Selective serotonin reuptake inhibitor, fluoxetine, impairs E-cadherin-mediated cell adhesion and alters calcium homeostasis in pancreatic beta cells

Selective serotonin reuptake inhibitors (SSRIs) are the most commonly prescribed drugs for mood disorders. Long term use of SSRIs is associated with an increased risk of diabetes, but the underlying mechanism(s) remains elusive. E-cadherin-mediated cell-cell adhesion and elevated [Ca2+]i are important for insulin release and pancreatic β cell functions. This study aims to investigate whether a SSRI, fluoxetine (Prozac), induces pancreatic β cell dysfunction through affecting E-cadherin and/or [Ca2+]i. Here we show that fluoxetine significantly reduces glucose stimulated insulin secretion (GSIS). MIN6 cells, an established murine immortalized β cell line, form smaller colonies of loosely packed cells with reduced cell-cell contact after fluoxetine treatment. Immunofluorescence staining reveals that fluoxetine increases cytoplasmic accumulation of E-cadherin and reduces the membrane-localized E-cadherin probably due to increase of its endocytosis. Fluoxetine inhibits spreading of β cells on E-cad/Fc coated slides and also disrupts E-cadherin-mediated actin filaments. Additionally, fluoxetine significantly suppresses endoplasmic reticulum (ER) calcium release and store-operated calcium entry (SOCE) activation, probably through reduction of ER calcium storage and inhibition of stromal interaction molecule 1 (STIM1) trafficking. These data suggest that exposure to fluoxetine results in impaired β cell functions, occurring in concert with reduction of E-cadherin-dependent cell adhesion and alterations of calcium homeostasis.


Fluoxetine treatment alters cell morphology, and reduces cell-cell adhesion.
To investigate the effects of fluoxetine on β cell function, MIN6, an established murine immortalized β cell line was employed as an in vitro model 32 . Cells were incubated with fluoxetine, a widely used SSRIs 33 , for 3 h. Our results showed that fluoxetine (30 μM) had no effect on cell proliferation and cell viability (Fig. S1A,B); however, it significantly inhibited GSIS (Fig. S1C). Next, we sought to understand the cellular and molecular events underlying this deleterious effect of fluoxetine on insulin secretion. Cell-cell adhesion plays an important role in regulating GSIS from pancreatic β cells 16,18 , so next we examined whether fluoxetine can affect cell morphology, and cell-cell adhesion. Our results showed that MIN6 cells grew in tightly packed colonies with close cell-cell contact in the control group, while cells formed smaller colonies of loosely packed cells with reduced cell-cell contact in the fluoxetine-treated group (Fig. 1A). To assess the role of adhesion molecules in mediating the alteration in cell morphology, MIN6 cells were immuno-stained with Alexa 488 (green) for E-cadherin and Alexa 594 (red) for β-catenin (Fig. 1B). We found control group with adjacent cells within each colony shared common boundaries demarcated by E-cadherin, but E-cadherin was reduced at area of cell contact and cell dispersed after fluoxetine treatment (Fig. 1B). Here we defined three characteristics of cell populations from our confocal images by performed z-section from top to bottom of cells (Fig. 1C). Combined cells stood for cells stick together at each stage, while separated cells represented that cells were totally disconnected from the top to bottom. Interestingly, there were some cells being associated to each other at the middle stage, but separated at the top and bottom stage. We defined this population as semi-separated cells. Quantification of these three characteristics of cell populations from confocal images stage-by-stage, as shown in Fig. 1D, 96.1 ± 2.7% of control cells combined to other cells, but only 67.2 ± 8.6% of fluoxetine-treated cells remained combined. The results indicated that fluoxetine altered cell morphology correlated with a loss of cell-cell adhesion.
Fluoxetine alters the structure of adherens junction and the distribution of E-cadherin. Adherens junction is the most important structure to maintain cell-cell adhesion. E-cadherin connects neighboring cells at outer membrane 34 , and is regulated by cytosolic protein β-catenin 35,36 . To investigate whether fluoxetine affects the adherens junction, E-cadherin and β-catenin were immuno-stained and visualized by confocal microscope. MIN6 cells treated with fluoxetine did not share a common boundary with their neighbors and were physically separated by distinctly visible gaps between E-cadherin demarcated membranes, indicating reduced cell-cell adhesion ( Fig. 2A). The 3D image constructed from confocal images clearly shows the location of E-cadherin expression at cell junction was altered by fluoxetine treatment (Fig. 2B). To our surprise, immunofluorescence staining showed that E-cadherin appeared to accumulate in cytosol after fluoxetine treatment ( Fig. 2A). Furthermore, our in vivo data revealed that the pancreatic islets from mice treated with fluoxetine showed a decrease in junctional E-cadherin (Fig. S2) Fluoxetine induces E-cadherin accumulation in Golgi apparatus, but not ER. Translated from ER, pro-E-cadherin is transported to the Golgi apparatus for protein trimming/maturation and then the mature E-cadherin is delivered to the plasma membrane 37 . Since fluoxetine induced accumulation of cytosolic E-cadherin in treated MIN6 cells ( Fig. 2A), we next sought to identify the cellular compartment where E-cadherin accumulates in the cytosol under fluoxetine treatment. Our immunostaining data showed that accumulated cytosolic E-cadherin did not co-localize with calnexin, an ER marker (Fig. 3A), suggesting that cytosolic E-cadherin did Scientific RepoRts | 7: 3515 | DOI:10.1038/s41598-017-03747-0 not accumulate in ER. This is confirmed by the fluorescence intensity profile, showing that the peak of E-cadherin expression did not overlap with that of calnexin (Fig. 3A). Golgin subfamily A member 5 (Golgin A5), which is responsible for vesicle tethering and docking at Golgi apparatus, serves as a marker of the Golgi apparatus. Immunofluorescence staining demonstrated that localization of Golgin A5 (red) was consistent with E-cadherin staining (green) under fluoxetine treatment, suggesting that cytosolic E-cadherin accumulated in the Golgi (Fig. 3B). Additionally, fluorescence intensity profile showed that the peak of Golgin A5 was correlated with that of E-cadherin (Fig. 3B). Thus, our data indicated that the fluoxetine induced the accumulation of E-cadherin in the Golgi apparatus, but not ER.
Fluoxetine reduces the membrane-localized E-cadherin by increasing E-cadherin endocytosis. Mature E-cadherin mediates adhesion function at the cell junction and its membrane localization is essential for adhesion activity 38,39 . To examine whether fluoxetine has an impact on cell surface expression of E-cadherin, we performed a cell surface protein isolation assay. Our data showed fluoxetine-treated cells displayed decrease E-cadherin expression at the cell surface when compared to control cells ( Fig. 4A and B). Of note, the total amount of E-cadherin remained unchanged after fluoxetine treatment ( Fig. 4A and B). The quantitative data showed a reduction of 33% in membrane-localized E-cadherin after fluoxetine treatment (Fig. 4B). Adherens junction is a highly dynamic structure 40 . The rate of endocytosis and exocytosis of E-cadherin regulates the activity of cell-cell adhesion 41 . We further asked whether the intracellular accumulation of E-cadherin in fluoxetine-treated cells was caused by an increase of its endocytosis. An early endocytosis marker, early endosome antigen1 (EEA1) was employed to test the endocytosis of E-cadherin. Fluoxetine increased co-localization of cytosolic E-cadherin and EEA1, which was not observed in control condition (Fig. 4C). The quantitative data of the co-localization of EEA1 and E-cadherin suggested that fluoxetine significantly increased the endocytosis of E-cadherin (Fig. 4D). Fluoxetine inhibits the spreading of β cells attached to E-cad/Fc. In order to directly determine the effect of fluoxetine on E-cadherin-mediated cell adhesion, excluding the effect of other junctional proteins, recombinant E-cadherin chimeric proteins (E-cad/Fc) were employed. First, we test the effect of E-cad/Fc on β cell spreading. MIN6 cells showed a round shape when attached to the control slides; however when attached to E-cad/Fc-coated slides, they tended to flatten and to change their shape, which is characteristic of the phenomenon hereafter referred to as cell spreading (Fig. 5A). The percentage of spreading cells was significantly higher after 6 hr or 9 hr of culture on E-cad/Fc-coated slides (Fig. 5B). To further investigate the effect of fluoxetine on cell spreading, cells were seeded on E-cad/Fc-coated slides, and then incubated with or without fluoxetine. The population of round-up cells is significantly increased when MIN6 cells were treated with fluoxetine ( Fig. 5C). Indeed, the quantitative data showed that fluoxetine also dramatically decreased the spreading cells (Fig. 5D). Taken together, our results showed that fluoxetine impaired E-cadherin-mediated cell adhesion, as assessed by cell morphology and spreading.

Fluoxetine disrupts E-cadherin-mediated actin filaments. Cortical filamentous actin (F-actin)
remodeling regulates insulin granule exocytosis in β cells 42,43 . We asked whether fluoxetine induced GSIS impairment observed in Fig. S1C, could be attributed to the impact of actin dynamics on insulin release. Staining with fluorescence-tagged phalloidin revealed that fluoxetine disrupted actin organization in cultured MIN6 cells (Fig. S3). To further confirm that cortical actin distribution was caused by disruption of E-cadherin, E-cad/ Fc-coated slides and fluorescence-tagged phalloidin were employed. In control group, cortical actin was highly expressed at the cell membrane ( Fig. 6A and B), but fluoxetine disrupted cortical actin structure ( Fig. 6C and D).

Fluoxetine suppresses ER calcium release and SOCE activation.
Calcium signaling plays an important role in multiple cellular functions, including insulin secretion 44 . We further studied whether fluoxetine affects calcium homeostasis in MIN6 cells, particularly focusing on ER calcium release and the SOCE activation. In fluoxetine treated cells, ER calcium release and SOCE activation were significantly suppressed, when compared with control groups (Fig. 7A and B). To determine whether the observed decrease in ER calcium release results from reduced ER calcium storage, we measured ER Ca 2+ level by using Mag-Fura-2/AM. Our data indicated that fluoxetine significantly decreased ER calcium storage ( Fig. 7C and D). STIM1 can sense the Ca 2+ loss during ER-calcium depletion, and aggregates by translocating to plasma membrane, to finally interact and open the SOC channels leading to Ca 2+ influx 45 . To investigate the mechanism how fluoxetine inhibits the activity of SOCE, we assessed STIM1 localization after ER Ca 2+ depletion. Here, we showed that thapsigargin, an inhibitor of sarco/endoplasmic reticulum Ca 2+ ATPase, induced endogenous STIM1 puncta formation with trafficking toward juxta-plasma plasma membrane (Figs 7, E3). Thapsigargin treatment also induced a similar effect on overexpressed eGFP-tagged STIM1 in control MIN6 cells (Figs 7, E7). In contrast, fluoxetine inhibited the thapsigargin-induced puncta formation and membrane trafficking of STIM1 (Figs 7, E4 and E8). These results indicated that ER calcium depletion-triggered STIM1 movement toward plasma membrane can be suppressed by fluoxetine.

Modulating intracellular Ca 2+ level disrupts spreading of MIN6 cells attached to E-cad/
Fc. Because fluoxetine inhibited ER calcium release and SOCE activation, we hypothesized that intracellular Ca 2+ might undergo certain stabilization of Ca 2+ -binding ectodomains. Here we designed a protocol to study whether intracellular Ca 2+ is also crucial for cell-cell adhesion (Fig. 8A). Treatment with ionomycin, an ionophore used to raise the intracellular level of Ca 2+ , significantly restored the E-cad/Fc-mediated β cell spreading that were inhibited by fluoxetine ( Fig. 8B and D). In addition, treatment with SKF-96365, a SOCE inhibitor, significantly decreased percentage of spreading β cells plated on the E-cad/Fc-coated slides ( Fig. 8C and E). Taken together, these data suggest that fluoxetine altered intracellular Ca 2+ level may affect E-cadherin-mediated adhesion of MIN6 cells.

Discussion
This study highlights the impact of acute exposure to fluoxetine on inhibition of GSIS in β cells, probably via impaired E-cadherin-mediated cell adhesion and altered calcium homeostasis. This conclusion is supported by following evidence. (a) Fluoxetine disrupted cell-cell adhesion; (b) fluoxetine induced E-cadherin accumulating in Golgi apparatus; (c) fluoxetine reduced membrane-localized E-cadherin via increased endocytosis; (d) Maintaining a tight cell-cell interaction of β cells in the islets of Langerhans is essential to form a syncytium to help the propagation and synchronization of the stimulus-secretion response from islet 46 . Indeed, studies shows that dispersed pancreatic β cells show decreased GSIS when compared to intact islets 19 . In this study, we observed definitive morphological changes in cell shape and colony phenotype upon fluoxetine treatment. Fluoxetine-treated cells lost their tightly packed arrangement and dispersed from each other with distinct intercellular gaps between adjacent cells. These changes in cell morphology may be attributed to changes in actin cytoskeleton, and E-cadherin-mediated cell-cell adhesion. In this study, we found that E-cadherin was relatively well expressed at the intercellular site of islet cells of control mice. In mice treated with fluoxetine, we observed a significant decrease in intercellular content of E-cadherin within pancreatic islets, which may be relevant to the T2DM pathogenesis. Furthermore, we also observed a significant decrease in E-cadherin in the pancreatic islets of HFD-fed mice, which is in consistent with the data of others 47 .
In this study, we defined three characteristics of cell populations from our confocal images by performing z-section from top to bottom of cells. Our results suggested that fluoxetine altered cell morphology correlated with a loss of cell-cell adhesion. The link between various cell population (combined, semi-separated, and separated) and GSIS remains to be further elucidated. Recombinant E-cadherin chimeric proteins (E-cad/Fc) were employed to directly examine the effect of fluoxetine on E-cadherin-mediated cell adhesion. Our results showed that the population of round-up cells is significantly increased when MIN6 cells were treated with fluoxetine. A previous study by Parnaud et al. reported that total insulin secretion was six times higher in spreading β cells

. Fluoxetine reduces the membrane-localized E-cadherin and increases E-cadherin endocytosis. (A)
After 3-hour incubation with or without fluoxetine (30 μM), cellular proteins were harvested by using a biotinsurface protein isolation assay, and then analyzed by western blotting. E-cadherin in total lysates, membrane fractions and cytosol fraction are shown. β-actin was used as a loading control for total protein. Hsp90 (heat shock protein 90) was used as a loading control for cytosol fraction. PMCA (calcium pump pan PMCA ATPase) was used as a loading control for membrane fraction. The representative data was from at least 3 independent experiments. Cropped blots have been presented. Full length blots are presented in Supplementary Fig. S6. (B) Quantitative analysis of total E-cadherin, membrane E-cadherin and cytosolic E-cadherin expression. Values (means ± SEM) are from three independent experiments. *P < 0.05 (C) After 3-hour fluoxetine (30 μM) treatment, MIN6 cells were fixed and then stained. The representative confocal images show the localization of E-cadherin (green) and EEA1 (red) in MIN6 cells. Nuclei were stained with Hoechst 33258 (blue). White arrows indicate the co-localization of E-cadherin and EEA1. Scale bar, 5 μm. The representative images were from three independent experiments. (D) Co-localization ratio between E-cadherin and EEA1 with pixelby-pixel analyses. Each value represents mean ± SEM from at least 30 different cells from three independent experiments. ***P < 0.001. compared with round β cells, suggesting that E-cadherin-mediated cell adhesion is very important for regulation of insulin secretion 18 . It has been known that insulin release from intact islets or aggregated islet cells is increased when compared with that of isolated islet cells 19 . Blockade of E-cadherin-mediated cell adhesion in pancreatic islets decreases GSIS, suggesting that loss of E-cadherin in β cells is related with impaired insulin secretion 18 . Our in vivo study demonstrated that pancreatic islets from mice treated with fluoxetine showed a decrease E-cadherin expression. The pancreatic islet contains different cell types (endocrine, neuronal, endothelial, mesenchymal and blood cells) that are interconnected by extracellular matrix, cell-to-cell adhesion molecules, cell-to-matrix adhesion molecules and gap junctions 48 . However, the phenomenon (various combined, semi-separated, and separated cell populations) has not been apparently observed in the pancreatic islets in this study due to the fact that many kinds of molecules and junctions are engaged in aggregated β cells and lack of technical expertise, but it is very interesting and worth of further investigation. E-cadherin, a cell surface protein normally associated with Ca 2+ -dependent adhesive functions, is shown to mediate adhesion of β cells and to affect their insulin secretory capacity 49 . Our results revealed that fluoxetine decreased E-cadherin at the cell surface but the total amount of E-cadherin was unchanged. The cytoplasmic accumulation of E-cadherin may reflect either reduced export to the cell surface or enhanced internalization to cytoplasm. Notably, immunofluorescence staining showed that after fluoxetine treatment, cytoplasmic E-cadherin accumulated mainly in Golgi rather than ER, suggesting that E-cadherin export is retarded at Golgi. Since E-cadherin is internalized constantly through an endocytic pathway, we assume the loss of cell surface E-cadherin is also due to increased endocytosis. As expected, the confocal images revealed a significantly increased co-localization of E-cadherin with EEA1 after fluoxetine exposure. The loss of adherens junctions is one of the earliest ultrastructural changes in diabetic isletopathy 42 . Therefore, the reduction of surface E-cadherin leading to loss of tight cell-cell adhesion represented a mechanism by which fluoxetine impaired β cell function and in vitro GSIS.
The actin cytoskeleton, a highly dynamic structure, plays essential roles in the regulation of numerous cellular processes, such as vesicle exocytosis 50 . Cortical actin serves as a mechanical barrier, and its dissolution helps insulin granule mobilization to plasma membrane and subsequent exocytosis. In β cells, actin filaments exist as a dense web beneath the plasma membrane, and are transiently depolymerized after glucose stimulation. Interestingly, we found that most β cells were kept in round shape, and cortical actin was located on the whole cell circumference after fluoxetine treatment, which is similar to the effect of E-cadherin-neutralizing antibodies on β cell morphology 18 , suggesting fluoxetine reduces E-cadherin-mediated cell junction leading to alteration of actin filaments and cell morphology.
Fluoxetine has been shown to alter Ca 2+ homeostasis in a variety of cell types, however, there is no study addressing this effect on pancreatic β cells. An increase in intracellular calcium ([Ca 2+ ] i ) is one of the crucial steps for insulin granule exocytosis in response to secretagogue stimulation 51 . A rise in [Ca 2+ ] i can also occur independently of voltage-dependent calcium channel (VDCC), by mobilization of Ca 2+ from intracellular stores and subsequent SOCE in β cells 20 . In this study, we demonstrated for the first time that exposure to fluoxetine significantly diminished ER calcium release and inhibited SOCE activation, leading to alteration in intracellular calcium homeostasis. The ER is a key source for [Ca 2+ ] i , particularly with regard to stimulus-secretion coupling 29 . Alterations in ER calcium handling have long been associated with islet dysfunction 28 . Here we directly measure ER calcium concentration in pancreatic β cells to clarify the role of [Ca 2+ ] i in the effect of fluoxetine. Our results indicated that fluoxetine altered resting [Ca 2+ ] ER , partially, which may be correlated with ER stress, that has emerged as an important event in the pathophysiology of diabetes. Furthermore, a recent study by Isaac et al. has demonstrated that long-term (16 hours) treatment with 30 μM sertraline (ZOLOFT ® ) induces ER stress and the initiates unfolded protein response 9 . In this study, our results showed that acute fluoxetine treatment did not affect expression of 78 kDa glucose-regulated protein (GRP78), suggesting that short-term (3 h) treatment with fluoxetine did not induce an ER stress (Fig. S4). SOCE plays a potential role in Ca 2+ entry in excitable pancreatic β cells. Recent work by Sabourin et al. has demonstrated that the dynamic STIM1 activation and their coupling to the plasma membrane Orai1 channels contribute to insulin secretion in rat β cells 31 . In this study, our results indicated that fluoxetine suppressed STIM1 movement toward plasma membrane, resulting in decreased SOCE activation. Another mechanism involved in the regulation of STIM1 is the phosphorylation of the protein 52 . Casas-Rua V et al. has described STIM1 activity is modulated by ERK1/2-dependent phosphorylation at residues Ser575, Ser608 and Ser621 53 . In this study, we found fluoxetine treatment inhibited the ERK1/2 activation and STIM1 phosphorylation at residues at Ser621 (Fig. S5). We thus propose that inhibition of ERK1/2 activation and STIM1 phosphorylation also involve in the inhibitory effect of fluoxetine on Ca 2+ entry. Furthermore, we found that inhibition of SOCE activation is clearly associated with the reduction of β cell spreading, suggesting a functional role of SOCE in β cells. Additionally, we observed that ionomycin-induced increase of intracellular Ca 2+ restored percentage of spreading β cells on E-cad/Fc-coated slides. Hence, calcium may be not only essential to elicit GSIS but also critical for stabilization of E-cadherin-mediated cell adhesion.
In conclusion, we demonstrated that in MIN6 cells that acute exposure to fluoxetine results in impaired GSIS, which is associated with alterations in E-cadherin and Ca 2+ homeostasis. The present work showed the first time that the distributions of E-cadherin as well as F-actin were significantly altered in fluoxetine-treated β cells, which may be relevant to the T2DM pathogenesis in patients treated with SSRIs.

Materials and Methods
Cell cultures, transfection. We obtained murine insulin-secreting MIN6 cells (passages 18) from Dr. Junichi Miyazaki (Osaka University Medical School) and cells were grown in Dulbecco's Modified Eagle's Medium (DMEM; Gibco) supplemented with 25 mM glucose, 10% heat-inactivated fetal bovine serum (FBS; Gibco), 100 U/ml penicillin, 100 μg/ml streptomycin (Gibco), 10 mM HEPEs (Gibco), 1 mM sodium pyruvate (Gibco) and 2 mM GlutaMax (Gibco). Cells were cultured at 37 °C in a humidified atmosphere containing 95% air and 5% CO 2 . Subculture and maintenance were performed as previously described 32 . Cultured cells were passaged every  MIN6 cells were seeded on glass slides coated with 5 μg/ml E-cad/Fc for 6 h, and then treated with or without fluoxetine for 3 h. After one-hour incubation with fluoxetine, cells were treated with or without ionomycin (5 μM) for 2 h. Cells were rinsed with PBS to remove unattached cells, and then visualized by EVOS FL Auto. At least 100 cells were analyzed for each cell population. Scale bar, 50 μm (C) MIN6 cells were seeded on glass slides coated with 5 μg/ml E-cad/Fc for 4 or,7 h, and then treated with or without SKF-96365 (25 μM or 50 μM) for 2 h. Cells were rinsed with PBS to remove unattached cells, and then visualized by EVOS FL Auto. At least 100 cells were analyzed for each cell population. Scale bar, 50 μm (D) Quantitative analysis of the percentages of spreading cells. Each value represents mean ± SEM from at least three independent experiments. ***p < 0.001 (E) Quantitative analysis of the percentages of spreading cells. Each value represents mean ± SEM from at least three independent experiments. ***p < 0.001.