Introduction

Components of the unfolded protein response (a cellular response to stress of the endoplasmic reticulum [ER]) act as beneficial regulators under physiological conditions or as inducers of beta cell dysfunction and apoptosis under chronic stress. In vitro studies showed that proinflammatory cytokines (IL-1β and IFN-γ) can induce beta cell ER stress [1, 2] suggesting that ER stress might contribute to beta cell loss in type 1 diabetes. These cytokines activate some, but not all, branches of the ER stress pathway and hamper beta cell defences by inhibiting ER chaperones [1, 2]. In the context of type 1 diabetes, ER stress might amplify proapoptotic pathways, augment inflammation and increase the presentation of beta cell antigens to the immune system.

There is evidence of enhanced ER stress in beta cells from both humans with type 2 diabetes and from rodent models of the disease [1, 3, 4]. Furthermore, in patients or in mouse models with mutations in key components of the unfolded protein response or in proinsulin, beta cell loss and diabetes develop [1]. However, it remains to be established whether ER stress occurs in the beta cells of humans with type 1 diabetes. To address this question, we have evaluated the presence of the ER stress markers C/EBP homologous protein (CHOP), immunoglobulin heavy chain (BIP) and X-box binding protein 1 (XBP-1) in pancreatic sections from type 1 diabetic and control individuals. Particular care was taken to correlate insulitis with levels of beta cell ER stress markers and to analyse the data in an unbiased and quantitative way.

Methods

Human pancreatic tissue

Pancreases from type 1 diabetic, adult and paediatric non-diabetic individuals from two different collections—UK autopsy cases [5, 6] and the Juvenile Diabetes Research Foundation-sponsored Network for Pancreatic Organ Donors with Diabetes (nPOD) programme, USA—were studied, with ethical permission. The series consisted of 13 patients with type 1 diabetes (seven female and six male); eight with recent-onset type 1 diabetes (age [mean±SD] 7.9 ± 8.7 years, range 1.5–27 years, time since diagnosis <3 months) and five cases of long-standing type 1 diabetes (age 24.6 ± 14.8 years, range 12–50 years, mean time since diagnosis 9.8 ± 7.1 years). Fifteen pancreases from non-diabetic individuals—seven children (four females and three males, mean age 5.5 ± 3.7 years, range 0.5–10 years) and eight adults (five females and three males, mean age 55.4 ± 15.7 years, range 30–72 years)–served as controls.

The pancreases had been fixed in formalin buffer and embedded in paraffin. Most of the type 1 diabetic pancreases retained some insulin-containing islets, and insulitis was also present in some of these (electronic supplementary material [ESM] Table 1). As positive controls, HeLa cells were treated with the ER stressor thapsigargin (1 μmol/l) for 24 h and then embedded in paraffin for subsequent immunostaining, or retrieved for evaluation of mRNA expression of selected ER stress markers [2].

Immunodetection

Serial sections of 4 μm thickness were processed and labelled for CHOP, BIP, XBP-1, CD45, insulin and glucagon using a standard immunoperoxidase and double immunofluorescence methods. Immunostaining was carried out as described in ESM Supplementary Methods.

Immunohistochemical quantification of ER stress markers

To quantify the immunostaining in an unbiased way, images of each slide were acquired at 20× magnification using a Hamamatsu NanoZoomer HT2.0 whole-slide scanner (Hamamatsu Photonics, Hamamatsu, Japan) as previously described [7]. The islets were manually selected using the annotation tool within NDP-view software (Hamamatsu Photonics), allowing superposition of the same islet in successive sections. A dedicated in-house tool was used to automatically extract each selected islet and to export it as a standard bitmap image file [7]. A routine was developed for Definiens Developer XD 1.2.1 (Definiens AG, Munich, Germany) to detect the selected islets from extracted bitmap images. Staining was quantified using Definiens TissueMap 3.0 (Definiens AG), which integrates with Definiens Developer XD-1.2.1. The ‘quick-score’ feature [7], defined as the ratio of the sum of stained pixel intensities over the islet area, was then calculated. Positive pixels were detected by a TissueMap 3.0 routine using a constant threshold manually optimised across the different slides.

The quantification results of immunostaining intensities from the UK and nPOD pancreas cases were initially examined and evaluated separately (data not shown). Since the results from each were similar, they are presented together.

Results

Validation of the CHOP, BIP and XBP-1 antibodies in ER-stressed HeLa cells

HeLa cells treated with thapsigargin (1 μmol/l for 24 h) were used as a positive control (ESM Fig. 1a). Compared with untreated HeLa cells, CHOP, BIP and XBP-1 levels were increased in response to thapsigargin. The increase in protein expression was paralleled by an increase in mRNA expression for each marker (ESM Fig. 1b).

Evaluation of CHOP, BIP and XBP-1 levels in human type 1 diabetic and non-diabetic pancreas

Immunostaining for insulin and CD45 (a marker of immune cells) in type 1 diabetic pancreases revealed three types of islet: insulin-containing islets (ICIs) without detectable inflammation (Fig. 1g), ICIs with immune cell infiltration (insulitis; defined as more than ten CD45-positive cells within the islet periphery [Fig. 1m, r]) and insulin-deficient islets (IDIs; Fig. 1s). No IDIs or insulitis-positive islets were detected in non-diabetic pancreases.

Fig. 1
figure 1

Expression of the ER stress markers CHOP, BIP and XBP-1 in non-diabetic control (a–f) and type 1 diabetic (T1D) human pancreas (g–x). Samples from patients with type 1 diabetes are separated into type 1 diabetic ICIs without (g–l) or with (m–r) insulitis, and IDIs (s–x). Serial pancreatic sections from control and type 1 diabetic pancreases were immunostained for insulin (a, g, m, s), CHOP (b, h, n, t), BIP (c, i, o, u), XBP-1 (d, j, p, v), glucagon (e, k, q, w) or CD45 (f, l, r, x). Type 1 diabetic islets with insulitis are identified by positive CD45 immunostaining as shown in (r). Magnification 400× (scale bar: 40 μm)

Full size image

Weak cytoplasmic immunostaining for CHOP was observed in the majority of islet cells in all samples. However, the intensity of this staining was increased in type 1 diabetic islets (with or without insulitis) in comparison with control islets (Fig. 1; compare h, n and t against b).

Immunostaining for BIP was heterogeneous, with some islet cells displaying strong cytoplasmic staining while others were immunonegative. This pattern was similar in control islets (Fig. 1c) and in insulitis-negative islets in type 1 diabetes (whether ICI or IDI; Fig. 1i, u). In contrast, insulitis-positive islets of type 1 diabetic patients displayed an enhanced intensity of BIP expression, and the protein was present in a larger proportion of cells (panel o).

XBP-1 immunostaining was mainly cytoplasmic or perinuclear and was present at equal intensity in the majority of endocrine cells in islets from both control and type 1 diabetic samples (Fig. 1d, j, p, v).

Unbiased quantification of immunostaining intensity revealed that CHOP staining was significantly elevated in type 1 diabetic islets compared with controls (Fig. 2a, b). The intensity of BIP immunostaining also tended to be greater in type 1 diabetic islets, but this achieved statistical significance only for insulitis-positive islets (Fig. 2c, d). No significant differences were observed in XBP-1 immunostaining intensity between type 1 diabetic and non-diabetic pancreases (Fig. 2e, f). The patterns of expression and intensity of staining of the three ER stress markers was similar in adult and paediatric non-diabetic pancreases (Fig. 2).

Fig. 2
figure 2

Quantification of the intensity of CHOP (a, b), BIP (c, d) and XBP-1 (e, f) immunostaining in pancreases from type 1 diabetic (n = 13, i = 264) or control adult (n = 8, i = 92) and paediatric (n = 7, i = 102) non-diabetic individuals (a, c, e). (b, d, f). Comparison of the intensities of CHOP, BIP and XBP-1 immunostaining (represented by optical density) in control (adult [A] + paediatric [P]) islets (n = 16, i = 194) vs type 1 diabetic ICIs (n = 10, i = 80), insulitis-positive islets (n = 11, i = 30) and IDIs (n = 11, i = 154). A. Ctrl, adult control; Ctrl, adult and paediatric control non-diabetic pancreases; i, number of islets analysed per group; P. Ctrl, paediatric control; T1D, type 1 diabetic. *p < 0.05; **p < 0.01; ***p < 0.001 vs control (Mann–Whitney U test)

Full size image

Double staining for CHOP or BIP and insulin or glucagon in selected cases indicated co-localisation between CHOP or BIP and insulin in many but not all cells (ESM Fig. 2).

Discussion

The role of ER stress in mediating beta cell loss in type 1 diabetes remains unclear, and few studies have addressed whether ER stress markers are present in islets from individuals with type 1 diabetes. Evidence of enhanced expression of activating transcription factor 3 in type 1 diabetic beta cells has been presented [8], whereas CHOP was not found [4]. Conversely, studies on pancreases from patients with type 2 diabetes revealed increased levels of CHOP, BIP and activating transcription factor ATF-3 [3, 4, 8] as well as ER dilation [9].

Such differences in islet expression of ER stress markers between type 2 and type 1 diabetes might reflect differences in the islet milieu in each condition. Thus, glucolipotoxicity is likely to underlie the changes in type 2 diabetes, whereas proinflammatory cytokines may play a role in type 1 diabetes [1]. If proinflammatory cytokines are responsible for mediating ER stress in islet cells in type 1 diabetes, it is probable that the expression of ER stress markers will be heterogeneous, with the highest levels of induction found in inflamed islets. These considerations, coupled with the scarcity of adequate samples for histology from individuals in the pre-clinical stages of type 1 diabetes and the difficulty of finding reliable antibodies for ER stress markers (as our own unpublished observations [I. Marhfour and D.L. Eizirik] and a study by Haataja et al. [10] show) may explain the conflicting results on the presence of ER stress markers in islets from patients with type 1 diabetes.

We have now studied pancreases from 13 individuals with type 1 diabetes (plus non-diabetic controls) obtained from two independent sources, and have performed an unbiased and quantitative evaluation of immunostained samples. The results indicate that islets from type 1 diabetic individuals display increased levels of CHOP and, in the case of insulitis-positive islets, BIP. No increase in the level of XBP-1 was detected in type 1 diabetes, although this marker was present in all the islets examined. These results suggest that islet cells from individuals with type 1 diabetes show a partial ER stress response, which is reminiscent of that described previously in purified beta cells exposed to proinflammatory cytokines in vitro [2]. Interestingly, increased CHOP expression is not limited to beta cells, since it is also present in IDIs.

These observations, taken together with the recent findings that islets isolated from prediabetic NOD mice show markedly increased levels of ER stress markers [11], suggest that ER stress may be a contributory factor for beta cell dysfunction/death and insulitis in early type 1 diabetes.