The Protective Action of Metformin against Pro-Inflammatory Cytokine-Induced Human Islet Cell Damage and the Mechanisms Involved

Metformin, a drug widely used in type 2 diabetes (T2D), has been shown to protect human β-cells exposed to gluco- and/or lipotoxic conditions and those in islets from T2D donors. We assessed whether metformin could relieve the human β-cell stress induced by pro-inflammatory cytokines (which mediate β-cells damage in type 1 diabetes, T1D) and investigated the underlying mechanisms using shotgun proteomics. Human islets were exposed to 50 U/mL interleukin-1β plus 1000 U/mL interferon-γ for 48 h, with or without 2.4 µg/mL metformin. Glucose-stimulated insulin secretion (GSIS) and caspase 3/7 activity were studied, and a shotgun label free proteomics analysis was performed. Metformin prevented the reduction of GSIS and the activation of caspase 3/7 induced by cytokines. Proteomics analysis identified more than 3000 proteins in human islets. Cytokines alone altered the expression of 244 proteins (145 up- and 99 down-regulated), while, in the presence of metformin, cytokine-exposure modified the expression of 231 proteins (128 up- and 103 downregulated). Among the proteins inversely regulated in the two conditions, we found proteins involved in vesicle motility, defense against oxidative stress (including peroxiredoxins), metabolism, protein synthesis, glycolysis and its regulation, and cytoskeletal proteins. Metformin inhibited pathways linked to inflammation, immune reactions, mammalian target of rapamycin (mTOR) signaling, and cell senescence. Some of the changes were confirmed by Western blot. Therefore, metformin prevented part of the deleterious actions of pro-inflammatory cytokines in human β-cells, which was accompanied by islet proteome modifications. This suggests that metformin, besides use in T2D, might be considered for β-cell protection in other types of diabetes, possibly including early T1D.


Introduction
Diabetes mellitus (DM) is a disorder of the metabolism of carbohydrate, fat, and protein, due to the interplay of genetic and environmental factors [1,2]. It is characterized by an absolute or relative shortage of insulin production and secretion by the pancreatic islet β-cells [1,3,4]. In 2021 there were 537 million people (age 20-79 yrs) with DM, which is expected to increase to 643 million by 2030 and 783 million by 2045 [2]. Morbidity and mortality in diabetic subjects are high, mainly due to the acute metabolic and chronic vascular complications of the disease, and it has been calculated that approximately 6.7 million diabetic adults died in 2021 [2]. In parallel, the direct costs of diabetes grew to USD 966 billion in 2021 [2]. Therefore, better strategies to prevent and treat this disease are needed.
Type 2 diabetes (T2D) is the most common form of DM, representing approximately 90% of all cases [1][2][3][4]. Several drugs are used to treat T2D, and metformin is the most widely employed [5,6]. Metformin is derived from galegine, a natural component of Galega Officinalis, a plant used in herbal medicine in medieval Europe and that was introduced into clinical use for the treatment of T2D in the 1950s [7]. Although its molecular mechanisms of action remain to be fully elucidated, metformin has been proven to be a safe and effective therapy, and it is now recommended as the first-line pharmacological treatment against T2D [8]. Metformin reduces blood glucose levels by decreasing hepatic glucose production, modifying the gut microbiome, and enhancing GLP-1 secretion [9][10][11][12]. In addition, the drug has anti-inflammatory properties, as indicated by its reduction of the neutrophil to lymphocyte ratio in subjects with T2D, anti-oxidative stress action, direct inhibitory effects on NF-kB signaling, and suppression of inflammatory cytokines in non-diabetic individuals [13][14][15].
Cytokines are small proteins produced by immune cells and other cell types that may have pro-inflammatory or anti-inflammatory effects, and which act via autocrine, paracrine, and/or endocrine mechanisms. A large body of evidence shows that pro-inflammatory cytokines (locally produced by immune cells in the course of insulitis) are involved in the pathogenesis of type 1 diabetes (T1D) [16][17][18][19][20][21]. At the level of the β-cells, they contribute to β-cell dysfunction and/or death during the early (particularly type 1 interferons, such as interferon-a (IFN-a)) and late (particularly interleukin-1b (IL-1b) and interferon-g (IFN-g)) phases of insulitis in type 1 diabetes.
Interestingly, a few studies have shown that metformin could have direct protective effects on β-cells under metabolic stress, including non-diabetic and T2D human islet cells [22][23][24][25][26][27][28][29][30]. However, it is not known whether metformin directly protects human β-cells against the damage induced by pro-inflammatory cytokines, nor the mechanisms possibly involved. Previous studies have indicated that the drug could shelter chondrocytes from IL-1β injury [31], reduce cytokine production in the cardiac muscle following ischemiareperfusion [32], and limit the damage induced by lipopolysaccharide exposure in human bronchial epithelial cells [33].
In the present study we evaluated if metformin can defend human islet cells from IL-1β + IFN-γ-induced dysfunction and death. The mechanisms possibly involved were investigated at the proteome level with the use of shotgun proteomics, a bottom-up technique that enables comprehensive protein identification and profiling [34]. Although different tissues and cell types have been extensively evaluated using this approach [35,36], shotgun proteomics have been used to analyze pancreatic islets, the key tissue in diabetes pathogenesis [37][38][39], in only a few studies [40][41][42].
We show that, in our experimental conditions, metformin was able to shield isolated human islets from part of the insults induced by the tested cytokines, which was associated with several changes at the proteomic level, with the involvement of pathways mainly regulating inflammation and oxidative stress.

Human Pancreatic Islets
Isolated islets were prepared by enzymatic digestion and gradient purification from the pancreas of 14 multiorgan donors (age: 71 ± 9 years; 5M/9F; BMI: 26 ± 3 kg/m 2 ) [43], with written consent by next-of-kin. Glands that were not suitable for clinical purposes were processed [43,44] with the approval of the local Ethics Committee (#2615 of 15 January 2014). We selected donors without a known history of diabetes. Following isolation, islets were cultured in M199 medium (Euroclone SpA, Milan, Italy) containing 5.5 mM glucose, supplemented with 10% (v/v) adult bovine serum, 100 U/mL penicillin, 100 µg/mL streptomycin, 50 µg/mL gentamicin, and 750 ng/mL amphotericin B (all from Sigma-Aldrich, St. Louis, MO, USA) at 37 • C in a CO 2 incubator. For the purpose of the present study, approximately 1000 islets were incubated with cytokines (50 U/mL IL-1β, 1000 U/mL IFN-γ) for 48 h, in the presence and absence of metformin (2.4 µg/mL) (Sigma-Aldrich). This is a therapeutic concentration of the drug, which has been used in our laboratory previously [28,29]. The cytokine concentrations were based on those used by us and others in previous experiments [17,[45][46][47]. Afterwards, isolated islets were used for functional, survival, and proteomics analyses, as described below.

Insulin Secretion Studies
Insulin release experiments were conducted as previously described [43,48,49]. After 45 min pre-incubation at 3.3 mM glucose, batches of 15 handpicked islets were challenged acutely (45 min) with 3.3-and 16.7-mM glucose. Then the islets were subjected to acidalcohol extraction for insulin content measurement, as previously reported [43,48,49]. Insulin was quantified using a radioimmunometric assay (DIAsource ImmunoAssays S.A., Nivelles, Belgium). Insulin release was expressed as a percentage of the total insulin content. Insulin stimulation index was calculated as the ratio of insulin release at 16.7 mM glucose over the release at 3.3 mM glucose.

Caspase 3/7 Activity Assay
A Caspase-Glo ® 3/7 assay kit (Promega Corporation, Madison, WI, USA) was used to detect caspase 3/7 activity, as described in [50,51]. Briefly, batches of 10 size-matched islets were seeded in a white solid 96-well plate, in a total volume of 100 µL/well. Then 100 µL of caspase 3/7 reagent, a solution containing luciferase and a tetrapeptide substrate linked to aminoluciferin, was added to each well and incubated for 1 h at room temperature. Following caspase cleavage of substrate, aminoluciferin was released and processed by luciferase, resulting in the production of light. Luminescence was recorded with a FLUOstar Omega microplate reader (BMG Labtech, Ortenberg, Germany).

Protein Extraction from Human Pancreatic Islets
The proteomic analysis was performed with islet preparations obtained from three different multiorgan donors (representing the biological replicates). Protein extraction from human pancreatic islets was performed as previously described [52]. Briefly, isolated islets were collected and washed twice with PBS (37 • C). Cells were suspended in the rehydration solution (7 M urea, 2 M thiourea, 4% CHAPS, 60 mM dithiothreitol (DTT), 0.002% bromophenol blue) containing 50 mM NaF, 2 mM Na 3 VO 4 , 1 µL/10 6 cells of protease inhibitors, 1 µM trichostatin A, and 10 mM nicotinamide. After stirring and sonication (4 s, 5 times) cells were allowed to rehydrate for 1 h at room temperature (RT) with occasional stirring. Thereafter, the solution was centrifuged at 17,000× g for 5 min at RT. The protein concentration of the resulting supernatant was determined using the Bio-Rad RC/DC-protein assay (Bio-Rad). BSA was used as a standard.

Protein Fractionation
For shotgun analysis [53], technical triplicate experiments were performed on each of the three human islet preparations. For each preparation, three different conditions were analyzed, i.e., control islets, cytokine alone treatment, and cytokine plus metformin treatment. For this purpose, approximately 1000 human islets were treated as described above, and protein extracts were processed as follows: aliquots (40 µg of proteins) were loaded onto 12% acrylamide resolving gel and subjected to 1D-electrophoresis, as previously performed by us and others [43,54]. After protein staining using Coomassie blue R-250, 16 gel bands, matched for each lane, were excised and washed twice with wash buffer (25 mM NH 4 HCO 3 in 50% acetonitrile). Afterwards, proteins were reduced with 10 mM dithiothreitol (45 min, 56 • C) and alkylated with 55 mM iodoacetamide (30 min at RT in the dark). After two washes with the washing buffer, protein bands were completely dried in a centrivap vacuum centrifuge. Then the dried pieces of gel were rehydrated for 30 min at 4 • C in a porcine trypsin (Promega, Madison, WI, USA) solution (3 ng/µL in 100 mM NH 4 HCO 3 ) and incubated overnight at 37 • C. The reaction was quenched by adding 10% trifluoroacetic acid. The samples were stored at −20 • C before being analyzed by LC-MS/MS.

Shotgun Label Free Analysis
The resulting peptides, 48 samples for each subject (16 controls, 16 treated with cytokines, and 16 treated with cytokines + metformin), were grouped by band and analyzed in technical triplicates using LC-MS/MS using a Proxeon EASY-nLCII (Thermo Fisher Scientific, Milan, Italy) chromatographic system coupled to a Maxis HD UHR-TOF (Bruker Daltonics GmbH, Bremen, Germany) mass spectrometer equipped with a nanoESI spray source. Peptides were loaded on an EASY-Column C18 trapping column (2 cm L, 100 µm I.D, 5 µm ps, Thermo Fisher Scientific) and subsequently separated on an Acclaim PepMap100 C18 (75 µm I.D., 25 cm L, 5 µm ps, Thermo Fisher Scientific) nano scale chromatographic column. The flow rate was set to 300 nL/min, and the gradient (mobile phase A: 0.1% formic acid in H 2 O) was from 3 to 35% of mobile phase B (1% formic acid in acetonitrile) in 80 min, followed by 35 to 45% in 10 min and from 45 to 90% in 11 min. The mass spectrometer was operated in positive ion polarity and Auto MS/MS mode (data dependent acquisition-DDA), using N2 as a collision gas for CID fragmentation. Precursors in the range of 350 to 2200 m/z (excluding 1220.0-1224.5 m/z) with a preferred charge state +2 to +5 (excluding singly charged ions) and absolute intensity above 4706 counts were selected for fragmentation in a maximum cycle time of 3 s. After acquiring one MS/MS spectrum, the precursors were actively excluded from selection for 30 s. Isolation width and collision energy for MS/MS fragmentation were set according to the mass and charge state of the precursor ions (from 3 to 9 Da and from 21 eV to 55 eV). In-source reference lock mass (1221.9906 m/z) was acquired online throughout the runs. Altogether, 432 instrumental runs were performed. Each raw data file was converted to mzXML format and submitted to LFQ processing (see below).

Raw Data Processing and Quantitative Analysis
Raw mass spectrometry data were analyzed using the PEAKS ® Studio 7.5 software using the "correct precursor only" option. Spectra were matched against the neXtProt database (including isoforms as of June 2017; 42,151 entries), and the false discovery rate (FDR) was set to 0.1% at the peptide-spectrum matches (PSM) level. The post-translational modification (PTM) profile was set as follows: fixed cysteine carbamidomethylation (∆Mass: 57.02), variable methionine oxidation (∆Mass: 15.99), and glutamine and asparagine deamidation (∆Mass: 0.98). Non-specific cleavage was allowed to one end of the peptides, with a maximum of 2 missed cleavages and trypsin enzyme specificity. The highest error mass tolerances for precursors and fragments were set at 10 ppm and 0.05 Da, respectively. After processing every single raw data point, the label free quantification (LFQ) tool of PEAKS Studio was used to detect differentially expressed proteins. Parameters for LFQ were set as follows: quantification type as label free quantification; mass error tolerance, 10.0 ppm; retention time shift tolerance, 2.0 min; FDR threshold, 0.5%. The nine samples for each of the sixteen slices were allotted to 3 groups corresponding to Ctrl, Cyt, and Cyt + Met. For quantitative analysis, the significance threshold at the protein level was set to ≥ 20−10lgP with a fold change ≥2.0. Sixteen lists of differentially expressed proteins were obtained for each subject.

Pathway Analysis
Gene ontology and pathway analyses of differentially expressed proteins were performed with Metascape v3.5 (https://metascape.org/, accessed on 7 July 2022) [55] and Ingenuity Pathway Analysis (IPA, QIAGEN Redwood City, www.qiagen.com/ingenuity, Build version: 321,501 M, Content version: 21249400, accessed on 7 July 2022), respectively. IPA core analysis provides not only gene functional annotation, canonical pathway, and network discovery, but also estimates the status of upstream regulators and downstream effects associated with canonical pathways, diseases, and functions. The upstream regulator analysis highlights the expected effects between the transcriptional regulators and their target genes [56]. The predicted activation or inhibition of each transcriptional regulator is inferred by the z score, which in turn is derived from the protein ratios in the dataset (z scores >2.0 indicate that a molecule is activated, whereas z scores <−2.0 indicate the inhibition of target molecules). SwissProt accession numbers with corresponding ratios were imported into the software, and the analysis was performed selecting only direct relationships among genes and molecules in all species and confidence settings were set to high predicted or experimentally observed. An IPA comparison analysis between the results of the different sections for every condition was also performed.

Statistical Analysis
Data are expressed as means ± SEM. The ANOVA test followed by Tukey correction was applied to assess the difference between groups in the insulin secretion and caspase activation experiments. A p value less than 0.05 was considered statistically significant. For proteomics analyses, we used the results generated with the islet preparations obtained from three different multiorgan donors (biological replicates), with peptides from each fraction analyzed in technical triplicates. The false discovery rate (FDR) was set to 0.1% at the peptide-spectrum matches (PSM) level, resulting in an average protein FDR lower than 1.0%. Expression analysis for the relative abundance of identified proteins was performed at the band-slice level using the label-free quantification module PEAKS-Q, part of PEAKS Studio v. 7.5. This quantification method is based on the MS1 ion peak intensity of the extracted chromatograms of peptides detected in multiple samples and applies an expectation-maximization algorithm to detect and resolve overlapping features. A highperformance retention time alignment algorithm was also used to align features of the same peptide from multiple samples. The significance of the LFQ proteomics data provided directly from the software PEAKS Studio was calculated using the PEAKS Q method, which is similar to the significance B, as previously defined [57]. Briefly, protein ratios are calculated as the median of peptide ratios, minimizing the effect of outliers and normalizing the protein ratios, to correct for unequal protein amounts. An outlier significance score for log protein ratios is computed and a p-value for detection of significant outlier ratios is defined. Peptide ratios are calculated using the XIC of three different peptides. The differentially expressed proteins were calculated for each band using 0.5% PSMs FDR, and the resulting ID lists were filtered, considering only peptides confidently identified in at least 1 sample with significance ≥20 (−10lgP) and quality factor ≥0.5, and by considering only proteins identified with a significance ≥20 and fold change ≥2. The lists of the resulting dysregulated proteins of each band-slice were merged and manually curated to remove proteins identified in multiple adjacent bands, non-distinguishable isoforms, and keratins.

Effects of Metformin on Cytokine-Induced Damage
The first aim of the present work was to assess whether metformin could protect human islets from cytokine toxicity, as previously observed for lipotoxic and glucotoxic damage [28,30]. The insulin content in cytokine-exposed islets was lower than in control cells and was not significantly modified by the presence of metformin ( Figure 1A). As expected, the insulin stimulation index in response to glucose was significantly reduced after cytokine treatment ( Figure 1B). However, with metformin added to the cytokines, the β-cell responsiveness to glucose stimulation was comparable to that of control islets ( Figure 1B). As shown in Figure 1C, the presence of metformin led to a significant decline of cytokine-induced caspase 3/7 activity, a marker of cell apoptosis. These results show that metformin could counteract part of the deleterious actions of proinflammatory cytokines on human islet cells. Insulin content, reduced after cytokine-treatment, was marginally affected by metformin. (B) Insulin stimulation index was reduced after 48 h treatment with cytokines (Cyt) and tended to return to the control values (Ctrl) in islets treated with metformin (Cyt + Met). (C) Cytokines induced a significant activation of caspase 3/7, while metformin significantly reduced this activation. One to three replicates from three to four independent islet preparations were studied. The different groups were compared with One-way ANOVA followed by the Tukey correction. **** p < 0.0001, ** p < 0.01, * p < 0.05.

Identification of Islet Proteins Using Multidimensional Shotgun Proteomics
Isotope free shotgun proteomics analysis, performed after 1D-PAGE separation, was carried out in control and treated islets, to assess and identify differentially expressed proteins in cytokine-and cytokine plus metformin-exposed vs. control samples. A global view of the experimental workflow is shown in Figure 2A. In 1D gels, the 16 extracted bands were highlighted and paired to the corresponding average mass of the proteins identified in each band ( Figure 2B). By merge-processing the 16 bands of each gel lane, altogether 3115 proteins were identified (Table S1): 1857 in subject 1; 2471 in subject 2; and 2585 in subject 3. The proteins identified across all samples were 1525 and those present in at least two preparations were 2271. Figure 2C shows a Venn diagram of the three series. We then assessed the proteins significantly affected by the addition of cytokines and cytokines plus metformin. Taking into consideration the proteins present in at least two preparations, we found 244 proteins significantly affected by cytokines (145 up-and 99 down-regulated), of which 32 were exclusively detected in the cytokine-treated islets (Table S2). The addition of metformin to cytokine treatment significantly altered the expression of 231 proteins (128 up-and 103 downregulated), compared to control samples (Table S3). Of these, 19 were only detected in the cytokine-treated islets.
As shown in Figure 3A, 212 differentially expressed proteins were found in common between the two different comparisons (Table 1), mostly regulated in the same direction (98 up-up and 88 down-down, Figure 3B and Table S4). Interestingly, 26 proteins showed an inverse regulation compared to control islets (Table S4). Of these, 11 were downregulated by cytokine treatment and upregulated after the addition of metformin ( Figure 3B). Most proteins were involved in vesicle motility (transgelin, Ras-related protein Rab-14), defense against oxidative stress (peroxiredoxins, PRDX2 and PRDX5) and metabolism (flavin reductase, mitochondrial ATP synthase subunit O). Among the 15 proteins upregulated by cytokines and downregulated by metformin ( Figure 3B), we detected proteins involved in protein synthesis (40S ribosomal proteins S3, S6, S9, eukaryotic translation initiation factor 4E), glycolysis, or glycolysis regulation (triosephosphate isomerase, pyruvate kinase); protein folding and secretion (peptidyl-prolyl cis-trans isomerase FKBP2, protein disulfideisomerase); and cytoskeletal proteins or proteins interacting with the cytoskeleton (myosin light polypeptide 6, Ras-related protein Ral-A, coactosin-like protein).

Western Blot (WB) Analysis of ERAP2 and IFI30 in Human Islets
The different expression of IFI30 and ERAP2 (proteins involved in antigen processing) in human islets treated with cytokines and cytokines plus metformin was also evaluated using WB analysis. A specific 28 KDa immunoreactive band was detected for IFI30 ( Figure 6A), while four specific immunoreactive bands, with apparent weights of 110, 105, 70, and 45 kDa (the first 2 reported in Figure 6A), were detected for ERAP2, corresponding to different isoforms of this protein. In our shotgun experiments, ERAP2 was identified in band 3 of the 1DE gel, corresponding to the highest molecular weight isoforms 1 and 3 of 110 and 105 KDa, respectively. According to the shotgun proteomic analysis, WB showed a significantly higher expression of both IFI30 and ERAP2 in islets treated with cytokines alone than in control samples ( Figure 6B,C). The changes induced by the addition of metformin, as observed by proteomic evaluation (significant decrease of IFI30 and significant increase of ERAP2 expression), remained as an apparent, although not significant, trend.

Discussion
The present study reports the analysis of human pancreatic islets after 48 h exposure to pro-inflammatory cytokines (50 U/mL IL-1β and 1000 U/mL IFN-γ), with or without the presence of metformin, using a multidimensional shotgun proteomics approach. We observed protective effects of metformin on citokine-induced β-cell functional damage and increased capase 3/7 activity and investigated the underlying molecular mechanisms by assessing the related proteome modulation.
The previously observed deleterious effects of cytokines on human islet β-cell function and survival [21,58] were confirmed in the present study. Interestingly, the presence of metformin partially prevented β-cell dysfunction and activation of caspase 3/7 (a marker of apoptosis), similarly to the protective action that the compound exerts on human islets exposed to lipoglucotoxicity [28,30] and on islets isolated from type 2 diabetic patients [29]. This suggests that metformin has broad beneficial effects on stressed human islet cells, regardless of the insulting condition, possibly due to pleiotropic, and so far poorly understood, mechanisms.
A few previous studies have investigated the effects of pro-inflammatory cytokines on human islet gene and protein expression [45,46,[59][60][61]. Recently, Nakayasu et al. used tandem mass tags and the label-free technique to study islet proteomics in depth [42]. In our study, 3115 proteins were identified in control samples (untreated islets), of which 3014 were also reported by Nakayasu et al. [42], indicating the good reproducibility of the results between the two approaches. After cytokine exposure, we found that 244 proteins were differentially expressed compared with the control islets, mainly pertaining to the cytoskeleton, immune response signaling, apoptosis signaling, energy metabolism, protein metabolism, and RNA metabolism. Of them, 57 were also reported in ref [42]. All but two of these 57 proteins were modulated in the same direction in both studies and were mostly upregulated compared to control islets.
Interestingly, the addition of metformin to cytokine-treated islets inhibited several canonical pathways and upstream regulators related to inflammation, such as interleukin and HMGB1 (high mobility group protein B1) signaling and IL15 [62,63]. An anti-inflammatory effect of metformin has been described in a few models, including pancreatic islets [64][65][66][67]. The mechanisms by which metformin dampens inflammation are still unclear. However, the drug can reduce oxidative stress [29,68], which is linked to the promotion of inflammatory processes [69,70]. Accordingly, in metformin-treated islets, we also found a significantly increased expression of proteins involved in the defense against oxidative stress [71,72], such as glutathione S-transferase α1 and 2 (GSTA1 and 2); thioredoxin 1 (TRX1); and peroxiredoxins 2, 3, and 5 (PRDX2, 3, and 5). The peroxiredoxin/thioredoxin antioxidant system has been described in rodent β-cell lines and pancreatic islets as a relevant protective mechanism against oxidative damage [73][74][75][76]. The specific role of each peroxiredoxin is still debated. PRXD1 has been recently described as the main isoform involved in protection against hydrogen peroxide and peroxynitrite [77]. Interestingly, we presently observed that the two mitochondrial peroxiredoxin isoforms, PRDX3 and PRDX5, resulted as upregulated in metformin-treated islets, suggesting that the mechanism of metformin action in mitochondria could go beyond the inhibition of complex I [78].
Metformin was also able to reduce cytokine-induced immune response through the inhibition of pathways such as the systemic lupus erythematosus in T cell signaling pathway, PKCθ signaling in T lymphocytes, role of NFAT in regulation of the immune response, iCOS-iCOSL signaling in T helper cells, and acute phase response signaling. Some immunesuppressive action of metformin via modulation of T lymphocyte activity has been described [79]. In our study, the addition of metformin to cytokines reduced the expression of components of major histocompatibility complex (MHC) class I antigen presentation, such as HLA class I histocompatibility antigen B alpha chain and C alpha chain (HLA-B and HLA-C) and beta-2-microglobulin (B2M), which is critical for the expression of functional HLA-I molecules on the cell surface [80,81]. These changes might have protective effects on β-cells in T1D, by decreasing the presentation of b-cell β-cell neoantigens for the immune system [82]. Conversely, the expression of HLA-A and G α chains increased in the islets exposed to cytokines and metformin. HLA-G, a non-classical HLA class I molecule having immunomodulatory properties, has previously been found to be constitutively expressed in human pancreatic islets [83] and has been reported to be involved in the attenuation of autoimmune and inflammatory processes [84].
Our proteomic data also showed variations in the expression of IFI30 and ERAP2, enzymes involved in antigen processing [85,86] and that are required for peptide binding to MHC class I antigens and for generating MHC class II-restricted epitopes from disulfide bond-containing proteins, respectively. IFI30 has been implicated in the pathogenesis of some autoimmune diseases, and genetic polymorphisms of ERAP2 and its paralog ERAP1 are associated with increased susceptibility to autoimmune/chronic inflammatory disorders [87,88]. In agreement with findings previously reported with different models [89,90], we observed that cytokines induced overexpression of both enzymes in human islets, which was confirmed by Western blot analysis. Interestingly, the shotgun proteomics results showed significantly decreased expression of IFI30 and increased expression of ERAP2 when metformin was added to the cytokines, which was tendentially confirmed by Western blot analysis.
Among the proteins exclusively upregulated in metformin-treated islets, calcium/ calmodulin-dependent serine protein kinase (CASK) and transcription factor JUNB deserve a special mention, since previous work has implicated them in protecting β cells against cytokines-induced damage [91,92]. CASK, which plays a role in synaptic transmembrane protein anchoring and ion channel trafficking, seems to be also involved in insulin secretion from β cells [93]. It has been reported that IL-1β reduces CASK expression in INS-1 cells and rat islets, while its overexpression counteracts the cytokine-induced β cell dysfunction, by improving insulin secretion [91]. Although, in our experiments, we did not detect a decrease of CASK expression following cytokine exposure, the addition of metformin induced the expression of this protein kinase, suggesting a possible protective mechanism of the drug at this level. Furthermore, it has been shown that the proinflammatory cytokines IL-1β and INF-γ induce an initial and transient upregulation of JUNB in INS-1E cells [92] and that JUNB overexpression reduces cytokine-induced β cell death [94]. Our data therefore suggest that metformin could contribute to keeping active a self-defense system that is otherwise transient.
Another effect of metformin on cytokine-treated islets inferred from our proteomics results is the inhibition of the upstream regulator RAPTOR (regulatory-associated protein of mTOR) and hence the mTOR (serine/threonine-protein kinase mTOR) signaling pathway. It is known that metformin downregulates mTOR signaling through either 5 -adenosine monophosphate-activated protein kinase (AMPK)-dependent-or independent mechanisms [95,96]. We found a decreased expression of proteins involved in mTOR activity, specifically eukaryotic translation initiation factor 4E (eIF-4E), a regulator of translation, and the 40S ribosomal proteins S3, S6, and S9, involved in protein synthesis [97]. Interestingly, mTOR inhibition is associated with increased autophagic fluxes [98], which, in turn, could favor the function and survival of stressed β-cells [99][100][101][102].
Of interest, the senescence pathway was also activated by cytokine-treatment and inhibited by metformin addition. The beneficial action of metformin in mitigating aging hallmarks has previously been reported [95]. Notably, cellular senescence has been identified as a key process in both T1D and T2D development [103][104][105]. In a murine model of pf T2D, it was found that insulin resistance accelerates β-cell senescence, while removal of senescent β-cells (senolysis) improves β-cell function and glucose homeostasis, leading to a better disease outcome [103]. Moreover, it has been reported that during T1D development a subset of β-cells apparently acquire the senescent phenotype and thus contribute to immune-mediated β-cell destruction [104].
This study has some limitations. Although the concentration of metformin that we used is in the therapeutic range, the actual levels of the drug in the pancreatic islet microenvironment in vivo are currently unclear. In addition, pancreatic islets are heterogeneous within the same pancreas and between subjects [38,106], and it is unknown if cytokines and metformin have different effects on islets from different individuals. Nevertheless, the islet functional and proteomics results were generated with cutting edge methodologies applied in experienced laboratories and analyzed using strict statistical assessment. This allowed confirming and supporting data from previous studies and, more importantly, to add new knowledge to the field.
In conclusion, the present study shows, for the first time, that metformin prevents, at least in part, the deleterious actions of pro-inflammatory cytokines on human β-cell function and islet cell caspase 3/7 activation, which is accompanied by several modifications of the islet proteome. These modifications include pathways involved in inflammation, immunity, mTOR signaling, and cellular senescence, all known to impact β-cell health. This evidence suggests that metformin, a widely used drug for the treatment of T2D, might be repurposed for β-cell protection in early T1D.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/cells11152465/s1. Table S1: List of total proteins found in isolated human islets. Table S2: List of differentially expressed proteins after cytokine treatment with respect to controls (in blue color those exclusive of either condition). Table S3: List of differentially expressed proteins after cytokine plus metformin treatment with respect to control (in blue color those exclusive of either condition). Table S4: List of differentially expressed proteins in common between the