Gene expression profiles for the human pancreas and purified islets in Type 1 diabetes: new findings at clinical onset and in long-standing diabetes

Abstract

Type 1 diabetes (T1D) is caused by the selective destruction of the insulin-producing β cells of the pancreas by an autoimmune response. Due to ethical and practical difficulties, the features of the destructive process are known from a small number of observations, and transcriptomic data are remarkably missing. Here we report whole genome transcript analysis validated by quantitative reverse transcription–polymerase chain reaction (qRT–PCR) and correlated with immunohistological observations for four T1D pancreases (collected 5 days, 9 months, 8 and 10 years after diagnosis) and for purified islets from two of them. Collectively, the expression profile of immune response and inflammatory genes confirmed the current views on the immunopathogenesis of diabetes and showed similarities with other autoimmune diseases; for example, an interferon signature was detected. The data also supported the concept that the autoimmune process is maintained and balanced partially by regeneration and regulatory pathway activation, e.g. non-classical class I human leucocyte antigen and leucocyte immunoglobulin-like receptor, subfamily B1 (LILRB1). Changes in gene expression in islets were confined mainly to endocrine and neural genes, some of which are T1D autoantigens. By contrast, these islets showed only a few overexpressed immune system genes, among which bioinformatic analysis pointed to chemokine (C-C motif) receptor 5 (CCR5) and chemokine (CXC motif) receptor 4) (CXCR4) chemokine pathway activation. Remarkably, the expression of genes of innate immunity, complement, chemokines, immunoglobulin and regeneration genes was maintained or even increased in the long-standing cases. Transcriptomic data favour the view that T1D is caused by a chronic inflammatory process with a strong participation of innate immunity that progresses in spite of the regulatory and regenerative mechanisms.

Introduction

Type 1 diabetes (T1D) is an organ-specific autoimmune disease that results from autoimmune destruction of insulin-producing β cells in the pancreatic islets of Langerhans [1]. Due to ethical and practical problems that limit access to these tissues, current understanding of the destructive processes in the diabetic pancreas is limited to morphological and immunopathological observations made in a small number of autopsies [2–9] and biopsies [10] of T1D patients, including a previous study from our laboratory. These studies have shown β cell destruction and islet infiltration by lymphocytes, macrophages, natural killer and dendritic cells and over-expression of a short list of immune system genes and, in some cases, of Coxsackie virus proteins [8,11]. Unfortunately, there is no information on the global changes at the molecular level that occur in situ in T1D patients, from studies of peripheral blood [12,13] and sera [14]. To interpret the pathology more clearly and to advance in our understanding of the processes that lead to islet destruction it would be important to have detailed information of all the changes occurring at the molecular level in situ.

During the last 20 years, and in part through a programme of clinical islet transplantation and sample exchange with other groups, we have collected pancreases from multi-organ donors and from four T1D patients who died at different stages of the disease.

Here we report whole genome transcript analysis validated by quantitative reverse transcription–polymerase chain reaction (qRT–PCR) and correlated with immunohistological observations for these four pancreases and for purified islet preparations from two of them. This study provides an insight into the complex pattern of the immune gene expression disturbances that occur in the diabetic pancreas, in addition to confirming the predicted immunopathological mechanisms that highlight a number of natural immunity–inflammatory, immunoregulatory and regenerative pathways, some of which have received relatively little attention so far.

Materials and methods

Patients

Pancreases were obtained from four patients who died at different T1D stages. Case 1 (clinical onset) was a 19-year-old woman who died 5 days after T1D diagnosis [5], human leucocyte antigen (HLA) type HLA-A*02, *68; B*39, *49; Cw*03, *12; DRB1*04, *13; DQB1*0302, *0603; case 2 (recent onset) was a 16-year-old man who died 9 months after T1D onset, HLA type HLA-A*02, *24; B*39, *-; Cw*07, *-; DRB1*07, *08; DQB1*0202, *0402; case 3 (long-standing) was a 29-year-old man who died 8 years after T1D onset, HLA type HLA-A*01, *29; B*07, *08; Cw*07, *-; DRB1*03, *-; DQB1*0201, *-; and case 4 (longstanding) was a 26-year-old man who died 10 years after T1D onset [5], HLA type HLA-A*01, *30; B*18, *40; Cw*03, *05; DRB1*03, *04; DQB1*0201, *0302. No clinical features associated with other autoimmune diseases were present in these patients. As controls, pancreases from seven organ donors (four men and three women, age range 19–53 years, negative for islet cell antibodies) were used. The body mass index (BMI) was between 20 and 25 for all patients and controls (data are summarized in Table 1). The protocols were approved by the Ethical Committee of the University Hospital Germans Trias i Pujol, in accordance with the Declaration of Helsinki.

Table 1.

Clinical data for Type 1 diabetes (T1D) patients and controls used in gene expression studies.

Subject	T1D	Age	Gender	COD	BMI	ICA	Glic	Risk alleles	Pancreas	Islets/purity
Case 1	5 days	19	Female	BE	20·9	+	648	B39 DR4–DQ8	Yes	Yes > 90%
Case 2	9 months	16	Male	BE	< 25	n.d.	285	B39	Yes	n.a.
Case 3	8 years	29	Male	BE	21·9	n.d.	146	DR3–DQ2	Yes	n.a.
Case 4	10 years	26	Male	HT	23·4	n.d.	n.a.	DR4DQ8/DR3DQ2	Yes	Yes > 90%
Control 1	Healthy	30	Female	HT	22·9	NEG	257	NEG	Yes	n.a.
Control 2	Healthy	21	Male	HT	23·1	NEG	133	DR4–DQ8	Yes	n.a.
Control 3	Healthy	20	Female	CA	20·2	NEG	271	NEG	Yes	n.a.
Control 4	Healthy	20	Female	HT	< 25	NEG	NA	n.a.	n.a.	Yes > 80%
Control 5	Healthy	19	Male	HT	23·5	NEG	299	n.a.	n.a.	Yes > 90%
Control 6	Healthy	48	Male	HT	< 25	NEG	126	n.a.	n.a.	Yes > 90%
Control 7	Healthy	53	Male	BE	24·1	NEG	NA	n.a.	n.a.	Yes > 90%

Age is given in years. COD: cause of death; HT: head trauma; BE: brain oedema; CA: cardiorespiratory arrest; n.a.: not available; n.d.: not done; NEG: negative; Glic: gycaemia in mg/dl.

Pancreas processing and islet isolation

As soon as the tissue samples reached the laboratory they were divided into two portions: approximately three-quarters of each gland underwent digestion, while one-quarter was cut into small cubes (0·5 cm³) that were snap-frozen. Islets from four of the controls and from cases 1 and 4 were isolated from the fresh pancreas tails by enzymatic digestion using a standard automated method [15], selected manually under a stereoscopic microscope and snap-frozen. Samples were kept in liquid nitrogen until RNA extraction.

Sample processing and chip hybridization

RNA was obtained from each T1D sample and controls (whole pancreas and purified islets) using RNeasy Micro (Qiagen, Hilden, Germany). RNA quality (2100 Bioanalyzer, Agilent Technologies Inc., Santa Clara, CA, USA) was optimal for microarray experiments (RNA integrity number 6·2–7·9). cDNA was synthesized with 3·5–5 mg of total RNA using the One Cycle cDNA Synthesis Kit (Affymetrix, Inc., Santa Clara, CA, USA), purified (GeneChip® Sample Cleanup Module; Affymetrix), fragmented and checked to verify the integrity. Human Genome U133 Plus 2·0 Arrays (38 500 genes) were hybridized and scanned by an Affymetrix G3000 GeneArray Scanner and processed with Microarray Analysis Suite 5·0.

Statistical analyses for microarrays

Pre-processing

Raw expression values obtained from .CEL files were pre-processed using the robust multi-array average (RMA) method (Affymetrix) [16]. These normalized values were used for all subsequent analyses. Data were subjected to non-specific filtering to remove low signal and low variability genes. Conservative (low) thresholds were used to reduce possible false negative results. Selection of differentially expressed genes was based on a linear model analysis with empirical Bayes modification for the variance estimates, as described previously [17]. This method is similar to using a ‘t-test’ with an improved estimate of the variance. To account for multiple testing probability effects arising when many tests (one per gene) are performed simultaneously, P-values were adjusted to obtain stronger control over the false discovery rate [18]. Genes were considered to be expressed differentially based on the following criteria. Genes with a P-value ≤ 0·05, adjusted P-value ≤ 0·26 and fold change (FC) ≥ 1·75 (log₂FC ≥ 0·8) were considered up-regulated, and genes with FC ≤ −1·46 (log₂FC ≤ −0·54) were considered down-regulated. The chosen criterion was P-value ≤ 0·05. The ingenuity pathway analysis (IPA) (Ingenuity Systems®, Redwood City, CA, USA) was used to identify the canonical pathways from the IPA library that were most significant to the data sets.

qRT–PCR

RNA (1 µg for whole pancreas and 200 ng for purified islets) was reverse transcribed with Moloney murine leukemia virus (MMLV) (Promega, Madison, WI, USA). qRT–PCR was performed on a LightCycler® 480 (Roche Diagnostics, Mannheim, Germany) using SYBR® Green I Master mix (Roche Diagnostics) and specific primers (Invitrogen, Life Technologies, Gaithersburg, MD, USA) (Table 2). We selected 21 genes with P-values < 0·05 on the basis of their levels of differential expression and biological relevance: NMES1, CRP, REG3A, TLR3, CD44, HLA-E, CD36, S100A8, TNFSF10 (TRAIL), IFNAR2, IL8, CXCL1, INS, PNLIP, ERAP2, PPY, REG4, LYZ, CXCL12, HSPA6 and INSR. Relative quantification was determined by normalizing the expression for each gene to the HPRT gene following the 2^−ΔΔCt method [21]. HPRT was selected as a housekeeping gene among seven candidate genes because it showed the most constant expression levels for both normal and pathological samples. All measurements were performed in triplicate in three separate runs and expressed as mean ± standard error of the mean (s.e.m.). Statistical analysis used a t-test for pancreas samples and a one-sample t-test for islet samples.

Table 2.

List of primers used for quantitiative reverse transcription–polymerase chain reaction.

Gene	Sense	Anti-sense	Amplimer length
NMES1	5′-ATTCCCTTGGTGGTGTTCAT-3′	5′-TTTTTTTTTCGATCAAGGATCA-3′	101
CRP	5′-TCCTATGTATCCCTCAAAGCA-3′	5′-GGTGGCATACGAGAAAATACT-3′	117
REG3A	5′-CCTGTCGAGAAGCACAGC-3′	5′-GGGAGGAAGAAACAGAAGAAA-3′	232
TLR3	5′-ACACCATCTCAAAACTGGA-3′	5′-AAACACCCTGGAGAAAACT-3′	397
CD44	5′-TGAGCATCGGATTTGAGAC-3′	5′-GTGTCATACTGGGAGGTGTT-3′	139
HLA-E	5′-TCTCCGAGCAAAAGTCAAA-3′	5′-GAGATGGGGTGGTGAGTC-3′	160
TRAIL	5′-GTCTCTCTGTGTGGCTGTAA-3′	5′-CTCAAAATCATCTTTCTAACG-3′	201
IFNAR2	5′-AACACGAACTACTGTGTATCT-3′	5′-GTGTCACTATGGTGCTTG-3′	184
IL-8	5′-TACTCCAAACCTTTCCACC-3′	5′-AACCCTACAACAGACCCAC-3′	288
CXCL1	5′-GTTAATATTTCTGAGGAGCCTGC-3′	5′-AAACAGTTACAAAACAGATGTGC-3′	128
INS†	5′-AAGCGTGGCATTGTGGAAC-3′	5′-CAAGGGCTTTATTCCATCTCTC-3′	138
PNLIP	5′-GAGCAGTAGCAGGAAAAGAAGT-3′	5′-TTAGTATATAGGAGGAAGCGGG-3′	145
CD36	5′-GACAGTTTTGGATCTTTGATG-3′	5′-CTTTGCTTAACTTGAATGTTG-3′	76
ERAP2	5′-TCTACTATCCACTCTCCAAACT-3′	5′-GTCTCCCTATATGTAATGAGG-3′	126
S100A8	5′-GATGACCTGAAGAAATTGCT-3′	5′-ATATCCAACTCTTTGAACCAG-3′	86
PPY	5′-TGCCCAGGGAGCCCCACTG-3′	5′-CTAGGCCTGGTCAGCATGTTGA-3′	114
HSPA6	5′-GAGATGAACTTTCCCTCCAAAGC-3′	5′-TTGAAGCAGAAGAGGATGAACCG-3′	98
LYZ	5′-GGGCTTGTCCTCCTTTCTGTT-3′	5′-GTTGTAACCACTCTCCCATTTG-3′	150
CXCL12	5′-AACGTCAAGCATCTCAAAATT-3′	5′-CTTGTCTGTTGTTGTTCTTCA-3′	82
REG4	5′-GGGTGGGAACAAGCACTG-3′	5′-GTCTCTAAGCCTAAAAAAGCC-3′	260
INSR	5′-GGAACTACAGCGTGCGAAT-3′	5′-GAAGAAGCGTAAAGCGGTC-3′	229
HPRT‡	5′-TGACACTGGCAAAACAATGCA-3′	5′-GGTCCTTTTCACCAGCAAGCT-3′	94

^†INS and

^‡HPRT primers are described in Sabater L et al.[19] and Vandesompele J et al.[20].

Immunohistological analysis

Frozen sections were evaluated using a double immunofluorescence (IFL) technique. Sections (5 mm) were first incubated with primary monoclonal antibodies (mAb): anti-CD44 (hCD44H; R&D Systems, Abingdon, UK), anti-HLA-E (clone 3D12; Dr D. Geraghty, Fred Hutchinson Cancer Research Center, Seattle, WA, USA), anti-C-reactive protein (CRP) and anti-REG3A (Santa Cruz Biotechnology, Santa Cruz, CA, USA), anti-CD68 (Dako, Glostrup, Denmark), anti-CD36 and anti-CD209 (BD Pharmingen, San Diego, CA, USA), anti-CD3 and anti-CD4 (Professor P. Beverley, London, UK), anti-CD45, anti-CD4 and anti-CD19 (Dr R. Vilella, Hospital Clinic, Barcelona, Spain) or guinea pig anti-insulin (ICN, Aurora, OH, USA). Primary antibodies were followed by either Alexa 488 goat anti-mouse immunoglobulin (Ig)G1 or IgG2a, Alexa 488 rabbit anti-goat Ig, Alexa 546 goat anti-mouse IgG2b (Molecular Probes, Leiden, the Netherlands) or by tetramethyl rhodamine isothiocyanate (TRITC) goat anti-guinea pig (ICN). Second staining was performed using a mAb either to glucagon Glu001 (Dr P. Jorgensen, Novo Nordisk A/S, Bagsvaerd, Denmark) or anti-glutamic acid decarboxylase (GAD) (GAD6; Hybridoma Bank, Baltimore, MD, USA) and either TRITC goat anti-mouse IgG1 or IgG2a, respectively (SBA, Birmingham, AL, USA). The preparations were examined in a fluorescence microscope and analysed using software Openlab version 2·0 (Improvision, Coventry, UK). To phenotype the leucocytes that infiltrated the islets in the four diabetic cases, sections from five different areas from two to three different tissue blocks were stained. The staining for CD45 served as the indicator of the number of islets with insulitis. Peri-insulitis or insulitis were considered when lymphoid cells infiltrate the pancreas but remain outside the islets or infiltrate the islets, respectively. The numbers and percentages of CD3⁺, CD4⁺, CD8⁺, CD19⁺, CD68⁺ and CD209 -DC-SIGN-⁺ cells were counted when at the periphery or inside the islet. Between 50 and 100 islets per case were examined and all the islets (with or without infiltrating leucocytes) were scored. Analysis of variance used Tukey's multiple comparisons test, with P-values < 0·05 were considered significant.

Results

Immunohistological analysis of insulitis

Most of the islets (91–100%) in the pancreas of the four cases showed mild insulitis or peri-insulitis that was less intense but present even in cases 3 and 4 (Table 3). Phenotypes among the infiltrating cells followed the hierarchy CD3⁺CD8⁺ >> CD68 ≥ CD19 ≥ CD3⁺CD4⁺ > CD209⁺. Case 1 insulitis showed a significantly higher number of CD3⁺ cells when compared to case 2 (P < 0·05) and to cases 3 and 4 (P < 0·001 for both) and of CD8⁺ cells in case 1 versus case 4 (P < 0·05). By contrast, the number of CD4⁺ and B cells, macrophages and dendritic cells was maintained overall. Cases 1 and 4 have already been partially characterized and reported [5]. These immunopathological data confirm that the four cases had typical T1D and provide the framework for the interpretation of the transcriptomic analysis.

Table 3.

Insulitis characteristics in the islets from four Type 1 diabetes (T1D) patients.

	CD3+		CD4+		CD8+		CD19+		CD68+		DC-SIGN+
Case	Positive islets	Cells/islet, mean ± s.e.m.	Positive islets	Cells/islet, mean ± s.e.m.	Positive islets	Cells/islet, mean ± s.e.m.	Positive islets	Cells/islet, mean ± s.e.m.	Positive islets	Cells/islet, mean ± s.e.m.	Positive islets	Cells/islet, mean ± s.e.m.
1	100%	12·17 ± 2·89	69%	1·69 ± 0·51	93%	9·28 ± 2·81	71%	1·66 ± 0·35	100%	6·85 ± 1·16	24%	0·41 ± 0·19
2	91%	8·06 ± 0·89	74%	2·95 ± 0·95	95%	4·93 ± 0·73	79%	2·88 ± 0·47	98%	4·66 ± 0·55	27%	0·39 ± 0·10
3	81%	2·09 ± 0·19	57%	1·31 ± 0·27	91%	2·04 ± 0·27	80%	2·47 ± 0·32	82%	3·06 ± 0·48	28%	0·38 ± 0·07
4	91%	4·09 ± 0·93	28%	0·64 ± 0·16	63%	1·44 ± 0·31	76%	1·68 ± 0·25	100%	4·11 ± 0·99	67%	1·00 ± 0·26

s.e.m.: Standard error of the mean.

Transcriptomic profile of diabetic pancreas and purified islet cells

Four cRNA preparations from different blocks from each of the diabetic pancreases, and one from each the three control pancreases were hybridized to Affymetrix arrays (U133 Plus version 2·0). cRNA from purified islet preparations corresponding to cases 1 and 4, and from three control islet preparations were also hybridized. Following recommended procedures for the analysis of small number of samples [22], the average gene expression level of three blocks from each diabetic pancreas was compared to the average of the three control pancreases and cases 1 and 4 islet gene expression levels were compared to the average of the control islet preparations.

The number of differentially expressed genes, both up- and down-regulated, is given in Table 4. A total of 635–1444 differentially expressed genes were detected in the T1D pancreases, 149 shared in the four cases. Among these, 44 (29%) genes fell into the category of immune system and five (3·4%) into endocrine system. Approximately 900 differentially expressed genes were detected in T1D islet preparations, of which 423 were shared in the two cases. Among these, only 20 (4·7%) were classified as immune system while 66 (15·6%) were related to endocrine function (Fig. 1). Raw data are available at the ArrayExpress repository, European Bioinformatics Institute (http://www.ebi.ac.uk/arrayexpress, ArrayExpress accession ID: E-MEXP-1140).

Table 4.

Number (percentage) of differentially expressed genes in the pancreas and islets of Type 1 diabetes (T1D) patients.

	Pancreas					Purified islets			Pancreas and islets
	Case 1	Case 2	Case 3	Case 4	Shared	Case 1	Case 4	Shared	Shared
Differentially expressed	851	1058	1444	635	149	902	975	423	4
Up-regulated	779 (92%)	1003 (95%)	1157 (80%)	613 (97%)	146	322 (36%)	178 (18%)	65	2
Down-regulated	72 (8%)	55 (5%)	287 (20%)	22 (3%)	3	580 (64%)	797 (82%)	358	2

Fig. 1.

Distributions of differentially expressed genes in the Type 1 diabetes (T1D) pancreas and islets by functional categories. The numbers of differentially expressed genes in each category for each sample are shown. Only the 15 most represented categories are given. Each stacked bar corresponds to a different sample and is divided into up-regulated genes (red) and down-regulated genes (green). From left to right: pancreas case 1, pancreas case 2, pancreas case 3, pancreas case 4, islets case 1, islets case 4. (a) Up- and down-regulated genes in general categories and (b) up- and down-regulated genes of immune response subcategory.

Collectively, these results show a divergent pattern in the changes of gene expression in pancreas and islets; while in the former most changes are increases in immune system genes, in the islets the predominant change is a reduction in the expression of endocrine and neural function genes. This divergence is also qualitative, as few of the differentially expressed genes are shared between pancreases and islets.

Analysis of differentially expressed genes in T1D pancreases and islets

The genome localization of the genes in the top rank of the over- and under-expressed genes were included in two genetic regions linked more strongly to T1D and known as insulin-dependent diabetes class 1 (IDDM1) (HLA) and IDDM2 [insulin (INS)]. However, no other significant relationship between differentially expressed genes and defined IDDM predisposing loci were found consistently in the four cases.

The category that contained the largest number of differential expressed genes in the diabetic pancreas versus controls was immune response [between 15% and 21% while 6% in the gene ontology (GO) biological process immune system]. GO categories related to immune system were enriched in T1D pancreases, e.g. GO:0006955, GO:0006959, GO:0006935 and GO:0006954 (P < 0·01). Within it, the most over-represented subcategories were: (i) antigenic presentation; (ii) chemotaxis; (iii) innate immunity and inflammation; (iv) complement; (v) immunoregulation; (vi) adhesion molecules; (vii) interferon (IFN) responsive; and (viii) leucocyte (Fig. 1b). The heatmap analysis for each of them is represented in Fig. 2a–h.

Fig. 2.

Heatmaps of gene expression profiles of the immune system in pancreas and purified islets from Type 1 diabetes (T1D) patients. Rows are for differentially expressed genes and columns are for pancreases and purified islets from T1D patients. Data were transformed to log₂ ratios relative to the mean of the normal controls and subjected to hierarchical clustering. The colour gradient key reflects relative expression on a log₂ scale. The most over-represented immune response subcategories are antigen presentation (a), chemotaxis (b), innate immunity and inflammation (c), complement (d), immunoregulation (e), adhesion molecules (f), interferon responsive (g) and leucocytes (h). P1, pancreas from case 1; P2, pancreas from case 2; P3, pancreas from case 3; P4, pancreas from case 4; I1, islets from case 1; I4, islets from case 4.

Changes in gene expression affected 19 of the 47 genes included in the Type 1 diabetes pathway of the Kyoto Encyclopaedia of Genes and Genome (http://www.genome.jp/kegg/pathway/hsa/hsa04940.html), thus supporting the specificity of the changes observed.

The ingenuity pathway analysis (IPA) identified 28 different canonical pathways among the ‘top 10’ pathways for each pancreas (Table 5). Antigenic presentation, allograft rejection, hepatic fibrosis, complement system, acute phase response signalling and autoimmune thyroid disease signalling were associated significantly in most cases. Altered pathways in purified islets from cases 1 and 4 were related to nervous system, signalling and endocrine functions.

Table 5.

Top 10 canonical pathways identified by ingenuity pathway analysis (IPA) for each case (pancreas and islets).

Sample	Top 10 IPA canonical pathways (P-value)
Pancreas case 1	Antigen presentation pathway (6·31E-14)
	Complement system (3·98E-12)
	Allograft rejection signalling (1·41E-06)
	Autoimmune thyroid disease signalling (2·57E-06)
	Acute phase response signalling (7·41E-06)
	Graft-versus-host disease signalling (7·41E-06)
	Glutathione metabolism (3·98E-05)
	Metabolism of xenobiotics by cytochrome P450 (0·0001)
	Systemic lupus erythematosus signalling (0·0003)
	Coagulation system (0·0004)
Pancreas case 2	Antigen presentation pathway (2·00E-15)
	Complement system (3·98E-12)
	Hepatic fibrosis/hepatic stellate cell activation (3·16E-11)
	Dendritic cell maturation (5·89E-08)
	Acute phase response signalling (8·32E-08)
	Type I diabetes mellitus signalling (3·72E-06)
	Caveolar-mediated endocytosis signalling (4·79E-06)
	Allograft rejection signalling (7·41E-06)
	Autoimmune thyroid disease signalling (1·32E-05)
	NF-κB signalling (1·55E-05)
Pancreas case 3	Complement system (1·26E-12)
	Antigen presentation pathway (6·31E-12)
	Hepatic fibrosis/hepatic stellate cell activation (1·10E-07)
	Acute phase response signalling (2·45E-05)
	Leucocyte extravasation signalling (3·16E-05)
	Role of NFAT in regulation of the immune response (7·41E-05)
	Allograft rejection signalling (8·51E-05)
	Dendritic cell maturation (0·0001)
	Autoimmune thyroid disease signalling (0·0001)
	NRF2-mediated oxidative stress response (0·0002)
Pancreas case 4	Hepatic fibrosis/hepatic stellate cell activation (5·25E-05)
	Glucocorticoid receptor signalling (0·0006)
	MIF regulation of innate immunity (0·0007)
	Acute phase response signalling (0·0008)
	Chemokine signalling (0·0008)
	p38 MAPK signalling (0·0011)
	CCR5 signalling in macrophages (0·0017)
	Growth hormone signalling (0·0023)
	IL-17 signalling (0·0038)
	HMGB1 signalling (0·0044)
Islets case 1	Cardiac β-adrenergic signalling (6·03E-06)
	Maturity onset diabetes of young (MODY) signalling (7·41E-05)
	NRF2-mediated oxidative stress response (0·0004)
	cAMP-mediated signalling (0·0004)
	CCR5 signalling in macrophages (0·0008)
	CXCR4 signalling (0·0011)
	Relaxin signalling (0·0012)
	Corticotrophin releasing hormone signalling (0·0013)
	G protein signalling mediated by tubby (0·0013)
	α-adrenergic signalling (0·0029)
Islets case 4	Cardiac β-adrenergic signalling (1·07E-06)
	Axonal guidance signalling (3·89E-06)
	cAMP-mediated signalling (3·98E-05)
	MODY signalling (0·0001)
	Cellular effects of sildenafil (0·0002)
	Synaptic long-term potentiation (0·0002)
	Melatonin signalling (0·0002)
	G-protein coupled receptor signalling (0·0002)
	CREB signalling in neurones (0·0003)
	Calcium signalling (0·0003)

cAMP: cyclic adenosine-5′-monophosphate; CCR5: chemokine (C-C motif) receptor 5; CXCR4: chemokine (CXC motif) receptor 4; CREB: cAMP responsive element binding protein; HMGB1: high mobility group box 1; IL: interleukin; MAPK: mitogen activated protein kinase; MIF: macrophage migration inhibitory factor; NFAT: nuclear factor of activated T cells; NFkB: nuclear factor kappa B; NRF2: nuclear factor-like 2.

Total pancreas from T1D patients

Changes were analysed in order of magnitude, as they appear in Figs 2 and 3. Immune system differentially expressed genes are shown in Fig. 2a–h and Table 7 as follows.

Fig. 3.

Heatmaps of islet (a) and nervous system (b) gene expression profiles in pancreas and purified islets from Type 1 diabetes (T1D) patients. Rows are for differentially expressed genes and columns are for pancreases and purified islets from T1D patients. Data were transformed to log₂ ratios relative to the mean of the normal controls and subjected to hierarchical clustering. The colour gradient key reflects relative expression on a log₂ scale. P1, pancreas from case 1; P2, pancreas from case 2; P3, pancreas from case 3; P4, pancreas from case 4; I1, islets from case 1; I4, islets from case 4.

Table 7.

Differentially expressed genes of the immune system in the pancreas and islets of Type 1 diabetes (T1D) patients.

			Log2 FC pancreas	Log2 FC pancreas	Log2 FC pancreas	Log2 FC pancreas	Log2 Fc islets	Log2 FC islets
Function	Locus ID	Symbol	Case 1	Case 2	Case 3	Case 4	Case 1	Case 4
Antigen presentation	972	CD74	1·93	2·47	2·74	–	–	–
	1 510	CTSE	–	1·88	1·43	–	–	–
	1 515	CTSL2	–	–	−1·17	–	–	–
	1 520	CTSS	2·20	3·59	2·17	0·89	–	–
	51 752	ERAP1	1·34	1·23	1·91	1·37	–	–
	64 167	ERAP2	2·82	3·88	−0·54	2·72	–	–
	3 105	HLA-A	1·49	1·72	1·26	–	–	–
	3 106	HLA-B	1·68	1·66	1·43	–	–	−1·33
	3 107	HLA-C	1·79	1·31	1·21	–	3·46	–
	3 108	HLA-DMA	2·47	2·29	2·12	1·23	–	–
	3 109	HLA-DMB	1·39	1·26	1·47	–	–	–
	3 113	HLA-DPA1	1·17	2·38	1·91	–	–	–
	3 115	HLA-DPB1	1·72	2·22	2·08	–	–	–
	3 117	HLA-DQA1	2·58	1·77	2·50	1·85	–	–
	3 119	HLA-DQB1	1·30	–	1·07	–	–	–
	3 122	HLA-DRA	1·99	2·35	2·65	–	–	–
	3 123	HLA-DRB1	1·28	2·08	2·12	–	–	–
	3 126	HLA-DRB4	–	1·89	−1·29	1·08	–	–
	3 127	HLA-DRB5	1·19	1·65	2·26	–	–	–
	3 920	LAMP2	1·32	1·09	–	–	–	–
	27 074	LAMP3	1·41	–	–	–	1·80	1·66
	10 279	PRSS16	–	–	–	–	1·43	1·53
	5 699	PSMB10	−2·95	–	–	–	–	–
	5 696	PSMB8	1·38	1·09	–	–	–	–
	5 698	PSMB9	2·06	1·67	1·11	1·06	–	–
	5 718	PSMD12	–	–	−0·94	–	–	–
	5 994	RFXAP	1·81	–	1·22	1·27	–	–
	6 890	TAP1	–	–	−0·62	–	–	–
	6 892	TAPBP	1·11	2·24	1·61	–	–	–
Chemotaxis	6 356	CCL11	1·01	2·18	1·97	0·89	–	–
	6 363	CCL19	–	2·04	2·10	–	–	–
	6 347	CCL2	2·90	4·68	3·15	3·41	–	–
	6 364	CCL20	–	3·12	–	–	–	–
	56 477	CCL28	1·82	–	–	–	–	–
	6 348	CCL3	–	1·99	0·89	3·19	–	–
	6 351	CCL4	–	1·65	0·56	1·31	–	–
	1 230	CCR1	–	1·72	–	–	–	–
	1 231	CCR2	1·36	2·74	–	–	–	–
	1 234	CCR5	–	–	0·50	–	–	–
	51 554	CCRL1	–	2·20	–	1·25	–	–
	51 192	CKLF	1·14	–	–	–	–	–
	54 918	CMTM6	–	1·20	–	–	–	–
	6 376	CX3CL1	–	2·06	0·94	1·06	–	–
	1 524	CX3CR1	1·51	2·11	1·98	1·35	–	–
	2 919	CXCL1	2·14	3·02	1·57	3·94	–	–
	3 627	CXCL10	–	2·95	–	–	–	–
	6 373	CXCL11	–	1·64	–	2·11	–	–
	6 387	CXCL12	–	–	1·86	2·19	2·10	2·81
	284 340	CXCL17	–	–	1·21	–	–	–
	2 920	CXCL2	–	2·63	1·82	2·87	–	–
	2 921	CXCL3	–	–	–	2·25	–	–
	6 372	CXCL6	2·06	5·00	2·11	2·00	–	−1·53
	4 283	CXCL9	1·24	1·95	0·58	–	–	–
	7 852	CXCR4	2·24	4·49	1·99	2·22	–	–
	57 007	CXCR7	–	–	1·17	–	1·16	1·21
	2 357	FPR1	–	1·62	–	–	–	–
	3 576	IL8	–	3·87	–	4·01	–	–
Innate immunity	60 489	APOBEC3G	–	1·43	–	–	–	–
	929	CD14	–	1·95	1·09	–	–	−1·59
	64 581	CLEC7A	1·87	2·80	2·05	–	–	–
	10 321	CRISP3	–	3·93	3·32	–	–	–
	1 755	DMBT1	–	2·50	1·65	–	–	–
	2 219	FCN1	–	3·38	0·89	1·09	–	–
	3 656	IRAK2	–	1·68	0·63	–	–	–
	3 934	LCN2	3·39	–	–	–	–	–
	3 959	LGALS3BP	–	–	1·20	–	–	−1·19
	4 057	LTF	1·45	1·52	1·85	–	–	–
	9 450	LY86	–	1·56	1·09	–	–	–
	23 643	LY96	1·78	2·47	2·15	1·37	–	–
	4 069	LYZ	5·05	3·31	3·41	2·39	3·93	–
	4 582	MUC1	2·67	1·54	1·55	1·44	2·03	–
	4 360	MRC1	–	–	–	–	–	−2·43
	5 284	PIGR	2·40	2·22	1·90	–	–	−1·66
	5 806	PTX3	2·22	–	–	2·16	–	–
	5 068	REG3A	2·70	2·80	3·13	2·06	–	–
	130 120	REG3G	–	2·89	2·68	–	–	–
	7 096	TLR1	2·06	1·26	0·83	–	–	–
	7 097	TLR2	–	1·48	0·79	–	–	–
	7 098	TLR3	1·99	1·40	1·12	–	–	–
	7 187	TRAF3	–	1·66	0·66	–	–	–
	85 363	TRIM5	1·08	1·34	–	–	–	–
Inflammation	183	AGT	–	1·97	1·01	–	−2·44	−2·97
	199	AIF1	–	–	1·36	–	–	–
	240	ALOX5	–	–	–	–	–	−1·55
	241	ALOX5AP	1·26	1·32	1·11	–	–	–
	325	APCS	–	1·28	1·49	–	–	–
	8 455	ATRN	1·67	–	0·73	–	–	–
	1 116	CHI3L1	3·15	1·69	1·73	–	–	–
	1 401	CRP	4·76	3·23	3·50	–	–	–
	1 906	EDN1	1·11	0·93	–	3·17	–	–
	2 150	F2RL1	–	–	−0·88	–	–	–
	2 244	FGB	1·46	–	–	–	–	–
	2 266	FGG	5·99	2·20	2·29	1·17	–	–
	2 267	FGL1	–	–	1·12	–	1·40	–
	10 875	FGL2	–	2·19	1·99	–	–	–
	11 251	GPR44	–	–	–	–	−2·01	−1·53
	57 817	HAMP	–	2·02	−2·02	−1·81	–	–
	3 146	HMGB1	1·27	–	–	–	–	–
	3 700	ITIH4	1·18	2·20	0·89	–	–	–
	145 741	NLF1	–	1·20	–	–	–	–
	10 562	OLFM4	3·75	–	0·91	–	–	–
	81 579	PLA2G12A	–	–	1·22	–	–	–
	5 320	PLA2G2A	–	–	0·97	1·31	–	–
	8 399	PLAG2G10	–	–	–	–	–	1·59
	6 281	S100A10	–	–	–	–	2·31	–
	6 283	S100A12	1·21	2·11	–	2·36	–	–
	6 275	S100A4	1·48	3·49	1·00	–	1·51	–
	6 279	S100A8	3·64	3·46	1·58	3·64	–	–
	6 280	S100A9	1·36	1·83	0·64	1·15	–	–
	55 829	SELS	–	–	−0·87	–	–	–
	5 054	SERPINE1	–	–	−1·76	–	–	–
	6 696	SPP1	−1·19	1·87	1·23	–	–	–
	7 078	TIMP3	1·35	1·77	1·47	1·57	–	–
Complement	712	C1QA	–	–	2·03	–	–	–
	713	C1QB	1·74	0·94	2·16	–	–	–
	714	C1QC	1·29	1·49	2·59	–	–	–
	715	C1R	1·69	1·93	1·44	1·12	–	–
	716	C1S	2·62	2·10	1·92	1·59	–	–
	718	C3	2·69	3·02	2·29	–	–	–
	719	C3AR1	–	1·66	–	–	–	–
	720	C4A	2·32	3·53	2·38	–	–	–
	721	C4B	–	–	4·00	–	–	–
	722	C4BPA	4·96	1·05	0·60	–	2·26	−1·47
	725	C4BPB	3·38	–	0·52	–	3·05	–
	727	C5	−2·33	–	–	–	–	–
	728	C5AR1	–	2·81	0·80	–	–	–
	729	C6	–	–	0·86	–	–	–
	730	C7	1·39	1·20	1·15	1·45	1·72	–
	966	CD59	1·07	1·36	0·59	–	–	–
	629	CFB	2·33	3·66	2·16	–	–	–
	3 078	CFHR1	–	1·54	1·07	–	−2·01	−2·50
	3 426	CFI	1·42	1·87	2·06	–	–	–
Immunoregulation	4 345	CD200	–	–	–	–	−2·70	−2·56
	11 314	CD300A	–	1·58	0·88	–	–	–
	2 281	FKBP1B	–	–	−1·21	–	–	–
	2 289	FKBP5	1·67	–	−1·04	–	–	–
	3 133	HLA-E	2·23	1·05	1·01	–	–	–
	3 134	HLA-F	1·18	1·34	0·83	–	–	–
	3 135	HLA-G	1·27	1·40	1·13	–	–	–
	10 068	IL18BP	1·80	2·13	1·81	–	–	–
	7 850	IL1R2	1·43	–	–	–	–	–
	3 903	LAIR1	–	1·74	0·95	–	–	–
	3 956	LGALS1	–	1·60	–	–	–	–
	3 957	LGALS2	–	–	1·15	–	2·13	–
	3 965	LGALS9	–	1·32	0·64	–	–	–
	10 859	LILRB1	3·41	3·49	0·83	1·78	–	–
	149 041	RC3H1	–	1·61	–	1·15	–	–
	5 272	SERPINB9	–	–	1·73	–	–	–
	710	SERPING1	2·08	2·12	1·73	–	–	–
	140 885	SIRPA	–	1·85	0·89	–	–	–
	9 021	SOCS3	–	1·77	1·14	1·14	–	–
	10 312	TCIRG1	–	–	–	–	–	1·44
	7 072	TIA1	–	1·45	1·19	–	–	−1·45
	283 131	TncRNA	−1·60	1·67	–	1·39	–	–
	7 128	TNFAIP3	–	1·98	–	–	–	–
	11 326	VSIG4	1·92	1·68	0·92	–	–	–
	79 679	VTCN1	1·57	2·13	1·55	–	–	–
	6 935	ZEB1	−3·19	–	–	–	–	1·62
Adhesion molecules	214	ALCAM	1·30	–	–	–	−1·48	–
	9 936	CD302	1·28	–	1·64	1·64	–	–
	948	CD36	1·97	2·67	1·51	2·39	2·17	1·44
	960	CD44	3·86	1·24	0·94	–	–	–
	961	CD47	1·64	1·51	1·26	1·08	–	−1·25
	962	CD48	–	1·69	–	–	–	–
	4 267	CD99	–	–	1·28	–	−2·28	−2·21
	83 692	CD99L2	–	–	–	–	–	−1·27
	3 383	ICAM1	–	2·34	1·05	1·10	–	–
	3 684	ITGAM	–	1·83	0·97	–	–	–
	3 688	ITGB1	–	–	–	–	–	−1·63
	3 689	ITGB2	1·57	3·24	1·69	–	–	–
	23 421	ITGB3BP	1·27	–	–	–	–	–
	3 694	ITGB6	1·10	–	–	–	–	–
	3 696	ITGB8	–	1·34	–	–	–	–
	3 964	LGALS8	1·89	–	–	–	–	–
	4 162	MCAM	1·10	–	–	1·30	–	–
	5 175	PECAM1	1·90	1·87	1·29	–	–	–
	7 412	VCAM1	1·07	2·83	2·10	1·23	–	–
Interferon-responsive	23 586	DDX58	–	–	1·42	–	–	–
	2 633	GBP1	1·97	3·23	0·95	2·63	–	–
	2 634	GBP2	2·83	2·43	–	1·35	–	–
	2 635	GBP3	1·58	–	–	3·45	–	–
	51 191	HERC5	1·71	0·96	0·81	1·61	–	–
	3 428	IFI16	–	1·22	–	–	–	–
	10 437	IFI30	–	2·16	0·8	–	–	–
	10 561	IFI44	1·45	–	0·6	–	–	–
	2 537	IFI6	–	1·66	0·95	–	–	–
	10 581	IFITM2	1·29	1·02	0·73	–	–	–
	10 410	IFITM3	1·05	–	–	–	–	–
	3 455	IFNAR2	–	1·39	–	–	–	–
	3 659	IRF1	–	1·35	–	–	–	–
	3 660	IRF2	–	1·41	1·28	–	–	–
	3 664	IRF6	–	–	−1·12	–	–	–
	3 665	IRF7	–	–	−0·74	–	–	–
	10 379	IRF9	1·36	0·87	–	–	–	–
	4 332	MNDA	2·76	3·58	2·44	1·81	–	–
	4 599	MX1	1·00	1·36	–	–	–	–
Cytokine signalling	1 436	CSF1R	–	1·78	1·46	–	–	–
	3 587	IL10RA	–	2·29	1·63	–	–	–
	3 588	IL10RB	–	1·50	–	–	–	–
	3 597	IL13RA1	1·45	–	–	–	–	–
	55 540	IL17RB	–	1·25	−0·79	1·79	1·70	1·55
	3 554	IL1R1	–	1·76	1·13	–	–	–
	58 985	IL22RA1	–	–	–	1·24	2·20	2·46
	9 466	IL27RA	–	1·35	–	–	–	–
	9 235	IL32	–	–	1·85	–	–	–
	3 750	IL6R	–	–	–	–	–	−1·89
	3 572	IL6ST	–	2·26	−0·71	1·10	–	–
	3 609	ILF3	1·59	–	0·84	–	–	–
	4 254	KITLG	2·18	1·34	0·62	0·93	–	–
	3 977	LIFR	1·44	1·93	0·86	1·95	1·39	–
	10 135	PFEB1	–	1·29	–	1·76	–	–
	4 090	SMAD5	–	1·17	–	–	–	–
	4 093	SMAD9	–	–	–	–	−1·50	−1·64
	10 673	TNFSF13B	–	2·20	1·67	1·02
	7 042	TGFB2	–	2·48	1·40	0·97	–	−1·66
	7 048	TGFBR2	–	–	0·92	–	–	–
	9 839	ZEB2	–	3·13	1·62	1·59	–	–
Signalling	84 674	CARD6	–	–	−0·77	–	–	–
	22 900	CARD8	–	–	0·84	–	–	–
	963	CD53	–	1·81	0·89	–	–	–
	80 790	CMIP	–	–	1·02	–	−1·77	–
	28 514	DLL1	–	–	1·06	–	–	–
	1 910	EDNRB	1·54	–	–	1·32	–	–
	2 268	FGR	–	–	0·58	–	–	–
	2 359	FPRL2	–	–	0·84	–	–	–
	2 534	FYN	–	1·66	–	–	–	–
	3 055	HCK	–	1·60	0·57	–	–	–
	3 059	HCLS1	1·91	2·75	1·82	–	–	–
	182	JAG1	–	2·08	1·16	1·68	–	–
	3 716	JAK1	–	1·30	–	1·10	–	–
	9 404	LPXN	–	1·44	–	–	–	–
	4 067	LYN	1·08	1·14	–	–	–	–
	5 341	PLEK	–	2·29	0·84	–	–	–
	122 769	PPIL5	–	–	−1·26	–	–	–
	5 530	PPP3CA	–	1·59	0·81	–	–	–
	5 734	PTGER4	–	–	–	1·39	–	2·35
	5 795	PTPRJ	–	2·16	–	1·45	–	–
	10 254	STAM2	–	–	0·96	–	–	–
	353 376	TICAM2	–	–	–	–	–	−1·14
	92 610	TIFA	–	–	–	–	1·78	–
	8 792	TNFRSF11A	0·97	2·42	–	–	−2·61	−2·95
	7 188	TRAF5	–	–	–	–	1·21	–
B cells and immunoglobulins	29 760	BLNK	–	1·91	1·02	1·62	–	–
	8 857	FCGBP	–	1·83	1·74	–	–	–
	3 476	IGBP1	–	–	–	–	1·47	–
	3 493	IGHA1	–	–	1·72	–	–	–
	3 500	IGHG1	–	–	–	–	2·21	–
	3 502	IGHG3	3·47	2·06	1·26	1·81	4·35	−1·35
	3 507	IGHM	–	1·27	–	–	–	–
	3 514	IGKC	–	–	–	–	3·6	–
	28 902	IGKV1D-13	–	–	–	–	1·4	–
	3 535	IGL@	1·21	1·67	0·89	1·16	2·34	–
	28 831	IGLJ3	1·83	2·32	1·34	1·85	2·81	–
	91 353	IGLL3	–	–	–	–	1·33	–
	118 788	PIK3AP1	–	2·00	1·15	–	–	–
Transcription regulation	604	BCL6	–	2·27	–	1·61	–	–
	83 706	FERMT3	–	–	0·73	–	–	–
	2 624	GATA2	–	–	–	–	–	1·52
	9 935	MAFB	–	1·68	–	1·20	−1·37	–
	10 725	NFAT5	–	1·41	0·85	–	–	–
	4 775	NFATC3	–	−1·24	−1·12	−1·34	–	–
	64 332	NFKBIZ	–	–	1·47	–	1·63	–
	10 062	NR1H3	–	–	1·48	–	–	–
	5 971	RELB	–	1·67	0·84	–	–	–
	6 772	STAT1	1·27	1·94	–	–	–	–
	6 925	TCF4	1·26	1·80	0·92	1·40	–	−1·21
	1 831	TSC22D3	–	1·16	–	–	–	–
	222 643	UNC5CL	1·35	–	–	–	–	–
	26 137	ZBTB20	–	1·91	0·96	–	–	–
	7 763	ZFAND5	–	–	−1·09	–	–	–
Apoptosis	597	BCL2A1	1·05	3·65	–	–	–	–
	330	BIRC3	–	2·90	1·00	1·25	–	–
	834	CASP1	–	1·79	0·99	–	–	–
	55 179	FAIM	1·46	–	0·51	0·95	–	–
	55 303	GIMAP4	–	1·49	0·75	–	–	–
	5 552	SRGN	–	–	1·29	–	–	–
	8 797	TNFRSF10A	–	–	−1·38	–	–	–
	55 504	TNFRSF19	1·06	–	1·17	–	–	–
	8 743	TNFSF10	2·27	2·39	0·94	–	–	–
T cells	2 533	FYB	–	–	0·77	–	–	–
	84 868	HAVCR2	–	1·88	0·92	–	–	–
	9 452	ITM2A	1·97	–	–	–	–	–
	3 936	LCP1	0·94	2·08	1·31	–	–	–
	9 840	ONECUT2	–	–	1·25	–	−1·82	−2·67
	5 588	PRKCQ	–	–	0·93	1·38	–	–
	6 503	SLA	–	1·88	–	–	–	–
	7 070	THY1	–	1·91	0·88	–	–	–
	9 760	TOX	–	–	–	–	−2·91	−3·96
Macrophages	54	ACP5	–	1·43	–	–	–	–
	9 332	CD163	1·90	–	–	–	1·63	−2·02
	968	CD68	–	1·38	–	–	–	–
	23 531	MMD	–	1·14	–	–	–	–
	219 972	MPEG1	2·01	3·01	2·37	1·35	–	–
	80 149	ZC3H12A	–	–	–	–	1·90	–
Neutrophils	57 118	CAMK1D	1·21	–	–	–	–	1·64
	1 511	CTSG	–	–	1·11	–	–	–
	79 887	FLJ22662	–	–	1·75	–	–	–
	4 688	NCF2	–	2·60	1·83	–	–	–
Dendritic cells	57 556	SEMA6A	1·15	1·82	1·36	1·66	–	–
	28 959	TMEM176B	–	–	–	2·44	−1·40	–
	54 209	TREM2	–	1·19	–	–	–	–
Fc receptors	2 207	FCER1G	1·32	2·70	2·49	–	–	–
	2 209	FCGR1A	–	1·74	0·93	–	–	–
	2 210	FCGR1B	–	2·56	1·71	1·01	–	–
Leucocyte markers	1 043	CD52	0·92	4·01	1·53	0·93	–	–
	8 832	CD84	0·99	1·92	1·34	1·25	–	–
	5 788	PTPRC	1·64	3·55	2·25	1·14	–	–
Natural killer (NK) cells	22 914	KLRK1	–	1·27	–	–	–	–
	7 305	TYROBP	1·40	3·14	1·90	–	–	–
Mast cells	7 177	TPSAB1	–	–	1·46	–	–	–
	64 499	TPSB2	–	–	1·51	–	–	–
Miscellaneous
Cell migration	10 154	PLXNC1	1·40	3·35	2·84	2·13	−1·73	−1·28
	5 420	PODXL	1·42	–	–	–	–	1·46
Cytotoxicity	23 705	CADM1	–	–	1·37	–	−2·60	−2·33
Cytoskeleton	23 075	SWAP70	–	–	–	0·96	–	–
Degranulation	25 801	GCA	1·03	1·25	–	1·00	–	–
Lysosomal protein	7 805	LAPTM5	1·87	3·59	2·35	–	–	–
Cell activation	9 976	CLEC2B	–	–	–	–	1·57	–
Oxidative stress	1 536	CYBB	1·54	3·47	–	1·04	–	–
Phagocytosis	22 918	CD93	–	–	1·34	–	–	–
Platelets	10 461	MERTK	–	1·59	0·76	–	–	–
Proteolysis	290	ANPEP	–	–	0·93	1·49	–	1·84
Tissue repair	81 035	COLEC12	–	1·37	–	1·20	–	–
Vascular endothelium	10 085	EDIL3	–	–	−0·99	–	−3·13	−4·29
Unknown	7 920	BAT5	–	–	–	–	−1·27	–
	11 119	BTN3A1	–	1·41	–	–	–	–
	11 118	BTN3A2	1·33	2·23	1·37	–	–	–
	10 384	BTN3A3	1·74	2·41	1·14	–	–	–
	2 124	EVI2B	–	2·88	–	–	–	–
	168 537	GIMAP7	1·53	1·30	–	1·20	–	–
	10 261	IGSF6	0·93	3·04	1·25	–	–	–
	3 321	ISGF3	–	1·31	–	–	–	–
	4 033	LRMP	–	–	–	–	–	1·43
	7 940	LST1	1·14	2·24	1·29	–	–	–
	3 071	NCKAP1L	–	1·40	–	–	–	–

(a) Genes of antigen processing and presentation of both the endogenous and exogenous antigen processing pathways encoded within the HLA region (HLA-A,B-C, HLA-E, -F, -G and HLA-DR, -DP, -DQ, -DM and LMP2 and 7) and outside it [low molecular weight polypeptide proteasome subunit (LMP10), invariant chain-CD74 and cathepsin S] constituted the more over-expressed and over-represented of all these subcategories (Fig. 2a). Comparison with the immunostaining of the pancreas indicated that the changes for HLA class I were due to both modulation in parenchymal cells and leucocytic infiltration, while for HLA class II infiltration was the main cause.

(b) Chemotaxis genes were also over-expressed in the four cases. Remarkably, case 2, collected 9 months after diagnosis, showed the highest levels. Two patterns partially overlap, neutrophil attracting chemokines CXCL1, CXCL6 and IL8 (CXCL8) and monocyte and activated T lymphocytes attracting chemokines, CCL2, CCL3 and CCL4. This pattern does not fit completely with the type of infiltration seen by immunohistology. Angiogenic CXCL chemokines may explain the capillary enlargement and the increase in CD36 expression (Fig. 2b).

(c) Transcripts of inflammatory response, anti-microbial defence genes and acute phase proteins were also found increased in the four cases. The over-expression of pattern recognition receptors (PRR) i.e. TLR1, TLR2 and TLR3, CLEC7A (DECT1), scavenger receptor (CD36) and ficolin (FCN1) and of molecules belonging to the Toll-like receptor/interleukin-1 (TLR/IL-1) signalling pathway, e.g. CD14, Ly86 (MD-1), Ly96 (MD-2) and IRAK2, may explain the over-expression of message corresponding to acute phase reactants, e.g. CRP, lactoferrin (LTF), fibrinogen (FGG), pentraxin 3 (PTX3) the local inflammatory and anti-bacterial response genes, i.e. arachidonate 5-lipoxygenase-activating protein (ALOX5PA), MUC1 and REG (regenerating gene). Some changes may reflect the infiltration by cells of the monocyte/macrophage lineage, e.g. lysozyme LYZ and calprotectin (S100A8/9), rather than the activation of transcription (Fig. 2c).

(d) Twelve of the 30+ genes of the complement system were found increased in T1D pancreases. Of interest, effector and regulatory/inhibitor molecules were represented almost equally. This may reflect infiltration by activated macrophages and dendritic cells that are the main producers of complement factors and their regulators outside the liver [23] (Fig. 2d).

(e) Immunoregulatory over-expressed genes included a remarkable variety of ligands/receptor pairs capable of starting inhibitory pathways, i.e. HLA class I (classical and, more remarkably, non-classical)-LILRB1, B7 family members VSIG4- and B7-H4 (VTCN)-ligand(s) still uncertain, collagen-LAIR1 and CD47-SIRP. There was also an increase of transcripts corresponding to molecules that counteract activation of the IL-1/IL-18 pathway, such as IL1R2 and IL18bp, and of serpins, modulators of the cytotoxic and proteolytic activity of phagocytes and cytotoxic T cells [24]. The increase of transforming growth factor (TGF)-β2 and of collagen is probably linked to repair [25] (Fig. 2e).

(f) The expression of adhesion molecules VCAM1, CD36 and CD47 was increased consistently in the four T1D pancreases.

(g) Thirty-five per cent (16 of 46) of interferon responsive genes were over-expressed in the two of the four cases, but analysis of the pattern could not discriminate between the type I and IL/IFN pathways (Pathvision 1·1 and Keg Type I and IL/IFN pathways) (Fig. 2f).

(h) The changes in expression in the subcategory of leucocyte-specific genes were reflective of the insulitis present in the four glands but also of the scattered leucocytes seen in the exocrine areas of the four cases (and not in the controls). By number and intensity, the most up-regulated genes corresponded to B cells (Ig) followed by macrophage and dendritic cells. Few T cell signalling and activation – a few of them shared with NK cells – genes were over-expressed (Fig. 2h).

Of the T1D autoantigens, the only clear change in the total pancreas was a marked reduction of insulin message, but transcripts for the other islet hormone genes showed a large case-by-case variability (Fig. 3a and Table 6). Finally, eight exocrine function genes [amylase 1A (AMY1A), elastases, aquaporin (AQP12A) and zymogen granule protein (ZG16)] were diminished in case 1, but in case 2 only two genes were down-regulated: ZG16 and GP2 (Table 6).

Table 6.

Differentially expressed genes related to islet, exocrine and nervous system function in the pancreas and islets of Type 1 diabetes (T1D) patients.

			Log2 FC pancreas	Log2 FC pancreas	Log2 FC pancreas	Log2 FC pancreas	Log2 FC islets	Log2FC islets
Function	Locus ID	Symbol	Case 1	Case 2	Case 3	Case 4	Case 1	Case 4
Autoantigens	2 562	GAD2	–	–	–	–	−4·13	−4·58
	57 818	G6PC2	–	–	–	–	−5·10	−4·34
	130 026	ICA1L	–	–	–	–	–	−2·02
	3 630	INS	−6·57	−6·68	−7·38	−7·38	−2·96	−9·05
	5 798	PTPRN	–	–	–	–	−3·53	−4·46
	5 799	PTPRN2	–	–	–	–	−1·88	–
	169 026	SLC30A8	–	–	–	–	−3·09	−4·48
Hormones/receptors	116	ADCYAP1	–	–	–	–	−5·63	−5·93
	51 129	ANGPTL4	–	−1·59	–	–	–	–
	887	CCKBR	−1·94	−1·44	–	–	–	–
	1 113	CHGA	–	–	–	1·01	−1·79	–
	1 114	CHGB	–	2·21	2·11	1·72	−2·28	−3·26
	2 099	ESR1	–	–	–	–	−3·43	–
	2 641	GCG	1·15	–	−2·92	1·91	–	–
	170 589	GPHA2	−1·54	–	–	–	–	–
	3 375	IAPP	−5·84	−5·40	−5·68	−5·72	−8·95	−10·01
	3 643	INSR	1·63	–	–	–	2·01	–
	84 634	KISS1R	−3·31	−3·05	−2·95	−2·68	–	–
	5 241	PGR	–	–	–	–	–	2·35
	5 539	PPY	–	5·35	6·47	–	−4·63	−6·87
	7 857	SCG2	1·08	1·41	–	1·46	–	−4·07
	6 447	SCG5	–	–	–	–	−2·38	–
	6 750	SST	–	–	−1·33	1·55	−2·41	−2·29
	6 751	SSTR1	–	–	–	–	−2·31	–
	8 614	STC2	–	–	−1·09	–	−2·63	–
Insulin secretion	3 741	KCNA5	–	–	–	–	−1·77	−2·24
	27 445	PCLO	–	–	–	–	−2·11	–
	130 399	ACVR1C	–	–	–	–	−2·44	−2·40
	5 126	PCSK2	–	–	–	–	−2·98	−3·56
	10 590	SCGN	–	–	–	–	−3·08	−3·32
	57 393	TMEM27	–	–	–	–	−3·20	−2·93
	1 363	CPE	–	–	–	–	−3·24	−3·71
	2 864	FFAR1	–	–	–	–	−3·64	−3·52
	5 208	PFKFB2	–	–	–	–	−3·77	−4·02
	6 833	ABCC8	–	–	–	–	−3·77	−5·23
	5 122	PCSK1	–	–	–	–	−5·60	−7·91
	5 168	ENPP2	–	–	–	–	–	−3·11
	6 277	S100A6	–	–	–	–	–	−1·65
Endocrine transcription factors	3 642	INSM1	–	–	–	–	−3·62	−4·09
	3 760	ISL1	–	–	–	–	–	−3·47
	4 760	NEUROD1	–	–	–	–	−2·95	−4·78
	4 821	NKX2-2	–	–	–	–	−2·73	−4·23
	5 080	PAX6	–	–	–	–	−3·58	−4·44
	5 087	PBX1	1·01	–	1·37	2·40	–	–
	63 935	PCIF1	1·95	–	–	–	–	–
Exocrine genes	276	AMY1A	−2·34	–	–	–	−2·05	–
	375 318	AQP12A	−1·17	–	–	–	–	–
	5 407	PNLIPRP1	−1·29	–	0·78	–	–	2·58
	123 887	ZG16	−1·77	−0·92	–	–	–	–
	3 816	KLK1	−1·20	–	–	–	–	1·13
	342 898	SYCN	−2·55	–	–	–	–	–
	51 032	ELA2B	−4·00	–	–	–	−2·38	–
	1 990	ELA1	−1·78	–	−0·56	–	–	–
	2 813	GP2	–	−1·79	–	–	–	–
Nervous system	107	ADCY1	–	–	–	–	−1·64	–
	273	AMPH	–	–	–	–	−1·84	−2·60
	10 776	ARPP-19	–	–	–	–	–	−1·18
	10 882	C1QL1	–	–	–	–	−1·37	−1·34
	285 175	C2orf21	–	–	–	–	−2·78	−2·81
	773	CACNA1A	–	–	–	–	−1·82	−1·55
	1 131	CHRM3	–	–	–	–	1·22	1·50
	25 927	CNRIP1	–	–	–	–	−1·28	−1·71
	152 330	CNTN4	–	–	–	–	−1·69	−2·57
	1 690	COCH	–	–	–	−1·35	–	3·05
	1 400	CRMP1	–	–	–	–	−1·62	−1·67
	27 065	D4S234E	–	–	–	–	−1·95	−1·65
	1 627	DBN1	–	–	–	–	−1·14	–
	1 804	DPP6	–	–	–	–	−2·24	−3·77
	2 026	ENO2	–	–	–	–	−1·91	−3·22
	54 332	GDAP1	–	–	–	–	−1·37	−1·78
	2 676	GFRA3	–	–	–	–	–	1·35
	2 743	GLRB	–	–	1·36	–	–	–
	2 824	GPM6B	1·60	–	–	–	–	–
	9 118	INA	–	–	–	–	−2·46	−2·27
	3 797	KIF3C	–	–	–	–	−1·66	−1·30
	54 551	MAGEL2	–	–	–	–	−1·57	−2·11
	89 796	NAV1	–	–	–	–	−1·49	–
	4 692	NDN	–	–	–	–	–	−1·42
	65 009	NDRG4	–	–	–	–	−2·64	−3·42
	23 114	NFASC	–	–	–	–	−3·17	−4·25
	56 654	NPDC1	–	–	–	–	−1·62	−1·36
	4 885	NPTX2	–	–	–	–	−3·36	−4·98
	4 852	NPY	–	–	–	–	−2·62	−2·87
	4 902	NRTN	–	–	–	–	1·21	1·33
	9 378	NRXN1	–	–	–	–	−2·62	−3·35
	4 917	NTN2L	–	–	–	–	3·38	–
	56 137	PCDHA12	–	–	–	–	−2·07	−2·49
	5 121	PCP4	–	–	–	–	−3·82	−3·60
	5 376	PMP22	1·50	–	–	–	–	–
	84 152	PPP1R1B	–	–	0·70	–	–	–
	5 961	PRPH2	–	–	1·03	–	–	–
	9 444	QKI	1·35	2·56	1·60	1·58	–	–
	23 475	QPRT	–	–	0·81	–	2·04	1·75
	9 743	RICS	–	–	–	–	1·18	–
	57 142	RTN4	–	–	–	–	–	−1·28
	9 037	SEMA5A	–	–	–	–	−1·41	−1·44
	23 208	SYT11	–	–	–	–	−2·43	−2·07
	29 114	TAGLN3	–	–	–	–	−1·88	−1·88
Exocytosis	9 501	RPH3AL	–	–	–	–	−1·32	–
	10 640	EXOC5	–	–	–	–	−1·38	−1·87
	23 025	UNC13A	–	–	–	–	−1·50	−1·27
	9 699	RIMS2	–	–	–	–	−1·62	–
	9 900	SV2A	–	–	–	–	−1·65	−1·51
	25 837	RAB26	–	–	–	–	−1·73	–
	10 814	CPLX2	–	–	–	–	−1·86	−1·99
	6 804	STX1A	–	–	–	–	−2·25	−3·46
	9 066	SYT7	–	–	–	–	−2·47	−2·72
	11 075	STMN2	–	–	–	–	−2·75	−3·69
	114 088	TRIM9	–	–	–	–	−3·16	−3·73
	6 252	RTN1	–	–	–	–	−3·19	−3·94
	6 860	SYT4	–	–	–	–	−3·26	−4·16
	5 865	RAB3B	–	–	–	–	−3·89	−4·76
	57 586	SYT13	–	–	–	–	−4·52	−5·58
	29 106	SCG3	–	–	–	–	−4·58	−5·41
	23 335	WDR7	–	–	–	–	–	−1·26

FC: fold change.

Islets from T1D patients

Changes in gene expression in cases 1 and 4 islets were confined mainly to endocrine and neural genes, some of which are T1D autoantigens, i.e. insulin, GAD65 (GAD2), I-A2 (PTPRN), I-A2B (PTPRN2), IGRP (G6PC2) and ZnT8 (SLC30A8). Insulin secretion and endocrine transcription genes that are also highly β cell-specific were reduced markedly. Interestingly, nervous system gene transcripts, mainly neuronal-specific, were as reduced as β cell ones (Fig. 3a,b and Table 6).

By contrast, these islets showed only a few over-expressed immune system genes that were detected almost exclusively in case 1. The highest increases corresponded to lysozyme, complement, IL-1R1 and immunoglobulins, and they most probably reflect infiltration by macrophages, B and T lymphocytes (Fig. 2 and Table 7). Interestingly, Ingenuity software analysis pointed to chemokine (C-C motif) receptor 5 (CCR5) and chemokine (CXC motif) receptor 4 (CXCR4) chemokine pathway activation (Table 5).

Differentially expressed genes reflected in situ damage and regeneration

Oxidative stress genes over-expressed were metallothioneins, such as MT1M (in the four cases and in purified islets) and SOD2, ceruloplasmin and thioredoxin interacting protein (case 1, pancreas and islets). Heat shock proteins (HSP) were over-expressed mainly in the islets from case 1 (HSPA6 and several genes of the HSP40 family). Altered expression of both pro- and anti-apoptotic genes showed a distinct pattern in each case. Islets showed some common changes in gene expression, including down-regulation of pro-apoptotic genes: MLLT11, PRUNE2 and NLRP1.

Growth and proliferation genes showed distinct patterns of altered expression in the pancreases and in the islets. In cases 1 and 2, REG4 and PDGFRA were up-regulated in both the pancreas and the islets. Only a few genes were up-regulated in the total pancreases, e.g. receptor for fibroblast growth factor (FGFR2) and TBC1D1. In the islets TFF3 was down-regulated and MACROD1 was up-regulated. Stem cell-associated genes prominin 1 was up-regulated in cases 2 and 3 and nestin and CD133 in case 3.

Validation of the microarray results by qRT–PCR analysis

Using qRT–PCR, differential expression was confirmed in all 16 genes from different categories, i.e. NMES1, CRP, REG3A, TLR3, CD44, HLA-E, CD36, S100A8, TNFSF10 (TRAIL), IFNAR2, IL8, CXCL1, INS, PNLIP, ERAP2 and PPY, selected to validate the total pancreas arrays and in the eight genes selected to validate the purified islets arrays, i.e. CD36, INS, PPY, REG4, LYZ, CXCL12, HSPA6 and INSR) (Fig. 4).

Fig. 4.

Quantitative real-time polymerase chain reaction results for selected genes in the Type 1 diabetes (T1D) pancreas and purified islets. Gene expression signals were normalized to HPRT and calibrator sample. Results from three independent experiments (mean ± standard error of the mean) (*P < 0·05; **P < 0·01; ***P < 0·001). (a) Whole pancreas. White bars are means of three normal pancreases, clear grey bars are means of three to five blocks from case 1, grey bars are means of three blocks from case 2, dark grey bars are means of three blocks from case 3, black bars are means of three blocks from case 4. (b) Purified islets. White bars are means of four islets samples from different donors, clear grey bars correspond to islets from case 1 and black bars correspond to islets from case 4.

Tissue distribution of differentially expressed genes

A selection of the differentially expressed genes identified by microarray analysis was studied by indirect immunofluorescence in pancreatic tissue sections from cases 1 and 4 (Fig. 5). The number of insulin-positive β cells of the islets was reduced markedly, but some were detected even in case 4 (Fig. 5b) concomitantly with a relative increase of α-cells. Leucocytes (CD45⁺) were seen mainly around the islets (peri-insulitis), only a few in the islets and fewer still scattered all over the exocrine tissue. As already mentioned (see Immunohistological analysis of insulitis), leucocytes were more abundant in case 1 than in case 4 (Fig. 5c). B lymphocytes (CD19⁺) were seen in and around the islets and scattered over the exocrine tissue (Fig. 5d). Large areas of exocrine tissue were stained for CRP mainly in case 1 (Fig. 5e). CD44 expression was increased in exocrine areas of case 1 and, occasionally, in the islets (glucagon-negative cells) (Fig. 5f). Most CD44⁺ cells were CD45^- (data not shown). Control pancreases showed some CD44⁺ cells in the exocrine tissue. HLA-E expression was increased markedly in the exocrine tissue and in some islets for case 1 (Fig. 5g). CD36 expression was increased in endothelial cells (co-localizing with FVIII, data not shown) for cases 1 and 4 (Fig. 5h). REG3A was clearly increased in the exocrine tissue and in the islets from cases 1 and 4 (Fig. 5i).

Fig. 5.

Double immunofluorescence staining of pancreatic cryostat sections from cases 1 and 4 and control. Left column: control pancreas; middle column: case 1; right column case 4. From top to bottom, haematoxylin and eosin staining (a), insulin (INS)/glucagon (GCG) (b), GCG/CD45 (c), GCG/CD19 (d), GCG/CRP (e), GCG/CD44 (f), glutamic acid decarboxylase (GAD)/human leucocyte antigen E-related (HLA-E) (g), GCG/CD36 (h) and GAD/REG3A (i). INS, GCG and GAD are islet markers. Magnification ×200. Bars = 100 µm.

Discussion

The results reported here consist of the first validated data set of whole genome expression profiles for human T1D pancreases and islets and its analysis. Overall, the results of the microarray analyses confirm the accepted views of the physiopathology of T1D but also qualify some of them and add some new perspectives.

The study of only four cases compounded with the dispersion of the insular tissue and the uneven distribution of immunopathological changes in the pancreas makes interpretation of this type of study difficult. There are several arguments that endorse their validity for interpretation: (i) the unbiased nature of the transcriptomic analysis in itself; (ii) the excellent preservation of the starting material; (iii) the consistency of the results with the predictions based on the immunohistological and the bioinformatic analysis; and (iv) the concordance of most data with previous similar studies in other autoimmune diseases and animal models.

On one hand, changes in transcriptomic profiling of an unfractioned tissue may arise from actual modulation of gene expression or from changes in its cell composition [26]. On the other hand, fractionating the tissue also poses problems. We were well aware that is likely that most of the peri-insulitis was lost in the process of separation and that the islets purified from cases 1 and 4 represent those partially spared by the autoimmune process. We took these problems into account and interpreted the results in the context of the immunohistological data from the same tissue and by comparing the profiles of total pancreas and the islets.

Among the results we would like to highlight four aspects, as follows.

1. Collectively, the expression profile of immune response and inflammatory genes confirms the predictions from previous studies but suggests some nuances to current views on the immunopathogenesis of T1D: (i) the implication of innate immunity and inflammatory pathways seems broader and more maintained than expected as encompasses IL-1, IL-6, TLRs, type I and type II interferon pathways, this differently from the non-obese diabetic (NOD) mice where there the signature points clearly to the Type II IFN pathway [27]; (ii) the prominence of Ig transcripts in the profiling suggests that β cells may be activated by up-regulated B cell activating factor (BAFF). It is plausible that B cells play a regulatory rather than an effector role, a possibility that deserves further investigation [28]; (iii) the limited scope of adaptive response genes expressed – HLA and Ig but no transcripts for cytokines and T cell-specific genes – can be attributed to the low sensitivity of the arrays to detect cytokines [29] and to the limited transcription activity of the effector lymphocytes. The implication is that the activation and expansion of the T cell response did not occur in situ and underlines the need to analyse pancreas and pancreatic lymph nodes in parallel (when feasible) [30]; and (iv) the variety of transcripts corresponding to inhibitory pathways suggests that regulatory and repair responses co-exist within the autoimmune process. Finally, the over-expression of natural immunity and IFN-responsive genes do not prove the link of T1D and enterovirus infections but certainly justifies further effort to identify a viral agent in T1D pancreas.

2. The loss of β cell-specific transcripts was expected, but the detection of insulin message in islets 8–10 years after diagnosis and its confirmation by qPCR and IFL would agree with the concept that the destruction is a slow ongoing process. On the other hand, the over-expression of REG4 in the islets indicates that β cell destruction is probably compensated partially by regeneration [31,32]. There is evidence that REG proteins are involved in inflammation [33] and are putative autoantigens [32], and their over-expression may also contribute to the perpetuation of autoimmunity, resulting in a relapsing–remitting process [34].

3. The strong reduction of neural specific genes confirms that β cells are not the sole targets of the autoimmune response, fitting well with the reported responses to neural antigens and the destruction of peri-islet Schwann cells in NOD mice [35,36].

4. The reduction in exocrine-specific gene transcripts is, in fact, a confirmation [4,37], but gives support to little-noticed reports on a concomitant autoimmune response to exocrine pancreas in T1D with antibodies to carbonic anhydrase II and lactoferrin [38]. The detection of CRP and other changes in the acini supports this possibility.

In the last few years some genome expression profile studies that included the analysis of the target organ/tissue have been reported for the main autoimmune diseases, e.g. systemic lupus erythematosus [39], rheumatoid arthritis [40], autoimmune thyroiditis [41], multiple sclerosis [42] and Sjögren's syndrome [43,44]. A broad spectrum of immune response genes were found to be over-expressed in these diseases, mainly in the subcategories of IFN response, HLA, antigen presentation, complement, immunoglobulins and chemokines. The so-called ‘IFN signature’ was the most common finding [45]. Our results indicate that in T1D the general pathways of the immune response are also activated and that larger data sets of patients stratified by age, gender and disease stage will be required to identify the common and the specific pathways of the different autoimmune diseases.

Whole genome transcriptomic profiling has been applied widely to the animal models of autoimmune diseases, including the NOD mouse [27,46]. The results of our study are similar in many respects to previous studies in the pancreas and islets of the NOD mouse [47]. However, some differences are also observed, including the absence of the reduction in B lymphocyte gene transcripts seen in diabetic mice or the low levels of cytokines (but not chemokines) in our T1D samples [48]. These differences could be due to the very different design of the study or also to the differences between human and NOD mouse T1D. Studies using similar designs in both species would help in the interpretation of these type of experiments but also to establish which aspects of T1D can be addressed in the NOD model [49].

The data presented here favour the view that T1D is caused by a chronic inflammatory process with a strong participation of innate immunity that progresses in spite of the regulatory and regenerative mechanisms. Finally, and more importantly, these results constitute a very valuable data set, which should grow with the analysis of pancreas and islets from T1D patients using the approaches outlined in this paper, carried out by others in possession of the relevant tissues, in order that our work can be built upon, data pooled and a consensus reached. In these future analyses, important substudies will include whether it is better to use fractionated or unfractionated tissue and what different information the different preparations provide; and also, perhaps, to use additional approaches, such as laser microdissection, as a means to isolate tissue fractions of interest. Such resources will be a very valuable asset for diabetes research.

Disclosure

None.