Alterations of circulating innate lymphoid cells in recent‐onset type 1 diabetes

Abstract

Objectives

Type 1 diabetes (T1D) is driven by autoreactive T lymphocytes with contributions from innate immune compartments, yet early innate immune alterations associated with autoimmune amplification remain incompletely defined. We aimed at investigating whether circulating innate lymphoid cells (ILCs) contribute to these processes.

Methods

Circulating ILCs from patients with recent‐onset T1D (n = 12) and age‐matched healthy donors were analysed by spectral flow cytometry and targeted gene‐expression profiling of purified ILC progenitors (ILCP). Metabolic and inflammatory modulation of ILC activation was assessed in vitro. Antigen presentation to autologous memory CD4 T cells was functionally evaluated. Findings were examined across disease progression in NOD mice.

Results

Total ILC frequencies were comparable between groups; however, ILC1 and ILCP from T1D patients displayed increased activation markers (CD69, HLA‐DR). ILCP exhibited reduced expression of migratory (CCR7, CCR6) and regulatory‐associated genes (ICOS, CD58). Glycolytic inhibition selectively attenuated ILCP activation, supporting a metabolic contribution. IL‐1β, and to a lesser extent IFNγ, enhanced antigen presentation‐associated phenotypes in human circulating ILCs. Upon inflammatory pre‐activation, insulin peptide‐loaded circulating ILCs upregulated MHC class II–associated markers and induced antigen‐specific activation of autologous CD4⁺ memory T cells. NOD mice recapitulated ILC activation and MHC class II expression patterns during diabetes progression.

Conclusion

Circulating ILCs in recent‐onset T1D display a pre‐activated, metabolically regulated phenotype with functional antigen‐presenting capacity, supporting a potential role in autoimmune amplification. The findings of this exploratory study provide translational insight into innate immune dysregulation associated with recent‐onset T1D.

Circulating innate lymphoid cells (ILC) in patients with recent‐onset type 1 diabetes (T1D) show a chronic activation and a primed phenotype, with overexpression of HLA Class II (ILC1 and ILCP) and dysregulation of costimulatory and migratory mRNA transcripts. Glucose and cytokines (IL‐1β) mediate circulating ILC activation and HLA Class II expression. Circulating ILC from patients with T1D act as non‐classical APC activating antigen‐specific autologous memory CD4 T cells. NOD mouse ILC recapitulate human ILC alterations at T1D onset.

Introduction

Type 1 diabetes (T1D) is a chronic, metabolic and inflammatory disorder of autoimmune origin. It has a prevalence of 8,4 million people affected worldwide, according to epidemiological data from 2021. ¹ T1D is caused by the selective destruction of pancreatic islet β cells by β‐cell specific autoreactive lymphocytes. Like in other autoimmune disorders, genetic, environmental and immunological factors contribute to T1D development. In individuals with a basis of genetic predisposition under the influence of environmental triggers, immune tolerance mechanisms may break leading to self‐reactive responses against β‐cell antigens. As in other autoimmune disorders, autoreactive responses start long before clinical diagnosis, complicating the identification of early immune dysregulations. The first stage, characterised by β‐cell autoimmunity without symptoms and with normoglycemia, is followed by a second stage with dysglycaemia, ultimately progressing to a third stage where β‐cell autoimmunity and dysglycaemia are accompanied by clinical symptoms. Self‐reactive CD8⁺ and CD4⁺ T cells progressively infiltrate pancreatic islets and selectively destroy β cells in an antigen‐specific manner, progressively reducing the β‐cell mass and hampering the production of insulin, necessary to keep the homeostasis of glucose in our body. ² Yet, the innate immune events driving T1D at each stage of the natural history remain incompletely defined.

Innate lymphoid cells (ILC) are a group of lymphoid cells with innate features of relatively recent identified. ³ , ⁴ ILC cells lack the peptide‐specific rearranged T‐cell or B‐cell receptors (TCR, BCR) and do not undergo clonal selection. Instead, they are activated by microenvironmental signals, such as cytokines, innate immunity ligands, hormones or neuropeptides that bind to cell surface specific receptors. When activated, ILC are poised to rapidly respond producing high amounts of cytokines. Functionally, they have been classified in Groups 1, 2 and 3 functionally equivalent to their adaptive counterparts, the T helper (Th) 1, Th2 and Th17‐Th22 cells, based on lineage‐associated transcription factors and cytokine profile. Besides classical cytotoxic natural killer (NK) cells, the helper counterpart of group 1 ILC (ILC1) provide protection against pathogens producing cytokines, such as interferon (IFN)γ and tumour necrosis factor (TNF)α. ⁵ Group 2 ILC (ILC2) provide type 2 cytokines, such as interleukin (IL)‐5 and IL‐13. ⁶ They are involved in responses to parasites and helminths, but they also have a pathogenic role in asthma and allergy. Group 3 ILC is a more heterogeneous subgroup with a high degree of plasticity. ⁷ ILC3 include a subset of lymphoid‐tissue inducer‐like cells that express the C‐C chemokine receptor type 6 (CCR6), produce lymphotoxin and contribute to the formation of secondary and tertiary lymphoid structures. Recently, it has been observed that human ILC3 cells present in blood can give rise to ILC from all three subsets and have been renamed as innate lymphoid cell progenitors (ILCP). ⁸ Inherently to their first‐line defensive functions, ILC preferentially locate in barriers and tissues, ⁹ although they are also present in secondary lymphoid organs, such as spleen, tonsils and lymph nodes. Moreover, ILC are found in circulation although in minor proportions. Through their early cytokine production and interactions with cells of the adaptive immune system, ILCs play an essential role in pathogen defence, the maintenance of tissue homeostasis, and the regulation of immunological tolerance. The exact features that are shared or distinct in tissue‐infiltrating and circulating ILCs are only started to be described. Despite a clear influence of the tissue microenvironment in the phenotype and plasticity of ILC, ¹⁰ circulating ILCs serve as a snapshoot of underlying autoimmunity. ¹¹ Circulating ILC show dysregulated frequencies and subset distribution but far from a generic pattern in autoimmunity, disease‐specific differences have been observed. ¹¹

In T1D there is still limited, albeit growing, evidence regarding the role of ILCs. An accumulation and activation pattern of ILC1s has been observed in patients with T1D in the intestinal mucosa. ¹² Notably, studies have highlighted the relationship between microbiota, ILC3s and pancreatic islets, reporting that the microbiota can induce IL‐22 expression in intestinal ILC3s. Interestingly, this cytokine appears to act distally on β cells, promoting the expression of the antimicrobial peptide β‐defensin 14, which may exert a protective effect in disease by bridging innate immunity and the autoimmune response. ¹³ Therefore, it is possible that other costimulatory or antimicrobial molecules altered in intestinal ILC3 in the NOD model modulate susceptibility to T1D. Supporting this notion, a study in which the microbiota was depleted using antibiotic treatment reported a reduced frequency of IL‐2–producing ILC3s and regulatory T cells (Tregs) in the small intestine of diabetic NOD mice. ¹⁴ , ¹⁵ Of note, glucose metabolism serves as a key activator and functional regulator of ILC2 and ILC3. ¹⁶ , ¹⁷ Given that hyperglycaemia is a predominant feature in Stage 2 and 3 individuals with T1D, it is expected that phenotypic and functional alterations occur in ILCs in patients with recent‐onset T1D.

ILCs are crucial in providing fast cytokine release at first‐line defence but also interact with other immune system cells to guide adaptive immune responses. ¹⁸ A feature of ILCs, particularly ILC3 and ILC2, is their potential to express MHC Class II molecules and, under certain circumstances, to act as antigen‐presenting cells (APCs), functioning as non‐classical APCs. ¹⁹ , ²⁰ , ²¹ Models with ILC3‐specific MHC‐II deletion have demonstrated that these cells act as non‐classical APCs that limit the expansion of CD4⁺ T cells reactive to the microbiota under homeostatic conditions. ²² MHC‐II⁺ ILC3 select for and maintain microbiota‐specific Tregs. ²³ Despite the physiological immunoregulatory function of intestinal ILCs, several studies have shown that, under inflammatory conditions, dysregulation of the ILC3 subset at the gastrointestinal, systemic or local level can contribute to autoimmunity. ²⁴ In celiac disease, ILCP contribute to systemic inflammation and can present gliadin‐derived peptides, thereby activating specific CD4 T‐cell responses. Moreover, ILCP from patients respond to gliadin by producing pro‐inflammatory cytokines. ⁸ In experimental autoimmune encephalomyelitis (EAE), the murine model of multiple sclerosis, Grigg et al. ²⁰ reported an increase in ILC3 within the central nervous system, exhibiting antigen‐presenting characteristics, such as costimulatory molecule expression, proximity to infiltrating T cells, the ability to process and present myelin to autoreactive T cells in vitro, and a pathogenic role.

Together, these findings highlight the potential for ILC dysregulation to drive or enhance autoimmune processes in humans. To date, whether circulating ILCs in humans are functionally altered in early overt T1D, and whether this contributes to autoimmune amplification, remains unknown. Here, we investigated the phenotype, activation state and functional potential of circulating ILCs in recent‐onset T1D patients, integrating functional assays and murine validation to explore translational relevance.

Results

Circulating ILCs from recent‐onset T1D patients exhibit a distinct activation profile dominated by ILC1 and ILCP activation features

To investigate ILC alterations in T1D, we characterised circulating ILC subsets by spectral flow cytometry in recent‐onset patients compared with age‐matched healthy individuals (Supplementary figure 1). Although total ILC frequencies within the CD45⁺ leukocyte compartment were comparable between groups, the distribution of ILC subsets revealed disease‐associated changes. We observed a modest but significant reduction in circulating ILC1 frequencies in T1D patients compared to healthy donors (Figure 1) (P = 0.0386). ILC1 and ILCP displayed the most pronounced phenotypic deviations. Across the whole ILC compartment, T1D patients showed elevated expression of CD69, indicating enhanced baseline activation or tissue‐emigration signals (Figure 2). This trend was most evident in ILCP, although statistical significance was only reached for total ILCs. More notably, HLA‐DR expression was significantly increased in total ILCs, ILC1 and ILCP, but not in ILC2, pointing to a selective activation in the ILC1/ILCP axis (Figure 2). A positive correlation between HLA‐DR expression in ILC1 and ILCP was observed in the T1D subgroup (r = 0.7531, P = 0.0063, Spearman correlation test), suggesting a coordinated activation within these phenotypically related subsets (Figure 2c). This association was absent in healthy donors, indicating that this coordination may be disease‐induced. We did not find statistically significant correlations between ILC activation or HLA‐DR overexpression and age, gender, months after T1D diagnosis (data not shown).

Figure 1.

Frequency and distribution of innate lymphoid cell (ILC) subpopulations in peripheral blood of type 1 diabetes (T1D) patients and healthy donors. (a) Representative dot plot showing the identification and subclassification of ILC. To identify ILCs, we gated on viable lineage⁻ leukocytes (CD3, CD4, CD8, CD14, CD15, CD16, CD19, CD20, CD33, CD34, CD203c and FcεRI) and CD127⁺ (IL‐7Rα⁺) cells (left). ILC1 were defined as double‐negative for CD117 and CD294 (CRTh2); ILC2 as CD294⁺; and ILCP as CD294⁻ CD117⁺ (right). (b) histogram showing the percentage of total ILCs relative to total viable CD45⁺ leukocytes in peripheral blood from healthy donors (HD) (n = 23) and in patients with T1D (T1D) (n = 12). (c) Histogram showing the percentage of ILC1, ILC2 and ILCP within the total ILC population in peripheral blood from healthy donors (HD) (n = 23) and patients (n = 12). Mean and standard deviation is shown. Individual values are also shown as dots, grey dots for healthy donors and blue dots for patients with T1D. Student's t‐test was used for total ILC percentage comparison. ILC1, ILC2 and ILC3 frequencies were compared between donors and patients using a mixed‐effects two‐way ANOVA with Sidak's multiple comparisons test. *P < 0.05.

Figure 2.

Increased expression of activation markers in innate lymphoid cell (ILC) subpopulations from type 1 diabetes (T1D) patients. (a) Bar graphs showing the percentage of total ILC, ILC1, ILC2 or ILCP populations positive for each marker (CD69, HLA‐DR), compared with the fluorescence‐minus‐one (FMO) control in PBMC samples from T1D patients (n = 12) and age‐matched healthy donors (n = 12). Mean and standard deviation is shown. Individual values are also shown as dots, grey dots for healthy donors and blue dots for patients with T1D. *P < 0.05, **P < 0.01 ****P < 0.0001. (b) Representative contour plots depicting the expression of HLA‐DR in ILC1 (left) and ILCP (right) from healthy donors (HD) and patients with T1D (T1D). (c) Correlation between the percentage of HLA‐DR⁺ ILC1 cells and the percentage of HLA‐DR⁺ ILCP cells in all individuals (left), in the donor cohort (middle), and in the T1D cohort (right). The r‐value corresponds to Spearman's correlation coefficient. Student's t‐test was used for normally distributed data and the Mann–Whitney U‐test for all other comparisons in a and b; Spearman's correlation test was applied in c) *P < 0.05, ****P < 0.0001.

Transcriptional profiling confirms activation profile and reveals migration defects and reduced regulatory potential in circulating ILCPs from patients with T1D

To further delineate the activation state of ILCP—the subset showing the strongest phenotypic alterations—we performed quantitative gene‐expression analysis on purified ILCP from T1D patients and matched controls (Figure 3). Among chemotactic receptors, CCR7 expression was significantly reduced in ILCP from patients with T1D, with CCR6 showing a downward trend. CCR7 axis normally facilitate lymphocyte entry and retention in secondary lymphoid organs, raising the possibility that T1D ILCP fail to home efficiently to tolerogenic sites, such as lymph nodes. Such impaired migratory programming may increase their persistence within the circulation, potentially exposing them to systemic inflammatory cues.

Figure 3.

Gene‐expression analysis by qPCR in purified ILCP from type 1 diabetes (T1D) patients and healthy donors. Total RNA was isolated from purified total innate lymphoid cells (ILCs) from PBMCs, reverse‐transcribed to cDNA and amplified. SYBR Green and validated specific primers were used. Relative expression was calculated using the ΔΔCt method, with the 60S ribosomal gene as the housekeeping control and the mean of healthy donors as the calibrator sample. The mean Ct values from all the healthy donors were used as the calibrator sample for each gene. Mean and standard deviation are shown. Individual values are also shown as dots, grey dots for healthy donors and blue dots for patients with T1D. Student's t‐test was used for normally distributed data and the Mann–Whitney U‐test for all other comparisons. *P < 0.05.

Beyond migration, ILCP from T1D patients also displayed reduced expression of ICOS, a costimulatory molecule associated with regulatory and homeostatic functions that enhances ILC proliferation, and decreased CD58, an adhesion molecule involved in immune synapse formation and T‐cell interaction. These defects point to a broader dysregulation affecting ILCP communication with other immune cells. Notably, ILCP also showed a trend towards elevated HLA‐DRA transcript levels (Figure 3), consistent with the flow cytometry findings and reinforcing the notion of heightened antigen‐presentation potential.

Combined, these transcriptional alterations suggest that ILCP in T1D are not only activated but also functionally skewed away from their canonical regulatory, lymphoid‐tissue–associated role and towards a more pro‐inflammatory, circulation‐persistent phenotype.

ILCP activation is driven by glycolytic metabolism and enhanced under conditions mimicking inflammation

Elevated plasma glucose levels—as observed during hyperglycaemic episodes in clinical T1D—may directly enhance ILCP activation in vivo, since glycolysis is a central source of energy for several ILC subsets. Indeed, we found a trend for correlation between the percentage of HLA‐DR⁺ circulating total ILC and the levels of glycosylated haemoglobin (HbA1c) as shown in Figure 4a (r = 5394, P = 0.0733). Therefore, we evaluated how glucose metabolism regulates ILC activation. Using an in vitro model of ILC polarisation, enriched human peripheral blood ILCs were cultured with IL‐2 and IL‐7, and stimulated to acquire ILC1‐ or ILC3‐like phenotypes. Inhibition of glycolysis using 2‐deoxy‐D‐glucose (2‐DG) significantly reduced CD69 and HLA‐DR expression in ILCP, whereas ILC1 activation remained largely glycolysis‐independent (Figure 4a). These results were confirmed in cell sorting‐purified total ILC (n = 3, data not shown).

Figure 4.

Heightened responses to in vitro mitogen stimulation in innate lymphoid cell (ILC) from patients with type 1 diabetes (T1D). (a) Representative dot plots showing the identification of lin⁻ CD127⁺ ILC (left) on pre‐gated live CD45⁺ PBMC from healthy donors (HD) or patients with T1D (T1D) upon in vitro PHA stimulation for 48 h. Right contour plots show the expression of CD69 (x‐axis) and HLA‐DR (y‐axis) on Lin⁻ CD127⁺ ILC from HD and T1D. (b) Histograms showing the mean and standard deviation together with individual values (circles) corresponding to the percentage of lin⁻ CD127⁺ ILC positive for CD69, HLA‐DR, CD80 and CD25 within total circulating ILCs upon 48 h PHA stimulation of PBMC from HD and T1D analysed by spectral cytometry (n = 4–10). (c) Paired dot plots with lines connecting matched samples showing the percentage of ILC1 and ILCP that, express CD69 and HLA‐DR with or without glycolysis inhibition by 1 mm 2‐DG for 72 h (n = 7). PBMC have been enriched in ILC by magnetic cell sorting and enriched ILC have been cultured under polarisation conditions (IL‐12 for ILC1 and IL‐1β + IL‐23 for ILCP). Values of ILC1 (left) and ILCP (right) gated among live CD45⁺ lin⁻ CD127⁺ ILC are shown. *P < 0.05, ***P < 0.001. Student's t‐test was used for normally distributed data and the Mann–Whitney U‐test for all other comparisons in a and b. Paired Student's t‐test was used in C. *P < 0.05.

IL‐1β induces antigen‐presentation features in human circulating ILCs

Besides direct ex vivo activation phenotype, we evaluated the activation potential of ILC from patients with T1D and healthy donors upon in vitro mitogen stimulation. ILCs from T1D patients displayed a higher potential to upregulate CD69, CD25, HLA‐DR and CD80 when compared with ILCs from healthy donors, reaching significance for CD69 and CD25 and showing a trend for CD80 in response to mitogenic stimulation ex vivo (Figure 4b and c).

To identify inflammatory signals capable of inducing the activation phenotype observed in T1D, we exposed purified circulating ILCs from healthy donors to pro‐inflammatory cytokines implicated in pancreatic islet inflammation, such as IL‐1β, IFN‐γ and IFN‐β. Among the stimuli tested, IL‐1β emerged as the most potent driver of antigen‐presentation characteristic molecules, significantly upregulating HLA‐DR (Figure 5a). IFN‐γ induced a milder non‐significant yet detectable increase, while IFN‐β had minimal impact. IL‐1β not only upregulated HLA‐DR in purified ILCs from healthy donors at 5 days of culture but also slightly upregulated the costimulatory molecule CD80 (Figure 5b and c).

Figure 5.

Regulation of antigen‐presentation features by cytokines and glycolysis in human circulating innate lymphoid cell (ILC). (a) Representative staggered histograms showing the percentage of positive ILC after cell sorting purification from healthy donor buffy coat PBMC and 5‐day culture with IL‐2 only or with additional 10 ng mL⁻¹ of cytokines (IL‐1β, IFN‐γ, IFN‐β, all from Peprotech), left. On the right, the percentage of HLA‐DR⁺ cells is shown, both with individual values and mean ± standard deviation (n = 5), **P < 0.01. (b) Representative spectral cytometry contour plots showing HLA‐DR and CD80 expression vs SSC in purified ILCs (magnetically enriched and sorted), unstimulated or stimulated for 5 days with IL‐1β (10 ng mL⁻¹). (c) Paired dot plots with lines connecting matched samples showing the percentage of HLA‐DR and CD80 positive cells in purified ILCs after 5 days with IL‐1β (10 ng mL⁻¹) (n = 5). paired Student's t‐test was used, *P < 0.05.

IL‐1β is abundant in inflamed pancreatic islets and contributes to β‐cell apoptosis, may be a key environmental factor promoting HLA‐DR expression and costimulatory potential in ILCs in T1D. But systemically increased IL‐1β in T1D (together with other cytokines) could also be causative of HLA‐DR overexpression in T1D. To assess whether the altered ILC activation states could be driven by systemic cytokines, we quantified circulating cytokine levels in plasma using a cytometric bead assay. While some cytokines showed a trend towards increased levels in patients compared to donors (IL‐1β, TNF‐α and IFN‐γ), these differences were not statistically significant (Supplementary figure 2). Moreover, we did not observe any correlation between circulating cytokine levels and ILC activation markers, including HLA‐DR expression on total ILC, ILC1 and ILCP, or CD69 expression in total ILC (Supplementary figure 2), arguing against systemic cytokines as direct drivers of ILC activation.

Circulating ILCs from T1D patients loaded with islet self‐peptides activate insulin‐specific autoreactive CD4 ⁺ T cells

Given the overexpression of HLA molecules in ILCs from recent‐onset T1D patients and their activated phenotype, we sought to evaluate whether ILC activation in T1D translates into functional antigen‐presentation capacity. Therefore, we established an autologous co‐culture system using purified circulating ILCs and autologous memory CD4⁺ T cells. ILCs pre‐activated for 5 days with IL‐1β were loaded overnight with the immunodominant insulin peptide InsB9‐23. Peptide‐loaded or unloaded ILCs were subsequently co‐cultured for 72 h with autologous memory CD4⁺ T cells obtained by magnetic cell sorting. In three of the three T1D patients tested, peptide‐loaded ILCs induced a substantial increase in OX40⁺CD25⁺ memory CD4+ T cells compared with unloaded controls (Figure 6). This indicates that human circulating ILCs can indeed present islet autoantigens and activate autoreactive CD4⁺ T cells under inflammatory conditions. Of note, in co‐cultures of memory CD4⁺ T cells from three healthy donors with InsB9‐23 peptide‐loaded autologous pre‐activated ILCs we did not detect an increase in OX40⁺CD25⁺ T cells.

Figure 6.

Insulin peptide‐loaded innate lymphoid cells (ILCs) induce activation of autologous memory CD4 T cells in type 1 diabetes (T1D) patients. (a) Dot plots showing CD25 and OX40 expression in memory CD4 T cells purified by magnetic cell separation (left), co‐cultured with autologous purified ILCs pre‐activated for 5 days with IL‐1β (centre) and co‐cultured with autologous purified pre‐activated ILCs loaded with 10 μm InsB9‐23 for 72 h (right). A representative experiment from patients with T1D is shown in the upper panel (T1D) and from a healthy donor in the lower panel (HD). T cells are gated as viable CD45⁺ CD3⁺ CD4⁺. (b) Paired dot plot showing the percentage of OX40 and CD25 double positive cells in individual patients with T1D (T1D, left) or healthy donors (HD, right) under three conditions: cultured alone, co‐cultured with purified autologous ILC and co‐cultured with peptide‐loaded purified autologous ILC. Each dot represents one patient, and lines connect paired measurements across conditions (n = 3). Samples with less than 100 live memory CD4⁺ T cells were excluded. Data were analysed with the Friedman test followed by Dunn's multiple comparisons test (Bonferroni‐adjusted P‐values) (non‐significant).

ILC activation patterns are conserved in the NOD mouse model of autoimmune diabetes

To assess whether the activation patterns observed in circulating ILC from patients with T1D are recapitulated by the NOD mouse model of autoimmune diabetes, we analysed ILCs in the spleen and PLNs of female NOD mice at three differentiated timepoints in terms of islet infiltration and disease progression: 5 weeks (no insulitis to peri‐insulitis), 10 weeks (mild to established insulitis) and recent onset (moderate‐to‐severe islet‐infiltrative insulitis), as we previously reported ²⁵ (Supplementary figures 3 and 4). In the spleen, NOD ILCs displayed increasing frequencies of MHC‐II⁺ cells and MFI, expressing statistically higher levels of MHC‐II spleen ILC from diabetic mice than pre‐diabetic NOD female (Figure 7). Moreover, ILC showed enhanced activation, indicated by CD44 upregulation, in the spleen at the stage of clinically overt diabetes, paralleling the trajectory observed in circulating human ILCs. A trend of CD44 and MHC‐II upregulation in ILC from diabetic mice is also observed in pLN, with more heterogeneity in the diabetes condition (Figure 7). Overall, these conserved alterations in activation and antigen‐presenting molecules of ILC in human and murine T1D at recent onset reinforce their biological relevance in autoimmune diabetes.

Figure 7.

NOD mice innate lymphoid cell (ILC) from spleen and pancreatic lymph node (pLN) recapitulating antigen presentation and activation features identified in circulating ILC from patients with type 1 diabetes (T1D). (a) Representative dot plots showing a summarised version of the gating strategy of lin⁻ CD127⁺ (left) CD90⁺ (middle) ILC from spleen female NOD mice at 12 weeks of age (upper plots) and at clinical T1D onset (lower plots). In lin⁻ CD127⁺ CD90⁺ ILC, the expression of MHC‐II (x‐axis) vs FSC (y‐axis) is shown (right dot plots). (b) Histogram plots showing the mean, standard deviation and individual values corresponding to the percentage of ILC expressing MHC‐II (left), the mean fluorescence intensity (MFI) for MHC‐II in ILC (middle) and the MFI for CD44 (right) in ILC from spleen (upper plots) and pLN (lower plots) from NOD females at different timepoints of T1D progression: 5 weeks (5w), 12 weeks (12w) and clinical onset (T1D) (n = 5–7). Kruskal–Wallis test was applied. *P < 0.05, **P < 0.01.

Discussion

We identify a pattern suggestive of altered activation of circulating ILCs in T1D. Our results indicate that circulating ILCs—particularly ILCP—may exhibit features of pre‐activation, both phenotypically and functionally. They show increased MHC‐II expression, altered migratory and regulatory gene signatures, glycolysis‐associated activation and the capacity to present autoantigen and activate insulin‐specific autoreactive CD4⁺ T cells. The convergence of these features in both humans and NOD mice points to a potential and previously underappreciated contribution of circulating ILCs in recent‐onset autoimmune diabetes.

We observed a modest decrease in circulating ILC1 frequencies that may reflect redistribution to peripheral tissues or changes in relative subset composition, although absolute cell numbers were not assessed. Accordingly, the scarce previous data on the role for ILC in T1D points out to the accumulation of ILC1 cells in GALT from patients with T1D. ¹² Also in the GALT of NOD mice, ILC3 seem to have a T1D‐protective role. Two alternative mechanisms have been described in this sense. One mechanism is the activation of Tregs by IL‐2‐producing ILC3 in the GALT. ¹⁴ In addition, IL‐22 produced by GALT ILC3 distally enhances the production of antimicrobial peptides by β cells in NOD mice. ¹³ The GALT tissue is an important immunomodulatory site where ILC1 and ILC3 are enriched; it is in close proximity to the pancreas (gut‐pancreas axis) and is a key location where interaction between commensal bacteria and the immune system takes place. However, focusing uniquely on GALT could limit our understanding of how ILC contribute to T1D at the systemic level, highlighting the need to investigate functional and phenotypic alterations of circulating ILC and their migratory behaviour across immune and non‐immune tissues. This contrast suggests that tolerogenic functions of ILC described in barrier tissues such as the GALT may become uncoupled from systemic ILC responses in T1D, underscoring the context‐dependent nature of ILC‐mediated immune regulation.

While most ILC are tissue‐resident and have developed from migrating ILC precursors, some circulate from tissues to peripheral lymph nodes and through peripheral lymph nodes in a S1PR‐dependent manner. ²⁶ , ²⁷ In our study, we have identified a downregulation of CCR7, the receptor for chemokines (CCL19 and CCL21) that guide immune cells to secondary lymphoid organs, and CCR6 mRNA in ILCP from patients with recent‐onset T1D when compared to healthy controls. These results point to a lower capacity of ILCP migration to secondary lymphoid organs due to their activation status, favouring their retention in blood or tissues.

The features observed in circulating ILCP from patients with recent‐onset T1D—including pre‐activation, heightened responsiveness to secondary stimulation, phenotype dependent on glycolysis and sustained upregulation of activation and antigen‐presenting molecules—raise the possibility that these cells acquire a trained immunity–like phenotype. Trained immunity, classically described in myeloid cells, refers to the long‐term functional reprogramming of innate immune cells following inflammatory or metabolic cues. Trained cells are faster and more potently ‘ready to react’ when challenged a second time with the same or a different unrelated stimulus. ²⁸ , ²⁹ Besides myeloid cells, there is emerging evidence of innate memory in ILC in settings of chronic inflammation (i.e. ILC2 in allergy) or antiviral responses. ³⁰ , ³¹ In line with this concept, we observe that ILCP from T1D patients are more readily activated (probably intrinsically and bystander) upon in vitro stimulation, suggesting an imprinting effect of the in vivo inflammatory and hyperglycaemic environment. Systemic hyperglycaemia, together with pro‐inflammatory cytokines, such as IL‐1β and IFN‐γ, may therefore act as training signals that durably reprogram ILCP metabolic and functional states. Such trained‐like ILCP would be poised to respond excessively to local inflammatory cues, acquiring non‐classical antigen‐presenting functions and amplifying autoreactive CD4⁺ T‐cell responses. Although the epigenetic mechanisms underlying this potential training remain to be defined, our findings provide a framework to explore trained immunity in ILC as a previously unrecognised contributor to immune dysregulation and tolerance failure in T1D.

Glucose metabolism is a critical regulator of immune cell phenotype and effector function, as previously demonstrated in adaptive immune cells, including CD8⁺ T cells, where high‐glucose conditions promote activation and effector function. ³² Among ILC, distinct metabolic requirements have been described, with ILC2 and ILC3 being highly dependent on glycolysis to sustain cytokine production, proliferation and activation, whereas ILC1 appear to rely more on oxidative metabolism. In line with this concept, our data demonstrate that glycolysis is required for cytokine‐induced upregulation of the activation and antigen‐presenting markers CD69 and HLA‐DR in human circulating ILC precursors (ILCP), but not in mature ILC1. These findings indicate that ILCP exhibit a heightened metabolic sensitivity compared with differentiated ILC subsets. In the context of recent‐onset type 1 diabetes (T1D), where systemic hyperglycaemia is a defining feature, altered glucose availability may therefore directly modulate ILCP activation status and their acquisition of antigen‐presenting properties. Of note, we observe a trend for correlation between HLA‐DR positive ILC frequency and the level of glycated haemoglobin. Supporting the idea of circulating ILC as metabolic sensors, women with insulin resistance upon gestational diabetes show a higher activation of circulating ILC. ³³ This metabolic–immune axis could contribute to shaping early inflammatory responses and potentially influence autoreactive T‐cell activation during disease initiation or progression. In the future, it would also be of interest contrasting the functional phenotype of circulating ILCs in autoimmune T1D with other metabolic and pro‐inflammatory contexts beyond T‐cell–mediated autoimmunity, such as type 2 diabetes. Interestingly, despite the glycolysis‐dependent activation phenotype observed in ILCP, we did not detect differences in HIF1A mRNA expression in ILCP sorted from recent‐onset T1D patients compared with healthy donors (not shown). Although HIF‐1α is a key regulator of glycolytic programmes in immune cells, the absence of changes in HIF1A transcript levels in ILCP from T1D patients indicates that additional mechanisms, such as post‐translational modifications likely contribute to the metabolic regulation of ILCP activation in this context.

One of our main findings with more potential consequences for the failure of tolerance in T1D is the consistently higher expression level of Class II molecules in circulating ILC1 and ILCP in patients with recent‐onset T1D when compared with healthy controls. The expression of MHC Class II by ILC can be pathologically relevant, since in homeostatic conditions mature circulating ILC1 express negligible levels of MHC Class II. Other subsets such as circulating ILC3 can express MHC Class II. The quantitative and qualitative change in HLA Class II expressing circulating ILCP in recent‐onset T1D supports a functional alteration rather than a simple cell subset expansion. These findings support the idea of plasticity of the ILC compartment in inflammatory environments like in autoimmune disorders.

A positive correlation between the frequencies of HLA‐DR–expressing ILC1 and ILC3 subsets is observed in patients with T1D, but not in healthy donors. This could indicate a pathology‐driven process where a shared pro‐inflammatory environment co‐ordinately induces the overexpression of Class II in different ILC subsets. The recapitulating upregulation of MHC Class II in spleen lin⁻ CD127⁺ CD90⁺ ILC in diabetic NOD mice supports this hypothesis.

Circulating mature ILC1 express almost no detectable level of HLA Class II in healthy conditions. However, HLA Class II positive ILC1 have been identified in the context of autoimmunity. ²⁴ Two possible non‐excluding explanations might be behind this aberrant expression. One is a transcriptional induction by pro‐inflammatory cytokines, such as IL‐1β or IFN‐γ, for instance. An additional possible mechanism eliciting HLA Class II expression in ILC1 in T1D is the molecule acquisition by trogocytosis, as already documented in the context of autoimmunity for example in CD20⁺ T cells. ³⁴ Nevertheless, the global ILC1 activated profile supports the functional relevance of ILC1 HLA Class II expressing cells in T1D. ILC2 and ILC3 can express Class II molecules in humans and in mice. Their expression in ILC3 is regulated by cytokines, such as IL‐1β. ³⁵ We have confirmed the IL‐1β dependency for HLA Class II upregulation in human circulating ILCP. ILC3 that express HLA Class II but lack costimulatory molecules are considered tolerogenic in the context of tolerance to microbiota antigens. Recently, it has been observed that HLA Class II expressing ILC3 in the GALT provide tolerance to microbiota by STING‐dependent recognition of microbiota‐derived nucleic acids. ³⁶

We have observed that circulating ILCP from patients with recent‐onset T1D have a higher predisposition to upregulate HLA Class II and the costimulatory molecule CD80 upon stimulation than healthy donor‐derived ILCP. In T1D, under inflammatory and metabolic reprogramming, they seem to acquire a non‐classical APC phenotype. Our data are consistent with T1D autoimmunity‐intrinsic signals shaping ILC activation, rather than driven only by systemically increased cytokines. In this setting, non‐classical APC functions that could amplify pre‐existing autoreactive CD4 T‐cell responses. In a mechanistic in vitro assay, we demonstrate that in a pro‐inflammatory microenvironment with IL‐1β, ILCP from recent‐onset patients with T1D have the functional capacity to activate cognate insulin‐specific CD4 T‐cell responses. Since ILCs elicit rapid and abundant effector responses in a metabolism‐sensitive manner, they might act locally at times when classical APCs are not dominant, potentially amplifying or contributing to autoimmune pathology in certain contexts.

Despite these insights, several limitations of the present study should be acknowledged. First, the experimental design does not allow us to establish whether the ILC alterations observed in T1D patients are causal or secondary to the disease, and further studies will be required to clarify this question. Second, the human T1D cohort is relatively small. This are, however, well‐defined recent‐onset patients that provided statistically significant results. In addition, the low frequency of ILC in circulation (0.1–0.5% of lymphocytes) represents a challenge for human ILC research, especially in limited amount sample (i.e. patient‐derived blood). For this reason, we applied hypothesis‐driven transcriptomics with previous nucleic acid amplification has been applied, rather than more exploratory total or single cell RNA sequencing. Moreover, mitogenic activation of bulk PBMCs induces broad cytokine release and bystander activation, which may differ between T1D and healthy donors. Therefore, the increased CD69 expression observed on circulating ILCs from patients with T1D may reflect integrated signals from the global immune environment. Future studies using cytokine‐stimulated purified ILCs from T1D patients will be required to dissect cell‐intrinsic activation thresholds. Finally, the concordance regarding higher activation and MHC Class II expression between circulating ILC from human diabetes and splenic ILC from NOD mice with recent‐onset diabetes prompt us to continue exploring autoimmune‐prone features of ILC in this standard experimental model of autoimmune diabetes. ³⁷ However, additional studies in larger cohorts of patients with T1D at different stages and with different disease endotypes (paediatric, adult) will be very valuable.

Our study provides a novel understanding of the circulating ILC compartment functional and phenotypical alterations at recent‐onset autoimmune T1D and highlights several translational implications. Our data demonstrate that ILC1 and ILCP contribute to the peripheral inflammatory milieu observed in recent‐onset T1D. Circulating ILC1 and ILCP exhibit an activated phenotype, characterised by a reduced potential for migration to secondary lymphoid organs and impaired tolerogenic costimulatory potential. Chronic inflammation associated with T1D reshapes the circulating ILC compartment, including aberrant functional expression of HLA Class II molecules in cells that are not classical APCs. In this inflammatory and hyperglycaemic context, rather than promoting immune tolerance, ILCs facilitate the activation of autoreactive CD4⁺ T cells, thereby contributing to the breakdown of immune tolerance. Notably, circulating ILC1 and ILCP emerge as a potential therapeutic target in T1D. Future studies should aim at identifying activation biomarkers in circulating ILCs across different disease stages and endotypes of T1D. Ultimately, the development of therapeutic strategies that restore their tolerogenic state, counteract dysfunctional activation and tightly regulate their activity will be essential.

Methods

Human samples

Blood samples were obtained from adult (18–49) T1D patients (n = 12) recruited from the Endocrinology Department of the Hospital Germans Trias i Pujol (Badalona, Spain). T1D was diagnosed fulfilling the diagnostic criteria of the American Diabetes Association. ³⁸ The samples were obtained after a maximum of 2 years since clinical diagnosis. Blood samples from age‐matched healthy volunteers (n = 23) were obtained from volunteer healthy donors of the Blood and Tissue Bank. Demographic and clinical data from donors are detailed in Supporting Information (Supplementary table 1). Blood samples from T1D patients and healthy volunteers were collected in EDTA tubes. Peripheral blood mononuclear cells (PBMC) were obtained by density gradient cell centrifugation at 24 h after extraction and stored in liquid nitrogen. Plasma was separated by centrifugation and stored at −80°C. Due to limited cell availability, not all assays were performed on all donors. Consequently, sample sizes vary across experiments, and the exact number of donors included in each analysis is indicated in the corresponding figure legends.

NOD mice

Wild‐type NOD mice were originally obtained from the Jackson Laboratory (Bar Harbor, ME, USA) and then kept in the Animal Facility of the Centre de Medicina Comparativa i Bioimatge de Catalunya (CMCiB, Badalona, Spain) under specific pathogen‐free conditions. The colony was subjected to a 12 h dark/12 h light cycle and controlled temperatures between 19 and 23°C with 40–60% humidity and fed with ad libitum access to acidic water at pH 5 and irradiated Teklad Global 18% Protein Rodent Diet (Harlan, Indianapolis, IN, USA). Pre‐diabetic 5 and 10 weeks old and diabetic females were used. T1D was determined by weekly monitoring glycosuria levels using urine test strips (Combiscreen; Analyticon Biotechnologies AG, Lichtenfels, Germany) and confirming hyperglycaemia (> 300 mg dL⁻¹) with a glucotest control (OneTouch Verio Reflect®, LifeScan IP Holdings, LLC., Zug, Switzerland).

Ethics statements

Human: The study followed the principles of the Declaration of Helsinki for human research and was carried out upon approval by the Committee on the Ethics of Research of the Germans Trias i Pujol Hospital, the Bioethics Commission of the University of Barcelona and the Ethics Committee of the Blood and Tissue Bank. Signed informed consent was obtained in all donations.

Mice: This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the Generalitat de Catalunya. The procedures carried out with animal models were authorised by the Animal Experimentation Ethics Committee of the CMCiB and IGTP and by the Generalitat de Catalunya, and they followed the principles outlined in the Declaration of Helsinki for animal experimental investigation. All the protocols conducted followed the principles of the 3R, prioritising the welfare of animals used in research.

Ex vivo immunophenotyping of human ILC

For human ILC immunophenotyping by spectral cytometry, cryopreserved PBMC samples were thawed, washed and stained in PBS with the viability marker Ghost Dye Red 780 (Cytek Biosciences, Freemont, CA). Fc receptors were blocked with an Fc receptor blocking solution (Human Trustain FcX; BioLegend, San Diego, CA, USA). Surface proteins were stained with specific antibodies in 2% heat inactivated fetal bovine serum (FBS) 1 mm EDTA PBS for 20 min at 4°C and washed. Human ILCs were identified as CD45⁺ (Alexa Fluor 700, clone 2DI; BioLegend) lineage (lin)⁻, all FITC‐conjugated anti‐CD3 (clone UCHT1; Cytek Biosciences), anti‐CD4 (clone SK3, Cytek Biosciences), anti‐CD8 (clone RPA‐T8; Cytek Biosciences), anti‐CD14 (clone 61D3, Cytek Biosciences), anti‐CD15 (clone W6D3; BioLegend), anti‐CD16 (clone 3G8; STEMCELL Technologies Europe, Grenoble, France), anti‐CD19 (clone HIB19; Cytek Biosciences), anti‐CD20 (clone 2H7; Cytek Biosciences), anti‐CD33 (clone HIM3‐4; STEMCELL Technologies), anti‐CD34 (clone 581; STEMCELL Technologies), anti‐CD203c FceRI (clone NP4DE; BioLegend) and positive for CD127 (BV605, clone A019D5; BioLegend). Anti‐CD294 (BV421, clone BM16; BioLegend) and anti‐CD117 (BV711, clone 104D2; BioLegend) antibodies were used to distinguish circulating ILC1, ILC2 and ILCP subsets among CD45⁺, lineage⁻, CD127⁺ cells as follows: CD294⁻ CD117⁻ cells as ILC1, CD294⁺ cells as ILC2 and CD294⁻ CD117⁺ cells as ILCP. The following antibodies were used to phenotype total and subset‐specific circulating ILC: anti‐HLA‐DR (PE, clone L243; Cytek Biosciences), anti‐CD69 (PE‐Cy5, clone FN50; BioLegend), anti‐CD80 (PE‐Cy5, clone 2D10; BioLegend) and anti‐CD25 (PE‐Cy7, clone BC96; BioLegend). Samples were acquired on a Cytek Aurora cytometer (Cytek Biosciences), and the data were analysed using the FlowJo™ v10.10 software (BD Life Sciences, Madrid, Spain). Gates were set based on fluorescence‐minus‐one (FMO) controls (shown in Supplementary data 1).

Cell sorting of human ILC

For cell sorting, total, lin⁻ CD127⁺ ILC were first enriched by magnetic cell sorting using the Easysep Human Pan‐ILC enrichment kit (STEMCELL Technologies), cell staining with the lineage cocktail, a viability marker and anti‐CD127 for total ILC sorting (and additionally anti‐CD117 and anti‐CD294 for ILCP sorting) on BD FACSDiscover S8 Cell Sorter (BD Biosciences, Heidelberg, Germany).

Ex vivo immunophenotyping of mouse ILC

Spleens and pancreatic lymph nodes (pLN) harvested from 5 weeks old, 10 weeks old and clinical‐onset diabetic female NOD mice (five animals per group) were mechanically disrupted pressing with the plunger end of a syringe through a 40‐μm cell strainer with 5 mL of PBS with 2% FBS and washing the strainer with an additional 10 mL of PBS 2% FBS. Residual tissue was discarded by sedimentation after 5 min of incubation at room temperature. The supernatant was centrifuged for 5 min at 300 g to pool the cells. Spleen single cell suspensions were treated with 5 mL ammonium‐based erythrocyte lysis buffer and washed with PBS by centrifugation for 5 min at 300 g. Cells were washed additionally once in PBS and the cell pellet was treated with the viability marker Ghost Dye Red 780 (Cytek Biosciences) diluted 1/20.000 for 20 min at 4°C. After a wash with PBS 2% FBS by centrifugation, Fc receptors were blocked with an Fc receptor blocking solution (Mouse Trustain FcX anti‐CD16/anti‐CD3; BioLegend). Cells were stained with specific antibodies in 2% FBS 1 mM EDTA PBS for 20 min at 4°C with the following antibodies: anti‐CD3 (efluor 506, clone 17A2; Invitrogen), anti‐CD45 (Alexa Fluor 700, clone I3/2.3; BioLegend), anti‐CD4 (FITC, clone GK1.5; Cytek Biosciences), anti‐CD8 (FITC, clone 53/6.7; Cytek Biosciences), anti‐CD11b (FITC, clone M1/70; Cytek Biosciences), anti‐CD11c (FITC, clone N418; Cytek Biosciences), anti‐B220 (FITC, clone RA3‐6B2; Cytek Biosciences), anti‐FcεRIα (FITC, clone MAR‐1; Cytek Biosciences), anti‐CD127 (PE‐Cy7, clone A7R34; Cytek Biosciences), anti‐CD90 (APC, clone 30H12; BioLegend), anti‐CD44 (PE‐Dazzle594, clone IM7; BioLegend) and anti‐MHC‐II (PE‐Cy5, clone ME/114.15.2; BioLegend). The staining of the I‐A molecules of the NOD haplotype was confirmed with lineage‐positive cells as a clear and abundant population in comparison with the FMO control (Supplementary data 1). Samples were acquired on a Cytek Aurora cytometer (Cytek Biosciences) and the data were analysed using FlowJo™ v10.10 software (BD Life Sciences). ILC were defined as live singlet CD45⁺ lineage‐negative CD127⁺ CD90⁺.

In vitro stimulation

For bulk mitogen stimulation, 1 × 10⁶ of thawed PBMC were seeded in 48‐well plates in 8% human male AB serum (HS) (Sigma Aldrich) supplemented with 2 mm L‐Glutamine, Penicillin (100 U mL⁻¹)‐Streptomycin (100 mg mL⁻¹), non‐essential amino acids and 1 mm sodium pyruvate. Cells were stimulated for 48 h with 1 μg mL⁻¹ of phytohemmaglutinin‐L (PHA) (Merck Life Science S.L.U., Madrid, Spain) in the presence of IL‐2 (20 ng mL⁻¹) and IL‐7 (20 ng mL⁻¹). Stimulated cells were harvested, washed and stained for ILC immunophenotyping by spectral cytometry.

For stimulation of purified ILCs, PBMC from healthy donor buffy coats were enriched with the Easysep Human Pan‐ILC enrichment kit (STEMCELL Technologies) and sorted (see above) as live, lin‐ CD127⁺ CD294⁻ CD117⁺ ILCP HLA‐DR⁻. Cells were cultured in 96‐well plates (2000–5000 cells per well) in RPMI 8% HS with supplements in the presence of IL‐2 (20 ng mL⁻¹) (STEMCELL technologies) alone or with the IL‐1β, IFN‐γ or IFN‐β (10 ng mL⁻¹) (PeproTech Inc., Cranbury, NJ, USA) for 5 days.

ILC glycolysis inhibition with 2‐deoxyglucose (2‐DG)

PBMC from healthy donor buffy coats were enriched with the Easysep Human Pan‐ILC enrichment kit (STEMCELL Technologies) and cultured in 96‐well plates in RPMI 8% HS with supplements in the presence of IL‐2 (20 ng mL⁻¹) and IL‐7 (20 ng mL⁻¹) (both from STEMCELL Technologies) with or without 1 mm 2‐deoxyglucose (2‐DG) (Merck Life Sciences S.L.U.) for 72 h. Cells were analysed for expression of activation markers by spectral cytometry. A confirmatory batch of experiments was performed in the same conditions with fluorescence cell sorting purification of total lin⁻ CD127⁺ ILC.

Plasma cytokine measurement

Human plasma cytokine levels were measured with a bead‐based assay (Legendplex Human Inflammation Panel 1 Mix and Match, BioLegend) according to the manufacturer's instructions. Samples were acquired on a Gallios flow cytometer (Beckman Coulter, Brea, CA, USA), and data were analysed using the LEGENDplex™ Data Analysis Software Suite (BioLegend), using the Qognit cloud‐based platform.

ILC‐memory CD4 T‐cell co‐culture

Total, lin⁻ CD127⁺ ILC were first enriched by magnetic cell sorting using the Easysep Human Pan‐ILC Enrichment Kit (STEMCELL Technologies), cell staining with the lineage cocktail (see above), a viability marker and anti‐CD127 on BD FACSDiscover S8 Cell Sorter (BD Biosciences). Purified lin⁻ CD127⁺ total ILC from patients with T1D were cultured in vitro for 5 days (500–1000 cells per well) in the presence of IL‐2 (20 ng mL⁻¹) and IL‐1β (10 ng mL⁻¹) in 8% HS RPMI medium with supplements. On Day 4, ILCs were loaded or not with 10 μg mL⁻¹ InsB9‐23 peptide (Peptide Synthesis Service, Universitat Pompeu Fabra, Barcelona, Spain) overnight. On Day 5, loaded ILC were washed three times by centrifugation to eliminate traces of remainder peptide from the medium, thus avoiding potential direct peptide binding to the Class II expressed by activated memory T cells. In parallel, autologous memory CD4⁺ T cells were isolated by negative magnetic cell sorting with a Memory CD4⁺ T‐cell Enrichment Kit (STEMCELL Technologies) (purity > 90%). For the ILC‐CD4⁺ T‐cell co‐culture, 25.000 memory CD4⁺ T cells were seeded in 96‐well plates alone, or in the presence of peptide‐loaded or unloaded ILC in 8% HS RPMI medium with supplements for 72 h. Co‐cultured harvested, stained with anti‐Ghost Dye Red 780‐conjugated viability marker (Cytek Biosciences) anti‐CD3 (APC, clone UCHT1; Cykek Biosciences) and the antibodies specific for activation‐induced T‐cell markers (AIT) anti‐CD25 (PE‐Cy7, clone BC96; BioLegend) and anti‐OX40 (PE, clone ACT35; BioLegend). ³⁹ Samples with < 100 live T cells were excluded from analysis.

Gene‐expression analysis by quantitative RT‐PCR

RNA extraction and cDNA amplification was performed at the Functional Genomics Service at the IRB Barcelona. RNA was isolated by the Pico Profiling method for very small cell populations using magnetic beads from 750 to 2500 cells. The RNA extraction, cDNA synthesis, library preparation and amplification were performed as described. ⁴⁰ Amplification was performed for 23 cycles. The mRNA transcript level of selected immune genes for patient‐ and donor‐derived circulating ILCP were analysed by real‐time quantitative qPCR using the CFX384™ Real‐Time System (Bio‐Rad, El Prat de Llobregat, Spain). qPCR was performed with specific primers for CCR6, CCR7, CD58 and ICOS (KiCqStart primers predesigned, Merck Life Sciences, S.L.U.), for HLA‐DRA (Meyer et al., 2021). All the analyses were performed in triplicate wells using 384‐well plates with 5 μL of NZYSupreme qPCR Green Master Mix 2x (NZYtech, Lisboa, Portugal), 0.5 μL of forward and reverse primers (500 nm) and 1 μL of diluted cDNA for each sample in a final volume of 10 μL with RNAse‐free milliQ water. The expression level of each gene was calculated with the ΔΔCt method using as a reference gene the 60s ribosomal protein L6 (RPL6) (primer) and the geometric mean of the healthy donors Ct as a calibrator sample.

Statistical analysis

Statistical analyses were performed using the GraphPad Prism version 10 (GraphPad Software, San Diego, CA, USA). For a two‐variable comparison, datasets with normal distribution according to Shapiro–Wilk distribution test were compared with a t‐test, while the non‐parametric Mann–Whitney U‐test was applied for data without normal distribution. For more than two variables, a mixed‐effects two‐way ANOVA with Sidak's multiple comparisons test was used. For cytokine‐stimulation of sorted ILC and 2‐DG in vitro treatment data analysis, paired t‐test was applied, since sample distribution passed normality (Shapiro–Wilk test) in all cases. For T‐cell activation by co‐culture with autologous ILC data analysis, the non‐parametric Dunn's multiple comparison test was used. For correlation analyses, the Spearman correlation test was used. Results were considered statistically significant when reaching P‐values equal to or lower than 0.05.

Author contributions

Eva Aguilera: Resources. Sebastián Pérez: Methodology; investigation. Writing – review and editing; resources; conceptualization; funding acquisition. Raquel Planas: Conceptualization; investigation; funding acquisition; writing – original draft; formal analysis; project administration; supervision. Marta Vives‐Pi: Investigation; resources; writing – review and editing; funding acquisition. David Perna‐Barrull: Investigation; resources; writing – review and editing; funding acquisition. Ariadna Roig: Investigation; methodology; writing – review and editing; formal analysis.

Conflict of interest

Unrelated to the work herein presented, M.‐V.‐P. is named inventor in the patent titled: Liposome‐based immunotherapy and Tolerogenic composition (patent nos.: WO2015107140 and WO2024110516A1) for autoimmune diseases and is co‐founder and CSO of Ahead Therapeutics SL. All the other authors declare no competing financial interests.

Supporting information

Supplementary data 1

Data availability statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.