Abnormal neutrophil signature in the blood and pancreas of presymptomatic and symptomatic type 1 diabetes

Abstract

Background: Neutrophils and their inflammatory mediators are key pathogenic components in multiple autoimmune diseases, while their role in human type 1 diabetes (T1D), a disease that progresses sequentially through identifiable stages prior to the clinical onset, is not well understood. We previously reported that the number of circulating neutrophils is reduced in patients with T1D and in presymptomatic at-risk subjects. The aim of the present work was to identify possible changes in circulating and pancreas-residing neutrophils throughout the disease course to better elucidate neutrophil involvement in human T1D.

Methods: Data collected from 389 subjects at risk of developing T1D, and enrolled in 4 distinct studies performed by TrialNet, were analyzed with comprehensive statistical approaches to determine whether the number of circulating neutrophils correlates with pancreas function. To obtain a broad analysis of pancreas-infiltrating neutrophils throughout all disease stages, pancreas sections collected worldwide from 4 different cohorts (i.e., nPOD, DiViD, Siena, and Exeter) were analyzed by immunohistochemistry and immunofluorescence. Finally, circulating neutrophils were purified from unrelated nondiabetic subjects and donors at various T1D stages and their transcriptomic signature was determined by RNA sequencing.

Results: Here, we show that the decline in β cell function is greatest in individuals with the lowest peripheral neutrophil numbers. Neutrophils infiltrate the pancreas prior to the onset of symptoms and they continue to do so as the disease progresses. Of interest, a fraction of these pancreas-infiltrating neutrophils also extrudes neutrophil extracellular traps (NETs), suggesting a tissue-specific pathogenic role. Whole-transcriptome analysis of purified blood neutrophils revealed a unique molecular signature that is distinguished by an overabundance of IFN-associated genes; despite being healthy, said signature is already present in T1D-autoantibody-negative at-risk subjects.

Conclusions: These results reveal an unexpected abnormality in neutrophil disposition both in the circulation and in the pancreas of presymptomatic and symptomatic T1D subjects, implying that targeting neutrophils might represent a previously unrecognized therapeutic modality.

Funding: Juvenile Diabetes Research Foundation (JDRF), NIH, Diabetes UK.

Keywords: Autoimmune diseases; Autoimmunity; Immunology; Innate immunity.

Figure 1. Reduced pancreas β cell function correlates with reduced circulating neutrophils.

(A) Scatterplot and Spearman’s correlation analyses between peripheral neutrophil (upper) or lymphocyte (lower) counts and fasting C-peptide, early C-peptide response (30-minute – 0-minute C-peptide), stimulated C-peptide (mean C-peptide AUC) HOMA-β and Index60 collected from TN-intervention studies (TN10, TN18, TN20). Each data point represents 1 donor (n = 298). Data distribution, Spearman’s rank correlation coefficient (r), and P value are shown for each analysis. As this is a nonparametric correlation analysis, the regression lines could be added. Neutrophil and lymphocyte counts, fasting and stimulated C-peptide, and HOMA-β were log transformed to perform the analysis, but they are shown as original measures. (B) Final linear regression models for predicting neutrophil counts on the basis either of fasting or stimulated C-peptide with additional consideration for the potential effects of age, sex, and BMI percentile as well as the respective interactions between said effects are shown. The data used are from TN-intervention studies (n = 298) and the final models are plotted with the original scale of the variables. (C) Final linear mixed-effects models for predicting neutrophil counts on the basis of either fasting or stimulated C-peptide, when also considering the potential effects of age, sex, and BMI percentile as well as interactions with them are shown. The data used are from the Milan-TN01 study (n = 109 subjects; n = 303 observations) and the final models are plotted with the original scale of the variables.

Figure 2. Intact and NETting neutrophils infiltrate the pancreas of presymptomatic and symptomatic T1D donors.

(A) One representative image of an immunohistochemical (IHC) analysis of a formalin-fixed, paraffin-embedded (FFPE) section (out of 12 total sections analyzed) from an autoAb-positive nPOD donor (no. 6197) stained with hematoxylin and anti-MPO (neutrophil-specific) antibody (brown staining). Scale bar: 100 μm. (B) Representative images of immunofluorescence (IF) analyses of frozen OCT-embedded nPOD sections (out of 13 total sections analyzed) stained with anti-MPO antibody (green, left column) and Hoechst 33342 for DNA detection (white, middle column). Expression signals were then merged (right column). Donor ID and characteristics are shown on the left. Images are represented as Z-stacked following projection. Insets were cut (dotted square) and magnified ×1.5 at the bottom of each panel and they highlight the occurrence of decondensed DNA colocalized with MPO, which in turn suggests the presence of pancreas-residing NETting neutrophils. Scale bars: 20 μm. (C) Quantification of MPO-positive cells on FFPE pancreatic sections analyzed by IHC (empty dots, and green triangles for the Siena cohort) and IF (red dots) are shown in box-and-whisker plots as number of cells per squared millimeter. Numbers indicate donor ID. Each symbol represents 1 section analyzed (n = 26 donors; n = 34 sections). Comparisons between groups were performed with a linear mixed-effects model followed by post hoc analysis (see supplemental methods for details). (D) One representative image of an IF analysis of an FFPE section (out of 22 total sections analyzed) from one DiViD T1D donor (no. 5) stained with anti-MPO (green), anti-citrullinated histone H3 (red) antibodies and Hoechst 33342 for DNA detection (white). Scale bars: 10 μm.

Figure 3. Neutrophils infiltrating the pancreas of presymptomatic and symptomatic T1D donors are not merely localized close to β cells.

Representative images of IHC analyses of FFPE sections stained with hematoxylin and anti-MPO antibody (out of 12 total sections analyzed). Donor ID and characteristics are shown at the bottom. Arrows indicate MPO-positive cells (brown), while islets are highlighted by dotted lines (left panel), glucagon staining (middle panel), or immune cell infiltration (right panel). Scale bars: 10 μm.

Figure 4. Peripheral neutrophil transcriptomic signature from at-risk autoAb-negative, autoAb-positive, and new-onset T1D patients form a single large cluster distinct from that of unrelated nondiabetic controls.

Heatmap showing 2,058 genes differentially expressed in neutrophils freshly isolated from pediatric nondiabetic controls without relatives with T1D (unrelated nondiabetic controls, gray, n = 16), autoAb-negative relatives of patients with T1D (at-risk autoAb negative, blue, n = 13), autoAb-positive relatives of patients with T1D (at-risk autoAb positive, green, n = 8), and patients with T1D at disease onset (T1D new onset, orange, n = 5). Scale bar represents the relative gene expression and ranges from the minimum log expression (0.0) to the maximum log expression (1.0).

Figure 5. IFN-inducible gene expression signature characterizes circulating neutrophils during all stages of T1D.

(A) The most highly enriched terms from Gene Ontology Biological Processes in genes differentially expressed between at-risk subjects and unrelated nondiabetic controls are shown (terms shown have significant enrichment threshold of 0.1 for the Benjamini-Hochberg–adjusted P value). (B) Median expression of the Molecular Signatures Database (MSigDB) hallmark gene sets for response to IFN-α and IFN-γ, by donor group; points show values for individual donor (in the at-risk donor category: red dots represent autoAb-negative subjects, black dots autoAb-positive subjects). To test variation between the 3 groups, Kruskal-Wallis test was applied, and results are shown below each IFN set. To test differences between groups the Mann-Whitney test was applied (IFN-α: at-risk subjects vs. unrelated nondiabetic controls P = 0.059, T1D new onset vs. unrelated nondiabetic controls P = 0.032, T1D new onset vs. at-risk subjects P = 0.41; IFN-γ: at-risk subjects vs. unrelated nondiabetic controls P = 0.0015, T1D new onset vs. unrelated nondiabetic controls P = 0.032, T1D new onset vs. at-risk subjects P = 0.71). (C) Heatmap showing the 39 IFN-related genes differentially expressed in neutrophils freshly isolated from pediatric nondiabetic controls without relatives with T1D (unrelated nondiabetic controls, gray, n = 16) versus relatives of patients with T1D who are either autoAb negative or positive (at risk, green, n = 21). Patients with T1D at disease onset are shown for comparison (T1D new onset, orange, n = 5). Scale bar represents the relative gene expression and ranges from the minimum log expression (0.0) to the maximum log expression (1.0).

Figure 6. The IFN pathway is active during all stages of T1D.

Pathway diagram was constructed from the “Influenza A pathway” (KEGG PATHWAY Database hsa05164) and literature on IFN-related genes (described in detail in the supplemental methods section). Dysregulated transcripts are shaded in green (influenza pathway) or in red (IFN-related genes).