Type 1 Diabetes (T1D) is considered to be a T-helper- (Th-) 1 autoimmune disease; however, T1D pathogenesis likely involves many factors, and sufficient tools for autoreactive T cell detection for the study of this disease are currently lacking. In this study, using gene expression microarrays, we analysed the effect of diabetes-associated autoantigens on peripheral blood mononuclear cells (PBMCs) with the purpose of identifying (pre)diabetes-associated cell processes. Twelve patients with recent onset T1D, 18 first-degree relatives of the TD1 patients (DRL; 9/18 autoantibody positive), and 13 healthy controls (DV) were tested. PBMCs from these individuals were stimulated with a cocktail of diabetes-associated autoantigens (proinsulin, IA-2, and GAD65-derived peptides). After 72 hours, gene expression was evaluated by high-density gene microarray. The greatest number of functional differences was observed between relatives and controls (69 pathways), from which 15% of the pathways belonged to “immune response-related” processes. In the T1D versus controls comparison, more pathways (24%) were classified as “immune response-related.” Important pathways that were identified using data from the T1D versus controls comparison were pathways involving antigen presentation by MHCII, the activation of Th17 and Th22 responses, and cytoskeleton rearrangement-related processes. Genes involved in Th17 and TGF-beta cascades may represent novel, promising (pre)diabetes biomarkers.
Type 1 Diabetes (T1D) is considered to be a T-helper- (Th-) 1 autoimmune disease and is characterised by a lack of insulin, which is caused by the autoimmune destruction of insulin-producing pancreatic beta cells [
In patients presenting with recent T1D onset, there are various interventions that may stop, or at least delay, pancreatic beta cell destruction; however, these therapies are unable to reverse the patient’s lifelong dependency on insulin injections because beta cell proliferation and their capacity for regeneration are limited. To save sufficient beta cell masses, these therapies should be used in the clinically silent prediabetes phase; however, it is difficult to identify suitable candidates for such immunointervention [
The preclinical period is marked by the presence of autoantibodies against beta cell antigens, including insulin, glutamic acid decarboxylase-65 (GAD65), insulinoma-associated tyrosine phosphatase (IA-2), and zinc transporter 8 (ZnT8). The presence of these autoantibodies in the serum is highly predictive of T1D development [
The preclinical disease stage is characterised by the generation of activated, self-reactive lymphocytes that infiltrate the pancreas and selectively destroy the insulin-producing beta cells present in the islets [
Autoreactive T lymphocytes are present in peripheral blood at extremely low frequencies, and methods for their detection are still used for scientific, rather than clinical, purposes [
The last decade has ushered in a boom of “array techniques” that enable complex analyses of gene expression or protein production. These methods have also been used in T1D research to improve the prediction of T1D and increase the general knowledge of T1D pathogenesis [
In this paper, we report the identification of cell processes that may be important for the progression of prediabetes to diabetes. We isolated peripheral blood mononuclear cells (PBMCs) and subsequently stimulated these cells with a mixture of “T1D-associated” autoantigens. We compared the expression profiles of stimulated PBMCs and PBMCs that were cultivated for the same period in the absence of autoantigens to determine the effect of autoantigens on gene expression. We describe, at the level of gene expression, the differences in the immune responses among the tested groups that are predicted to be important in T1D pathogenesis. Genes involved in these cascades, or in the activation of these cascades, may serve as promising potential prediabetes biomarkers. In our analyses, we primarily concentrated on functional pathways and attempted to reveal differences in gene expression among the multitude of signalling pathways within which these genes operate.
The study population is described in Table
Study population.
Study group | No. of individuals | Age (years) |
Age |
Sex |
---|---|---|---|---|
T1D recent onset | 12 | 12; 3–41 | 12 |
M |
| ||||
First-degree relatives |
9 | 19; 5–52 | 5 |
F |
| ||||
First-degree relatives |
9 | 7; 3–21 | 7 |
F |
| ||||
Controls | 13 | 27; 14–42 | 14 |
M |
The sampling of patients with a recent T1D onset was performed after the metabolic stabilisation phase on the seventh day after diabetes diagnosis in the morning hours. Metabolic stabilisation is defined as the establishment of normoglycaemia and the normalisation of acid-base balance, biochemical parameters (such as ions and pH), and blood count parameters. Patients with severe diabetic ketoacidosis (
Approximately 17 mL of peripheral blood was obtained from the test subjects. PBMCs were isolated from whole venous blood by Ficoll density gradient centrifugation (Amersham Biosciences, Uppsala, Sweden) and were used in all
Total RNA from cultured cells was extracted using TRIzol reagent and a RiboPure kit (Invitrogen, Carlsbad, CA, USA), dissolved in 60
Microarray data processing and statistical analysis of differential gene expression was performed using the limma package in the R statistical environment (
We examined differences in gene expression and affected cellular pathways between all combinations of the three groups: normal controls, diabetic patients, and their relatives, who were divided according to their autoantibody statuses. We compared basal gene expression with gene expression following stimulation with the diabetogenic autoantigens. The top table genes according to limma analysis (
MetaCore is a proprietary, manually created database that analyses human protein-protein, protein-DNA, and protein-compound interactions, metabolic and signalling pathways, and the effects of bioactive molecules. This software generates interactive networks between user inputs and proteins and/or genes stored in the database. The software enables a user to analyse the distribution of canonical pathways, networks, GeneGo, and Gene Ontology processes, as well as the relevance of disease biomarkers in the tested samples. Canonical pathway maps represent a set of approximately 2,000 signalling and metabolic maps, comprehensively covering human biology. The content of approximately 110 cellular and molecular processes has been defined and annotated as GeneGo processes, and each process represents a preset network of interactions characteristic to the process. In this database, there are also over 500 human diseases with gene content annotated by GeneGo and organised in disease-specific folders, which are further organised into a hierarchical tree (
qRT-PCR was used to verify microarray data. Differences in the expression levels of CD4, signal transducer and activator of transcription 3 (STAT3), and TGF-beta 1 between RNA samples from PMBCs collected from an independent cohort of T1D and the controls were assessed. Specifically, expression was analysed in a cohort of 14 newly diagnosed patients with T1D (7M/7F, mean age 8,6 years, median 9,1, range 1,7–17,2 years) and 12 control volunteers (5 M/7F, mean age 10,7 years, median 11,2, range 2,1–18,7 years) using TaqMan Gene Expression assays (Lifetechnologies, Carlsbad, CA, USA). Total RNA was extracted using TRIzol reagent, according to the manufacturer’s recommendations (Lifetechnologies, Carlsbad, CA, USA). cDNA was synthesised according to recommendations by Lifetechnologies using the High Capacity RNA-to-cDNA Master Mix (Lifetechnologies, Carlsbad, CA, USA). Experiments were analysed using a LightCycler 480 Real-Time PCR System (Roche, Basel, Switzerland). A comparative ΔΔ cycle threshold (Ct) was used for quantification of relative mRNA levels. The expression of CD4 using commercially available primers (cat. no. Hs01058407_m1), STAT3 (cat. no. Hs00427259_m1), and TGF-beta (cat. no. Hs00171257_m1) was normalised to beta-glucuronidase (GUSB, cat. no. Hs99999908_m1).
Data from the qRT-PCR were analysed using the R programme. An unpaired, two-tailed Student’s
Table
The number of identified genes with different expression levels when the various test groups were subjected to pair group comparisons.
Comparison | Total no. of sign. differentially activated genes | No. of sign. |
No. of sign. |
---|---|---|---|
DRL versus D | 2222 | 1513 | 709 |
DV versus D | 1318 | 896 | 422 |
DRL versus DV | 1347 | 955 | 392 |
D: T1D patients; DRL: first-degree relatives of T1D patients; DV: controls (healthy volunteers).
Verification of gene microarray data. Relative expression of TGF beta1, STAT3, and CD4 by qRT-PCR (data were obtained from an independent cohort of 14 newly diagnosed patients with T1D and 12 healthy volunteers).
Interestingly, the highest number of differentially expressed genes (2,222;
An enhanced gene expression heatmap was constructed using probe signal intensities that had a log fold change that was greater than +1 or less than −1 (Figure
Genes differentially activated in each group (cluster formation). The enhanced gene expression heatmap was constructed using probe signal intensities that had a log fold change that was greater than +1 or less than −1. Genes that were significantly altered in the relatives group clustered into specific gene families.
The top ten canonical pathways that changed most significantly in the pair-wise comparisons are listed in Table
(a) GeneGo pathway top 10 maps (“immune response pathways” are in bold). (b) The complete list of significant “immune response pathways” for each pair comparison; pathway rankings (the position of each pathway within the list) are indicated.
T1D (D) patients versus healthy controls (DV) | Relatives of T1D patients (DRL) versus healthy controls (DV) | T1D (D) patients versus relatives of T1D patients (DRL) |
---|---|---|
|
|
(1) Cytoskeleton remodeling_CDC42 in cellular processes |
T1D (D) patients versus healthy controls (DV) | Relatives of T1D patients (DRL) versus healthy controls (DV) | T1D (D) patients versus relatives of T1D patients (DRL) |
---|---|---|
1: Antigen presentation by MHCII |
1: MIF-JAB1 signalling |
24: Cytokine production by Th17 cells |
D: T1D patients; DRL: first-degree relatives of T1D patients; DV: controls (healthy volunteers).
The greatest number of differences for pathways that were altered, specifically 69 pathways, was observed when relatives were compared to controls. Of these pathways, 15% belonged to “Immune response pathways.” However, the highest percentage (24%) of significant differences in immune response-related pathways was observed when patients with T1D were compared with healthy controls (11 out of 46 pathways), with “Antigen presentation by MHCII” as the highest scoring pathway. An important variable appeared to be Th17 lymphocyte activation, as we observed a difference in “Th17 cell differentiation” among the groups. Specifically, differences in Th17 polarisation were observed when relatives were compared with patients. The Th17 cell differentiation pathway is shown in Figure
Immune response and Th17 cell differentiation. Differences in Th17 polarisation were observed when controls were compared with T1D patients using microarray data. Genes of interest were analysed by qRT-PCR and were found to be upregulated in T1D patients. STAT3 and TGF-beta were chosen as representatives of Th17 cell differentiation. Microarray data demonstrated that CD4 was one of the most significantly upregulated molecules in T1D patients.
Immunologic responsiveness in relatives was similar to the responsiveness observed in patients with T1D. However, only 7% of the differentially activated pathways could be classified as “immune response-related” (i.e., 4 pathways out of 54 differentially activated pathways). Within these pathways, cell cascades related to Th17 polarisation and the action of the immunoregulatory cytokine TGF-beta were also identified.
Upon activation and expansion, naive CD4+ T cells develop into different Th cell subsets that exhibit different cytokine profiles and effector functions to protect the body against different types of pathogens. Until recently, T cells were divided into Th1 and Th2 cells, depending on the cytokines they produced (e.g., IFN-gamma and TNF-beta versus IL-4, -5, and -13, resp.).
A third subset of IL-17-producing effector Th cells called Th17 cells has recently been discovered. The participation of TGF-beta in Th17 cell differentiation places the Th17 lineage in close relationship with CD4+CD25+Foxp3+ regulatory T cells (Tregs) [
T1D is an autoimmune disease that results from the selective destruction of pancreatic beta-cells by T cells, and the development of this disease is most likely due to the interaction between environmental and genetic factors. CD4+ T cells are largely implicated in the pathogenesis of this disease, and T1D is believed to be a predominantly Th1-driven disease. Moreover, increased IL-17 expression has been detected in the sera and target tissues of patients with various autoimmune diseases, and in animal models, IL-23, a Th17 stabilisation factor, is involved in the development of autoimmune diabetes. The differentiation of Th17 cells is initiated by TGF-beta, IL-6, and IL-21, which activate STAT3 and induce the expression of transcription factors, including retinoic acid related orphan receptor (RORgamma t). In humans, Th17 activity seems to cause multiorgan inflammation, contributing to the manifestation of rheumatoid arthritis, inflammatory bowel disease, and celiac disease [
In this unique study on gene expression and functional analysis, we demonstrated that the “Th17 differentiation,” “IL-22 signalling,” and “Development of TGF-beta receptor signalling” pathways were among the most significantly different pathways identified when patients with T1D were compared with healthy controls. A difference in “Th17 signalling” pathway activation was also observed when we compared T1D patients with relatives. Consistent with these data, we previously reported that a bias in IL-10 and TGF-beta production at the protein level is typical of the prediabetes phase [
Using a murine model of the disease, two groups previously reported that the transfer of islet-specific Th17 cells induced diabetes, although this effect was apparent only after the cells had converted to IFN-producing cells [
IL-9 is a T cell-derived cytokine that was initially characterised as a Th2 cytokine. The secretion of IL-9 was recently attributed to a novel CD4+ T cell subset termed Th9 cells in mice. However, IL-9 can also be secreted by mouse Th17 cells and may mediate aspects of the proinflammatory activities of Th17 cells [
Not surprisingly, the highest scoring pathway in the comparison of patients with T1D versus their healthy counterparts was “Antigen presentation by MHCII”; indeed, it is well known that genes encoding HLA class II molecules are the most important “T1D-associated genes” [
Glucose-stimulated insulin secretion from islet beta-cells involves secretory granule transport, a highly coordinated process that involves changes in cytoskeletal architecture with the help of G proteins and their respective effector molecules. Small G proteins include Cdc42, Rac1, and ARF-6, with corresponding regulatory factors including GDP/GTP-exchange factors and GDP-dissociation inhibitors. In addition to their positive modulatory roles, certain small G proteins also contribute to the metabolic dysfunction and the demise of islet beta-cells that has been observed in
The bone morphogenic protein (BMP) signalling pathway also appeared on the list of differentially activated pathways when patients were compared with controls and also with relatives. It is well known that diabetic nephropathy is a leading cause of end-stage renal disease. Additionally, the TGF-beta-BMP pathway has been implicated in the pathogenesis of diabetic nephropathy. The BMP2, BMP4, and BMP7 genes are located near linkage peaks for renal dysfunction, and it was hypothesised that genetic polymorphisms in these biological and positional candidate genes may constitute a risk factor for diabetic kidney disease; however, common BMP gene polymorphisms do not strongly influence genetic susceptibility to diabetic nephropathy in white individuals with T1D [
There have been only a limited number of T1D gene expression studies. One example is the report by Kaizer and colleagues [
One potential drawback to our study is the limited number of samples tested. However, we believe that approximately ten subjects per group are sufficient to reveal genes with statistically significant alterations in their gene expression levels when high-density microarray chips are used. In this context and in many other aspects, the results of this study parallel our previous work [
In conclusion, we can summarise that important differences were observed when the activation of cell processes following artificial exposure to diabetes-related autoantigens was compared among T1D patients, their first-degree relatives, and healthy controls. Important immune response-related pathways were involved, with a high degree of variability observed for these pathways when either patients with T1D or their relatives were compared with healthy controls. These important immune response-related processes largely included the induction of Th17 and Th22 responses, as well as cytoskeletal rearrangements, MHCII presentation, and the upregulation of CD4, TGF-beta, and STAT3. These findings potentially suggest that these processes could be utilised as predictive markers for the development of T1D or as molecular targets for the repression of specific immunocompetent cell populations for the treatment of diabetes.
T1D patients
First-degree relatives of T1D patients
Relatives of T1D patients who are autoantibody(ies) negative
Relatives of T1D patients who are autoantibody(ies) positive
Controls (healthy volunteers)
Fold change
Antiglutamic acid decarboxylase (GAD65) autoantibodies
Antityrosin phosphatase (IA-2) autoantibodies
Insulin autoantibodies.
Radek Blatny, Zbynek Halbhuber, Michal Kolar, Dominik Filipp, and Katerina Stechova contributed equally to this work.
This work was supported by project NPVII 2B06019 from the Czech Ministry of Education and partially by Grant RVO: 68378050 from the Academy of Sciences of the Czech Republic, as well as by the Ministry of Health of the Czech Republic through the conceptual development of research organisation 00064203 (University Hospital Motol, Prague, Czech Republic) and by IPL 699 001 (Department of Paediatrics, 2nd Faculty of Medicine, Prague, Czech Republic). The authors also like to acknowledge Miluse Hubackova, Vendula Stavikova, and Barbora Obermannova for the sampling of study subjects and samples processing.