Pancreatic INS-1 β-Cell Response to Thapsigargin and Rotenone: A Comparative Proteomics Analysis Uncovers Key Pathways of β-Cell Dysfunction

Abstract

Insulin-secreting β-cells in the pancreatic islets are exposed to various endogenous and exogenous stressing conditions, which may lead to β-cell dysfunction or apoptosis and ultimately to diabetes mellitus. However, the detailed molecular mechanisms underlying β-cell’s inability to survive under severe stresses remain to be explored. This study used two common chemical stressors, thapsigargin and rotenone, to induce endoplasmic reticulum(ER) and mitochondria stress in a rat insuloma INS-1 832/13 β-cell line, mimicking the conditions experienced by dysfunctional β-cells. Proteomic changes of cells upon treatment with stressors at IC₅₀ were profiled with TMT-based quantitative proteomics and further verified using label-free quantitative proteomics. The differentially expressed proteins under stress conditions were selected for in-depth bioinformatic analysis. Thapsigargin treatment specifically perturbed unfolded protein response (UPR) related pathways; in addition, 58 proteins not previously linked to the UPR related pathways were identified with consistent upregulation under stress induced by thapsigargin. Conversely, rotenone treatment resulted in significant proteome changes in key mitochondria regulatory pathways such as fatty acid β-oxidation, cellular respiration, citric acid cycle, and respiratory electron transport. Our data also demonstrated that both stressors increased reactive oxygen species production and depleted adenosine triphosphate synthesis, resulting in significant dysregulation of oxidative phosphorylation signaling pathways. These novel dysregulated proteins may suggest an alternative mechanism of action in β-cell dysfunction and provide potential targets for probing ER- and mitochondria stress-induced β-cell death.

Graphical Abstract

Introduction

Pancreatic β-cells regulate the homeostasis of blood glucose levels by synthesizing, processing, and secreting insulin into the circulation.¹ This condition is highly disrupted in both type 1 and type 2 diabetic patients.² Chiefly among the potentially important mechanisms that cause β-cell dysfunction or death are induction of the endoplasmic reticulum (ER)^2–6 or mitochondria stress.⁷

ER is responsible for protein folding, storage of intracellular Ca²⁺, transport of synthesized proteins, and degradation of misfolded proteins.³ This normal function of ER may be disturbed by a variety of factors including environmental toxins, viral infection, and inflammation, causing ER stress.³ The main cause of ER stress is an accumulation of misfolded or unfolded proteins in the ER lumen. For example, in the case of type 1 diabetes (T1D), increased insulin synthesis in pancreatic β-cells exceeds the folding capacity and causes an accumulation of unfolded insulin in the ER.^{3, 4} The accumulation of defected proteins in the ER leads to an activated unfolded protein response (UPR), a β-cell protective mechanism to restore the normal function of ER.⁸ In response to ER stress, the UPR first arrests protein translation, thereby slowing down protein synthesis, and then increases the expression level of molecular chaperones that assist protein folding or degradation of misfolded proteins.⁹ However, under severe ER stress, the UPR mechanism switches from pro-survival to pro-apoptotic, leading to β-cell dysfunction or death.^{3, 10} The UPR signaling pathway can be initiated by three sensory proteins: endoplasmic reticulum kinase (PERK), the inositol-requiring enzyme 1 (IRE1), and the activating transcription factor 6 (ATF6).^{3, 4} Studies have shown that under normal conditions, these ER-membrane associated proteins remain inactive by conjugating with the ER chaperon GRP78/Bip but are released during stress which triggers the UPR pathway and ER stress.³

A few studies reported the proteomic changes associated with ER stress in INS-1 cells.^11–13 D’Hertog et al. characterized the INS-1E cell proteome upon treatment with a combination of inflammatory cytokines (IL-1β + IFN-γ) and reported 158 proteins with altered expression levels involved in ER-associated pathways.¹¹ A subsequent study demonstrated that exposure of INS-1E cells to cyclopiazonic acid, another ER-stress inducer, resulted in alterations of key UPR proteins involved in β-cell dysfunction and apoptosis.¹² Despite this advancement, the underlying mechanism to the UPR regulatory networks and signaling events that cause β-cell dysfunction or apoptotic remains elusive.

On the other hand, mitochondrial dysfunction has been implicated as the main cause of several diseases including diabetes.^{7, 14} Mitochondria play a vital role in controlling cellular metabolism pathways such as the tricarboxylic acid (TCA) cycle, fatty acid metabolism, and amino acid metabolism. These pathways are involved in critical processes such as channeling nicotinamide adenine dinucleotide (NADH) and flavin adenine dinucleotide (FADH) into the electron transfer chain from the mitochondrial respiratory complexes (I - IV) which ultimately generate adenosine triphosphate (ATP).¹⁵ Studies also indicated that glucose metabolism drives β-cell glycolysis and pyruvate generation; where the latter enters mitochondria to further oxidize in the TCA cycle to produce NADH and FADH2 in ATP synthesis.⁷ Inhibition of these pathways has been reported to link with metabolic changes that lead to β-cell dysfunction and death. For example, inhibition of mitochondrial complex I by rotenone resulted in incomplete electron transfer within the complex chain that leads to ATP depletion and increased reactive oxygen species (ROS).^{15, 16} Previous studies have also demonstrated that mitochondrial inhibition in mouse pancreatic islets greatly affects insulin secretion and increases the future risk of type 2 diabetes.¹⁷. A proteomics study of complex I deficient HeLa cells treated with rotenone also highlighted proteins and metabolites highly enriched in the respiratory chain and TCA cycle.¹⁸

While existing as spatially separated organelles, ER and mitochondria function in various aspects of cellular metabolism through intimate interactions. The interplay between ER and mitochondria under stress conditions also indicated proteome changes at the ER-mitochondria contact sites, highlighting alternative pathways in scenarios where ER stress cannot be resolved by the UPR system.¹⁹ Although several studies contributed to increasing potential target and signaling pathways in β-cell dysfunction, the underlying mechanism of β-cell destruction or dysfunction in response to ER and mitochondria stress remains to be explored.

In the present study, we employed isobaric labeling-based quantitative proteomics to evaluate the proteome changes associated with ER and mitochondria stress in INS-1 832/13 β-cells, a widely used rat insulinoma cell line for pancreatic islet β-cell function studies. We first evaluated the biological effects to INS-1 cells by two classical chemical stressors, namely thapsigargin, a sarco/endoplasmic reticulum Ca²⁺ - ATPase inhibitor.²⁰ and rotenone, a strong inhibitor of complex I of the mitochondria respiratory chain.²¹ We next profiled the changes of the β-cell proteome induced by these two stressors. Our data not only provide a broad view of the effects of these stressors to differentially targeted cell compartments but also uncover key novel perturbed pathways specific to ER or mitochondria stress.

Experimental Procedures

Cell Culture and Treatment

The INS-1 rat pancreatic β-cell line (832/13; a gift from Dr. Christopher Newgard, Duke University, Durham, NC) was cultured in Roswell Park Memorial Institute (RPMI) 1640 medium (catalog number: 11875135) supplemented with 10% heat-inactivated fetal bovine serum (FBS) (catalog number: 10438026), 1 mM sodium pyruvate (catalog number: 11360070), 2 mM L-glutamine (catalog number: 25030081), 0.05 mM 2-mercaptoethanol (catalog number: 21985023), 100 U/mL pencillin-streptomycin (catalog number: 15140–122), all from Thermo Fisher Scientific, and 10 mM HEPES buffer (Sigma-Aldrich, catalog number: H0887). Cells were seeded in a 75 cm² tissue culture flask in five replicates and placed at 37 ⁰C under 5% CO₂ in a humidified incubator. Cells were grown until ~80% confluent before treatment with rotenone (Catalog number: 83-79-4, Tocris, Minneapolis, MN) or thapsigargin (catalog number: 67526-95-8, Tocris) at half maximal inhibition concentration, that is, concentrations resulting in 50% cell death (IC₅₀), which were predetermined using cell viability assays described below. All treatments were carried out for 24 h.

Cell Viability Assay

After treating with different concentrations of rotenone or thapsigargin, the cell viability was determined using the cell counting kit-8 (CCK-8) reagent in triplicate (Dojindo Molecular Technologies, Rockville, MD). Briefly, the cells were seeded in 96-well plates at a density of 2 × 10⁵ cell/mL and incubated overnight under culturing conditions. Cells were then treated with rotenone or thapsigargin at 0 – 1 μM concentration ranges and incubated for 24 h. After the indicated incubation time, 10 μL of CCK-8 reagent was added into each well and incubated for 4 h under culturing conditions. The absorbance of samples was measured at 450 nm using BioTek’s Synergy LX multimode microplate reader (BioTek Instruments, Inc.). The IC₅₀ values were calculated using the online tool (https://www.aatbio.com/tools/ic50-calculator) (accessed on 2021-06-11).

ATP Assay

The ATP assay was performed using a commercial ATP Colorimetric/Fluorometric assay kit (Catalog number: ab83355, Abcam, UK) according to the manufacturer’s instructions. In brief, cells were seeded in 75 cm² tissue culture flask at a density of 2 × 10⁵ cell/mL and incubated overnight under culturing conditions. Cells were then treated with rotenone or thapsigargin at IC₅₀ concentrations and incubated for 24 h. Untreated cells were included as control. Next, cells were harvested and homogenized in 100 μL of ATP assay buffer. Samples were then mixed with ATP reaction mix and incubated at room temperature for 30 min. ATP level of each well was measured using a microplate reader at OD 570 nm, and quantitation was performed against a standard calibration curve generated using known amounts of ATP. The data was normalized to the percentage of control before statistical analysis.

Determination of Mitochondrial ROS Generation

Intracellular determination of mitochondrial superoxide production was performed using MitoSOX Red (Molecular Probes, Eugene, OR) as previously described.²² MitoSOX Red is a fluorescent dye that permeates live cells to selectively target mitochondria superoxide which can be visualized using fluorescent microscopy. Briefly, the INS-1 cells grown on a 12-well plate were washed twice with Hank’s balanced salt solution (HBSS) (catalog number: 14065–056, Thermo Fisher Scientific) to remove the medium and incubated with 2 μM MitoSOX Red dye in the dark under culturing conditions. To confirm mitochondrial localization of MitoSOX Red, cells were incubated with 200 nM MitoTracker-Green (Molecular Probes, Eugene, OR) for 20 min following the removal of excess fluorescent dye with HBSS. Cells were then washed gently three times with warm PBS buffer and imaged (excitation/emission: 510/580 nm) immediately under a fluorescence microscope (TH4–100, Olympus). The fluorescence intensities from cells plated in 96-well plates were quantified using a BioTeck fluorescent plate reader and expressed as the mean ratio of MitoSOX to MitoTracker to compensate for differences in mitochondrial mass and unequal MitoSox loading.

Western Blot Analysis

Whole-cell lysates of treated and untreated INS-1 cells were lysed using 100 μL RIPA buffer (catalog number: R0278, Sigma-Aldrich) and centrifuged at 14,000 g for 15 min at 4 °C. The protein concentration of the supernatant was determined using a BCA protein assay kit (catalog number: 23252, Thermo Fisher Scientific). Equal amounts of the protein samples (~30 μg) were separated by 12% sodium dodecyl sulfate-polyacrylamide gel along with molecular weight standards and transferred to polyvinylidene difluoride membranes (Bio-Rad, Hercules, CA). Next, membranes were blocked with 5% milk and incubated at room temperature for 2 h. After blocking, the membrane was probed overnight at 4 °C with primary antibodies against 78 kDa glucose-regulated protein (GRP78, catalog number: MAB4846, R&D system), C/EBP-homologous protein (CHOP, catalog number: 5554, Cell Signaling Technology), β-actin (catalog number: NBP1–47423, Novus Biologicals). Membranes were subsequently washed and incubated with secondary antibodies: goat antimouse immunoglobulin G (catalog number: ab205719, Abcam) or goat antirabbit immunoglobulin G (catalog number: ab205718, Abcam). Blots were developed using enhanced chemiluminescence substrate kit (catalog number: 34076, Thermo Fisher Scientific). Actin was used as the loading control. Band’s intensity, reflecting relative protein expression, was determined with ImageJ software.

Protein Extraction and S-Trap Based Protein Digestion

Cells were lysed in lysis buffer containing 5% SDS, 50 mM triethylammonium bicarbonate (TEAB) pH 8.0 and sonicated for 30 min using a probe sonicator (Thermo Fisher Scientific, Model 100). Protein concentration was determined using a BCA assay kit (catalog number: 23252, Thermo Fisher Scientific). Protein digestion was performed using previously described S-trap micro columns.²³ Briefly, 50 μg of proteins from each sample were reduced with 20 mM dithiothreitol (catalog number: 43819, Sigma-Aldrich) for 10 min at 95 °C followed by alkylation with 40 mM iodoacetamide (catalog number: I1149, Sigma-Aldrich) incubated in the dark at room temperature for 30 min. Next, 12% aqueous phosphoric acid was added to each sample with 1:10 ratio. S-trap binding buffer (90% methanol, 100 mM TEAB, pH 7.1) was added into the sample at a 1:6 ratio. Sample contents were then transferred into S-trap column (PROTIFI, Farmingdale, NY) and cleaned three times using S-trap binding buffer. Next, 20 μL of digestion buffer (50 mM TEAB, pH 8.5) containing trypsin-LysC mix (catalog number: V5072, Promega) were added to the S-trap at 1:10 weight to weight ratio and incubated overnight for digestion. Peptides were finally eluted with 50 mM TEAB, 0.2% aqueous formic acid (FA), and 50% acetonitrile in 0.2% FA. Eluted peptides were dried, and peptide concentration was determined using Pierce fluorometric peptide assay kit (catalog number: 23290, Thermo Fisher Scientific) after reconstituted in 0.1% FA. Eluted peptide samples were aliquoted for TMT-labeled and label-free based analysis, respectively.

TMT 16-Plex Labeling and High pH-Reversed Phase StageTip Fractionation

Samples from each group were labeled with TMTpro 16plex kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. After labeling, samples were combined and fractionated by using high-pH reversed-phase stop-and-go extraction (Hp-RP StageTips). In brief, a C8 membrane was inserted into Gilson 200 μL pipet tips used as a frit. A portion of C18-AQ beads (5 μm) was packed properly into the StageTips followed by washing and conditioning by centrifugation at 1500 g for 2 min. Peptides (30 μg) were resuspended in 200 mM NH₄COOH (pH 10). Next, peptide samples were dissolved in 50 μL of 200 mM NH₄COOH (pH 10) and transferred into the StageTips followed by centrifugation at 15000 g for 2 min to elute the peptides in 6 fractions with increasing acetonitrile concentration. The fractionated samples were then dried using Speed vac and reconstituted in 0.1% FA. The resulting peptides were desalted and concentrated using EvoTips (EvoSep, Odense, Denmark).

Fractionation and Library Construction for Label Free Quantitative Proteomics

To build the label-free quantitation (LFQ) spectrum library, equal peptide amounts (0.4 μg) were pooled from each digested sample (total pooled peptide amount: 6 μg). The pooled peptides were fractionated using the same fractionation method (Hp-RP StageTips) as described above to collect six fractions. After drying, the fractionated peptide samples were reconstituted in 0.1% FA and subjected to Evotip loading.

LC-MS/MS Analysis

All peptide samples were separated on the Evosep One LC system (EvoSep) on a 15 cm × 150 μm i.d. capillary column (1.9 μm C18 particles) using the preprogrammed gradient of 15 samples per day method with gradient length of 88 min. The Evosep One system was coupled online to Orbitrap Exploris 240 mass spectrometer (Thermo Fisher Scientific) equipped with an easyspray source. For the TMTpro-labeled peptides, the instrument was performed in DDA mode with full MS scan settings: resolution 60k, mass range m/z 350 – 1600, RF lens: 70%, normalized AGC target: 300%, followed by the top 20 MS/MS scans with resolution 45k at m/z 110, standard AGC target, isolation window of 0.7 m/z at HCD collision energy of 31, and dynamic exclusion of 25 s.

For label-free quantification acquisition method, the spectrum libraries were built using the following MS setting: 120k resolution, mass range 375 – 1500 m/z, 25 ms injection time, and 300% of AGC target. The top 20 precursor ions in 30 s exclusion duration were selected with 1.5 m/z isolation window and 5E3 minimum intensity, and fragmented with HCD collision energy 30. In the case of running individual samples, the top 10 precursor ions with 1E5 minimum intensity were selected for fragmentation.

Database Search

For TMT data, the MS/MS raw files were processed with Proteome Discoverer (PD, version 2.5.0.400, Thermo Fisher Scientific). Briefly, the Sequest HT search engine was applied to search the raw data against a Uniprot Rattus norvegicus protein database (downloaded on July 2021 with 8131 reviewed entries) supplemented with commonly observed MS contaminants (containing 246 entries). Searches were configured with static modifications on lysine and N terminus (+304.207 Da) for the TMTpro reagents, carbamidomethyl on cysteines (+57.021 Da), dynamic modifications for oxidation of methionine residues (+15.995 Da), precursor mass tolerance of 20 ppm, and fragment mass tolerance of 0.5 Da. Trypsin was used as digestion enzyme with maximum of two missed cleavages. The minimum and maximum peptide lengths were set as 7 and 144, respectively. For high confidence results, protein identification was filtered to 1% false discovery rate (FDR) in peptide spectra match (PSM), peptide, and protein levels. The FDR was calculated using the Percolator algorithm embedded in PD.

For label-free data, the raw files were processed using the same software (PD, 2.5.0.400). In brief, two-stage Sequest HT search engine was applied and matched against the same protein database and MS contaminants as used in the TMT data processing. In this analysis, the Minora feature detector was used as a match between runs to increase identification. The search also performed intensity-based rescoring of PSMs using INFERY to enhance confident identifications. The modifications for label-free data were set as static modification as carbamidomethyl on cysteines (+57.021 Da), and variable modifications as oxidation of methionine residues (+15.995 Da), while precursor mass tolerance was set as 10 ppm and fragment mass tolerance of 0.02 Da. Other parameters were the same as described in the TMT data.

All the mass spectrometry raw files and database search results from this study were deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the data set identifier PXD030711.

Statistical Analysis

The exported protein abundance values were analyzed and visualized using Perseus software (version 1.6.14.0).²⁴ To ensure high confidence in statistical analysis, data were further filtered to include: (1) only proteins identified without any missing values in all the 15 biological samples; (2) quantified with more than 2 unique peptides; and (3) excluded potential contaminants. The quantitative protein data were log₂ transformed and further normalized using median centering. Principal component analysis (PCA), Pearson’s correlation coefficient, and boxplots were performed to evaluate the reproducibility of samples. One-way ANOVA (FDR < 0.05, by permutation-based FDR) was used to determine if treatment groups were significantly different from the control group. Two-tailed student’s t-test was also applied for additional comparisons between two conditions. Unsupervised hierarchical clustering analysis was performed to generate a visual heat map. GraphPad Prism (version 9.3.0, GraphPad software, San Diego, CA) was applied for statistical analysis of bioassay results with a p value < 0.05 considered as statistically significant.

Function Enrichment and Network Analysis

Enrichment of functions and signaling pathways of differentially expressed proteins (DEPs) perturbed by thapsigargin and rotenone was performed using Metascape (http://metascape.org),²⁵ a powerful tool integrating several functional databases such as Gene Ontology, KEEG, Reactome, etc., to explore the cognition of protein functions. In brief, the upregulated or downregulated proteins of each treatment group were submitted to Metascape software (version 3.5.20211101). Then, the enriched pathways and biological processes were identified based on statistically significant p value < 0.01, with a minimum count of three and an enrichment factor > 1.5. The top enriched terms were displayed as a heatmap (bar color intensity) and further extrapolated in a table to show the number of proteins associated with the enriched biological terms in each cluster. The protein-protein interaction network was analyzed using the molecular complex detection (MCODE) module, which further clusters the initial interaction map into subgroup modules. Pathview maps (https://pathview.uncc.edu/) were also applied for additional data integration and visualization of altered proteins.

Results

Rat pancreatic INS-1 β-cell line demonstrates glucose-stimulated insulin secretion and is considered as a good pancreatic islet cell-model in various functional studies.²⁶ The aim of our present study was to investigate the changes at the proteome level in the INS-1 cell line upon exposure to commonly known stress inducers. To this end, INS-1 832/13 cells were cultured and treated with rotenone, a selective mitochondrion complex I inhibitor,²¹ and thapsigargin, an ER stress inducer,²⁰ and the resultant biological and proteomic changes were profiled as discussed in the succeeding sections.

Effect of Rotenone and Thapsigargin on INS-1 Cell Viability

Optimal cell culture conditions were first established before the proteomic studies. INS-1 832/13 cells were exposed to different concentrations of rotenone or thapsigargin, and cell viability was evaluated using the CCK-8 assay after 24 h treatment. As expected, a concentration-dependent decline in cell viability occurred during the 24 h period in both treatment conditions (Figure 1A). The IC₅₀ value of rotenone and thapsigargin were determined as 30 nM and 22 nM, respectively, and were used as final treatment conditions of INS-1 832/13 cells in this study. The IC₅₀ values of rotenone agreed with previously reported results of the same cell type,²⁷ while that of thapsigargin was reported as ~ 4 μM in the SW-13 and NCI-H295R cell lines.²⁸ Clearly the IC₅₀ of thapsigargin is cell line dependent, which reflects each cell line’s potential vulnerability to stress.²⁷

Figure 1.

Biological effect of thapsigargin (Tpg) and rotenone (Rot) treatment on INS-1 cells. (A) Dose-response analysis of Rot and Tpg to the viability of INS-1 832/13 cells (n = 3); (B) Representative cropped western blot (left) and quantification analysis (right) results showing increased expression levels of GRP78/Bip and CHOP as a response to Rot and Tpg treatment. Quantification of immunoblotting data was performed after normalized for β-actin levels in each sample. ***P < 0.005 compared to controls.

Biological Effects of Rotenone and Thapsigargin on INS-1 Cells

Rotenone and thapsigargin cause cell death with distinct classical biological pathways perturbed in their mode of actions. Rotenone acts as a strong and selective inhibitor of complex I of the mitochondria respiratory chain by inhibiting electron transfer and blockage of oxidative phosphorylation, leading to reduction of ATP synthesis and accumulation of mitochondria reactive oxygen species (ROS);²¹ while thapsigargin, a widely used toxin in studies of ER stress-induced β-cell death, inhibits SERCA-ATPase pathway causing the depletion of ER Ca²⁺ stores.²⁰ Our bioassay results indicate that both rotenone (30 nM and 24 h) and thapsigargin (22 nM and 24 h) stimulated ROS production in INS-1 cells, as quantified using MitoSOX red fluorescence assay (Figure S1A, B). We also measured the level of ATP production to assess the productivity of the electron transfer chain and found significant inhibition compared to control in both treatment conditions (Figure S1C), suggesting both toxins can induce mitochondria dysfunction. Our Western blot results further confirmed that both stressors increased expression levels of GRP78 and CHOP (Figure 1B), two protein markers characteristic of ER and mitochrondria related stresses, which is in agreement with previous reports.^{13, 29}

Comprehensive Analysis of INS-1 Cell Proteomic Change upon Exposure to Rotenone and Thapsigargin

Based on the above treatment effect results, we applied a peptide level isobaric labeling-based quantitative proteomics strategy to accurately profile the proteomic changes in INS-1 cells under three different culture conditions: control (Cont), rotenone (Rot), and thapsigargin (Tpg), each with five biological replicates (Figure 2). In total, we identified 2453 protein groups at 1% FDR using a reviewed rat database (containing 8131 entries) after strictly filtering the processed data matrix as described in the Experimental Procedures section (Supplementary Table S1). Among these protein groups, 2286 (~93%) were quantified without missing values in any of the 15 samples; of which 1766 were quantifiable with ≥ 2 unique peptides and used for further analysis.

Figure 2.

Experimental workflow for quantitative proteomic analysis of INS-1 cells treated with Tpg or Rot (biological replicates n = 5). After treatment, protein digestion was performed using S-trap micro columns, and samples from each group were labeled with TMTpro 16plex reagents. The high-pH reversed-phase StageTips were used to fractionate the pooled TMT-labeled peptide mixtures before Evotip loading and LC-MS/MS analysis.

We next performed PCA of the quantifiable proteins. The results showed a clear separation of samples in each treatment condition (Figure 3A, left), which indicates these two chemicals resulted in distinct proteomic changes in INS-1 cells. In addition, as visualized in the loadings plot, the proteins driving the separation of the clusters in the PCA plot are clearly aligned with the well-known toxicological effects of the chemicals.³⁰ For example, the ER stress related proteins ER chaperone Hspa5, chromogranin (Chgb), and low-density lipoprotein receptor (Ldlr) contributed the most to the thapsigargin treatment (Figure 3A, right), and mitochondrial stress enriched proteins such as 2-oxoglutarate dehydrogenase (Ogdh), an important mitochondria redox sensor, were also among the main contributors to the separation of rotenone and thapsigargin treatment.

Fig. 3.

Effects of Rot and Tpg treatment to INS-1 cell proteome. (A) Left, scores plot showing the clustering of samples and separation of treatment conditions; right, loadings plot indicating the major proteins contributed to the separation of groups in the scores plot. (B) Workflow for identifying DEPs, clusters, and network analysis. (C) Unsupervised hierarchical clustering of the 835 DEPs identified in three treatment conditions by one way ANOVA analysis. The log₂ normalized abundances of each protein were Zscore transformed across the treatment conditions. Profiles of each cluster and corresponding functional enrichments are shown on the right.

To explore the biological functions and pathways of proteins associated with the stressed cell systems, we performed a stepwise bioinformatics analysis (Figure 3B). We first performed one-way ANOVA analysis (Permutation-based FDR < 0.05) comparing the abundance of the three distinct culturing conditions (Cont, Rot, and Tpg), which resulted in the identification of 835 differentially expressed proteins (DEPs) (Supplementary Table S2). Unsupervised hierarchical clustering analysis of these DEPs further revealed four major clusters of proteins predominantly determined by treatment conditions, which was supported by high reproducibility of biological replicates in the clustering analysis (Figure 3C, left): (i) upregulated proteins in Tpg-treated group (279 proteins), (ii) downregulated proteins in Tpg-treated group (231 proteins), (iii) upregulated proteins in the Rot-treated group (162 proteins), and (iv) downregulated proteins in the Rot-treated group (163 proteins) (Figure 3C, right and Supplementary Table S3–6).

To gain more functional insights of these clusters, we first evaluated the specific groups of proteins by gene ontology (GO) and annotated pathways using the Metascape visualization tool (integrated functional annotations). This identified biological functions and pathways that were dysregulated given the observed protein expression changes in the data set. The top biological processes or pathways that were elevated in thapsigargin treatment included protein processing in the ER, Golgi vesicle transport, carbon metabolism, and protein exit from ER (Figure 3C, right, Figure S2A). Other interesting pathways elevated in the Tpg-treated group included the ERAD pathway, pyruvate metabolism, pentose phosphate pathway, antigen presentation (folding, assembly and peptide loading of class I MHC), NADH metabolic process, tRNA aminoacylation for protein translation, p97-Ufd1-Npl4 complex, and post-translational protein phosphorylation (Supplementary Table S7). In addition, we also identified significantly increased proteins in the Tpg-treated group involved in lipid metabolism, including the lipid biosynthetic process, cholesterol biosynthetic process, and steroid biosynthetic process. While upregulation of proteins involved in protein processing and ER associated protein degradation is expected as a response to stress-induced accumulation of unfolded proteins, increased expression of lipid synthesis proteins maybe due to more phospholipids and cholesterols are needed in expansion of the ER membrane, as ER expansion is a well-known phenomena in the ER stress response. In contrast, the top biological processes that were inhibited in thapsigargin treatment included mRNA processing, actin cytoskeleton organization, SUMO E3 ligases SUMOylate target proteins, and signaling by Rho GTPases (Figure 3, right, Figure S2B, Supplementary Table S8). In this respect, downregulation of proteins in global mRNA processing and Sumo-mediated transcription regulation is a natural response to excessive amounts of unfolded proteins in the ER, for reducing initial synthesis of protein precursors and therefore less workload for ER-based protein folding.

On the other hand, rotenone treatment highlighted the top enriched biological functions and pathways related to mitochondrial dysfunction. These included elevated proteins involved in citric acid (TCA) cycle and respiratory electron transport, carboxylic acid catabolic process, cellular amino acid metabolic process, and mitochondria transport proteins (Figure 3C, right, Figure S2C, Supplementary Table S9). A Majority of these proteins are enzymes that predominantly reside in mitochondria and control TCA cycle, electron transport, and fatty acid metabolism. Upregulation of these enzymes indicates bypassing of complex I for respiration and energy production need is an adaptation mechanism to rotenone resulted inhabitation of respiratory complex I. On the other hand, L13a-mediated translational silencing of ceruloplasmin expression, translation initiation complex formation, intracellular signaling by second messengers, and TNF-alpha NF-kB signaling pathway were reduced under the same treatment (Figure 3C, right, Figure S2D). Dominance of ribosomal, RNA binding and translation related proteins in downregulation in general indicates slowed cell division, protein biosynthesis, and DNA replication. With rotenone treatment, we also identified other interesting pathways and biological processes including upregulation of proteins involved in the regulation of mitochondrial membrane potential, mitochondrial biogenesis, and apoptotic mitochondrial changes, neutrophil degradation, and Cpt1a-ACSl1-Vdac1 complex, while regulation of the pancreatic β-cell apoptosis process, protein folding, chromatin modifying enzymes, and signaling by RhoGTPases, positive regulation of DNA biosynthetic process, clathrin-mediated endocytosis proteins were reduced under the same treatment (Supplementary Table S10). Detailed characterizations of the specific perturbation to biological processes and pathways by these two stressors are illustrated in the succeeding sections.

Thapsigargin Perturbed Key ER-Related Pathways in INS-1 Cells

We used the molecular complex detection (MCODE) module in Metascape to analyze the protein-protein interaction (PPI) networks. The 88 upregulated proteins identified in the Tpg-treated group were highlighted in 11 interaction modules (Figure 4A, Supplementary Table S11a). One of the well-annotated modules was protein processing in the ER pathway (Figure 4A (ii)), which was among the top upregulated pathways, consisting of several UPR target proteins known to overexpress during ER stress such as Hspa5, Hsp90b1, and various isomers of protein disulfide isomerase (eg., Pdia4, Pdia3). In this pathway, 25 proteins were identified having consistently elevated expression levels in the Tpg-treated group compared to control (Figure S3A); 11 of them (Hspa5, Hyou1, Hsp90b1, Calr, Ppib, Ufd1, Pdia4, Vcp, Plaa, Nploc4, Pdia3) were enriched and shown in the PPI functional module with higher interaction scores (Figure 4A (ii)). Most of these 11 proteins have been reported as apoptosis contributors in different cell types. We also found that ER-associated protein degradation (ERAD) related proteins, such as Vcp (transitional ER ATPase), had increased expression levels in Tpg group indicating the persistent ER stress leads to cell death in ERAD’s ability to remove unwanted cellular proteins (misfolded or mutated proteins in ER).³¹ Through Pathview analysis, we further demonstrated that the 25 aforementioned upregulated proteins could trigger associated signaling processes such as protein recognition by luminal chaperone, ERAD, and ubiquitin ligase complexes (Figure S3B).

Fig. 4.

Network enrichment analysis of proteins significantly dysregulated in Tpg-treated INS-1 cells. PPI networks of upregulated proteins (A) and downregulated proteins (B). List of upregulated or downregulated proteins were analyzed using Metascape software, and enriched pathways were indicated in the interaction modules.

Earlier reports demonstrated that pretreatment of INS-1 β-cells with a variety of stress inducers such as palmitate leads to elevated expression levels of proteins associated with ER stress.³² Likewise, recent studies have also shown that severe ER stress is directly involved in the pathogenesis of T1D, where excessive misfolded or unfolded proteins accumulated in the ER lumen, and the UPR pathway promotes pancreatic β-cells death due to this severe stress.³³ Itzhak et al.³⁰ also demonstrated the enrichment of the same pathway (protein processing in ER) using different ER stressors. In that study, tunicamycin was used as one of the ER stress inducers in HeLa cells, and 74 UPR target (58 conventional and 16 novel) proteins were reported.³⁰ Our data indicated 26 proteins were among these reported UPR target proteins with consistent upregulated expression levels in thapsigargin-treated INS-1 cells, of which 15 (Calr, Ppib, Pdia4, Hyou1, Hsp90b1, Hspa5, Pdia3, Sec61a1, Sec31a, Lman1, Uggt1, Pdia6, Dnajb11, Canx, Ccdc47) were involved in protein processing in the ER pathway. Our PPI network analysis further indicated that 10 of these 15 proteins were highly interconnected (≥ 3 proteins) with these conventional UPR targets in the functional interaction module (Figure 4A (ii)).

Insulin-like growth factor (IGF) and its binding proteins (IGFBPs) are closely related to insulin and thus have been explored as biomarkers for diabetes millitus.³⁴ The expression levels of IGFBP1 and IGFBP2 were reported to be significantly increased in T1D.³⁵ In the present study, regulation of IGF transport and uptake by IGFBPs pathway was also among the most enriched annotations in the Tpg-treated group with seven significantly increased proteins involved (Chgb, Scg2, P4hb, Vgf, App, Scg3, and Tgoln2) (Figure 4A (vi)). Of note, most of these proteins belong to the granin family including chromogranin (Chgb or Scg1), secretogranin 2 (Scg2), and secretogranin 3 (Scg3), which recently were demonstrated as key contributors to diabetes mellitus.³⁶ A study reported secretion of Scg3 from dysfunctional β-cells and indicated the expression level as upregulated in T1D.³⁷ In addition to these canonical pathways, we also identified novel pathways which were significantly upregulated in Tpg-treated group, including COPI-mediated anterograde transport (Figure 4A (iv)), COPII-mediated vesicle transport (Figure 4A (viii)), cargo recognition for clathrin-mediated endocytosis (Figure 4A (x)), class I MHC mediated antigen processing and presentation (Figure 4A (vii)), butanoate metabolism (Figure 4A (iii)), translation (Figure 4A (i)), glycolysis (Figure 4A (v)), glutathione metabolism (Figure 4A (xi)), and ribosomal small subunit biogenesis pathways (Figure 4A (ix)) (details in Supplementary Table S11a).

On the other hand, proteins significantly downregulated in the Tpg-treated group were highlighted in six functional interaction modules including mRNA splicing-major pathway, oxidative phosphorylation, SUMOylation of DNA replication proteins, signaling by Rho GTPases actin filament, depolymerization, actin cytoskeleton organization, and innate immune systems (Figure 4B, supplementary Table S11b). Among these downregulated pathways, studies have shown that SUMOylation and ubiquitination are involved in the pathogenesis of diabetic nephropathy,³⁸ and also linked to several critical pathways associated with T1D such as nuclear factor kB (NF-kB), transforming growth factor-B (TGF-B), and Nrf2-oxidative stress signaling pathways.³⁹ Despite the involvement of SUMOylation and ubiquitination in diabetic mellitus, the exact mechanism remains to be elucidated. For example, Davey et al.⁴⁰ observed that treatment of INS-1E cells with palmitate significantly attenuates GSIS and increases the protein expression level of SUMOylation (Sumo1). In contrast to previous report, our proteomic data indicated that five proteins (Sumo1, Ube2i, Nup210, Nup62, Xpo1) in SUMOylation pathway had consistently reduced expression levels in the Tpg-treated group (Figure 4B (iii)). This shows that palmitate and Tpg have different action mechanisms in toxicity, and Tpg decreases nucleocytoplasmic transport activities for reduced initial protein synthesis to cope with the UPR stress in the ER.

It was shown that activated inositol-requiring ER-to-nucleus signaling kinase 1 (IRE1), a type I transmembrane protein and the most conserved ER stress sensor of the UPR signaling axes, triggers cleavage of mRNAs from X-box binding protein 1 (XBP1) to reduce the misfolded protein load on the ER.^{4, 41} During ER stress, the RNase domain of IRE1 also undergoes degradation of a subset of ER-localized mRNAs, a process called regulated IRE1-dependent decay of mRNA (RIDD).⁴² In line with this, we identified 15 proteins (Cpsf7, Ctnnbl1, Cwc15, Hnrnpa2b1, Hnrnpk, Hnrnpm, Prpf19, Rbmx, Wbp11, Srsf9, Elavl1, Hnrnpu, Snrpb, Smndc1, and Elavl1) with lower expression levels in the Tpg-treated group that were involved in mRNA splicing-major pathway (Figure 4B (i)). Studies also reported the mRNA splicing processes linked to diabetes mellitus. For example, reduced expression level of Cpsf7, cleavage and polyadenylation specific factor 7, was reported previously in T1D patients.⁴³ Dysregulation of many RNA-binding proteins such as Hnrnpk, Hnrnpm, and Elavl1 has also been reported previously in human islets in response to a variety of stressors including chronic hyperglycemia, proinflammatory cytokines, and palmitate,⁴⁴ indicating the importance of the mRNA splicing-major pathway in regulating β-cell dysfunction.

Rotenone Significantly Disrupted Mitochondrial Proteome and Associated Pathways in INS-1 Cells

Various studies have shown that mitochondria dysfunction is associated with a wide spectrum of human diseases including diabetes.⁴⁵ Excessive production of ROS is deemed as the main cause of cellular damage where the existing antioxidant of the system cannot effectively neutralize it. As a result, several proteins and mitochondrial DNA are affected, causing failure of enzymatic chains that can impair mitochondria function, which in turn leads to abnormal cell signaling and apoptosis.⁴⁶ In particular, inhibition of complex I, an enzyme in the mitochondrial respiratory chain by rotenone, has been reported to induce apoptosis in a variety of cell models.¹⁶

In the present study, we identified 40 upregulated proteins potentially involved in 8 interaction modules using PPI network analysis (Figure 5A, Supplementary Table S12b). Among the most annotated modules is the citric acid (TCA) cycle and respiratory electron transport (Sod2, Nop58, Aco2, Gfm1, Acaa2, Ppif, Ndufs2, Uqcrfs1, Acad9, Hibch, Ndufs1, and Ndufs4) (Figure 5A (i)). This is an essential pathway encompassing a series of protein complexes and electron transfer chain proteins involved in the oxidative phosphorylation system for ATP synthesis in the mitochondria.¹⁴ In contrast to our observations, the expression level of selected mitochondrial proteins (Ndufs1, Ndufs2, and Ndufs4) was reported as significantly downregulated at both mRNA and protein levels in islets isolated from 2-week diabetic βV59 M mice and INS-1 cells treated with high glucose.⁷ This indicates that rotenone and hyperglycemia have distinct mechanism of toxicity, in which rotenone’s inhibition of respiratory complex I stimulates expression of other respiratory and oxidative phosphorylation related proteins to meet the cellular energy need in ATP production, while hyperglycemia mainly bypasses the mitochondria in production of ATP through upregulated glycolysis.

Fig. 5.

Network enrichment analysis of significantly dysregulated proteins in Rot treated INS-1 cells. PPI networks of upregulated proteins (A) and downregulated proteins (B). List of upregulated or downregulated proteins were analyzed using Metascape software, and enriched pathways are indicated in the interaction modules.

The fatty acid β-oxidation inside mitochondria produces acetyl-CoA which ultimately generates ATP.^{14, 47} We identified mitochondrial fatty acid β-oxidation of saturated fatty acids as a significantly enriched pathway in Rot-treated group, with increased expression levels of four associated proteins (Acadl. Acadvl, Hadha, Etfa) (Figure 5A (iv)). In addition to these known pathways we also found upregulated novel pathways such as the innate immune system (Ttr, Ggh, Grn, Dpp7, Aga, Hexb, and Tollip) (Figure 5A (ii)), 2-oxoglutarate metabolic process (Got1, Idh1, Got2, Mdh2, Idh2) (Figure 5A (iii)), cellular respiration (Trap1, Atp5f1a, Atp5f1b) (Figure 5A (v)), mitochondrion organization (Phb, Vdac1, Vdac2) (Figure 5A (vi)), and metabolic process (Bckdk, Suclg1, Dld) (Figure 5A (viii)). Our network analysis of downregulated proteins in Rot-treated group also highlighted several interaction modules including: SRP-dependent cotranslational protein targeting to membrane (Figure 5B (i)), PTEN Regulation (Figure 5B (ii)), regulation of the cellular catabolic process (Figure 5B (iv)), formation of cytoplasmic translation initiation complex (Figure 5B (v)), synaptic vesicle endocytosis (Figure 5B (vi)), and protein folding (Figure 5B (vii)).

Label-Free Proteomic Quantification Validated Findings in TMT-Based Proteomics Study

As many of the altered proteins in our TMT data are novel, we performed additional validation analysis with a LFQ method using samples collected under the same conditions. By using a spectrum library generated from fractionated pooled samples, we identified 2860 proteins across all 15 individual samples (FDR < 1%, ≥ 2 unique peptides) (Supplementary Table S13) and 2249 proteins were shared between the TMT and label-free data sets (Figure S4A). Of note, applying the peptide spectra library increased the overall protein idneitifications in LFQ compared to TMT data. Of the 835 DEPs identified in our TMT data, 158 were redetected in the LFQ data set. The Pearson’s correlation between LFQ and TMT data of commonly quantified proteins (n = 1727) were high (R ~ 0.8, Figure S4B–D), while these commonly identified DEPs (n = 158) were even higher, R > 0.9 (Figure 6A, B). Unsupervised clustering analysis of these commonly identified DEPs indicated that these proteins were identified with consistent expression patterns in both data sets (Figure 6C). Importantly, the target proteins selected from our TMT data using PPI network analysis are among these 158 proteins with the same expression trends. In brief, as shown in Figure 6C, the clustering analysis of these 158 commonly detected DEPs clearly divided the TMT and LFQ data into specific subgroups in each cultured condition. We were able to extract three major subclusters in both data sets, namely: (1) 39 proteins with consistent upregulation in thapsigargin-treated group, (2) 20 proteins with consistent upregulation in rotenone-treated group, and (3) 67 proteins downregulated in rotenone-treated group compared to control. Of note, seven (Hsp90b1, Hspa5, Pdia3, Hyou1, Calr, Pdia4, and Ppib) and eight proteins (Mdh2, Sdha, Dld, Idh2, Got1, Hadha, Sod2, and Oxct1) identified from protein processing in ER and TCA cycle pathways, respectively, were among those with consistent upregulated expression trends in both data sets. These confirmed proteins may serve as potential targets for further functional study.

Figure 6.

Label-free quantitative proteomics validation of TMT-based proteomics results. The Pearson’s correlation between LFQ and TMT data of commonly identified DEPs in (A) Tpg-treated group (B) and Rot-treated group. Log₂ fold changes (t-test, FDR < 0.05) from TMT and LFQ data set were used. (C) Heatmap showing consistent dysregulation trends between the LFQ and TMT data (ANOVA, FDR < 0.05). Cluster 1: 39 upregulated proteins in Tpg treatment; cluster 2: 20 proteins upregulated in Rot treatment; and cluster 3: 67 proteins downregulated in Rot treatment.

Discussion

In this work, we generated a comprehensive quantitative proteomics data resource for stressed β-cells by comparing the effects of thapsigargin and rotenone on INS-1 β-cells. While thapsigargin and rotenone treated cell lines have been extensively used as in vitro models of diabetics, optimizing the doses of each drug for the expected treatment-induced biological effects is needed before any proteomics analysis. In our initial experiment, we first determined the optimal concentration of these two stressors to mimick the conditions experienced by dysfunctional β-cells. Following these treatments, our proteomic results highlighted that thapsigargin treatment (IC₅₀ value: 22 nM) specifically perturbed UPR related pathways such as activation of protein processing in the ER pathway (Figure 4A (ii)). These findings are supported by previous proteomic studies that reported the dysregulation of ER associated proteins upon exposure to a variety of ER stressors.^{13, 30} Kim et al. analyzed proteomic data of thapsigargin-treated INS-1 rat insulinoma cells and reported 20 altered cellular proteins involved in metabolic process and protein folding,¹³ of which 10 proteins (Calr, Gpd1, Acly, Actr1a, Hspa8, Eno1, Eif3i, Pkm, Rplp0, and Gnb1) were commonly identified with consistently increased expression levels in our thapsigargen-treated group (Supplementary Table S3). Most of these identified proteins have been reported to be involved in T1D. For example, calreticulin (Calr), the ER and calcium regulatory protein, has been shown as critical for TGF-β stimulation of T1D.⁴⁸ Consistent with our study, Gd1, the glycerol phosphate shuttle enzyme, was reported as significantly upregulated at both mRNA and protein levels in islets isolated from 2-week diabetic mice.⁷ A recent study by Itzhak et al. also demonstrated the activation of UPR pathway upon exposure to thapsigargin or tunicamycin treatment on HeLa cell lines.³⁰ By comparison of our current data with previously reported proteomic studies on ER stress, we were able to identify 58 additional novel UPR targets in thapsigargin-treated INS-1 cells (Supplementary Table S14).

Of particular note, most of these identified novel UPR proteins are related to chaperons, which are essential in the proper folding of secretory and membrane proteins, and significant expression changes of these proteins are likely contributing factors in the pathogenesis of diabetes.⁴⁹ As an example, glucose-regulated protein (GRP94 or Hsp90b1) is one of the important players in the interaction module of protein processing in the ER pathway in our data (Figure 4A (ii)). Hsp90b1 is involved in regulating ER quality control by interacting with other ER chaperons and assisting in the clearance of misfolded proteins for ER-associated degradation.⁵⁰ Hsp90b1 also networks with other conventional UPR targets (Bip, PDI, CHOP, etc.) to alleviate ER-stress.⁴⁹ Together, these results suggest the identified novel proteins may contribute to the regulation of UPR signaling pathways in protecting the pancreatic β-cell development in response to ER stress.

The pathway analysis also revealed that glycolysis was among the most significantly upregulated pathways in our thapsigargin-treated group, with a marked increase in most glycolytic enzymes (Figure 4A (v)). Consistent with our observations, upregulation of selected glycolytic enzymes (Gpi, Pfkl, Eno1) has also been reported in diabetic islets (βV59M, in vivo) and in INS-1 β-cell lines cultured at high glucose.⁷ In the present study, lone protease homologue 2 (Lonp2) was the most upregulated protein followed by alpha-enolase (Eno1) in INS-1 β-cells exposed to thapsigargin treatment (Figure S5A). Another novel aspect of our current study is the highlight of additional UPR-related pathways such as the protein complex I (COPI) and protein complex II (COPII) pathways, which mediate multiple pathways between the ER and the Golgi.⁵¹ In the present study, thapsigargin treatment specifically enriched proteins involved in these pathways. As selected proteins (Arf4, Tmed9, and Arfgap3, Uso1, and Actr1a) were previously reported as upregulated UPR targets in a variety of cell types,^{13, 30} this implicates these identified proteins may be involved in regulation of UPR signaling pathway in response to ER stress.

In contrast to the toxicity effect of thapsigargin, rotenone (IC₅₀ value: 30 nM) demonstrated the strongest effect on the induction of several dysregulated proteins mainly involved in mitochondria dysfunction pathways (Figure 5). It is well-known that mitochondrial dysfunction has been attributed as the major cause of several diseases including diabetes.⁵² Marked changes in the expression levels of proteins involved in critical metabolic pathways in the mitochondria causes β-cell death.¹⁴ In line with our data on ATP synthesis measurement and ROS production, rotenone treatment reveals significantly increased expression levels of proteins involved in the TCA cycle and respiratory electron transport pathway (Sod2, Nop58, Aco2, Gfm1, Acaa2, Ppif, Ndufs2, Uqcrfs1, Acad9, Hibch, Ndufs1, and Ndufs4), which could be a mechanistic feedback in response to rotenone-induced stress ultimately leading to reduced mitochondria energy supply.¹⁵ In our study, aconitate hydratase (Aco2) was the most upregulated protein in INS-1 β-cells exposed to rotenone treatment (Figure S5B). Conversely, the protein expression level of Aco2 has been reported as significant downregulation in diabetic islets (βV59M, in vivo) and in INS-1 β-cell line cultured at high glucose.⁷ We have also identified significant downregulation of many proteins in response to rotenone treatment that were involved in several key pathways. One of the most important signaling cascades is mitogen-activated protein kinases (MAPKs) that play a critical role in intracellular signal transduction in response to environmental stimuli including inflammation, oxidative stress, and apoptotic processes.⁵³ In our study, we found significant downregulated proteins (Psmc2, Hspa14, Cdk1, and Psmd2) involved in MAPK6/MAPK4 signaling pathways (Figure 5B (iii)) upon exposure to rotenone treatment. The inhibition of this pathway may be a result from increased ROS production.

PTEN (phosphatase and tensin homologue deleted on chromosome 10) is a critical regulator of glucose and fatty acids metabolism.⁵⁴ Previous studies showed that changes in the expression of PTEN are involved in the regulation of muscle protein degradation in diabetes.⁵⁴ Recent reports suggest that the expression and activity of PTEN varies in different forms of diabetes. For example, reduced PTEN signaling was found in acute T1D, whereas in chronic models of db/db mice with insulin resistance upregulated expression was reported.⁵⁴ This variability was supported mainly by either post-translational modifications⁵⁵ of PTEN or insulin deficiency leading to PTEN signaling instability and degradation.⁵⁶ Although great improvements of PTEN regulation to cell survival have been thoroughly investigated, the underlying mechanism for PTEN regulation to metabolism especially in mitochondrial metabolism in response to variety of stressors remains unclear. In our proteomic analysis, we found reduced expression of several proteins (Psma2, Psmc3, Rbbp7, Csnk1a1, Phf5a, Dnajc10, Hdac1, Usp10, Tomm34, Morf4l2, Srsf6, and Ybx1) involved in the PTEN regulation pathway in rotenone treatment (Figure 5B (ii)), reflecting the acute conditions leading to β-cell death.

Our study also provided novel insight into the effect of significant dysregulation of oxidative phosphorylation (OXPHOS) signaling pathway induced by both thapsigargin and rotenone. It has been shown that changes of redox homeostasis in the ER generates ROS which in turn links to the UPR pathway.⁵⁷ Studies also demonstrated the tight regulatory link between oxidative protein folding in ER and mitochondrial OXPHOS.⁶ However, the mechanism of how these oxidative stresses interact in many pathologic states is still unclear. In our study, we found significantly increased ROS production and depleted ATP synthesis in both thapsigargin and rotenone treatments. Our proteomic analysis highlighted decreased expression levels of nine proteins (Cox2, Cox4i1, Cox6c, Cox5b, Uqcrc2, and Uqcrc1, Atp6v0a1, Atp5f1e, and Sdhb) in thapsigargen treatment, while increased expression levels of seven proteins (Atp5f1a, Sdha, Atp5f1b, Ndufs2, Uqcrfs1, Ndufs1, and Ndufs4) in rotenone treatment involved in OXPHOS signaling pathways. These results may reflect that both stressors induce β-cell dysfunction in distinct mechanism of action.

In conclusion, we presented a comprehensive resource for study of the protein expression changes in INS-1 pancreatic β-cells induced by thapsigargin and rotenone stressors. By mining this new data set, we showed that specific and novel features of UPR related pathways were activated by thapsigargin, mitochondrial respiratory electron transport chains were perturbed by rotenone, and oxidative phosphorylation pathways were affected by both stressors. As INS-1 rat insulinoma β-cells are extensively used as a model in the mechanistic investigation of β-cell dysfunction, these highlighted features would facilitate an in-depth understanding of the mechanisms of ER and mitochondrial stresses-induced cell death and may provide further insights for the underlying molecular mechanism of pathophysiology of diabetes and diabetic complications. However, β-cells in vivo are exposed to additional stressors during the initiation and progression stages of diabetes. For example, a recent study demonstrated significant proteome changes in β-cells upon exposure to cytokines.⁵⁸ While it is warranted to do additional validation and functional studies of these dysregulated novel pathways and associated proteins, a better understanding of the mechanisms underlying β-cell death in vivo would benefit from a thorough investigation of the synergism between these chemical stressors and the inflammatory cytokines produced under stressing conditions.

Supplementary Material

Supplemental figures

Figure S1: Biological effect of thapsigargin and rotenone treatment. Figure S2: Functional enrichment results of upregulated and downregulated proteins. Figure S3: Upregulated proteins involved in protein processing in ER pathway. Figure S4: Overlap and correlation analysis of proteins between LFQ and TMT data. Figure S5: Selected proteins of critical pathways significantly upregulated under stress (PDF)

Supplemental tables

Table S1: Proteins identified in TMT data. Table S2: Differentially expressed proteins from ANOVA analysis. Table S3: Upregulated proteins in thapsigargin treatment. Table S4: Downregulated proteins in thapsigargin treatment. Table S5: Upregulated proteins in rotenone treatment. Table S6: Downregulated proteins in rotenone treatment. Table S7: Functional enrichment of upregulated proteins in thapsigargin treatment. Table S8: Functional enrichment of downregulated proteins in thapsigargin treatment. Table S9: Functional enrichment of upregulated proteins in rotenone treatment. Table S10: Functional enrichment of downregulated proteins in rotenone treatment. Table S11: Top three PPI network modules of dysregulated proteins. Table S12: Upregulated proteins in PPI functional modules. Table S13: Proteins identified in label-free data. Table S14: Manual annotation of the upregulated UPR target proteins (XLSX).