Full-length wild-type mouse insulin 2 cDNA was amplified by PCR from pIns2 (WT)-EGFP, a fusion construct obtained from Dr. Seiichi Oyadomari (New York University School of Medicine) 12 . The PCR fragment was cloned into pCRII-TOPO vector (Invitrogen). Using the QuikChange II XL Site-Directed Mutagenesis kit (Stratagene), a C96Y mutation was introduced into the insulin 2 cDNA. The mutation was confirmed by DNA sequencing. Ins2 (C96Y) was cloned into the pEGFP-N1 vector (BD Biosciences), the stop codon was removed and the resulting fusion construct pIns2 (C96Y)-EGFP was cloned into the NheI/NotI site of a pTRE-Tight vector (Clontech).

Initially a stable pTet-ON INS-1 clone was generated by transfecting INS-1 cells (obtained from Dr. Claes Wollheim, University of Geneva) 13 with the pTet-ON plasmid and stable clones were selected following the protocol provided (Clontech). Several clones were isolated and tested by transient transfection using a luciferase reporter assay and the clone with the lowest basal expression and highest doxycyline induction (INS1 pTet-ON #46) was used for generation of the double stable cell line.

Double-stable Ins2 (C96Y)-EGFP expressing cells were generated by co-transfection with a hygromycin resistance plasmid and the pTRE -Ins2 (C96Y)-EGFP plasmid into pTet-ON INS-1 #46 cells and selection by addition of antibiotics: 200 μg/ml geneticin and 50 μg/ml hygromycin B (Invitrogen). Individual clones were isolated and tested for Ins2 (C96Y)-EGFP expression by western blotting and immunofluorescence before and after induction with doxycycline (2 μg/ml) for 24 h. A positive clone was identified and fluorescence-activated cell sorting was used to isolate the highest 20% of EGFP-positive cells in this clonal population after induction with doxycycline for 48 h. The cells were expanded in media without doxycycline and referred to as clone #4S2.

Clone #4S2 cells were maintained in RPMI 1640 (11.1 mM glucose, 1 mM sodium pyruvate, 10 mM HEPES) supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 55 μM β-mercaptoethanol with antibiotics (100 units/ml penicillin and 100 μg/ml streptomycin) and selection drugs (200 μg/ml G418 and 50 μg/ml hygromycin).

Cells were washed in PBS and lysed in ice-cold lysis buffer (1% Triton X-100, 20 mM HEPES, pH 7.4, 100 mM KCl, 2 mM EDTA, 1 mM PMSF, 10 μg/ml leupeptin, and 10 μg/ml aprotinin, 10 mM NaF, 2 mM Na₃VO₄, and 10 nM okadaic acid) for 15-20 min on ice. Lysates were centrifuged at 13,000 rpm for 10 min at 4°C and the protein concentration in the supernatant was determined using the BCA protein assay (Pierce). Equal amounts of protein were resolved by SDS-PAGE and immunoblotted as described previously 14 . For detection of insulin and cleaved caspase-3 by immunoblotting, the samples were resolved using 4-12% NuPAGE gels (Invitrogen). The following primary antibodies were used: Phospho-eIF2α (Cell Signaling, #9721), GADD153/CHOP (Santa Cruz, sc-575), KDEL (StressGen, SPA-827), insulin (Santa Cruz, sc-9168), insulin (Dako, A0564), γ-tubulin (Sigma, T6557), polyclonal anti-GFP (obtained from Dr. James E. Rothman, Yale University), monoclonal anti-GFP (Clontech, 632381), cleaved caspase 3 (Cell Signaling, #9661S), ubiquitin (Dako, Z0458), GM130 (Transduction Laboratories, G65120), Hsp90 (Transduction Laboratories, 610418). The polyclonal Herp antibody was provided by Dr. Linda Hendershot (St. Jude Children's Hospital, Memphis, TN).

Cells were treated as described in the figure legends, washed with PBS and fixed with 4% paraformaldehyde in PBS for 20 minutes, washed with PBS and mounted on glass coverslips using Fluoromount G (EM Sciences). GFP fluorescence was visualized with a Zeiss laser scanning confocal microscope. Clone #4S2 INS-1 cells were treated as described in the figure legend, washed with PBS and fixed with 4% paraformaldehyde in PBS for 20 minutes. The samples were processed for immunofluorescence microscopy as reported previously 15 and immunostained with anti-ATF6 (TO13/14) antibody at a 1:100 dilution 14 . Quantitation of ATF6 staining in the nucleus (marked by DAPI staining) was measured using ImageJ Software. The average intensity in arbitrary units in the nuclear area was determined. A minimum of 18 cells for all experimental condition were quantified and the mean relative nuclear intensity is reported.

Following doxycycline treatment for 72 hours, clone #4S2 cells were washed with PBS and resuspended in homogenization buffer: (0.25 M Sucrose, 4 mM HEPES, 1 mM MgCl₂, 1.5 mM EDTA containing a complete protease inhibitor tablet (Roche), 1 mM PMSF, 10 mM NaF, 2 mM Na₃VO₄, 10 nM okadaic acid). Cells were homogenized on ice with 12 strokes using a ball-bearing cell homogenizer. The resulting homogenate was either centrifuged at 100 000×g for 1.5 h to generate membrane (pellet) and cytosol (SN) fractions, or subjected to a 0.45 M to 2 M sucrose gradient for 18 h at 30 000 rpm, 4°C as outlined in 16 . After ultracentrifugation, 0.5 ml fractions were taken from the top of the gradient.

Clone #4S2 cells were fixed in 2% glutaraldehyde (EM Sciences) in 0.15 M Sorensen's Phosphate Buffer (pH 7.4) (EM Sciences) for 30 min at room temperature. The cells were washed twice with PBS and collected into new tubes. Samples were prepared for transmission electron microscopy by the electron microscopy facility at Mount Sinai Hospital (Toronto). Briefly, the cell pellets were fixed in 2% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.3) for 2 h, rinsed with the same buffer for 10 min, and post-fixed for 1.5 h in 1% OsO₄. After rinsing with the same buffer for 10 min, the cells were dehydrated through a graded ethanol series up to 100% ethanol. The cells were then embedded in Spurr resin in an oven at 65°C overnight. Thin sections (~100 nm) were cut with an RMC MT6000 ultramicrotome. The sections were placed on copper grids and stained with uranyl acetate (20 min) and lead citrate (10 min). The grids were examined in a FEI Tecnai 20 transmission electron microscope and images were captured using a Gatan Dualview digital camera.

For real-time PCR analysis, total RNA was isolated from cells untreated or treated with 2 μg/ml doxycycline for the indicated times using TRIzol Reagent (Invitrogen) followed by purification with RNeasy Mini Kit (Qiagen). Total RNA was reverse transcribed to single-stranded cDNA using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). The resulting cDNA was used for real-time PCR analysis by the TaqMan Gene Expression system (Applied Biosystems) as described previously 14 . The following primers were obtained from Applied Biosystems: Chop (Rn00492098_g1), Atf4 (Rn00824644_g1), Grp78 (Rn00565250_m1), Pdi (Rn00564459_m1), Ero1α (Rn00593473_m1), Ero1β (Rn01520960_m1), Sel1 (Rn00710081_m1), Erdj4 (Rn00562259_m1), Erdj5 (Rn01486444_m1), β-actin (4352931E).

Total RNA was isolated and rat XBP-1 cDNA was amplified by RT-PCR (QIAGEN OneStep RT-PCR kit) using primers that flank the intron excised by IRE1 exonuclease activity as described previously 14 .

Clone #4S2 cells were reverse transfected using Lipofectamine RNAiMAX with 10 nM Herp siRNA (Invitrogen) or a control siRNA (directed to bacterial β-galactosidase) as described previously 17 . Twenty-four hours later the cells were incubated or not with doxycycline for 48 h then either lysed for western blot analysis or for measurement of apoptosis using the cell death ELISA kit (Roche).

Cell apoptosis was measured using the cell death detection ELISA kit (Roche) according to the instructions provided in the kit. Briefly, clone #4S2 cells were seeded in 24-well plates (200 000 cells/well) and treated as indicated in the figure legends. Following the treatments, the cells were lysed and oligonucleosomes in the cytoplasm (indicative of apoptosis-associated DNA degradation) were quantified according to the manufacturer's instructions.

Terminal deoxynucleotidyl transferase-mediated dUTP nick-end labeling (TUNEL) assays were performed using the APO-BrdU TUNEL Assay Kit (Invitrogen). Clone #4S2 cells were treated as described in the figure legend, washed with warm PBS, trypsinized, then stained with Fluor-647 dye-labeled anti-BrdU antibody, as described by the manufacturer's instructions. After staining, the samples were analyzed in a FACSCalibur flow cytometer (Becton Dickinson). To standardize each sample and set the quandrants, a negative control (INS-1 cells) and a positive control (provided by the APO-BrdU TUNEL assay kit) were used. GFP and the Fluor-647 dye-labeled anti-BrdU antibody were excited using the FL1 (530/30) and FL4 (661/16) lasers, respectively.

Microarray expression profiling was used to assess the global transcriptional changes in response to Ins2 (C96Y)-EGFP expression in clone #4S2 cells. Cells were induced or not with doxycycline for 24 h, 48 h or 5 days and total RNA was isolated using TRIzol Reagent (Invitrogen) followed by isolation using an RNeasy Mini Kit (Qiagen). Assessment of RNA quality and microarray analysis was performed at the University Health Network (Toronto) microarray centre. Briefly, following RNA quality assessment with an Agilent BioAnalyzer, samples were reverse transcribed to cDNA. cDNA was purified with a cDNA purification module from Affymetrix. Biotin was incorporated during in vitro transcription and purified cRNA was then fragmented with a chemical reaction. 15 μg of labeled and fragmented cRNA was hybridized to Rat Genome 230 2.0 arrays (Affymetrix Genechip) for 17 h at 45°C at 60 rpm. The arrays were stained and washed using fluidic stations with antibody and streptavidin phycoerythrin conjugate solutions. All arrays are scanned using an Affymetrix 3000 7G scanner. The data sets were analyzed using GeneSpring software (Version 7.1) (Agilent Technologies) and statistical analysis was performed using GeneSpring software according to the manufacturer's specifications. Genes with a minimum 2-fold difference between control and doxycycline-treated cells in at least two independent experiments at 24 h, 48 h and 5 days are listed in the Additional file 1 and Additional file 2. The raw microarray data has been deposited in the Gene Expression Omnibus (GEO) database (GSE22537).

Results are presented as mean ± SE. Statistical significance between two experimental conditions was analyzed using a two-sample t-test assuming equal variance. Data from several or more groups was analyzed by ANOVA, followed by Tukey post hoc test. P < 0.05 was considered statistically significant.

We first generated a stable pTet-ON INS-1 cell line that allows for doxycycline-inducible protein expression. The cell line was tested by transient transfection of the Ins2 (C96Y)-EGFP fusion construct in the presence or absence of doxycycline. Fusion protein expression was monitored by western blotting and fluorescence microscopy. Expression of the fusion protein was doxycyline-dependent and the fusion protein was excluded from the nucleus and localized throughout the cell (Figure 1A, B). Interestingly, the expression of this construct also resulted in the production of a lower migrating band, which may be due to proteolytic cleavage of the fusion construct in the cell (Figure 1A, arrow). A stable Ins2 (C96Y)-EGFP pTet-ON cell line was generated by co-transfection with a hygromycin resistance plasmid. We isolated a clone with inducible mutant insulin expression and fluorescence-activated cell sorting was used to isolate EGFP-positive cells (top 20%) after 48 h of doxycycline induction. The sorted cells (designated clone #4S2) was expanded in media without doxycycline. This clone exhibited low basal Ins2(C96Y)-EGFP expression and a marked induction following doxycycline treatment (Figure 1C, D).

In addition to the full-length fusion protein other GFP immunoreactive bands were observed in both TX-100 detergent whole cell lysates or lysates prepared by direct lysis of the cells in SDS sample buffer (Figure 2A). These fragments are likely degradation products of the full-length fusion protein and most of the fragments could not be efficiently immunoprecipitated using an anti-GFP antibody (Figure 2B). Cells were homogenized and subjected to membrane and cytosol fractionation. Whereas the full-length Ins2 (C96Y)-EGFP fusion protein could be detected in the membrane fraction, the degradation fragments were present exclusively in the cytosol fraction (Figure 2C). The full-length fusion protein was detected in ER fractions by sucrose density fractionation while the lower migrating degradation fragments were present exclusively in the cytosol fractions (Figure 2D). Since no degradation fragments were present in microsome fractions it is likely that degradation of the full-length fusion protein occurs in the cytosol. The presence of some soluble ER lumenal GRP78 (and Ins2 (C96Y)-EGFP) in cytosolic fractions in the sucrose gradient may be due to some leakage during cell homogenization.

We examined the morphology of the stable cell line by electron microscopy before and after doxycycline induction. In non-induced cells a normal ER morphology was observed (Figure 3A, B), whereas cells expressing Ins2 (C96Y)-EGFP for three days had an altered ER morphology (Figure 3C, D). The ER was expanded and dilated compared to control cells. The ER tended to be even more severely dilated in many of the cells after six days of mutant insulin expression (Figure 3E, F). This morphology is characteristic of unfolded protein accumulation and ER stress 2 and has been observed in β-cells of the Akita mouse 9 . In addition, apoptotic cells were observed in the population expressing the fusion protein (Figure 3C, arrowheads).

The morphological analysis indicated that the expression of Ins2 (C96Y)-EGFP was causing an accumulation of this protein in the ER and causing ER stress. We thus examined whether doxycycline-induced expression of Ins2 (C96Y)-EGFP was activating UPR signaling pathways. Ins2 (C96Y)-EGFP expression was induced for various times and activation of the ER stress-sensing pathways was monitored. Spliced XBP-1, indicative of IRE1α activation, was detected after 24 h of doxycyline induction and similar levels of spliced XBP-1 were present throughout the induction period for up to 7 days (Figure 4A, shows a representative example up to 5 days of induction). The levels of phospho-eIF2α, indicative of PERK activation, were also increased (Figure 4B), as were the levels of active ATF6-p50 (Figure 4C). Thus, expression of the mutant insulin causes measurable activation of the ER stress pathways mediated by IRE1α, PERK and ATF6.

To examine the global transcriptional response to mutant insulin expression we performed microarray analysis of the gene expression changes following Ins2 (C96Y)-EGFP expression. Total RNA was prepared after 24 h, 48 h and 5 days of doxycycline treatment from three independent experiments, reverse transcribed and hybridized to Affymetrix rat genome arrays. Gene expression changes were examined and those that were upregulated or downregulated by 2-fold after doxycyline-induced expression of Ins2 (C96Y)-EGFP are listed in the Additional files 1 and 2. Following 24 h induction 45 genes were upregulated and 5 were downregulated. After 48 h induction 86 genes were upregulated and 37 downregulated, whereas 68 genes were induced and 56 reduced after 5 days induction. These numbers include genes (both well-substantiated and transcribed loci) that were identified as statistically significant in at least two of three experiments.

Shown in Table 1 are a selected list of genes induced >2-fold in at least two independent experiments at the various time points. This list includes genes involved in the UPR response, protein folding and modification, ERAD and protein transport. Several patterns emerged from this analysis. After 24 h of mutant protein expression, the majority of the genes induced were ER localized chaperone genes (Erdj4/Dnajb9, P58^IPK /Dnajc3, Erdj3/Dnajb11, Fkbp11) as well as the ERAD-associated genes (Herp and Sel1). The Sdf2l1 gene was the most abundantly induced gene present in all three experiments. In addition, the Atf4 and Chop transcription factors, target genes of the PERK-eIF2α pathway, were also induced in two out of three experiments at this early time point. Trib3 (Trb3), a putative target gene of the CHOP transcription factor 18 , was also induced in two out of three experiments.

Following 48 h of mutant protein expression a greater number of genes were induced compared to 24 h (Table 1 and Additional files 1 and 2). Additional chaperone genes (Pdia2, Pdia3 and Pdia4) were upregulated. The levels of the Chop transcription factor were also significantly higher than at the 24 h time point. However, the most abundantly induced gene at 48 h of mutant protein expression was Trib3 (Trb3), a gene whose protein product has been implicated in apoptosis among other cellular processes 18 19 . At this time point other genes including protein transport genes and transcription factors were also upregulated. By 5 days of mutant protein expression the gene expression changes as monitored by microarray was more variable between experiments than at the 48 h time point. However, all the chaperone genes observed at earlier time points were still elevated and additional PDI family member genes (Pdia6 and Txndc4) were induced. The levels of Trib3 were lower compared to the 48 h time point, but still elevated. Interestingly, the levels of Sdf2l1, Creld2 and Armet remained highly induced throughout the time course of mutant protein expression.

Although some of the gene expression changes found were expected based on the ER stress response induced in other models (discussed later), some genes predicted to be increased were not observed, such as the ER chaperone genes Grp78 and Pdi. Microarray expression profiling can lead to false negatives due to inefficient probes. We therefore examined the mRNA and protein levels of several well established targets of the UPR by real-time PCR and western blotting. An increase in the mRNA levels of Grp78, Pdi, Chop and Sel1 was observed in response to doxycycline-induced mutant insulin expression (Figure 5A-D). The latter two changes substantiate the microarray results. However, both Grp78 and Pdi are significantly upregulated early (within 24-48 h) of mutant insulin expression. This was not detected by microarray, which does produce false negative results in this case. Consistent with the gene expression data, GRP78 and CHOP protein levels were also induced by Ins2 (C96Y)-EGFP expression (Figure 5E-F). Interestingly, thapsigargin treatment for 6 h induced higher expression of CHOP protein in clone #4S2 than in parental INS-1 cells, suggesting that clone #4S2 are more susceptible to ER stress (Figure 5F).

We performed additional real-time PCR analysis of other putative ER stress response genes to validate some of the microarray results (Figure 6). Consistent with the microarray data (Table 1) we observed an induction of Erdj4, but not the related gene Erdj5/Dnajc10, despite the fact that the latter has been shown to be induced by ER stress in other cell types 20 21 . In addition, although not detected by microarray, we observed induction of Ero1β, but not Ero1α (Figure 6C, D). The former, but not the latter, has been reported to be ER stress-inducible, which is consistent with our results 22 . It is important to note that doxycyline at the same concentration used in these studies added to a parental INS-1 cell line does not induce any changes in Grp78 or Chop mRNA or protein levels (results not shown), indicating that the changes observed are due to Ins2 (C96Y)-EGFP expression and not the small molecule.

Overall, the microarray and real-time PCR results show that Ins2 (C96Y)-EGFP expression causes the induction of many established ER stress response genes, although there is clearly some selectivity in the genes induced.

As shown in Table 1, genes associated with apoptosis such as Chop and Trib3 were already upregulated after 24 h of mutant insulin expression and the expression levels were even greater after 48 h. We thus examined whether doxycycline-induced expression of Ins2 (C96Y)-EGFP induces apoptosis. By light microscopy many of the cells appear rounded in comparison to the control cells at about 72 h of mutant protein expression, indicative of dead or dying cells (Figure 7A, arrows). To confirm that cell death was due to apoptosis we measured cytoplasmic DNA-associated histone complexes using a sensitive ELISA assay and found that 48 h to 7-day treatment of doxycycline causes apoptosis in a time-dependent manner (Figure 7B). Western blot analysis also revealed a detectable increase in cleaved caspase-3 by ~3 days of mutant protein expression (Figure 7C). To identify the relative percentage of apoptotic cells in the mutant insulin expressing population we monitored apoptosis by TUNEL assay using FACS analysis, which also allowed for the detection of mutant protein expression by the GFP fluorescence. Addition of doxycycline increased GFP fluorescence depicted on the y-axis, compared to control cells not treated with doxycycline (Figure 7D). The percentage of apoptotic (TUNEL positive) cells is low and comparable to control cells following 24 h doxycycline induction (Figure 7D, left panels). However, mutant protein expression following 5 days of doxycycline treatment leads to the induction of apoptosis in approximately 14% of the population (Figure 7D, right panel).

The extent of apoptosis observed in the population expressing Ins2 (C96Y)-EGFP even for extended periods (>5 days) seemed rather low. We hypothesized that the ER stress response in these cells, particularly the induction of ERAD genes, might be responsible for maintaining cell survival by degrading the misfolded insulin and preventing accumulation to toxic levels. We therefore examined whether inhibiting the ERAD system might sensitize the cells to cell death. As a first approach we inhibited proteasome function using lactacystin. Cells were induced with doxycycline for 48 h then treated or not with lactacystin and cell apoptosis was monitored. Inhibition of the proteasome resulted in a significantly higher percentage of apoptotic cells in the population (Figure 8A). This was also accompanied by an increase in the cellular levels of ubiquinated proteins (Figure 8B, left panel) and an increase in the levels of Ins2 (C96Y)-EGFP fusion protein and degradation fragments (Figure 8B, right panel). These results suggest that inhibiting ERAD increases misfolded protein levels and enhances ER stress-induced apoptosis.

As an alternative approach for inhibiting ERAD we targeted the Herp protein which was found to be induced early by mutant insulin expression (Table 1). Herp has been implicated in ERAD of certain misfolded proteins 23 , although whether it is required for misfolded insulin degradation has not been examined. We first confirmed that the Herp protein is induced in response to mutant insulin expression (Figure 9A). To perturb ERAD function, we blunted the induction of Herp expression using siRNA (Figure 9B, C). Reducing Herp expression by about 40-50% resulted in an increase in the steady-state levels of the mutant insulin protein, even in the absence of doxycycline induction (Figure 9B). This also increased apoptosis levels in the population relative to control siRNA transfected cells (Figure 9D). With doxycyline induction for 48 h the levels of Ins2 (C96Y)-EGFP in Herp siRNA-treated cells was increased relative to control siRNA transfected cells, but the effect was small (Figure 9B). In some experiments we also observed a reduction in the degradation fragments in Herp siRNA-treated cells (see Figure 9B, Expt.#2, low exposure), but this was not a consistent finding (Figure 9B, Expt.#1). The inconsistent results may be due to the inefficient knock-down of the Herp protein achieved in these studies. Regardless, the levels of apoptosis in the population depleted of Herp expression was significantly higher compared to control siRNA-transfected cells as measured by cleaved caspase 3 levels (Figure 9E). Altogether, these results suggest that the mutant insulin is an ERAD target and that Herp function is required for its degradation and that reducing Herp levels increases mutant insulin levels and sensitizes the cells to apoptosis.