High Expression Rates of Human Islet Amyloid Polypeptide Induce Endoplasmic Reticulum Stress–Mediated β-Cell Apoptosis, a Characteristic of Humans With Type 2 but Not Type 1 Diabetes

  1. Peter C. Butler1
  1. 1Larry Hillblom Islet Research Center, University of California, Los Angeles, Los Angeles, California
  2. 2Endocrine Research Unit, Mayo Medical College, Rochester, Minnesota
  1. Address correspondence and reprint requests to Peter C. Butler, Larry Hillblom Islet Research Center, UCLA, Los Angeles, CA 90024-2852. E-mail: pbutler{at}


OBJECTIVE—Endoplasmic reticulum (ER) stress–induced apoptosis may be a common cause of cell attrition in diseases characterized by misfolding and oligomerisation of amyloidogenic proteins. The islet in type 2 diabetes is characterized by islet amyloid derived from islet amyloid polypeptide (IAPP) and increased β-cell apoptosis. We questioned the following: 1) whether IAPP-induced β-cell apoptosis is mediated by ER stress and 2) whether β-cells in type 2 diabetes are characterized by ER stress.

RESEARCH DESIGN AND METHODS—The mechanism of IAPP-induced apoptosis was investigated in INS-1 cells and human IAPP (HIP) transgenic rats. ER stress in humans was investigated by β-cell C/EBP homologous protein (CHOP) expression in 7 lean nondiabetic, 12 obese nondiabetic, and 14 obese type 2 diabetic human pancreata obtained at autopsy. To assure specificity for type 2 diabetes, we also examined pancreata from eight cases of type 1 diabetes.

RESULTS—IAPP induces β-cell apoptosis by ER stress in INS-1 cells and HIP rats. Perinuclear CHOP was rare in lean nondiabetic (2.6 ± 2.0%) and more frequent in obese nondiabetic (14.6 ± 3.0%) and obese diabetic (18.5 ± 3.6%) pancreata. Nuclear CHOP was not detected in lean nondiabetic and rare in obese nondiabetic (0.08 ± 0.04%) but six times higher (P < 0.01) in obese diabetic (0.49 ± 0.17%) pancreata. In type 1 diabetic pancreata, perinuclear CHOP was rare (2.5 ± 2.3%) and nuclear CHOP not detected.

CONCLUSIONS—ER stress is a mechanism by which IAPP induces β-cell apoptosis and is characteristic of β-cells in humans with type 2 diabetes but not type 1 diabetes. These findings are consistent with a role of protein misfolding in β-cell apoptosis in type 2 diabetes.

Both type 1 and type 2 diabetes are characterized by deficits in β-cell mass and increased β-cell apoptosis (16). The mechanism that initiates β-cell apoptosis in type 1 diabetes is believed to be autoimmune-mediated cytokine production (5). Several mechanisms have been proposed for increased β-cell apoptosis in type 2 diabetes, including oxygen free radicals (7), free fatty acid toxicity (8), interleukin-1β (9), and formation of islet amyloid polypeptide (IAPP) toxic oligomers (1012).

Programmed cell death, or apoptosis, is important in multicellular organisms to permit organ development and remodeling (13). In disease states, apoptosis permits selective removal of cells that are damaged, particularly in relation to cell cycle, so that damage is not propagated (3,14). Apoptosis may be initiated by a wide variety of cellular insults, which are currently thought to act through at least three pathways that converge to accomplish irreversible destruction of the cell's chromosomes. These three major pathways have been designated as the extrinsic and intrinsic pathways and endoplasmic reticulum (ER) stress pathway (15,16). The extrinsic pathway is classically exemplified by cytokine-induced cell death, mediated through cell surface death receptors (17). The intrinsic pathway is most often described as a response to mitochondrial disruption, for example, secondary to oxygen free radicals (18). ER stress–induced apoptosis is classically ascribed to aggregates of misfolded protein that are believed to compromise the ER membrane (15).

The human pancreatic β-cell is vulnerable to all three forms of apoptosis. Cytokines are recognized as important in the pathophysiology of type 1 diabetes (5) and have been proposed as potential mediators of glucose toxicity in type 2 diabetes (9). The β-cell bears a large burden of protein synthesis, protein folding and processing, and regulated protein secretion, with the primary client proteins being insulin and IAPP. Islet amyloid derived from IAPP is a characteristic of the islet in type 2 diabetes (3). Both mice and rats with high expression rates of human IAPP have been reported to develop diabetes because of loss of β-cells through increased β-cell apoptosis (11,12). There is increasing evidence that induction of apoptosis by amyloidogenic proteins is mediated by membrane-disrupting oligomers that are distinct from amyloid fibrils (19). Moreover, the pathway initiating apoptosis in several other diseases characterized by accumulation of unfolded proteins has been reported to be ER stress (10,20).

In the present studies, we first sought to establish whether the mechanism subserving human IAPP–induced β-cell death is ER stress. To this end, we used adenoviral expression of human IAPP versus rodent IAPP in a β-cell line (INS-1 cell). To affirm these findings in primary β-cells in an animal model, we studied islets from the human IAPP (HIP) rat, a transgenic rat model that develops islet pathology comparable with that in humans with type 2 diabetes (12). To place these findings in the context of human disease, we then examined pancreata from both lean and obese humans with type 2 diabetes and from nondiabetic control subjects. Recognizing that high glucose per se could theoretically induce ER stress by the actions of oxygen radicals inducing mitochondrial dysfunction and subsequent deprivation of energy required to maintain appropriate protein folding, we also examined pancreata from humans with type 1 diabetes, where sufficient β-cells were still present to make a meaningful evaluation.


Adenovirus generation and INS-1 cell line studies.

The complementary cDNA encoding the full-length human and rat IAPP precursor protein (preproIAPP) was ligated into pEGFP-N2 vector (Clontech, Palo Alto, CA), and the sequences were verified. PreproIAPP–enhanced green fluorescent protein (EGFP) construct was digested with XhoI and NotI, ligated into pENTR2B (Invitrogen, Carlsbad, CA), and subsequently inserted into pAd/cytomegalovirus/DEST adenovirus vector (Invitrogen). Recombinant adenovirus-expressing human and rat preproIAPP (Ad-hIAPP and Ad-rIAPP, respectively) fused to EGFP were generated and purified according to the manufacturer's instructions (Invitrogen). PreproIAPP is 36 kDa, but after the signaling peptide is removed when IAPP reaches ER, the fusion protein is 34 kDa (unprocessed proIAPP EGFP). Cleavage by prohormone convertase PC1/3 at the COOH-terminus creates a 6-kDa NH2-terminal proIAPP and a 28-kDa processed COOH-terminal proIAPP plus EGFP. Finally, prohormone convertase 2 cleaved at the NH2-terminal creates a fully processed 4-kDa mature IAPP. Adenovirus-expressing green fluorescent protein (GFP) was kindly provided by Dr. Christopher Rhodes (University of Chicago, Chicago, IL).

INS-1 cells were grown in RPMI medium, supplemented with 10% FBS, 50 μmol/l β-mercaptoethanol, 10 mmol/l HEPES, and 1 mmol/l sodium pyruvate. One day after plating, INS-1 cells were transduced at multiplicity of infection (MOI) = 100 with adenovirus-expressing GFP, Ad-rIAPP-EGFP, Ad-hIAPP-EGFP, or nontransduced control. For Western blotting experiments, cells were washed with PBS and lysed in 2× Laemmli sample buffer after transduction. As a positive control to induce ER stress, cells were cultured overnight in the presence of 0.5 μg/ml Tunicamycin (Sigma, St. Louis, MO). Protein concentrations were determined using Bio-Rad protein assay reagents (Hercules, CA). Proteins (10 μg) were separated on a 4–12% Bis-Tris NuPAGE gel (Invitrogen) and transferred to polyvinylidine fluoride membranes (Bio-Rad). Membranes were blocked with 5% nonfat dry milk in Tris-buffered solution (TBS)/0.1% Tween-20 and incubated overnight at 4°C with anti-GFP (1:1,000; Zymed Laboratories, San Francisco, CA), anti–β-actin, anti–caspase-3 (1:1,000; Cell Signaling Technology, Beverly, MA), or anti–C/EBP homologous protein (CHOP) (1:500; Santa Cruz Biotechnology, Santa Cruz, CA) antibodies. Membranes were washed with TBS/0.1% Tween-20 and incubated with horseradish peroxidase–conjugated secondary antibodies (1:3,000; Jackson Laboratories, Bar Harbor, ME) for 1 h. After washes, proteins were visualized using enhanced chemiluminescence (Bio-Rad).

To investigate the time course of apoptosis and CHOP protein expression, INS-1 cells were plated in a chamber slide. One day after plating, INS-1 cells were transduced (MOI = 100) with Ad-rIAPP-EGFP or Ad-hIAPP-EGFP and then cultured for 8, 16, 24, and 48 h. At the end of each indicated time, cells were washed and fixed with 4% paraformaldehyde, followed by staining for transferase-mediated dUTP nick-end labeling (TUNEL) (cell death detection kit TMR Red; Roche Diagnostics, Mannheim, Germany) according to the manufacturer's instructions. For CHOP staining, cells were permeabilized with 0.2% Triton X-100 in PBS at room temperature for 10 min after fixation and incubated with anti-CHOP antibody (1:100; Santa Cruz Biotechnology) overnight, followed by secondary antibody conjugated to Cy3, diluted in 1:200 (Jackson Laboratories). Fluorescent slides were viewed using a Leica DM6000 microscope and images acquired using Openlab software (Improvision, Lexington, MA). Experiments were repeated four times for each time point.

Small interfering RNA and transfection.

CHOP protein levels were reduced by the small interfering RNA (siRNA) technique. The sequences of the siRNA duplexes were chosen from CHOP coding region (Ddit3, NM_024134). The sense strand (regions 5–23) of siRNA was GAA UCU AAU ACG UCG AUC AdTdT (dT is deoxythymidine). The specific small interfering CHOP (siCHOP) and nonspecific control siRNA were synthesized by and purchased from QIAGEN (Valencia, CA). Transfection of cultured INS-1 cells with siCHOP and nonspecific control siRNA was carried out by the chemical transfectant Lipofectamine (Invitrogen) according to the manufacturer's instructions. Transduction (MOI = 150) of the INS-1 cells by the Ad-hIAPP preceded transfection of siCHOP versus nonspecific control siRNA by 1 h. Cell lysates were collected 48 h later. Immunoblotting was conducted as described above. CHOP and cleaved caspase-3 expression were quantified by measuring the optical density of the desired bands in Western blots using Labwork software (UVP Bioimaging Systems, Upland, CA).

Animal model and islet isolation.

The generation of the HIP has previously been described in detail (12). Briefly, the transgene is the fusion of the rat insulin gene II promoter linked to the cDNA encoding for IAPP. Sprague-Dawley–expressing human IAPP was designed as HIP rats and without the human IAPP as Sprague-Dawley wild-type nontransgenic controls. Both were bred and housed at an animal housing facility at the University of California, Los Angeles. The University of California, Los Angeles institutional animal care and use committee approved all surgical and experimental procedures in these studies. Five-month-old rats were studied because, by this age, HIP rats have increased β-cell apoptosis but do not yet have diabetes, avoiding the confounding effects of glucose toxicity. To obtain rat islets, 5-month-old HIP (n = 12) and wild-type (n = 12) rats were killed by intraperitoneal injection of pentobarbital (50 mg/kg). The bile duct was cannulated, and Hanks’ balanced salt solution (HBSS) (Flow Labs, Irvine, Scotland) containing 7.5 mmol/l calcium chloride, 20 mmol/l HEPES buffer, and 1 mg/ml collagenase (type II; Sigma, St. Louis, MO) was injected to uniformly distend the pancreas. The pancreas was then removed and incubated for 20 min in HBSS at 37°C, followed by transfer into HBSS containing 5 g/l BSA and 20 mmol/l HEPES buffer at 4°C. The pancreas was dispersed by gentle shaking and washed four times. After being handpicked three times, the islets from two to three animals were pooled, washed with PBS, and immediately lysed in 2× Laemmli sample buffer, aided by 10 strokes of a plastic micropestler. Antibodies used were as follows: anti–caspase-12 (rat monoclonal, 1:1,000; Sigma), anti-CHOP (rabbit polyclonal, 1:500; Santa Cruz Biotechnology), anti–β-actin (rabbit polyclonal, 1:1,000), and anti–caspase-3 (rabbit polyclonal, 1:1,000; Cell Signaling). The Western blotting procedure of islet extracts was the same as that of INS-1 cells, described above.

Human subjects and pancreatic tissue.

Institutional review board approval was obtained from both the Mayo Clinic (institutional review board no. 1516-03) and the University of California, Los Angeles (no. 06-04-021-01). We obtained human pancreatic tissue at autopsy from 7 lean and 12 obese nondiabetic humans and from 14 obese humans with type 2 diabetes. We also obtained pancreata at autopsy from three humans with long-standing type 1 diabetes and four with relatively recent-onset type 1 diabetes who died of diabetic ketoacidosis (see Table 1). In addition, we obtained pancreas from one 89-year-old man with recent-onset type 1 diabetes by surgery as previously described (6). The three autopsy cases of long-standing type 1 diabetes were selected as those with the largest number of β-cells per islet from a prior study of 42 cases of pancreas from patients with long-standing type 1 diabetes (1).


Clinical characteristics of human case subjects

Potential autopsy cases were identified by retrospective analysis of the Mayo Clinic autopsy database. To be included, cases were required to 1) have had a full autopsy within 12 h of death; 2) have had a general medical examination, including at least one fasting blood glucose documented within the year before death (the exception being those with recent-onset type 1 diabetes, who died at the first admission shortly after diagnosis); and 3) pancreatic tissue stored that was of adequate size and quality. Cases were excluded if 1) potential secondary causes of diabetes were present, 2) subjects had been exposed to chronic glucocorticoid treatment, or 3) pancreatic tissue had undergone autolysis or showed evidence of pancreatitis. The definitions of lean and obese were BMI <25 and >27 kg/m2, respectively.

Pancreas processing and immunohistochemistry

Rat pancreas.

To obtain pancreata for immunohistochemistry (IHC), 5-month-old HIP and wild-type rats (n = 5 for each group) were perfused and fixed with 4% paraformaldehyde. The tissue was paraffin embedded and cut into 4-μm sections. The IHC procedure for rat pancreatic tissue was the same as that described for the human pancreatic tissue.

Human pancreas.

At autopsy, pancreas was resected from the tail and, with a sample of spleen, fixed in formaldehyde and embedded in paraffin for subsequent analysis. The surgical specimen was a distal pancreatectomy, as recently described (6). Sections (4 μm) were cut from these paraffin blocks and mounted on Fisherbrand Ink Jet White IJL-6109-Plus-600621 charged slides (Fisher category no. 12550109; Fisher Scientific, Pittsburgh, PA). Sections were stained by immunofluorescence for either insulin and CHOP (two slides per case) or glucagon and CHOP (one slide per case), using methods as previously described (6), totaling three slides per case. Pancreatic sections were deparaffinized in toluene, rehydrated in grades of alcohols, washed in H2O followed by antigen retrieval using antigen-unmasking buffer (Vector Laboratories, Burlingame, CA), permeabilized in 0.4% Triton X-100/TBS, and blocked with 0.2% Triton X-100 and 3% BSA/TBS. Primary antibodies used were as follows: rabbit polyclonal anti-CHOP (Santa Cruz Biotechnology), guinea pig polyclonal anti-insulin (Zymed Laboratories), and mouse anti-porcine monoclonal glucagon (Sigma). The working dilution of the primary antibodies was 1:100 of the antibody solution with 0.2% Tween-20 and 3% BSA/TBS. Donkey-derived secondary antibodies conjugated to Cy3 and fluorescein isothiocyanate were diluted to 1:200 (Jackson ImmunoResearch Laboratories). All slides were mounted with Vectashield (Vector Laboratories) with DAPI and coverslipped. As a positive control for nuclear CHOP to identify ER stress, INS-1 cells were treated with 2.5 μm thapsigargin (Sigma) versus DMSO for 16 h; then, the INS-1 cells were fixed in 4% paraformaldehyde at room temperature for 30 min and permeabilized with the 0.2% Triton X-100 in PBS for 10 min. Then, the same anti-CHOP IHC procedure described above was used for detection of CHOP expression. Nuclear CHOP staining was detected in 73.2 ± 5.9 vs. 0.8 ± 0.1% (n = 4) of INS-1 cells treated with thapsigargin vs. DMSO, respectively. To assure specificity of primary antibodies, both a mouse monoclonal anti-CHOP (Abcam, CA) and a rabbit polyclonal anti-CHOP (Santa Cruz Biotechnology) were used, side by side, in the immunofluorescence protocol. The same pattern and frequency of nuclear CHOP staining was observed with both primary antibodies in human pancreatic tissues (Table 1) and INS-1 cells. We subsequently used the rabbit polyclonal anti-CHOP antibody in the human and rat studies.

Image analysis using an epifluorescent microscope and a confocal laser scanning microscope.

Fluorescent slides (three slides per case) were viewed using a Leica DM6000 microscope (Leica Microsystems, Wetzlar, Germany) connected to an Apple computer, and images were acquired using Openlab software (Improvision).


The number of islets with at least 20 β-cells per islet section on each slide varied from case to case, ranging from 71 to 157 in lean and obese controls and obese cases of type 2 diabetes. For detailed evaluation of CHOP expression, 15 islets per case (with a minimum of 20 β-cells per islet in plane of section) were selected at random in these cases. In the recent-onset type 1 diabetes case, there were 8–110 islets per section; four sections were studied, with detailed morphometic analysis of CHOP expression performed in 100 islets. As expected, insulin-positive β-cells were much less frequent in the three cases of long-standing type 1 diabetes. In these cases, 1–8 islets with >145 β-cells per plane of section were present in each slide (28 islets evaluated). The nuclear localization of CHOP was confirmed by use of a Leica spinning-disc laser confocal microscope (DMIRE2; Leica, Dearfield, IL) with a digital camera (Hamamatsu, Japan), and images were acquired using Volocity software (McBain Instruments, Chatsworth, CA). The series (z) sections were acquired with 0.4-μm step size. The images (512 × 512 pixels) were saved as TIF files, and the contrast levels of the images were adjusted with Adobe Photoshop (Adobe, Mountain View, CA). All morphometric analysis was evaluated independently in a blinded manner by two observers.

HIP rats.

Ten islets per tissue section were photographed and analyzed from each pancreas. The total number of insulin-positive cells that were CHOP positive or caspase-12 positive were counted and expressed as a fraction of the β-cells. The detailed procedure was the same as that for humans.

Statistical analysis.

To test specific hypotheses posed, we used ANOVA, and post hoc testing was performed when a significant difference existed. A P value <0.05 was considered statistically significant.


Apoptosis induced by human IAPP was mediated by CHOP in INS-1 cells.

To establish the mechanism of endogenously expressed human IAPP-induced apoptosis, we generated adenoviruses expressing human and rat proIAPP EGFP. Human and rat IAPP were expressed at similar levels in INS cells when transduced at MOI = 100 (Fig. 1A). Expression of CHOP, a mediator of ER stress (21), was increased in response to transduction with Ad-hIAPP EGFP but not Ad-rIAPP EGFP (Fig. 1B). Consistent with the actions of CHOP to mediate ER stress–induced apoptosis, increased CHOP expression and nuclear translocation preceded DNA fragmentation (TUNEL) in human IAPP–expressing cells (Fig. 1C vs. D). Moreover, nuclear CHOP was colocalized with TUNEL in Ad-hIAPP EGFP cells (Fig. 1E). Furthermore, after siRNA was used to knockdown CHOP expression by 81%, Ad-hIAPP EGFP–induced cleavage of caspase-3 was reduced by 68% (Fig. 1F).

FIG. 1.

Human IAPP–induced CHOP expression precedes apoptosis in human IAPP–expressing INS-1 cells. A: INS-1 cells were transduced with adenovirus-expressing GFP, Ad-rIAPP EGFP, Ad-hIAPP EGFP, or nothing. Cell lysates were prepared after 24 h. Equal amounts of protein samples were separated by SDS-PAGE and transferred. Membranes were probed with anti-GFP (A), anti-CHOP (B), and anti-actin antibodies. Unprocessed IAPP EGFP is 34 kDa (upper band), whereas processed COOH-terminal IAPP plus EGFP is 28 kDa (lower band). After 8, 16, 24, or 48 h, cells were fixed and immunostained for TUNEL (apoptosis) (C) or CHOP (D). Values are TUNEL- or CHOP-positive cells per 100 cells (means ± SEM) with three experiments. *P < 0.05 relative to INS-1 cells expressing GFP, rIAPP, or nontransduced, by ANOVA. E: Representative microphotograph shows the colocalization of CHOP (panel A) (blue) and TUNEL (panel B) (red) of INS cells (panel C) (EGFP positive) transduced with Ad-hIAPP EGFP (MOI = 100) for 36 h. The bar scale is 10 μm. F: The representative Western blot shows that CHOP siRNA reduced CHOP levels by 81% and decreased caspase-3 cleavage by 68% as measured by optical density and normalized with actin.

Human IAPP–induced expression of caspase-12 and CHOP in HIP rats.

To affirm that human IAPP–induced ER stress occurs in primary β-cells in vivo, we examined the HIP rat, a well-characterized model of type 2 diabetes, characterized by islet pathology comparable with that in humans with type 2 diabetes (12). The HIP rat is transgenic for human IAPP, expressing high levels of human IAPP. Activated caspase-12, a marker of ER stress in rodents (22,23), was much more abundant in HIP than in wild-type islets (Fig. 2B and Table 2). Caspase-12 was present in a perinuclear pattern and colocalized with insulin-positive β-cells but not other islet cell types (Fig. 2A and Table 2). Functional caspase-12 protein is not expressed in human tissues, owing to mutations leading to nine alternative splicing sites (24), so caspase-12 was not examined in human tissue. The IHC caspase-12 findings in the HIP rat were confirmed by Western blot analysis (Fig. 2B). Furthermore, islet lysates from HIP rats had a high expression level of cleaved caspase-3 compared with that in islet lysates from wild-type rats (Fig. 2C). Specificity for caspase-12 in IHC and Western blots was confirmed by using two different anti–caspase-12 antibodies, one monoclonal (22) and one polyclonal (21). TUNEL assays showed positive nuclear staining in HIP rats (Fig. 2D, panels DF), which was rarely seen in wild-type rats (Fig. 2D, panels AC).

FIG. 2.

Caspase-12 is highly expressed and activated in β-cells of HIP rats. A: Representative microphotographs of rat islets show that caspase-12 (red)–positive cells were insulin-staining (green) β-cells by fluorescence IHC of paraffin-embedded pancreatic sections. For quantification, please see Table 2. B: Representative Western blots confirm the presence of activated caspase-12 (four cleaved bands) in the islets from HIP rats but only a small amount of procaspase-12 (51 kDa) presence in islets from wild-type rats. Three independent Western blots had similar results. C: Western blot shows the cleaved caspase-3 (17 kDa) in HIP rats but a much lower level in wild-type rats (four animals per group with similar results). D: TUNEL assays show positive TUNEL nuclei (panels DF) in HIP rat islets but rarely in wild-type controls (panels AC). Activated caspase-12 can cleave procaspase-3, as reported by Hitomi et al. (ref. 43) in 2004. Bar scale represents 10 μm.


Percentage of β-cells that expressed ER stress markers in pancreatic islets from HIP and wild-type rats at 5 month of age

CHOP expression was increased in β-cells of HIP versus wild-type rats, predominantly in a perinuclear pattern (Fig. 3A and Table 2). In HIP rats, CHOP expression was occasionally nuclear (Fig. 3A, inset and arrow), with a frequency of 0.5 ± 0.2% of β-cell nuclei vs. none in wild-type rats (Table 2). While the frequency of β-cell apoptosis was increased in the HIP versus wild-type rat, only a minority of β-cells are at any time apoptotic (Fig. 2D) in the HIP rat at a frequency consistent with nuclear CHOP staining observed here. This finding of occasional nuclear CHOP is also consistent with prior reports of ER stress–induced apoptosis in tissue rather than isolated cells in culture (25). Western blot analysis of protein lysates from isolated islets confirmed increased CHOP protein expression in HIP rats (Fig. 3B).

FIG. 3.

ER stress marker GADD153/CHOP is induced in the pancreatic β-cells of HIP rats. A: IHC staining demonstrates the colocalization of CHOP (arrow) with insulin-positive β-cells. The arrow and inset indicate the nuclear CHOP in HIP but not wild-type rats. The arrow head shows a perinuclear cytosolic CHOP. For quantification, see Table 2. B: Western blot analysis (three independent immunoblotting) confirms abundant CHOP (29 kD) in islets from HIP rats and only a very low expression level in wild-type rats. Bar scale represents 10 μm. M, molecular weight.

CHOP expression in human pancreas

Type 1 diabetes.

Perinuclear β-cell CHOP expression was rarely detected in cases of type 1 diabetes or in lean nondiabetic humans (Figs. 4 and 5). This was true both in the recent-onset case of type 1 diabetes (Fig. 4AC) and in the cases of long-standing type 1 diabetes (Fig. 4DF). The frequency of occasional CHOP-positive cells in the nonendocrine pancreas was comparable in all groups, including the cases of type 1 diabetes (Fig. 4).

FIG. 4.

CHOP is not expressed in type 1 diabetic human pancreas. AC: Representative microphotograph of pancreatic islet from a human with recent-onset type 1 diabetes, with T-cell infiltrate marked by an arrow head. DF: Representative micrographs of a human with long-standing type 1 diabetes. In contrast with type 2 diabetes, CHOP immunostaining was rarely seen in islets of recent-onset or long-standing type 1 diabetes and no more frequently than in nondiabetic controls. Occasional CHOP-positive cells (B and C, white arrows) were present in parenchymal tissue at a frequency comparable with that in nondiabetic pancreas. Scale bar = 10 μm.

FIG. 5.

Expression of perinuclear CHOP and insulin in human pancreas. Representative microphotographs of pancreatic islets from lean nondiabetic (LN) (AC), obese nondiabetic (OB) (DF), and obese type 2 diabetic (GI) subjects obtained at autopsy and stained for insulin (A, D, and G, green), CHOP (B, E, and H, red), and nucleus (DAPI, blue). Perinuclear CHOP expression is increased in β-cells of obese nondiabetic compared with lean nondiabetic cases but is most abundant in type 2 diabetic subjects. See Fig. 6 for the quantification. Scale bar = 10 μm. The arrows show perinuclear staining.

Impact of obesity.

Perinuclear cytoplasmic CHOP expression was rarely detected in β-cells of lean nondiabetic humans (Fig. 5AC) but more frequently in obese nondiabetic humans (Fig. 5DF) (2.6 ± 2.0 vs. 14.6 ± 3.0%, P < 0.05 [see Fig. 6]). Nuclear CHOP was not detected in β-cells from lean nondiabetic cases and was only very rarely detected in obese nondiabetic cases.

FIG. 6.

Expression of CHOP in human pancreatic β-cells. Percentage of β-cells with CHOP expression (A: cytoplasmic CHOP; B: nuclear CHOP) detected by immunofluorescence in lean nondiabetic (LN) (n = 7), obese nondiabetic (OB) (n = 12), obese type 2 diabetic (T2DM) (n = 14), and type 1 diabetic (T1DM) (n = 7) cases obtained at autopsy. See Figs. 4, 5, and 7 for the images of perinuclear and nuclear CHOP staining, respectively. *P < 0.05 compared with lean nondiabetic; **P < 0.01 compared with lean nondiabetic; and #P < 0.05 compared with obese nondiabetic subjects.

Type 2 diabetes.

The frequency of perinuclear cytoplasmic CHOP expression was nonsignificantly increased in obese cases with type 2 diabetes compared with obese nondiabetic controls (18.5 ± 3.6 vs. 14.6 ± 3.0%, respectively; P = 0.2) (Fig. 6). Within the islet, this perinuclear cytoplasmic CHOP expression was confined to β-cells and specifically not detected in α-cells (data not shown). β-Cell nuclear CHOP was detected six times more frequently (0.49 ± 0.17 vs. 0.08 ± 0.04%, P < 0.05) in type 2 diabetic cases compared with obese nondiabetic controls (Fig. 7AE) (by high-power light microscopy and by laser confocal microscopy) (Figs. 6 and 7F).

FIG. 7.

Expression of perinuclear and nuclear CHOP in human pancreatic β-cells. Pancreatic tissue from an obese case with type 2 diabetes was stained for insulin (A, green), CHOP (B, red), and nuclei blue (C, blue) with DAPI. D: Merged image (40×). E: High-power (63×) image of pancreatic β-cells. F: Confocal image of β-cells from another obese case with type 2 diabetes, stained for insulin (green) and CHOP (red). White arrows indicate CHOP-positive nuclei in β-cells. See Fig. 6 for quantification. Scale bar = 10 μm.


In the present study, we report that the β-cell ER stress characterized by increased expression and nuclear translocation of CHOP is a characteristic of type 2 but not type 1 diabetes in humans. Moreover, we report that high expression levels of human IAPP induces ER stress–mediated β-cell apoptosis that can be overcome by inhibition of CHOP expression. These data imply that different mechanisms initiate β-cell apoptosis in type 1 and type 2 diabetes and suggest that toxic oligomers of IAPP may play a role in ER stress–induced apoptosis in type 2 diabetes.

The adaptive mechanisms that protect the ER from accumulation of unfolded proteins are often collectively referred to as the unfolded protein response (UPR) and have been studied for the most part in cell culture (2529). The initial UPR depends on the property of three sensory proteins—PERK, Ire1α, and ATF6—to detect the presence of unfolded proteins in the ER. This activates a sequence of events that globally decreases translation of major ER client proteins, increases transcription and translation of ER chaperone proteins (e.g., BiP), and increases expression of proteins involved in clearance of misfolded ER proteins. The importance of these protective mechanisms for pancreatic β-cells is illustrated by the increased β-cell apoptosis initiated by ER stress in PERK−/− mice (30). Degenerative disease states characterized by ER stress are typically characterized by formation of protein oligomers (20). Once protein oligomers form, it appears that cells transition from the protective UPR to programmed cell death or apoptosis through ER stress. One of the most well-characterized mediators of ER stress–induced apoptosis is the transcriptional regulator CHOP. For example, ER aggregates of mutant insulin in the Akita mouse induce ER stress–mediated β-cell apoptosis mediated by CHOP (31).

We report that in nondiabetic humans, obesity was characterized by increased perinuclear CHOP expression, although this was not accompanied by an increase in the frequency of nuclear CHOP translocation. Another recent report has shown increased cytoplasmic CHOP detected by IHC in humans with type 2 diabetes, but nuclear CHOP was not reported (32). The β-cell workload has been shown to predict vulnerability to apoptosis from several proapoptotic stimuli, and β-cell rest has been shown to delay or prevent onset of diabetes (33), perhaps mediated in part by differential expression of CHOP. The perinuclear expression of CHOP was increased to an even greater extent in obese individuals with type 2 diabetes, but this was now also accompanied by an increase in the frequency of β-cells with CHOP nuclear translocation. Taken together, these data imply that obesity increases transcription of CHOP but that nuclear translocation of CHOP is provoked by a factor (or factors) present in type 2 diabetes but not in obesity. Based on these data, the transcriptional regulation of CHOP expression and the factors(s) that lead to its nuclear translocation are apparently distinct. Further studies are required to establish what factors mediate the translocation of cytoplasmic to nuclear CHOP and to establish the relative roles of cytoplasmic CHOP versus those of nuclear CHOP in promoting ER stress–induced apoptosis. Most studies of ER stress–induced apoptosis have been carried out in cell lines where the ER stress–induced apoptosis appears to be related to nuclear CHOP (as we indeed observe in the INS-1 cell studies presented here). If, as we postulate, nuclear translocated CHOP is required to mediate ER stress–induced apoptosis, then in the setting of increased CHOP expression in obesity, the factor(s) that trigger nuclear translocation of CHOP would presumably be amplified.

Several mechanisms have been proposed as initiators of β-cell apoptosis in type 2 diabetes, including exposure to high glucose or fatty acids (7,8). Exposure of human β-cells to high glucose may initiate apoptosis through generation of oxygen free radicals or induction of expression of interleukin-1β (7,9). However, the absence of ER stress noted here in β-cells from humans with either long-standing or recent-onset type 1 diabetes in cases with documented increased β-cell apoptosis (1,6) argues against the primacy of glucose and/or cytokine toxicity as the underlying mechanisms subserving ER stress in type 2 diabetes. Also, although marked obesity appears to increase the β-cell expression of CHOP, neither β-cell apoptosis (3) nor nuclear translocation of CHOP was increased in islets of patients with morbid obesity without type 2 diabetes, implying that increased free fatty acids alone are unlikely to be responsible for increased ER stress or β-cell apoptosis in type 2 diabetes, although exposure of a β-cell to toxic free fatty acids in culture does initiate ER stress (32). In reality, the pathways that induce apoptosis are closely interrelated. For example, any damage to the mitochondria (through the so-called intrinsic pathway of apoptosis) will likely lead to failure of the ER to sufficiently fold proteins and potentially provoke the ER stress pathway. The ER stress pathway has been shown to induce activation of proinflammatory cytokines that may induce the so-called extrinsic pathways through death receptors (17,34). The ER and mitochondria exchange membranes, and thus formation of membrane permeant oligomers, such as those that form intracellularly in human IAPP transgenic mice (19), may lead to disruption of mitochondria and induce the intrinsic pathway of ER stress.

The most clearly established causes of ER stress–induced apoptosis involve protein mutations leading to protein misfolding, for example, the missense mutation in insulin 2 in the Akita mouse (35) and peripheral neuropathy as a result of mutations in the P0 protein (36). High expression rates of human IAPP in mice leads to the formation of toxic intracellular IAPP oligomers (37), implying misfolding and/or intracellular trafficking of human IAPP beyond a critical threshold of expression. The islet in type 2 diabetes is characterized by islet amyloid derived from IAPP (2,3), implying that protein misfolding is characteristic of this disease, but, of note, islet amyloid is not a feature of type 1 diabetes (2). Proof of principal of a potential role of IAPP misfolding playing a role in type 2 diabetes in humans is the rare S20G mutation in IAPP that increases its oligomeric and cytotoxic properties (38) and is associated with increased risk for type 2 diabetes (39). This is analogous to point mutations in AβP (Alzheimer's β protein)1–42 leading to hereditary forms of Alzheimer's disease (40). However, in most cases of type 2 diabetes, as in most cases of Alzheimer's disease, there are no mutations in either the IAPP transcript or promoter region (41) or, indeed, in the transcript or promoter region of AβP1–42, respectively (42). Also, only species with a primary sequence of IAPP with the propensity to form membrane-permeant toxic oligomers (humans, cats, and monkeys) spontaneously develop type 2 diabetes (10). However, it is not yet established why IAPP forms aggregates in humans who develop type 2 diabetes or whether these aggregates underpin the increased β-cell apoptosis in this disease. Moreover, while β-cell apoptosis is increased in type 2 diabetes, this only involves a relative minority of β-cells at any time, so loss of β-cell mass is gradual.

In contrast, a high proportion (the majority) of cells exposed to inducers of ER stress in culture have nuclear CHOP staining and undergo apoptosis within 24 h, confirmed by us in the positive control studies with INS-1 cells exposed to thapsigargin (an ER stress inducer), as presented here (25). If the frequency of ER stress–induced apoptosis in vivo approached the proportion of cells impacted in studies of ER stress induced in culture, the consequences would be short-lived and devastating. Fortunately, the loss of β-cell function (and presumably β-cell mass) in type 2 diabetes is a much more gradual process (11), and the frequency of nuclear CHOP observed here is comparable with that of β-cell apoptosis (3,4), as previously reported in type 2 diabetes. These findings emphasize the important protective mechanisms present in vivo to prevent ER stress–induced apoptosis in most humans (for example, obese individuals who do not develop type 2 diabetes).

In summary, we report that increased expression of human IAPP in INS-1 cells and human IAPP transgenic rats leads to ER stress–induced apoptosis. We note that this ER stress–induced activation of caspase-3 is overcome by knockdown of CHOP. We also report that obesity in humans is characterized by increased perinuclear cytoplasmic CHOP expression, although by very minimal nuclear translocation of CHOP. In contrast, in type 2 diabetes, increased cytoplasmic CHOP is accompanied by nuclear translocation of CHOP. These observations have led us to a novel hypothesis as to how obesity might predispose to β-cell ER stress. If, as we postulate, nuclear CHOP mediates β-cell apoptosis, then if the unknown signal that causes nuclear CHOP translocation were triggered, this signal would be amplified in the context of obesity. Future studies will be required to test that hypothesis. Meanwhile, the present data support the postulate that in patients with obesity and type 2 diabetes, ER stress is likely an important mechanism leading to increased β-cell apoptosis. In this context, the islet in type 2 diabetes is further characterized as revealing features of an unfolded protein disease, in common with most neurodegenerative diseases.


These studies were funded by the National Institutes of Health (grants DK 59579 [to P.C.B.] and DK29953 [to R.A.R.]) and by the Larry Hillblom Foundation.

We are grateful to Aleksey Matveyenko and Heather Gerber for their excellent technical support, Anil Bhushan and Kathrin Maedler for helpful suggestions, and Bonnie Lui for excellent editorial assistance.


  • Published ahead of print at on 2 May 2007. DOI: 10.2337/db07-0197.

  • The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

  • Received February 12, 2007.
  • Accepted April 10, 2007.


| Table of Contents

This Article

  1. Diabetes vol. 56 no. 8 2016-2027
  1. All Versions of this Article:
    1. db07-0197v1
    2. 56/8/2016 most recent