Ubiquitinated-Protein Aggregates Form in Pancreatic β-Cells During Diabetes-Induced Oxidative Stress and Are Regulated by Autophagy

INS1 832/13 β-cells were obtained from Dr. Christopher Newgard (18) and were cultured in RPMI 1640 with 10% fetal bovine serum, 10 mmol/l HEPES, 1 mmol/l pyruvate, l-glutamine, 50 μmol/l β-mercaptoethanol, and antibiotics and seeded in 24-well plates on glass coverslips 16–24 h before use. Cells were grown at 37°C in 5% CO₂. In high glucose studies, the medium was supplemented to 30 mmol/l glucose. Cells were transfected with GFP-LC3 (19) using FuGene 6. LCB is a protein marker of autophagosomes.

The following agents (all from Sigma-Aldrich except where noted) were used at the indicated concentration: cycloheximide (20 μg/ml), nocodazole (5 μmol/l), cytochalasin D (10 μmol/l), 3-amino-1,2,4-triazole (ATZ) (1 mmol/l), 3-methyladenine (3MA) (10 mmol/l), puromycin (5 μg/ml), epoxomicin (1 μmol/l; Biomol), NAC (1 mmol/l), taurine (1 mmol/l), and human insulin (100 nmol/l; Eli Lilly).

Male Zucker diabetic fatty rats (ZDF/Crl-Lepr^fa) and male Zucker lean rats (+/?) were obtained from Charles River Laboratories at 5 weeks of age with initial weights of 90–135 g. The animals were fed water and Purina 5001 chow ad libitum throughout the experiment. All experiments were approved by the Animal Care Committee of the University of Toronto in accordance with regulations set forth by the Canadian Council for Animal Care.

Within 10 min of killing, the pancreas was removed and blotted, and extraneous fat and lymph nodes were removed. The pancreas was then weighed before being placed in Bock's fixative. After fixation, tissue samples were cut and placed into tissue cassettes that were placed in 70% ethanol until paraffin embedding. Four-micrometer slices were cut on an Olympus microtome (Carsen Group, Markham, ON, Canada) from paraffin blocks and mounted onto 25- × 75-mm slides.

All frozen tissues were mounted using OCT mounting compound and were cut at 10 μm using a cryostat. Coronal brain sections containing the hippocampus and pituitary sections containing the anterior, intermediate, and posterior lobes were cut (10 μm) (20). Tissue sections were thaw-mounted onto VistaVision HistoBond adhesive slides.

After sectioning, frozen tissues were placed in 4% formaldehyde in PBS for 5 min at 4°C. Slides were washed in PBS, put in methanol for 10 min at −20°C, and allowed to air dry for 3 h at room temperature. They were subjected to graded ethanol rehydration (100–50%) and brought to water. For paraffin-embedded pancreata, sections were mounted onto slides and dried for staining. Glass-mounted sections were cleared from paraffin with xylene and rehydrated by sequential washings with graded ethanol solutions (95–70%). All slides were subjected to permeabilization for 3 h, and sections were incubated overnight at 4°C with the primary antibodies. The slides were washed with PBS, followed by the specific secondary antibodies for 2 h. The following antibodies were used for immunofluorescence: 1) mouse monoclonal antibody (mAb) FK2 (Biomol), 2) rabbit polyclonal antibody against insulin (Santa Cruz Biotechnology, Santa Cruz, CA), and 3) a rabbit polyclonal antibody against glucagon (Novocastra). The nuclei were stained using DAPI.

INS1 832/13 β-cells (18) were fixed in 2.5% paraformaldehyde in PBS pH 7.2 for 15 min at 37°C. Fixed cells were stained as previously described (21). The rabbit 20S proteasome antibody was from Biomol. Samples were analyzed using a Leica DMIRE2 fluorescence microscope (×63 and ×100 objectives) and OpenLab software. Images were imported into Adobe Photoshop and assembled in Adobe Illustrator for labeling.

A Leica DMIRE2 fluorescence microscope was used to visually quantify Ub-protein aggregates. In the tissue sections, Ub-protein aggregates were counted in random fields of cells stained for Ub-proteins with the FK2 antibody and insulin to visualize pancreatic islets and for the nucleus using DAPI. At least 100 cells were counted for each experiment. The average ± SD for at least three experiments is presented in each experiment. Statistical analysis was done using a Student's t test.

Hyperglycemia is thought to trigger ER stress (13). Through activation of ERAD, an increase in protein ubiquitination of misfolded proteins might be expected. To investigate protein ubiquitination in a diabetic model, tissues sections were prepared from 19-week-old Zucker diabetic fatty rats (ZDF/Crl-Lepr^fa). These animals exhibit hyperphagia, obesity, and the resulting diabetic phenotype due to a missense mutation of the COOH-terminal amino acid sequence of the leptin receptor (22–24). As controls for nondiabetic animals, 6-week-old Zucker diabetic fatty rats and 19-week-old Zucker lean control (+/fa, +/+, +/?) rats were also examined. Blood glucose levels were monitored in the Zucker diabetic fatty rats to verify they had diabetes compared with the control animals (Table 1). Tissue sections from each animal, including hippocampus, kidney, liver, muscle, pituitary, brain, spleen, and pancreas, were isolated and stained with the mAb FK2, which recognizes both mono–and poly–Ub-proteins but not free ubiquitin (25).

In muscle and pituitary tissue sections, we observed a diffuse cytosolic stain that did not change in the 19-week-old Zucker diabetic fatty rats compared with control animals (data not shown). However, tissue from the kidney and liver had Ub-protein aggregates in the 19-week-old Zucker diabetic fatty rats that were not present in the tissues from 6-week-old Zucker diabetic fatty and 19-week-old Zucker lean control rats (data not shown). Similar results were found in the hippocampus (data not shown), a site where neurodegeneration is associated with diabetes (26). These observations suggest that protein ubiquitination and aggregation occur as a result of diabetes in some tissues.

Pancreatic sections from each animal were isolated and stained as above and were stained in parallel for insulin to identify pancreatic β-cells. As shown in Fig. 1A, all pancreatic tissues contained a small number of infiltrating immune cells that stain nonspecifically with secondary antibodies (Fig. 1A, arrowheads). However, large Ub-protein aggregates were observed in the pancreas of the 19-week-old Zucker diabetic fatty rats (Fig. 1A, arrows). Because the pancreas exhibited the most robust ubiquitination phenotype and is affected during diabetes, we decided to further our investigations in this tissue. Closer analysis revealed that the Ub-protein aggregates were present in the cytosol of both β-cells and the surrounding acinar cells (Fig. 1A, inset). Ub-protein aggregates were present in acinar cells throughout the pancreas and were not found in higher abundance near islets. In contrast, the signal for Ub-proteins was diffuse in the cytosol of both of the nondiabetic control rats. The number of Ub-protein aggregates in β-cells and acinar cells was quantified in each group of rats. In the control rats, the number of aggregates was low, whereas the 19-week-old Zucker diabetic fatty rats showed a significant increase in the number of these structures (Fig. 1B). Ub-protein aggregates were also observed in α-cells of the 19-week-old Zucker diabetic fatty rats when pancreas sections were co-stained with glucagon to identify α-cells (data not shown).

We characterized the Ub-protein aggregates in INS1 832/13 cells, a pancreatic β-cell line (18). As shown in Fig. 2, aggregates morphologically similar to those observed in pancreatic β-cells in vivo were observed in INS1 832/13 cells when grown in 30 mmol/l glucose (a concentration used in studies of chronic hyperglycemia in β-cell models [6]). Aggregates were not observed when cells were grown in 11 mmol/l glucose (Fig. 2B) (18,27). Ub-protein aggregates increased with the duration of high glucose treatment, appearing as early as 8 h and being maximal in number at ∼48 h (Fig. 2A and B). Insulin-containing secretory granules were observed in INS1 832/13 cells, consistent with previous findings (27). The Ub-protein aggregates and insulin did not colocalize, suggesting that misfolded insulin is not likely targeted to these structures (Fig. 2A, arrows).

High glucose is known to induce insulin secretion by β-cells (1). Therefore, we wondered whether secreted insulin might induce Ub-protein aggregate formation by signaling through the insulin receptor. To test this, INS1 832/13 cells were treated for 8 and 24 h with normal medium (11 mmol/l glucose) containing exogenous insulin. Under these conditions, no Ub-protein aggregates were observed (data not shown), demonstrating that insulin signaling does not induce Ub-protein aggregate formation.

Formation and retention of Ub-protein aggregates was dependent on the continuous presence of high glucose. When cells treated with 30 mmol/l glucose for 48 h were returned to normal medium (11 mmol/l glucose), the number of Ub-protein aggregates decreased (Fig. 2A and B). Thus, high glucose induces the formation of Ub-protein aggregates in insulin-secreting INS1 832/13 β-cells, and these structures are subject to endogenous clearance mechanisms if returned to normal glucose levels.

When production of misfolded proteins exceeds degradation, these proteins can be sequestered into protein storage compartments. One such compartment, the aggresome, is localized at the microtubule organizing center (MTOC) (28,29). Aggresomes contain Ub-proteins and have been characterized in a number of cell lines (28,29), including the pancreatic cancer cell line L3.6pl (30). We investigated the possibility that the Ub-protein aggregates that we observed in vitro may be aggresomes. INS1 832/13 β-cells were grown in high glucose for 24 h, fixed, and costained for Ub-proteins and γ-tubulin, a marker of the MTOC. Ub-protein aggregates did not localize to the MTOC, suggesting that they are not aggresomes (Fig. 3A).

Recently, we and others have described the formation of aggresome-like induced structures (ALIS) in a variety of cell types in response to various cellular stresses, including microbial infection (21,31–36). Unlike aggresomes, ALISs do not localize to the MTOC and are typically found in greater numbers than aggresomes. ALIS are formed in a microtubule- and actin-independent manner. Conversely, aggresome formation requires both cytoskeletal structures to be intact (28,29). ALIS formation is thought to result from the aggregation of nascent proteins that are misfolded, so-called defective ribosomal products (DRiPs) (36). Consequently, under most conditions, ALIS formation is impaired by blocking protein synthesis with cycloheximide (21,34–36).

To examine the nature of the Ub-protein aggregates formed in response to high glucose, INS1 832/13 cells were subjected to pharmacological analysis during high glucose treatment. Ub-protein aggregates were compared with cells grown in medium without drug treatment. As shown in Fig. 3B, Ub-protein aggregates were observed at high levels when cells were treated with high glucose in the presence of either cytochalasin D or nocozadole. Thus, formation of these aggregates is independent of the actin and microtubule cytoskeleton.

To test whether protein synthesis is required for aggregate formation, cells were treated with cycloheximide, which inhibits ALIS formation under most conditions (34,35). Surprisingly, cycloheximide promoted Ub-protein aggregate formation at both normal and high glucose levels (Fig. 3B). To examine whether DRiPs contributed to aggregate formation, cells were treated with puromycin, which induces DRiP formation at normal glucose concentrations (36). Puromycin was shown to induce ALIS formation in dendritic cells (36) and HeLa cells (34). Puromycin treatment resulted in an increase of Ub-protein aggregates compared with the nontreated cells at 2, 4, and 6 h at 11 mmol/l glucose (Fig. 3C). These findings demonstrate that DRiP formation is sufficient to induce Ub-protein aggregates in INS1 832/13 cells. However, our results with cycloheximide demonstrate that protein synthesis, and consequent DRiP generation, is not required for Ub-protein aggregate formation during high glucose treatment.

Together these findings demonstrate that Ub-protein aggregates induced by high glucose in INS1 832/13 cells are distinct from aggresomes. In addition, INS1 832/13 Ub-protein aggregates display some similarities to ALIS, in that they are microtubule and actin independent. However, our studies suggest that newly synthesized protein is not targeted to or the main component of β-cell Ub-protein aggregates.

Glucose toxicity is associated with the formation of reactive oxygen species (3). Oxidative stress has been shown to induce ALIS formation in a number of different cell types (34). The possibility that oxidative stress could cause Ub-protein aggregates to form in β-cells was explored. INS1 832/13 cells grown in 11 mmol/l glucose were treated with ATZ, a drug that causes oxidative stress by inhibiting catalase (37). Cells were then fixed and stained for Ub-proteins. Treatment with ATZ for 10 h caused the formation of Ub-protein aggregates in ∼35% of cells, compared with 5% in the untreated cells (Fig. 4). Therefore, oxidative stress is sufficient to induce Ub-aggregate formation in INS1 832/13 cells.

Antioxidants are thought to protect β-cells from oxidative stress during hyperglycemia (rev. in 4). Here, we observed a significant decrease in the percentage of cells with Ub-protein aggregates when two different antioxidants were included with ATZ treatment (Fig. 4). These included NAC, which raises cellular glutathione levels (38), and taurine, a semi-essential amino acid that has the ability to scavenge the reactive oxygen species hypochlorite (39). These antioxidants have been used in Zucker diabetic fatty rats to prevent glucose toxicity (40). Because hyperglycemia can result in an increase in oxidative stress, we hypothesized that adding antioxidants would also effect formation of Ub-protein aggregates during high glucose treatment. The addition of either NAC or taurine to high glucose medium resulted in a significant decrease in Ub-protein aggregates after 10 h, compared with high glucose treatment alone (Fig. 4). Taken together, these data demonstrate that diabetes-induced oxidative stress plays an important role in the formation of Ub-protein aggregates in β-cells.

The clearance of many Ub-proteins and protein aggregates is mediated by proteasomal degradation (41–43). Some of the Ub-protein aggregates induced by high glucose treatment colocalized with the 20S proteasome, indicating proteasome recruitment (Fig. 5A, arrows). Colocalization was observed as early as 8 h and was maximal (∼75% of cells) after 24 h (Fig. 5B). Following recovery after 48 h of high glucose treatment, Ub-protein aggregates were no longer observed, and the proteasome was diffusely localized to the cytosol and nucleus, similar to untreated control cells (Fig. 5A, bottom). These data demonstrate that the proteasome is recruited to Ub-protein aggregates in INS1 832/13 cells during high glucose treatment and suggest that the proteasome may play a role in the degradation of Ub-proteins localized to these structures and subsequent clearance of the aggregates. However, as will be shown (see below), a role for the proteasome in clearance of Ub-protein aggregates will be excluded.

We examined the mechanisms that regulate Ub-protein aggregate formation and clearance. First, cells were treated with high glucose for 24 h, allowing Ub-protein aggregates to form, and then the cells were returned to normal medium to allow clearance of the aggregates. As shown in Fig. 6A, recovery led to the rapid clearance of Ub-protein aggregates; after 10 h of recovery, there was an ∼50% drop in the number of cells containing these structures. In contrast, maintaining cells in the high glucose medium (30 mmol/l) led to a further accumulation of Ub-protein aggregates. Next, cells were treated with epoxomicin (a potent and specific inhibitor of the proteasome [35,44]) during recovery. Surprisingly, epoxomicin did not affect Ub-protein aggregate clearance. We also treated control cells grown at normal glucose levels with epoxomicin and did not observe Ub-protein aggregate formation (Fig. 6B). Thus, proteasome inhibition is not sufficient to induce Ub-protein aggregate formation. This suggests that proteasome inhibition does not underlie the accumulation of Ub-proteins during high glucose treatment.

Autophagy is involved in the degradation of misfolded proteins, cytoplasmic aggregates, and entire organelles by mediating their delivery to lysosomes (16). Recent studies have demonstrated a role for autophagy in the degradation of Ub-proteins, including those in ALIS (34,45,46). To investigate a role for autophagy in the clearance of Ub-protein aggregates during chronic hyperglycemia, cells were allowed to recover in the presence of 3MA, an autophagy inhibitor (34,47). Surprisingly, we observed an increase in Ub-protein aggregates in cells treated with 3MA that was higher than when cells were grown in high glucose alone (Fig. 6A). Consistent with this, treatment of control cells with 3MA, without prior high glucose treatment, was sufficient to induce the formation of Ub-protein aggregates (Fig. 6B). GFP-LC3, a marker of autophagosomes (19), colocalized with Ub-protein aggregates in cells treated with high glucose for 48 h, indicating autophagy recruitment (Fig. 5C). These results demonstrate that autophagy plays a critical role in mediating clearance of Ub-protein aggregates and regulating their formation under normoglycemic conditions.

Previous data suggest that oxidative stress is a causative factor for the formation of Ub-protein aggregates (34). Oxidative stress is also a well-documented cause of the progression and development of diabetes. Notably, antioxidant strategies are being considered for the preservation of β-cell function after the onset of hyperglycemia (3). Here, we show that oxidative stress is a strong inducer of Ub-protein aggregates in INS1 832/13 β-cells. Chronic high glucose, which is known to promote oxidative stress, also causes an increase in Ub-protein aggregates. It is likely that high glucose stresses the cell in other, unrelated ways, but the addition of antioxidants and their subsequent prevention of Ub-protein aggregates strongly suggests that what we observed during high glucose was the result of oxidative stress.

We observed that when high glucose–treated INS1 832/13 β-cells were allowed to recover at a basal glucose level, Ub-protein aggregates were no longer evident. Although the 20S proteasome was localized to the structures, our data indicate that the proteasome does not play a major role in aggregate clearance. We show that autophagy is a major participant is regulation and clearance of Ub-protein aggregates formed during hyperglycemia. In this way, autophagy helps prevent cellular damage associated with misfolded proteins and is also known to protect the cell from other types of damage, including the clearance of damaged mitochondria (48). Together with our observations, one can now appreciate a central role for autophagy in development, immunity, and cellular stress responses, including responses to hyperglycemia during diabetes (16).

We show here that Ub-protein aggregates formed under high glucose stress are distinct from aggresomes. Another type of Ub-protein storage compartment, termed ALIS, are formed during cellular stresses, including microbial infection, oxidative stress, and heat (21,31,33–36,49). Based on our analysis, we propose the Ub-protein aggregates observed in INS1 832/13 β-cells during high glucose treatment are ALIS. Under most conditions studied to date, cycloheximide blocks ALIS formation, suggesting that DRiPs are primarily targeted to ALIS (34,36). However, during starvation, ALIS form in the presence of cycloheximide, presumably via recruitment of preformed, long-lived proteins (34). Such a mechanism would allow amino acid recycling and energy generation under these conditions. Our finding that cycloheximide did not block Ub-protein aggregate formation during high glucose treatment of INS1 832/13 cells (Fig. 3B) suggests that long-lived proteins are similarly being targeted to ALIS. The Ub-protein in these aggregates during diabetes is unknown, although insulin did not appear to colocalize with these structures. Furthermore, a reduction in total cellular proinsulin levels was observed in INS1 832/13 cells treated with high glucose for 24, 48, and 72 h (data not shown), the same time points when Ub-protein aggregates appear. A decrease in insulin mRNA levels during chronic high glucose treatment is well-documented in the INS1 832/13 cell line (6,50) and other experimental systems (rev. in 51). A drop in insulin expression is a hallmark of glucose toxicity and is consistent with the notion that insulin is not targeted to the Ub-protein aggregates. The identity of proteins localized to Ub-protein aggregates during diabetes is an important question to be addressed in future studies.

We observed Ub-protein aggregates in pancreatic and hippocampal tissues (and to a lesser extent in liver and kidney) but not in muscle, pituitary, or spleen tissue taken from the same diabetic animal. The difference in aggregate formation between tissues may be attributable to 1) organ sensitivity toward hyperglycemia, 2) differences in vascularization, 3) differences in protein misfolding rates or degradation (i.e., autophagy), and 4) different degrees of oxidative stress in each tissue. With regard to the latter possibility, there is evidence in a diabetic rat model that the hippocampus is very sensitive to oxidative stress (52,53).

The ER is particularly vulnerable to the occurrence of misfolded proteins because of its contribution in posttranslational modification, folding, and assembly of newly synthesized proteins. A highly developed ER is present in pancreatic cells because of their involvement in insulin and digestive enzyme secretion. The UPR and ERAD are two safety mechanisms that have evolved to help the ER cope with misfolded proteins (Fig. 7, left). ERAD targets damaged and misfolded proteins for proteasomal degradation by tagging them with ubiquitin. In our model, oxidative stress damages long-lived proteins, and these are subsequently targeted to ALISs for ubiquitination (Fig. 7, right). From here, autophagy delivers the Ub-proteins to the lysosome for degradation. As with ERAD, ALISs are likely a cytoprotective system, in that they sequester misfolded protein and target them for degradation. These studies may have future potential clinical consequences, in that factors that promote autophagy may protect β-cells from cellular damage caused by oxidative stress during hyperglycemia.

N.A.K. has received a postdoctoral fellowship from the Canadian Association of Gastroenterology (CAG)/Canadian Institutes of Health Research (CIHR)/Axcan Pharma administered by the CAG. M.K. has received Natural Sciences and Engineering Research Council of Canada and Banting and Best Diabetes Center Novo-Nordisk Scholarships. H.B. has received a Canada Graduate Scholarship Doctoral Research Award from CIHR. M.V. has received CIHR grant MT2197. A.V. holds a Tier II Canada Research Chair Award and has received research funding from the CIHR and the Dean's Fund Award (University of Toronto). J.H.B. holds an Investigators in Pathogenesis of Infectious Disease Award from the Burroughs Wellcome Fund.

We thank members of the Brumell laboratory and Dr. Amira Klip for critical reading of this manuscript.