Eukaryotic Translation Initiation Factor 3 Subunit E Controls Intracellular Calcium Homeostasis by Regulation of Cav1.2 Surface Expression

Inappropriate surface expression of voltage-gated Ca2+channels (CaV) in pancreatic ß-cells may contribute to the development of type 2 diabetes. First, failure to increase intracellular Ca2+ concentrations at the sites of exocytosis impedes insulin release. Furthermore, excessive Ca2+ influx may trigger cytotoxic effects. The regulation of surface expression of CaV channels in the pancreatic β-cells remains unknown. Here, we used real-time 3D confocal and TIRFM imaging, immunocytochemistry, cellular fractionation, immunoprecipitation and electrophysiology to study trafficking of L-type CaV1.2 channels upon β-cell stimulation. We found decreased surface expression of CaV1.2 and a corresponding reduction in L-type whole-cell Ca2+ currents in insulin-secreting INS-1 832/13 cells upon protracted (15–30 min) stimulation. This internalization occurs by clathrin-dependent endocytosis and could be prevented by microtubule or dynamin inhibitors. eIF3e (Eukaryotic translation initiation factor 3 subunit E) is part of the protein translation initiation complex, but its effect on translation are modest and effects in ion channel trafficking have been suggested. The factor interacted with CaV1.2 and regulated CaV1.2 traffic bidirectionally. eIF3e silencing impaired CaV1.2 internalization, which resulted in an increased intracellular Ca2+ load upon stimulation. These findings provide a mechanism for regulation of L-type CaV channel surface expression with consequences for β-cell calcium homeostasis, which will affect pancreatic β-cell function and insulin production.


Introduction
Voltage gated calcium channels (Ca V ) play a critical role in glucose-stimulated insulin secretion in pancreatic b-cells by activating Ca 2+ influx upon membrane depolarization [1,2]. Ca 2+ influx is important for activating several physiological events such as pancreatic islet development and phasic insulin secretion [3,4]. However, intracellular Ca 2+ overload has detrimental effects and causes endoplasmic reticulum (ER) stress and initiates cytotoxicity [5,6]. Dynamic Ca V channel expression in the plasma membrane could be an effective way to regulate intracellular Ca 2+ homeostasis and prevent adverse effects in the b-cell. Regulation of Ca V channel surface expression is a more dynamic process than previously assumed and can also be of importance for short-term variations in Ca V channel activity [7]. This could be of relevance for the respective phases of glucose-evoked secretion that in mouse are controlled by different Ca V isoforms [4,8]. For example, genetic ablation of Ca V 1.2, one of the L-type Ca V channels, strongly reduces first phase insulin release [8,9]. Human ß -cells have an L-type calcium current component and mRNA for both L-type Ca V 1.3 and Ca V 1.2 can be detected in human islets [10]. Ca V 1.2 denotes the Ca V subunit isoform a 1C , which determines the main electrophysiological and pharmacological properties of the channel and forms a heteromeric channel complex with the auxiliary subunits b, a2d and c [1]. Both b and a2d subunits have been implicated in Ca V channel transport to the plasma membrane [11,12,13]. eIF3e (Eukaryotic translation initiation factor 3 subunit E) is a subunit of the protein translation initiator complex that participates in the disassembly and recycling of posttermination ribosomal complexes and proteasome-mediated protein degradation [14,15,16]. eIF3e contains a highly conserved PCI domain, which binds the proteasome COP9 signaling complex that plays a central role in regulating ubiquitination and activation of proteolysis [17]. However, eIF3e has also been implicated in regulation of other cellular functions. For example, in neurons, eIF3e has been showed to influence Ca V 1.2 expression in the synaptic membrane [18]. In adipocyte and vascular smooth muscle cells, the eIF3 complex can interact directly with mTOR, a critical signal molecule in controlling intracellular trafficking of glucose transporters [19,20]. Whether eIF3e can affect Ca V 1.2 translocation to/from the b-cell membrane and regulate b-cell physiology is not known. To address this possibility, we investigated the trafficking of Ca V 1.2 in b-cells by a plethora of imaging and other methods and found that eIF3e is involved in depolarization-induced internalization of Ca V 1.2, with consequences for b-cell intracellular Ca 2+ homeostasis.

Ca V 1.2 Channel Clusters Internalize upon Glucose Stimulation in Insulin-secreting INS-1 832/13 Cells
To quantify the number of Ca V 1.2 clusters in the plasma membrane (PM) we first performed co-immunostaining of Ca V 1.2 and the PM marker Na + /K + ATPase ( Fig. 1A-C). Then we analyzed the ratio of Ca V 1.2 mean intensity in the PM over that in the cytosol to quantify internalization of Ca V 1.2 in single INS-1 832/13 cells. This ratio was significantly decreased upon stimulation by 20 mM glucose or 70 mM KCl (from 1.2660. 22 to 0.5860.08 or 0.5160.1, respectively; n = 12 in each group). The decreases in Ca V 1.2 surface expression were further confirmed by total internal reflection fluorescence microscopy (TIRFM) imaging ( Figure S1). This internalization was also observed in immunostaining experiments for Ca V 1.2 and the early endosome marker EEA-1 (Fig. 1D). In these experiments, colocalization of Ca V 1.2 and EEA-1 increased upon stimulation with either glucose or high K + (from 18.4263.21 to 3864.85 or 31.6363.01; n = 10 in each group; Fig. 1E and F). These results suggest that the decreases in Ca V 1.2 surface expression could be caused by endocytoic uptake. Indeed, we also observed increases in the colocalization of Ca V 1.2 with the recycling endosome marker Rab11 (from 5.9460.91 to 15.363.09; n = 9 in each group; Fig. 1G and H). We next measured voltage-gated Ca 2+ influx using the standard whole-cell configuration of the patch clamp technique. These experiments demonstrated that the integrated Ca 2+ current (Q Ca2+ ) elicited by voltage-clamp depolarizations decreased after 15-min exposure to high glucose (20 mM) ( Fig. 1I and J), and was recovered after restoring glucose levels to basal. This finding supports the possibility that alterations in surface expression of L-type Ca V channels may occur in response to glucose during protracted stimulation. This agrees with the data concerning Ca V 1.2 in Figure 1A-H, but the contribution of other L-type channels, e.g. Ca V 1.3 that is known to be expressed in beta-cells, can not be excluded. Evidence for activity-dependent trafficking of Ca V 1.2 was also obtained by immunoblotting in subcellular fractions. The Ca V 1.2 band detected in the PM is ,20 kD heavier than that detected in the cytosol. This has been reported previously, but at present we do not have a full explanation for this finding. Nevertheless, this experiment showed that Ca V 1.2 expression in the plasma membrane decreased when the cells were kept in 20 mM glucose for 30 min prior to fractionation, opposite to the intracellular expression of Ca V 1.2 expression that increased by the same treatment (Fig. 1K).
To observe dynamic internalization of Ca V 1.2 in single cells, we employed three-dimensional (3D) live imaging in EGFP-Ca V 1.2 transfected cells. This approach allows tracking of Ca V 1.2 distribution in the entire cells, thereby avoiding confounding effects by Ca V 1.2 transport out of the focal plane or by changes in cell shape during the experiment. First, Ca V 1.2 clusters were observed 36 h after transfection ( Fig. 2A). Interestingly, similarly EGFP-tagged P/Q-type Ca V 2.1 channels did not form such clusters, but were diffusely distributed (Fig. 2B). Next, real-time 3D imaging of EGFP-Ca V 1.2 was performed before and during stimulation with either high K + (70 mM) or high glucose (20 mM), as well as after wash-out. The plasma membrane location of EGFP-Ca V 1.2 location was determined using the marker CellMask (Fig. 2C). In untreated cells kept at basal (5 mM) glucose, the ratio of EGFP-Ca V 1.2 fluorescence in the PM over that in the cell interior (Ratio PM/C) remained largely unchanged during the entire experiment ( Fig. 2D and E, upper panels). By contrast, when the cells were treated with high K + , the Ca V 1.2 clusters internalized and the PM/C ratio decreased by ,60% after 15 and 30 min ( Fig. 2D and E, as indicated). Similarly, stimulation with high glucose (20 mM) reduced the PM/C ratio of EGFP-Ca V 1.2 fluorescence by ,50% after 15 and 30 min ( Fig. 2D and E, as indicated). Interestingly, after wash-out, surface expression of Ca V 1.2 was partially recovered (Fig. 2D and E), which suggests that this is a reversible physiological reaction. Moreover, the same results were observed by TIRFM imaging ( Figure S2).

Ca V 1.2 Cluster Internalization can be Inhibited by Blocking the Endocytotic Pathway
The data presented so far demonstrate that internalization of Ca V 1.2 clusters in the plasma membrane occurs upon stimulation. Specifically, the results in Figure 1D and 1G suggest that Ca V 1.2 is translocated to early and recycling endosomes. We next performed experiments to address the signals and mechanisms involved in this endocytotic process.
Endocytosis occurs by either the clathrin-dependent or caveolindependent pathways [21]. First, caveolin could not be detected (Fig. 3G). Next, we performed immunostaining of Ca V 1.2 and clathrin in the presence of either 5 mM or 70 mM KCl (Fig. 3A). By contrast, clathrin was highly expressed and 33.763.26% of Ca V 1.2 colocalized with clathrin (n = 9). Interestingly, the colocalization increased to 69.962.9% after stimulation with high K + (n = 9). Furthermore, dynamin that cuts off clathrin-coated endocytotic pits [22], also increased its colocalization with Ca V 1.2 after stimulation with high K + (from 12.563.0% to 41.765.9%; n = 9 in each group; Fig. 3B). Ca V 1.2 also revealed increased colocalization with the intracellular cytoskeletal component tubulin after stimulation, which increased from 20.664.5% to 45.767.2% (n = 9 in both groups; Fig. 3C). Next, we disrupted the endocytotic process using pharmacological inhibitors and then monitored the dynamic Ca V 1.2 distribution in living cells. Interestingly, the dynamin inhibitor dynasore specifically blocked the stimulation-dependent Ca V 1.2 internalization (Fig. 3E), as compared to control (Fig. 3D), and the microtubule synthesis inhibitor vinblastine (1 mM) [23] had the same effect (Fig. 3F). Moreover, immunoprecipitation experiments confirmed that Ca V 1.2 interacts with clathrin rather than caveolin, which supports the involvement of clathrin-dependent endocytosis in Ca V 1.2 internalization (Fig. 3G). Finally, in Figure 3H whole-cell Ca 2+ current I-V relations demonstrated the failure of long-term glucose treatment to reduce the currents after treatment with dynasore or vinblastine. Taken together with the results in Figure 1 and 2, these results support the view that decreased surface expression of Ca V 1.2, and perhaps other Ca V channels, occurs by dynamin and microtubule-dependent endocytic uptake.

Involvement of eIF3e in Glucose-evoked Ca V 1.2 Internalization
Previous studies on activity-dependent Ca V channel trafficking in neurons using yeast two-hybrid screening suggested eIF3e as a regulatory molecule in this process [18]. We therefore investigated eIF3e Controls Intracellular Calcium Homeostasis PLOS ONE | www.plosone.org the possible involvement of eIF3e in glucose-evoked Ca V 1.2 internalization in the insulin-secreting cells. To this end, we first demonstrated high expression of eIF3e in primary rat pancreatic beta cells and INS-1 cells ( Fig. 4A and B). Interestingly, the eIF3e distribution was also stimulation-dependent and the eIF3e ratio of PM/C expression dropped in response to stimulation with glucose or high K + (from 1.1760.11 to 0.6960.14 or 0.660.07, respectively; n = 12 in each group; Fig. 4C). Furthermore this reduction of eIF3e expression on the cell surface was detected by TIRFM images ( Figure S3). Previous reports demonstrate that eIF3e is not crucial for global protein synthesis [15,24], but a role in protein translation cannot be excluded. To rule out that such an effect could explain our results we next used cycloheximide (CHX) to see whether general suppression of mRNA translation also interferes with Ca V 1.2 trafficking. However, the PM/C ratio was unchanged compared to untreated cells both under basal and stimulated conditions. Furthermore, 20 mM glucose treatment effectively internalized Ca V 1.2 also in CHX-treated cells (Fig. 4D). This result supports the view that the regulation of cellular Ca V 1.2 distribution by eIF3e is based on the direct interaction with Ca V 1.2 rather than by affecting Ca V 1.2 mRNA translation. This view was reinforced by the fact that the two proteins partially colocalized (Fig. 4B). Direct evidence of their interaction was provided by co-immunoprecipitation experiments in which antibodies to Ca V 1.2 specifically precipitated eIF3e (Fig. 4E).
To further address the role of eIF3e in Ca V 1.2 internalization, we manipulated eIF3e expression by RNA interference in INS-1 832/13 cells. siRNAs against eIF3e silenced eIF3e gene expression (mRNA) by 79.463.9% (Fig. 5A), which on the protein level amounted to a 72.563.5% decreased expression, without significantly affecting Ca V 1.2 expression ( Fig. 5B and C). However, the most striking finding was that silencing of eIF3e almost fully prevented glucose-stimulated internalization of Ca V 1.2 clusters and the PM/C ratio remained unchanged upon exposure to 20 mM glucose when explored by immunocytochemistry ( Fig. 5D and E). To better visualize the dynamics of Ca V 1.2 in the PM we next performed experiments using total internal reflection microscopy (TIRFM; Fig. 5F, G and H). This imaging technique selectively visualizes fluorescent molecules within ,100 nm distance from the PM. First, these experiments showed that eIF3e silencing resulted in a slight, but non-significant, reduction in Ca V 1.2 cluster number at the PM. Second they confirmed the capacity of glucose stimulation to reduce Ca V 1.2 cluster number at the PM. Third, they demonstrated that glucose-induced internalization of Ca V 1.2 clusters at the PM effect was prevented by silencing of eIF3e. Interestingly, the whole-cell Ca 2+ currentvoltage relations (I-Vs) in Figure 5I show the failure of long-term glucose treatment to reduce voltage-stimulated Ca 2+ influx in eIF3e KD cells. These results underscore the physiological role of eIF3e to control surface expression of Ca V channels such as Ca V 1.2.

eIF3e Assists Cav1.2 Trafficking to the Plasma Membrane
Since the number of Ca V 1.2 clusters are affected by bi-directly intracellular trafficking of Ca V 1.2 we therefore investigated whether eIF3e is involved in the regulation of Ca V 1.2 trafficking to PM. To detect movement of Ca V 1.2 clusters to the PM, we designed a modified Fluorescence Recovery After Photobleaching (FRAP) protocol to track single cluster movements (Fig. 6A). After photobleaching, two possibilities for Ca V 1.2 fluorescence recovery in the bleached area exist: first, the possibility of lateral diffusion in the X-Y plane, implies that fluorescence recovery should first appear along the edges. The other option is that cluster replenishment occurs from the cytosol along the Z-axis, which means that the fluorescence increase should appear in a central spot. We obtained strong evidence for cluster recovery both by lateral diffusion, as well as from the cell interior (Fig. 6A). The latter type of events was selected to further study fluorescence recovery over time, which was fitted to an exponential function (Fig. 6B). This analysis clearly showed that the fluorescence recovery was faster after stimulation with 70 mM K + for 30 min, as demonstrated by the 4-fold increased velocity of recovery when compared to that observed at normal K + concentration (from 1.2560.29 s 21 to 4.8262.5 s 21 ; n = 27 events under each condition) (Fig. 6C). However, it is important to note that the amount of recovery was only half of that observed in unstimulated control cells (Fig. 6D). This suggests that depolarization increases the velocity of Ca V 1.2 recovery to the PM but also limits the amount of Ca V 1.2 clusters available for replenishment (recovery fluorescence intensity decreased from 35.560.09% to 14.960.07%; n = 27 events under each condition). Silencing of eIF3e slowed Ca V 1.2 trafficking and the velocity of recovery decreased under both resting (0.6460.16 s 21 ; n = 27) and activated conditions (1.6060.53 s 21 ; n = 27), respectively (Fig. 6C). The finding that eIF3e silencing decreases the speed and absolute amount of Ca V 1.2 trafficking to the PM, implies that eIF3e is a factor that assists Ca V 1.2 transport, whilst also controlling the size of the pool of Ca V 1.2 channel clusters available for recovery.

eIF3e Prevents Ca 2+ Overload
The finding that Ca V 1.2 clusters in eIF3e-silenced cells are not internalized suggests that this condition could potentially be associated with Ca 2+ overload during long-term stimulation. To explore this possibility, we used the low-affinity Ca 2+ dye Fluo-5F for confocal Ca 2+ imaging in both control and eIF3e-silenced cells (Fig. 7A). Upon stimulation with 70 mM K + an initial peak in intracellular Ca 2+ ([Ca 2+ ] i ) was followed by a rapid decline reaching a plateau after ,50 s (Fig. 7B). The initial [Ca 2+ ] i peak was higher in control cells treated with inactive siRNA, but from ,3 min after onset of the stimulation, the time integral of the Ca 2+ signal in eIF3e knock-down cells exceeded that observed in control cells ( Fig. 7B; inset). These results demonstrate that during long-  term stimulation, the total Ca 2+ influx in eIF3e knock-down cells is significantly elevated compared to control-treated cells (729693 vs 474646 AU*s, 27 cells in each group) (Fig. 7C). To determine whether the increase of Ca 2+ influx specifically reflects the localization of L-type channels, including Ca V 1.2, we used pharmacological channel inhibitors to detect the Ca V subtypedependence of the effect of eIF3e silencing (Fig. 7D and E). In cells stimulated by high K + , eIF3e siRNA caused long-term Ca 2+ influx and significantly raised the Ca 2+ load (from 517645 to 8556121 AU*s, 18 cells in each group). Interestingly, the L-type calcium channel blocker isradipine counteracted the increase of Ca 2+ load in eIF3e-silenced cells (342619 vs 333632 AU*s; n = 18 in each group). However, neither the R-type calcium channel blocker SNX-482, or the N-type calcium channel blocker w-cototoxin GVIA were able to prevent the increased Ca 2+ load caused by eIF3e silencing (Fig. 7D and E). We also measured whole-cell Ca 2+ current-voltage relations (I-Vs) to investigate the effects of the blockers on the capacity of long-term glucose treatment to suppress Ca 2+ influx (Fig. 7F, left). We first observed that the Ltype blocker isradipine overall decreased whole-cell Ca 2+ currents by ,50%. Interestingly, long-term glucose treatment now failed to further reduce Ca 2+ influx (Fig. 7F, middle). Secondly, a cocktail of non-L type Ca V channel blockers (R-channel blocker SNX-482, P/Q-channel blocker v-agatoxin IVA and N-channel blocker vcototoxin GVIA) reduced whole-cell Ca 2+ currents to a similar extent as isradipine. Important to note is the preserved capacity of glucose treatment to further reduce Ca 2+ influx under these conditions (Fig. 7F, right). These results support the view that in insulin-secreting cells L-type Ca V channels, e.g. Ca V 1.2, form the most important Ca V channel population in activity-dependent internalization.

Regulation of Ca V 1.2 Cluster Surface Expression
The present data suggest that L-type Ca V channel internalization, as exemplified by Ca V 1.2, is a physiological phenomenon, and one that is used as a mode of negative feedback upon stimulation by e.g. glucose. Ca V 1.2 cluster surface expression was significantly lowered after glucose stimulation for 15 min or longer ( Figs. 1 and 2). These results add another facet to the physiological function of glucose in b-cells. We wish to emphasize that although this study details the mechanisms whereby Ca V 1.2 is internalized under stimulatory conditions, we wish not to exclude the participation of other L-type Ca V channels in the same process. For example, Ca V 1.3 is likewise highly expressed in insulinsecreting cells (although functionality appears to be speciesdependent) and may very well contribute to the increased intracellular Ca 2+ load observed after interfering with Ca V trafficking. With present L-type blockers it is not possible to discriminate between different L-type channel species. Moreover, in neurons, also non-L-type Ca V channels appear to be subject to stimulation-dependent internalization [18]. However, in insulinsecreting cells this effect is difficult to discern, which perhaps is a consequence of their lower relative expression. Be that as it may, the ability to internalize specific Ca V channel subtypes offers a dynamic and rapid mode of regulating the physiology of the b-cell and could, for example, explain the shift from L-to R-type Ca V channel dependence during phasic insulin secretion in mouse [8].
The activity-dependent regulation of Ca V channel surface expression appears to be mediated by Ca 2+ -dependent signals, since stimulation with high K + has effects similar to glucose. However, many details remain unresolved, such as the fate of the internalized clusters. In general, two main possibilities exist, the first being that Ca V 1.2 is dispersed into single molecules and degraded by ubiquitin-dependent proteases through endoplasmic reticulum-associated protein degradation [25]. It is worthy of note that our unpublished experiments using the protease inhibitor Mg132 suggest that this is not the fate of the majority of internalized Ca V 1.2 clusters. The other possibility is that Ca V 1.2 clusters are internalized by the clathrin-dependent endocytotic pathway, which is expected to recycle the Ca V 1.2 channels via this standby pool of inactive Ca V 1.2 channels, back to their active state in the PM in response to environmental cues [26]. This alternative is supported by the observed interaction with clathrin and the early endosome marker EEA-1 ( Fig. 1D and 3A), as well as the experiments using inhibitors of microtubules or dynamin ( Fig. 3E and F). Taken together with the recovery of Cav1.2 expression in the PM (Fig. 2) after wash-out, these results collectively suggest that Ca V 1.2 is subject to dynamic regulation of its expression in the PM.

Ca V 1.2 Trafficking, Ca 2+ Homeostasis and b-cell Function
Like the Ca V 1.2 clusters, eIF3e internalizes in an activitydependent fashion. The physical interaction of eIF3e and Ca V 1.2 is evident from immunoprecipitation experiments (Fig. 4E), and probably occurs via a binding domain in the II-III loop of the Ca V 1.2 alpha subunit [18]. Furthermore, silencing of eIF3e prevents glucose-dependent Ca V 1.2 cluster internalization (Fig. 5D). These results collectively provide strong evidence for the physical and functional connection between Ca V 1.2 and eIF3e.
eIF3e is a regulatory subunit in the protein translation initiation complex. Therefore its silencing could be envisioned to affect Ca V 1.2 subcellular expression merely as a secondary effect to this fundamental function. However, it should be noted that eIF3e is not a required subunit in the complex and its silencing will not prevent protein synthesis [15,24]. Accordingly, silencing of eIF3e only resulted in a relatively modest reduction of the number of Ca V 1.2 clusters. Furthermore, the fact that insulin secretion was largely unaffected in eIF3e-silenced cells (data not shown) adds further impetus to the idea of a specific regulatory action exerted by eIF3e. Finally, activity-dependent Ca V 1.2 internalization (and its reversal) was totally unaffected in cells treated with the protein synthesis inhibitor cycloheximide (Fig. 4D). These data taken together strongly support the idea that eIF3e serves as a physiological modulator in the b-cell, controlling vesicular transport of Ca V 1.2 clusters, and possibly other L-type Ca V channels, to the PM and their internalization under periods of intense stimulation. Glucose-dependent insulin secretion involves activation of Ca V channels and requires substantial increases in local [Ca 2+ ] i in the nanodomains in which Ca V channels and insulin granules aggregate prior to release [27]. However, during marker dye, CellMaskH (Invitrogen). The region between the two white lines (0.5 mm distance) was defined as the surface region used for further analysis. D) Live 3D imaging of Ca V 1.2 cluster expression in the cells. The pseudocolored clusters in the upper (red), middle (green) and bottom (blue) parts of the cells are shown in unstimulated control cells or 0, 15 and 30 min after stimulation with 20 mM glucose or 70 mM KCl respectively. The 30min recovery experiments were performed by wash-out of the stimulation buffer. The color bar indicates real distance along the z-axis. E) The histograms show the average ratio PM/C in control cells (n = 16), or cells exposed to 70 mM KCl (n = 16) or 20 mM glucose (n = 12). * p,0.05, ** p,0.01 (ANOVA, F-test). doi:10.1371/journal.pone.0064462.g002 eIF3e Controls Intracellular Calcium Homeostasis PLOS ONE | www.plosone.org protracted stimulation spillover from the nanodomains could lead to alarmingly high global cellular [Ca 2+ ] i that could trigger excitotoxic effects and eventually apoptosis [5]. Therefore a mechanism whereby Ca V channel surface expression can be adjusted during intense activity offers an important cellular cut-off reaction for limiting Ca 2+ influx. In conclusion, the ability of the bcell to regulate cell surface expression of L-type channels such as Ca V 1.2 will have consequences for b-cell Ca 2+ homeostasis and further work will determine its relevance for progression of type 2diabetes.

Cell Culture and Transfection
Clonal INS-1 832/13 cells were cultured as previously described [28]. 0.5 mg N-terminally EGFP-tagged Ca V 1.2 and Ca V 2.1 [29,30] were transiently transfected using 1.5 ml Lipofectamine 2000 (Invitrogen, CA, USA) per well in 24-well plate. After 24 h transfection, the cells were re-seeded on a 0.175 mm thickness glass and followed up 12 h culture until images acquirement.

RNA Interference
INS-1 832/13 cells were seeded 1 day prior to transfection. 30 nM eIF3e RNA interference oligonucleotides (Applied Biosystems) or 30 nM control #1 (Applied Biosystems, USA) were used to silence eIF3e. The siRNA was transfected by using Dhamafect Kit (Thermo Scientific, USA). The transfection efficiency was estimated by BLOCK-It AlexaFluorRed (Invitrogen, CA, USA) which stains the transfected cells. After 48 h transfection, the cells were collected and total RNA was extracted by using a RNA extraction kit (Qiagen, Germany). 1 mg RNA was used for RT-PCR and real time PCR. Primers of eIF3e and housekeeping gene HPRT1 (Applied Biosystems, USA) which tagged FAM dyes were used for amplification detection.

Immunostaining
Cells were first washed twice and fixed with 3% PFA-PIPES and 3% PFA-Na 2 BO 4 for 5 min and 10 min respectively, followed by permeabilization with 0.1% Triton-X 100 for 30 min. The blocking solution contained 5% normal donkey serum in PBS and was used for 15 min. Primary antibodies against Ca V 1.2 (Sigma, USA), EEA-1 (BD transduction lab, USA), Na + /K + ATPase (Millipore, USA), Rab11 (BD transduction lab, USA), clathrin (BD transduction lab, USA), Dynamin (Millipore, USA), Tubulin (Sigma, USA), eIF3e (Santa Cruz, CA) were diluted in blocking solution and incubated overnight at 4uC. Immunoreactivity was done using fluorescently labeled secondary antibodies (1:200) and visualized by confocal microscopy (Carl Zeiss, Germany). Colocalization analysis was performed by using a ZEN2009 software based on Pearson's correlation coefficient analysis which recognizes the colocalized pair by comparison pixel by pixel intensity [31]. The internalization was indicated by ratio that is defined by mean intensity of plasma membrane to mean intensity in cytosol, according to the formula: Ratio = .
Where i 1 , i 2 and i 3 represent the intensities of whole cell, cytosol and nucleus, a 1, a 2 and a 3 represent the area of whole cell, cytosol and nucleus respectively. The specificities of Ca V 1.2 and eIF3e antibodies were validated by using synthesized peptides which totally blocked the signals. Others antibody stainings were performed following protocols provided by the vendor.

Electrophysiology
Whole-cell Ca 2+ currents were measured as described previously [28]. The extracellular solution (118 mM NaCl, 20 mM tetraethylammonium chloride, 5.6 mM KCl, 2.6 mM CaCl 2 , 1.2 mM MgCl 2 and 5 mM HEPES) was supplemented with 5 or 20 mM glucose as indicated. The pipette (intracellular) solution contained 125 mM Cs-glutamate, 10 mM CsCl, 10 mM NaCl, 1 mM MgCl 2 , 5 mM HEPES, 3 mM Mg-ATP, 0.1 mM cAMP and 0.05 mM EGTA (pH 7.2 with CsOH). L-type blocker isradipine (5 mM), R-channel blocker SNX-482 (200 nM), P/Qchannel blocker v-agatoxin IVA (100 nM) and N-channel blocker v-cototoxin GVIA (50 nM) were added as indicated in text or figures. Data was recorded on a HEKA EPC9 patch clamp amplifier with the Pulse Fit 8.64 software. The whole-cell configuration was used in voltage-clamp mode and pipettes had an average resistance of <5.5 MV. All the experiments were performed in bath-heated perfusion system which controls output temperature to 32uC.

Subcellular Fractionation
Subcellular fractionation of INS-1 832/13 cells was done as described previously [32]. Briefly, cells were scraped into 15 ml ice cold homogenization medium (HM; 250 mM sucrose, 5 mM HEPES, 0.5 mM EGTA, 0.2 mM PEFA Block and adjusted to pH 7.4 with KOH) and disrupted using a nitrogen bomb (350 psi). The homogenate was centrifuged at 7006g for 15 min in 4uC and postnuclear supernatant was separated. 15% (v/v) Percoll and 250 mM sucrose were added into the mix and super-centrifuged at 48,0006g for 25 min at 4uC. Two opaque bands were obtained at the top and bottom corresponding to the plasma membrane and vesicles, respectively. The fractions were sonicated and total protein concentration was measured using BCA protein assay kit (Pierce, IL, USA).

Live Cell Imaging
Cells seeded onto cover slips and mounted in the experimental chamber were perifused and temperature-controlled during the entire experiment. Confocal images were acquired using a Zeiss 510 Meta LSM and a 640 water immersion objective (NA = 1.2). EGFP-Ca V 1.2 was visualized by excitation at 488 nm and collected using a 500-530 nm bandpass filter. The pinhole was ,1 airy unit and the scanning frame was 5126512 pixels. Colocalization analysis was evaluated by Pearson's correlation coefficient analysis using the ZEN 2009 software (Zeiss, Germany). For z-stack acquisition, we used a low power laser output (,2%) and stack interval at 0.5 mm. The middle sections in a cell were selected for ratio (PM/C) analysis in which the area of the plasma membrane (PM) was defined using the PM marker CellMask (Invitrogen, USA). 3D reconstruction was performed by ZEN2009. For Fluorescence Recovery After Photobleaching (FRAP) experiments, we first adjusted the focus on the cell-cover slip interface and carefully shifted focus to the cell surface. Image acquisition in EGFP-Ca V 1.2 transfected cells was done using a minimum pinhole (section thickness 0.6 mm) to observe single Ca V 1.2 clusters in Ca V 1.2 overexpressing cells. Next, photobleaching performed by full argon laser outputs, as determined in the region of interests (ROIs) in the center of the visual fields. Finally the recovered fluorescent intensity was analyzed using ZEM 2009 and exponential fitting was performed using the Igor software (Portland, USA).

Ca 2+ Imaging
Fluo-5F (K d = 2.3 mM) (Invitrogen, USA) was used for measuring intracellular Ca 2+ . Cells were loaded using 1 mM Fluo-5F in room temperature for 30 min. The cells first were perfused in Krebs-Ringer bicarbonate (KRB) buffer containing 116 mM NaCl, 4.7 mM KCl, 2.6 mM CaCl 2 , 1.2 mM KH 2 PO 4 , 1.2 mM MgSO 4 , 20 mM NaHCO 3 , 16 mM HEPES, 2 mg/ml BSA, and supplemented with 5 mM glucose. Stimulation was carried out by a 70 mM KCl KRB buffer at room temperature. Images were acquired by confocal microscopy using a 406 water immersion objective.

INS-1 cells were seeded on glass bottom
MatTek dishes, and stainined for Ca V 1.2 by the normal immunocytochemistry protocol detailed above. TIRFM images were acquired by a high-aperture 1006 objective lens in an inverted epifluorescence microscope (Carl Zeiss). Before image acquisition, the penetration

Statistical Analysis
The data were shown present as averages6S.E.M. Evaluation of statistical significance was done by Student's T-test for experiments allowing paired comparisons, or by one-way analysis of variance (ANOVA) with the Friedman test for multiple comparisons.  Figure 7. eIF3e silencing disrupts intracellular Ca 2+ homeostasis. A) Images of intracellular Ca 2+ measured by the low affinity calcium dye, Fluo-5F upon 70 mM K + stimulation in control and eIF3e KD cells. B) Fluo-5F fluorescence signal ratio (Ratio) and its time integral (A.U.C.) were calculated to assess the Ca 2+ influx. Note that after 180-s stimulation the integrated Ca 2+ load in eIF3e KD cells exceeds that of the negative controltreated cells. The arrow indicates the onset of stimulation. C) Comparison of Ca 2+ load, expressed as the Area Under the Curve (A.U.C) 0-200 s after stimulation, between eIF3e KD and control cells (n = 27 cells, ***, P,0.001; Student's t-test). D) As in (a), but experiments performed in the presence of vehicle (DMSO), the L-type calcium channel inhibitor isradipine (5 mM), the R-type calcium channel inhibitor SNX482 (0.2 mM) or the N-type calcium channel inhibitor v-conotoxin GVIA (50 nM). Note that isradipine counteracts the effects of eIF3e KD on the sustained rise in cytosolic [Ca 2+ ]. E) Quantitative analysis of Ca 2+ load as in C for the conditions in D; (n = 18 cells for each condition, **, P,0.01; ***, P,0.001; Student's t-test). F) Currentvoltage (I-V) relations for whole-cell Ca 2+ currents after 30 min incubation in 5 or 20 mM glucose (left); same experiment but in the presence of 5 mM isradipine (middle); or a non-L-type blocker cocktail (0.1 mM v-agatoxin, 0.2 mM SNX-482 and 50 nM v-conotoxin GVIA; right). n = 11 or more for each condition. * P,0.05, ** P,0.01, *** P,0.001 (Student's t-test). doi:10.1371/journal.pone.0064462.g007