimulation in samples from individuals with type 1 diabetes (Fig. 6a). XCR1, which specifically marks the cross-presenting conventional type I dendritic cells (cDC1s) that are critical for inducing both cytotoxic CD8 T cells and Th1 responses, is upregulated after polyIC stimulation in type 1 diabetes vs healthy control participants (ESM Fig. 7). In addition, IL12RΒ1 is induced more in type 1 diabetes compared with the healthy control group with all three IFN-1-inducing stimuli (LPS, polyIC and IFN-β). As IL-12 signalling is critical for Th1 response induction, this may contribute to the Th1 bias described in type 1 diabetes [48, 52].

To assess whether the observed alterations in the IFN-1 response were due to differential induction of IFN-1 proteins, we measured IFN-α and IFN-β protein levels after polyIC stimulation by digital ELISA [36]. No significant differences in cytokine protein levels were observed between the type 1 diabetes and healthy control groups (Fig. 6f, g). Therefore, the observed differences may instead be due to alterations in IFN-1 signalling. Consistent with this hypothesis, several genes downstream of the IFNAR showed higher IFN-β-induced upregulation in type 1 diabetes, including JAK2, STAT1 and STAT2 (Fig. 6a and ESM Fig. 7). Together, these data show an altered IFN-1 response in individuals with type 1 diabetes, with a higher dynamic range than control participants due to both lower baseline expression and higher stimulated expression of IFN-1 genes.

Discussion

In this study, we show that measuring gene expression after whole blood ex vivo stimulation allows identification of significant alterations in innate immune responsiveness in type 1 diabetes that are not detected without stimulation. Increased IFN-1 signalling, observed after IFN-β, polyIC and LPS stimulation, was the response that was most strongly associated with type 1 diabetes. Although gene expression induced by type 1 and type 2 IFNs have considerable overlap, the underlying pathways that lead to induction of both types of IFN and the downstream biological effects can differ greatly. Therefore, separating the two types of IFN responses is important for understanding the pathogenesis of autoimmune diabetes. The increased IFN-1 responsiveness, together with a lack of change in IFN-γ signalling, suggests that individuals with type 1 diabetes are hyperresponsive to IFN-1, but not to type 2 IFN.

IL-1-induced responses after IL-1β or LPS stimulation were similar in individuals with established type 1 diabetes and healthy control participants. Although prior studies showed an IL-1-associated signature at the onset of type 1 diabetes, our study did not show significant differences in IL-1-induced responses between type 1 diabetes and healthy control participants. This may be due to differences in experimental design or to unique phenotypic characteristics present in the peri-diagnostic period [53, 54]. It is also possible that there is, in fact, a difference in IL-1-induced responses, but that our study was not sufficiently powered to detect them. Indeed, some IL-1-induced differences were observed and, although not significant, the effect was in the expected direction. It is noteworthy, however, that blocking IL-1 signals in new-onset type 1 diabetes showed no clinical effect, suggesting that this pathway may not be essential for diabetes pathogenesis [29]. The adaptive immune responses tested here using SEB were not significantly different between groups, perhaps due to an absence of antigen-driven autoreactive responses. Nevertheless, these results do show that individuals with type 1 diabetes do not have a substantive difference in responsiveness to global adaptive immune cell activation as compared with control participants.

Some, but not all aspects of innate immune response alterations in human type 1 diabetes are reflected in autoimmune NOD mice. ISG expression is lower at baseline in monocytes from prediabetic NOD mice with chronic autoimmunity, but higher after acute IFN-inducing stimulation. Interestingly, this phenotype of reduced tonic IFN-1 signals was stronger in mouse monocytes and monocyte-derived dendritic cells (moDCs) as compared with conventional dendritic cells (cDCs) and plasmacytoid dendritic cells (pDCs), which are in low numbers in the blood [16]. Our earlier work focusing on gene expression in specific dendritic cell subsets after CpG stimulation did not show increased ISG expression, but these populations are rare in whole blood and may have differential autoimmune-driven changes in IFN-1 responses [55]. Future studies isolating specific cell populations may identify type 1 diabetes-associated alterations in human cDC IFN-1 responses. Although NOD mice may model type 1 diabetes-associated IFN responses, our data suggest that IL-1 responses may differ between mice and humans because prediabetic NOD mice with ongoing autoimmunity displayed higher baseline IL-1 responses [16]. Increased IL-1 may be transient, being present at the time of onset but fading in human type 1 diabetes over time [23, 56]. The human participants in our study had established type 1 diabetes, with time from diagnosis ranging from 2 to 8 years. Further work is needed to resolve this discrepancy.

The ability to observe significant alterations with a relatively small number of individuals illustrates the advantage of using whole blood ex vivo stimulation with gene signature analysis, but further data with an independent cohort would help strengthen the conclusions of this study. Although the study size limits conclusions regarding participant subsets, heterogeneity in ISG responses was observed within both the type 1 diabetes and healthy control group. Assessing gene expression after ex vivo stimulation in larger participant cohorts would allow the integration of genetic susceptibility with induced gene expression phenotypes and may potentially help to identify type 1 diabetes immune subtypes for differential therapeutic interventions. Including first-degree relatives of individuals with type 1 diabetes in future studies could further facilitate understanding of how genetic risk contributes to altered innate immune response. Identification of the cell types involved and the regulatory mechanisms contributing to the differential IFN-1 response in type 1 diabetes could also enable more precise therapeutic targeting of this pathway. Because of the more robust gene expression changes occurring with stimulation, its use in the context of current interventional trials could facilitate novel biomarker identification.

We propose a model incorporating what we now know about the complex nature of IFN-1 responses in type 1 diabetes. Early in disease pathogenesis (before onset or even autoantibody seroconversion), baseline IFN-1 response genes are increased [18, 19], likely due to low-level increases in IFN-α and IFN-β expression. This chronic IFN exposure over time may suppress the tonic, unstimulated response, but prime a stronger acute response, possibly via epigenetic changes [57]. This could explain why unstimulated ISG expression is high in individuals at risk for type 1 diabetes [19, 26] and lower in established type 1 diabetes [58]. However, with strong acute stimulation, increased ISG expression is observed even after disease onset [59]. Additional work is needed to validate and better understand the changes in IFN response during the course of human type 1 diabetes. Using the TruCulture system at earlier disease stages, from before autoantibody seroconversion through to diabetes onset, could be informative for defining IL-1 and IFN-1 responses. Overall, this work shows the value in measuring both direct ex vivo and stimulated states for characterising how innate immunity is altered in autoimmune diseases such as type 1 diabetes.