Reticulon 4 Is Necessary for Endoplasmic Reticulum Tubulation, STIM1-Orai1 Coupling, and Store-operated Calcium Entry

Background: Store-operated calcium entry requires the redistribution of ER localized STIM1 to ER-PM junctions. Results: Altering ER morphology by modulating reticulon expression affects STIM1 redistribution and consequently calcium entry. Conclusion: Reticulon shaping of the ER is critical for store-operated calcium entry. Significance: Understanding how membrane shaping proteins participate in the regulation of organelle function is essential in elucidating the structure-function interdependence of organelles. Despite recent advances in understanding store-operated calcium entry (SOCE) regulation, the fundamental question of how ER morphology affects this process remains unanswered. Here we show that the loss of RTN4, is sufficient to alter ER morphology and severely compromise SOCE. Mechanistically, we show this to be the result of defective STIM1-Orai1 coupling because of loss of ER tubulation and redistribution of STIM1 to ER sheets. As a functional consequence, RTN4-depleted cells fail to sustain elevated cytoplasmic Ca2+ levels via SOCE and therefor are less susceptible to Ca2+ overload induced apoptosis. Thus, for the first time, our results show a direct correlation between ER morphology and SOCE and highlight the importance of RTN4 in cellular Ca2+ homeostasis.

The endoplasmic reticulum (ER) 2 is a structurally and functionally diverse organelle. It is composed of morphologically distinct domains with prominent functional specializations (reviewed in Refs. 1,2). The most noticeable of these domains are the nuclear envelope, ER sheets and ER tubules. ER sheets are relatively flat structures that are densely packed with polyribosomes, offer a large volume-to-surface ratio, and are enriched in proteins involved in the synthesis, folding, and sorting of integral membrane and secreted proteins (3). Accordingly, these structures are abundant in professional secretory cells, such as plasma cells and pancreatic cells. ER tubules on the other hand are long cylindrical structures with high membrane curvature that form an interlinked network via three-way junctions (4). The surface-to-volume ratio (5), polyribosome density (5,6), and molecular chaperones levels of ER tubules are lower than that of ER sheets (3). Hence, in contrast to ER sheets, ER tubules are often associated with lipid synthesis and calcium (Ca 2ϩ ) signaling, on the premise that they are enriched in cell types specialized in those functions. For example, muscle cells, specialized in Ca 2ϩ signaling, and adrenal cortex cells, that produce high volumes of steroids, have extensive tubular ER networks. Such cell-type specific enrichment of ER domains conveys how the morphology of the ER in a given cell type depends on which of its functions predominates and reinforces the theorem of functional specialization of ER domains. ER configuration can also have simpler justifications. For instance, a predominantly tubular ER configuration can be advantageous as it enables the extensive and uniform distribution of ER membranes throughout the cytoplasm without interfering with the diffusion and trafficking of cytoplasmic elements. Furthermore, diffusion of proteins to a target becomes more efficient as dimensionality is reduced (7). Therefore ER tubules (1D) are intrinsically more advantageous than sheets (2D) for processes relying on targeted diffusion, such as store-operated Ca 2ϩ entry (SOCE).
Ca 2ϩ depletion from the ER prompts the influx of extracellular Ca 2ϩ via SOCE; a process essential to sustain prolonged Ca 2ϩ transients and refill internal stores. The core molecular components of SOCE have recently been identified and include the ER-resident Ca 2ϩ sensor STIM1 (8,9) and the plasma membrane (PM)-resident Ca 2ϩ channel Orai1 (10 -12). The processes underlying SOCE activation require the structural rearrangement and dynamic redistribution of ER and PM components and their interaction at ER-PM junctions (reviewed in Ref. 13). Briefly, under replete conditions Ca 2ϩ binds and keeps dimeric STIM1 in a compact, inactive conformation. ER depletion alters STIM1 conformation, inherent to Ca 2ϩ dissociation, resulting in oligomerization of STIM1 and its activation via extension of its cytosolic domain. This domain harbors the Orai1 interacting domain (known variably as CAD (14), SOAR (15) or OASF (16)) and a C-terminal lysine-rich repeat. In the active, extended conformation, the lysine-rich repeat, through interaction with PM lipids, concentrates oligomerized STIM at ER-PM junctions. At ER-PM junctions, extended STIM1 oligomers can bridge the junctional gap and interact with and activate Orai1 channels, leading to Ca 2ϩ influx. STIM1-Orai1 complexes are disassembled, and the system is reset once Ca 2ϩ transients cease and internal stores are replenished. Concurrently to these mechanistic insights, other studies have demonstrated the paramount importance of SOCE in numerous cell processes, including apoptosis (reviewed in Ref. 17). Despite such advancements in the field, the role of ER structure in SOCE has received little attention.
The structure of the ER is established through the use of a wide array of proteins involved in ER membrane fusion, anchorage, and sculpting (reviewed in Refs. 1,2,18). The most studied and best characterized of these are the ubiquitously expressed ER-sculpting proteins of the reticulon (RTN) family. Mammals have four RTN paralogs (RTN1-4) that each generates multiple isoforms (19). The N-terminal region of these isoforms is highly divergent and is associated with biological functions believed to be initiated from the cell surface (20,21). RTNs, however, almost exclusively localize to the ER (21,22). Unlike most ER-resident proteins, they do not possess a canonical ER localization signal, and their ER retention seems to be due to structural elements in their C-terminal region (22,23). This C-terminal region, called the reticulon homology domain (RHD), is evolutionarily conserved and common to all RTN family members (19). The RHD consists of 150 -200 amino acid residues and contains a pair of hydrophobic regions separated by a hydrophilic loop and followed by a short C-terminal tail. The hydrophobic segments are longer than typical single pass transmembrane domains and are believed to form hairpins (24). These hairpins, too short to completely span the ER membrane, displace more of the outer leaflet than the inner leaflet. Additionally, RTNs have also been shown to form arc-shaped oligomers (25). These "wedging" and "scaffolding" mechanisms bestow RTNs with membrane curvature generating/stabilizing properties. This is best demonstrated in in vitro reconstitution experiments, using yeast homologues, where it was demonstrated that RTNs are capable of transforming proteoliposomes into tubules (26). Accordingly, in vivo, these proteins are necessary and sufficient for sculpting ER elements with high membrane curvature, such as ER tubules (24,27). Therefore, overexpression of RTNs leads to the enrichment of ER tubules at the expense of ER sheets, while the converse is true upon depletion of RTNs (24).
Coincidental evidence from recent studies collectively suggests that changes in ER morphology may play a role in Ca 2ϩ signaling. Studies investigating the role of the cytoskeleton in SOCE reported a decrease in Ca 2ϩ influx following disruption of the microtubule cytoskeleton (28,29). Interestingly, microtubule depolymerization has also been reported to lead to the collapse of the ER network and conversion of peripheral ER tubules to extended ER sheets (30). Additionally, SOCE is also disrupted during cell division (29,31,32), when similar changes in ER structure are also known to take place (33,34). Despite such examples of association and tremendous interest in the field, insufficient understanding of ER morphogenesis has so far limited efforts to directly test the long suspected importance of ER structure in SOCE. This limitation has recently been allevi-ated with the identification and functional characterization of ER shaping proteins. These findings offer the opportunity to use these proteins as a novel tool to modulate ER morphology and directly assess the interdependence of ER structure and SOCE. Here we demonstrate in a simple cell system that the loss of RTN4 results in ER morphological changes that selectively disrupt certain structure-sensitive ER processes such as SOCE and highlight the importance of RTN4 in Ca 2ϩ homeostasis.
Plasmids and Cloning-GFP-Sec61b (plasmid 15108), STIM1-YFP (plasmid 19754), Orai1-Myc (plasmid 21638), Orai1-YFP (plasmid 19756), and Orai1-CFP (plasmid 19757) were obtained from Addgene. mCherry-STIM1 was kindly provided by Dr Richard S Lewis (Stanford University, Palo Alto, CA). To generate GFP-RTN4b and GFP-RHD, RTN4b and RHD were amplified by PCR from pcDNA3 (20). The PCR reaction also added an HA tag to the C terminus. The PCR products were cloned into pEGFP-C1 (Clontech) using EcoRI and BamHI. TM2mut was generated using GFP-RTN4b as a template and according to Ref. 35. Initially, a small C-terminal fragment of RTN4b was PCR amplified to insert the extra nucleotides into TM2. This fragment was then used as a reverse primer to generate full-length TM2mut, which was then cloned into pEGFP-C1 using EcoRI and BamHI. To generate mCherrytagged constructs, eGFP in pEGFP-N3 (Clontech) was substituted by mCherry to generate pmCherry-N3. mCherry was PCR amplified from mCherry-Myc-KDEL kindly provided by Dr. Pietro De Camilli (Yale University, New Haven, CT). To generate Orai1-mCherry, Orai1 was amplified by PCR from Orai1-Myc and cloned into pmCherry-N3, using EcoRI and KpnI. STIM1 fragments CAD and D7 were generated by PCR based on Ref. 14 and cloned into pmCherry-N3 to generate CAD-mCherry and D7-mCherry using BglII and XmaI. Plasmids for lentiviral transduction were generated by cloning the appropriate RTN4b related PCR product into pHR-SIN-SFFV using BamHI and XhoI.
Cell Culture, Transfection, and Transduction-WT and RTN4-KO MEFs were isolated from C57Bl6 and congenic RTN 4a/b KO mouse embryos (36,37) and immortalized by serial passaging. Briefly, embryos were harvested at E13.5 and primary MEFs were generated following decapitation and evisceration of embryos and trypsinization of the bodies after mincing. The resulting slurry was plated in 75-cm 2 flasks. Cells were maintained in DMEM (Invitrogen) supplemented with 10% FBS and 2 mM glutamine in the presence of 100 units/ml penicillin/ streptomycin. Cells were passaged and plated at 1 ϫ 10 6 cells in 100-mm dishes every 3 days. After crisis, cells were expanded and used for experiments as pooled immortalized clones. For transient expression, constructs were transfected into cells for 24 h using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. To establish RTN4-KO MEFs stably expressing RTN4b, RHD, and TM2mut, lentiviruses were packaged by co-transfection of the appropriate constructs with the 3rd generation packaging plasmids in HEK293T cells in 100-mm dishes with Lipofectamine 2000. Medium was replaced the next day with OptiMEM (Invitrogen), viral particles were collected for 24 h, centrifuged at 200 ϫ g for 5 min, passed through a 0.45-m filter, and stored at 4°C until used. Ready to use RFP lentiviral particles were purchased from AMS Biotechnology. Cells were transduced with the viral suspensions in the presence of 8 mg/ml Polybrene (Sigma-Aldrich) in 6-well plates. Following 24 h, cells were passaged to 100-mm dishes and allowed to reach confluency. Transgene expression was confirmed by Western blot analysis and immunofluorescence by blotting against HA tag, and cell lines with physiologic levels were kept.
Western Blot Analysis-For immunoblot assays, cells were harvested in ice-cold lysis buffer (50 mM Tris-HCl pH 7.4, 0.1% SDS, 0.1% sodium deoxycholate, 0.1 mM EDTA, 0.1 mM EGTA, 1% Nonidet P-40, 1.5 mg/ml protease inhibitor mixture (Roche), 0.25 mg/ml AEBSF (Roche). Proteins were extracted for 45 min at 4°C, resolved by SDS-PAGE, and transferred onto nitrocellulose membranes for 2 h at 100 V. Membranes were blocked using 0.1% (w/v) casein (Bio-Rad) in TBS for 1 h and incubated with primary antibodies overnight at 4°C. Next, membranes were washed with TBS-T (TBS containing 0.1% Tween 20) and incubated with conjugated secondary antibodies for 1 h at room temperature. Finally, membranes were washed with TBS-T and developed using the Odyssey system (Li-Cor) and further analyzed using ImageJ software (NIH).
Immunofluorescence Microscopy Studies-For the detection of endogenous RTN4a/b, cells were grown on glass coverslips, fixed, and permeabilized in 100% methanol at Ϫ20°C for 10 min and blocked with 1% BSA (Sigma). Samples were incubated with the primary antibody overnight at 4°C, and then incubated with the secondary antibody for 1 h at room temperature. Washes were performed using PBS. Nuclei were stained with DAPI (4,6-diamidino-2-phenylindole; Sigma). Coverslips with stained cells were washed and mounted onto glass slides in Vectashield medium (Vector Laboratories) for microscopy. Images were acquired using a Leica TCS SP5 confocal laser-scanning microscope with a ϫ63/1.49 NA oil-immersion objective. The quantification of ER sheets and the detection of exogenous proteins were carried out using live-cell imaging on 35 mm glass-bottom dishes (MatTek). For the quantification of ER sheets, cells were stained with ER tracker and more than 50 cells per group were blindly analyzed from two independent experiments. Image analysis was performed with ImageJ software (NIH) and expressed as percentage of cytosol (total cell area minus the nucleus). For the detection of exogenous proteins, following 16 to 24 h transfection, culture medium was replaced with HEPES buffer, and cells were imaged at room temperature for no longer than an hour. For the quantification of endogenous STIM1 clusters a threshold was applied to collected images in ImageJ and only data from pixels above the threshold were quantified. The threshold was set at the minimum pixel intensity ϩ 10% of the range of pixel intensities of the untreated WT image (threshold ϭ min ϩ 10%[max Ϫ min]) and was used as the global threshold for the rest of the images. Cells were outlined in thresholded images before conversion to binary images. Cell areas in binary images were analyzed using the analyze particles tool to collect area values for STIM1 signal and displayed as frequency distribution.
Electron Microscopy-For transmission electron microscopy, cells were grown to 90% confluence in a 100-mm dish. For the depletion of internal Ca 2ϩ stores, cells were washed and bathed in Ca 2ϩ -free buffer followed by treatment with TG for 15 min at room temperature. Cells were then fixed in 2.5% glutaraldehyde/2% paraformaldehyde/2% sucrose in 0.1 M sodium cacodylate buffer, pH 7.4, for 30 min at room temperature followed by another 30 min at 4°C. Subsequently, cells were rinsed, postfixed in 1% osmium tetroxide for 1 h at room temperature, stained in 2% uranyl acetate in maleate buffer, pH 5.2, for a further hour then rinsed again. Finally, cells were dehydrated in ethanol, infiltrated with Epon resin and baked overnight at 60°C. Blocks were cut using a Leica UltraCut UCT. 60-nm sections were stained using 2% uranyl acetate and lead citrate and viewed using a FEI Tencai Biotwin transmission electron microscope at 80 kV and spot size of 1. Images were taken using a Morada CCD and iTEM software (Olympus). Quantification of ER elements was carried out on 10 randomly selected cell sections for each group. ER tubules diameter was estimated from 40 -45 measurements collected from 10 randomly selected cell sections from each group.
Ca 2ϩ Measurements-For Ca 2ϩ measurements, MEFs were incubated in loading medium (HBSS, 20 mM HEPES) containing 4 M Fluo-4 (Invitrogen) for 30 min at 37°C and another 30 min at room temperature. Cells were then washed and bathed in HEPES-buffered saline solution (140 mM NaCl, 1.13 mM MgCl 2 , 4.7 mM KCl, 2 mM CaCl 2 , 10 mM D-glucose, and 10 mM HEPES, adjusted to pH 7.4 with NaOH) for 15 min before Ca 2ϩ measurements. In Ca 2ϩ -free solution, CaCl 2 was omitted, and MgCl 2 was increased to 2 mM. All experiments were conducted at room temperature. Images were acquired every 3 s for a period of 3.5 to 5 min using a Carl Zeiss scanning microscope Axiovert 200 M imaging system (100 -400) using 488-nm (excitation) and 540-nm (emission) filters. The data were expressed as the change in fluorescence relative to baseline (F/F 0 ) as a function of time. Relative mobilizations of Ca 2ϩ levels were estimated from the area under the curve (AUC) of the Ca 2ϩ traces. These values were obtained with ImageJ software (NIH) and further analyzed with Prism software (GraphPad Software).
Quantitative Real-time Polymerase Chain Reaction-Total RNA was extracted using Trizol (Invitrogen) and further purified using the RNeasy kit (Qiagen). Reverse transcription of 2 g of total RNA was carried out using the AMV First Strand cDNA synthesis kit (Roche) according to the manufacturer's instructions. Quantitative real-time PCR was carried out on an iCycler iQ detection system using the iQ SYBR green Supermix kit (Bio-Rad). Samples were assessed in triplicate and normalized to the expression level of ␤-actin RNA. Relative mRNA expression was calculated versus a reference sample. The following primers were used: for RTN1, Fw  Flow Cytometry Analysis-Flow cytometry analysis was carried out to detect apoptosis via annexin-V and PI double labeling using the Vybrant Apoptosis Assay Kit #2 (Molecular Probes) according to the manufacturer's instructions. In brief, cells cultured in 6-well plates were subjected to apoptotic stimuli for the appropriate time. After treatment, floating and adherent cells were harvested, washed with PBS, resuspended in annexin-V binding buffer (10 mM HEPES, 140 mM NaCl, and 2.5 mM CaCl 2 , pH 7.4), and stained with annexin-V and PI for 15 min at room temperature. Ten thousand events were acquired per sample using a FACSCalibur flow cytometer using CellQuest software (Becton Dickinson), and data were analyzed using FlowJo analysis software (Three Star). Negative and positive controls were used to adjust instrument settings.
Confocal Microscopy-For the detection of endogenous RTN4a/b, cells were grown on glass coverslips, fixed, and permeabilized in 100% methanol at Ϫ20°C for 10 min and blocked with 1% BSA (Sigma). Samples were incubated with the primary antibody overnight at 4°C, and then incubated with the secondary antibody for 1 h at room temperature. Washes were performed using PBS. Nuclei were stained with DAPI (Sigma). Coverslips with stained cells were washed and mounted onto glass slides in Vectashield medium (Vector Laboratories) for microscopy. Images were acquired as single XY planes using a Leica TCS SP5 confocal laser-scanning microscope with a ϫ63/ 1.49 NA oil-immersion objective. The quantification of ER sheets and the detection of exogenous proteins were carried out using live-cell imaging on 35-mm glass bottom dishes (Mat-Tek). For the quantification of ER sheets, cells were stained with ER tracker and more than 50 cells per group were blindly analyzed from two independent experiments. Image analysis was performed with ImageJ software (NIH) and expressed as percentage of cytosol (total cell area minus the nucleus). For the detection of exogenous proteins, following 16 to 24 h transfection, culture media was replaced with HEPES buffer, and cells were imaged at room temperature for no longer than an hour.
For the quantification of endogenous STIM1 clusters a threshold was applied to collected images in ImageJ, and only data from pixels above the threshold were quantified. The threshold was set at the minimum pixel intensity ϩ 10% of the range of pixel intensities of the untreated WT image (threshold ϭ min ϩ 10%[max Ϫ min]) and was used as the global threshold for the rest of the images. Cells were outlined in thresholded images before conversion to binary images. Cell areas in binary images were analyzed using the analyze particles tool to collect area values for STIM1 signal and displayed as frequency distribution.
FRET Imaging and Analysis-FRET analysis was done using a CFP/YFP pair on a Leica TCS SP5 confocal laser-scanning microscope with a ϫ63/1.49 NA oil-immersion objective and its FRET Acceptor Bleaching module. Emission bands for the fluorescence channels were as follows: CFP (left: 465 nm; right: 499 nm), YFP (left: 530 nm; right: 628 nm). Co-transfected cells grown on glass-bottom dishes were left untreated or treated with TG in Ca 2ϩ -free buffer to induce Orai1 channel assembly. Cells were subsequently fixed with 4% paraformaldehyde for 15 min at room temperature and preserved in 90% glycerol. Controls were always run in parallel with experimental samples. Photobleaching was carried out using a 514 nm laser beam at 80% intensity (90 milliwatt laser power at the specimen) for 6 steps at 200 Hz. The FRET efficiency (FRET eff ) was automatically calculated for the user defined regions of interest by the software according to the equation: FRET eff ϭ [D post Ϫ D pre ]/ D post . Readings with altered fluorescence profile maps in regions not exposed to bleaching were eliminated from analysis. FRET efficiency was acquired from 10 randomly selected cells per sample from three independent experiments using the same experimental parameters.
TIRF Microscopy and Analysis-TIRFM microscopy was carried out with an Olympus IX-70 microscope equipped with argon (488 nm) and argon/krypton (568 nm) laser lines, a TIRFM condenser (Olympus or custom condenser), a 60 ϫ 1.45 NA TIRF objective (Olympus), an iXon887 EMCCD camera (Andor Technologies), and controlled using Andor iQ software as described (38,39). All live cell microscopy was done at 37°C (using a custom incubation chamber) in HEPES-buffered saline solution (pH 7.4). Calibration of the evanescent field penetration depth was done as described above. Images were further analyzed using ImageJ software (NIH). This ensured that cells were in close contact with the substrate, analysis of mCherry-STIM1 clustering was carried out using 20 m 2 regions applied to sections of cells with visible STIM1 clusters following TG challenge. Regions where no STIM1 clusters were visible were not included in the final analysis. Quantification of mCherry-STIM1 puncta formation was automated using the find maxima tool in ImageJ to detect STIM1 puncta as local spots of high fluorescence intensity above a noise tolerance of 10. Accuracy of puncta quantification was verified by independent blind counting.
Triacylglycerol Accumulation, Lipid Extraction, and Thinlayer Chromatography-To build up triacylglycerol in cells, WT and RTN4-KO MEF were incubated overnight in DMEM with 0.5 mM oleic acid (Cayman) complexed to 1% fatty acid free BSA. Cells were then washed with PBS and lysed with hypotonic lysis butter (20 mM Tris pH 7.4, 20 mM NaF, and 50 M EDTA) followed by Dounce with loose tissue grinder (Wheaton). Equal amounts of total proteins were transferred to glass tubes. Lipids were extracted by adding 4 ml of chloroform/methanol (2:1 v:v) followed by addition of 750 l of water. The lipid phase (organic phase) was collected and evaporated under N 2 . Lipid residues were resolved in 50 l of chloroform and loaded onto silica gel HTLC plate (Whatmann). Separation of neutral lipids were performed by developing the plate in a neutral lipid solvent (heptane/isopropyl ether/acetic acid, 60:40:4, v:v:v) (65). Neutral lipids were visualized by Hanessian's stain.
Visualization of Lipid Droplets (LD)-Subconfluent cells grown on coverslips were loaded overnight with 0.5 mM oleic acid (Cayman) complexed to 1% fatty acid free BSA. Cells were then rinsed with PBS and fixed with 3.7% paraformaldehyde. Cells were stained with BODIPY 493/503 (Invitrogen) for LD and DAPI for nuclei. Coverslips were then mounted by mounting medium (Vectashield) and imaged by confocal microscopy.
Statistical Analysis-Data are expressed as the mean Ϯ S.E. unless otherwise noted. Statistical analysis was performed using Prism software (GraphPad Software). Statistical significance between groups was determined by Student's t test. Significance between multiple groups was determined by one-way ANOVA with Tukey's post hoc analysis. In every case, p Ͻ 0.05 was considered statistically significant.

RTN4-KO Cells Display Altered ER Morphology with Loss of ER Tubulation and Enrichment of ER Sheets-
To study the structure-function relationship of the ER, we established MEFs (mouse embryonic fibroblasts) from either wild-type (WT) C57Bl6 mice or C57Bl6 congenic mice lacking both Rtn4a and Rtn4b as previously described (36,37). Disruption of a common exon shared by these isoforms results in complete deletion of both Rtn 4a/b mRNAs (called Rtn4-KO throughout this report). These cells provide the ideal system for the following reasons. First, MEFs abundantly express the ϳ45 kDa RTN4b protein ( Fig. 1A) with very little RTN4a. This is consistent with non-neural cells such as endothelial, smooth muscle, and macrophages expressing higher levels of RTN4b compared with RTN4a (20,36). Second, the importance of this isoform is emphasized by the extensive, predominantly tubular, ER network present in these cells (Fig. 1B). Third, the loss of RTN4a/b does not lead to compensatory gene expression by other rtns, assessed by Q-PCR (Fig. 1C), and does not alter the overall cellular ER content, based on the expression of several ER luminal markers (BiP, calreticulin, and GRP94) and the ER integral membrane protein, calnexin (Fig. 1D). Finally, in contrast to previous reports (6,7), the loss of the RTN4a/b isoforms is sufficient to significantly alter ER morphology without the deletion of additional RTNs or other curvature-inducing proteins (Fig. 1E, right panels). The morphological changes are characterized by a reduction of ER tubulation and enrichment of persistent ER sheets (designated by arrowheads). These changes were visualized by confocal microscopy using ER-resident GFP-Sec61b (Fig. 1E, top panels) or, to circumvent the possibil-ity that overexpression of GFP-Sec61b alters ER structure, ER tracker, a cell-permeant dye highly selective for the ER (Fig. 1E, bottom panels). Quantitative analysis of confocal images of ER tracker revealed a ϳ67% increase in ER sheets in RTN4-KO compared with WT MEFs (Fig. 2A).
These findings were also corroborated by high-resolution transmission electron microscopy. Ultrastructural analysis of cross-sections of WT cells revealed a cytoplasm rich in short ER elements (ER profiles Ͻ1 m, highlighted in green), reminiscent of ER tubules (Fig. 2B, top panels). On the contrary, RTN4-KO cell cross-sections displayed an abundance of longer ER elements (ER profiles Ͼ1 m; highlighted in red) characteristic of ER sheets (Fig. 2B, bottom panels). The ER elements observed in RTN4-KO cells were also less constricted. The average ER tubule diameter reported in mammalian cells varies from 60 to 100 nm (2). The average diameter of ER elements observed in WT MEFs fell within this range at ϳ80 nm, whereas that of RTN4-KO MEFs was unusually high at ϳ110 nm (Fig.  2C). Taken together, these results indicate that this simple cell system, where ER morphology can be altered through the manipulation of a single protein, can be used as a novel tool to study the dependence of ER functions on morphology.

The Structural Alteration of the ER in RTN4-KO Cells Does
Not Lead to Generalized ER Dysfunction but Reduces SOCE-Next, we sought to investigate whether the loss of RTN4a/b and the associated structural changes of the ER would affect general ER functions. Assessment of triglyceride synthesis and lipid droplet formation in WT and RTN4-KO MEFs revealed no significant differences (Fig. 3, A-C). Likewise, no indication of ER stress was detected in RTN4-KO cells, suggesting proper functioning of the protein folding machinery. Upon pharmacological induction of ER stress, ER-to-nucleus signaling through the unfolded protein response (ATF6, IRE1, and PERK pathways; Fig. 4, A-F) was also unaltered in RTN4-KO when compared with WT MEFs. These results imply that although substantial rearrangement of the ER is incurred by the loss of RTN4a/b, it does not result in overt dysfunction of the organelle.
The loss of RTN4a/b did, however, alter ER-regulated Ca 2ϩ dynamics (Fig. 5). The generation of short and long Ca 2ϩ transients was assessed in physiological levels of extracellular Ca 2ϩ (2 mM), using PDGF (platelet-derived growth factor, 10 ng/ml) and TG (thapsigargin, 1 M), a selective inhibitor of SERCA (sarco-endoplasmic reticulum Ca 2ϩ -ATPases), respectively. PDGF produced rapid Ca 2ϩ transients that mobilized similar amounts of Ca 2ϩ (measured via Fluo-4; Fig. 5A, bar graph), albeit with different kinetic profiles (Fig. 5A, line graph). A sharper decline in cytosolic Ca 2ϩ levels was observed following the initial peak in RTN4-KO compared with WT cells. Differences were further apparent when Ca 2ϩ transients were induced with TG. TG induced sustained Ca 2ϩ transients in both cell types; however, the magnitude of Ca 2ϩ mobilized was significantly reduced in RTN4-KO MEFs (Fig. 5B). Because both internal and external sources of Ca 2ϩ contribute to these transients, to gain further insight into the underlying cause of altered Ca 2ϩ dynamics in RTN4-KO cells, these two components were uncoupled and studied independently.
To assess the amount of ER Ca 2ϩ content, the major internal source of Ca 2ϩ , WT and RTN4-KO MEFs were treated with TG in the presence of nominally calcium-free buffer (0 mM extracellular Ca 2ϩ ). Inhibition of SERCA activity with TG lead to a passive efflux of ER Ca 2ϩ with no significant difference in the amount of Ca 2ϩ mobilized in RTN4-KO compared with WT MEFs, indicating that internal Ca 2ϩ stores are unaffected (Fig.   5C). To assess SOCE, internal stores were initially depleted with TG (in the absence of extracellular Ca 2ϩ ), and external Ca 2ϩ was added back to measure Ca 2ϩ influx. The addition of extracellular Ca 2ϩ prompted a robust influx in WT MEFs that was substantially reduced in RTN4-KO MEFs (Fig. 5D). Addition-  ally, proposed inhibitors of SOCE, 2-APB (2-aminoethoxydiphenyl borate, 50 M) and low concentrations of Gd 3ϩ (gadolinium, 10 M) blunted calcium influx in WT MEFs to levels close to those observed in RTN4-KO MEFs (Fig. 5D). Finally, when PDGF induced Ca 2ϩ transients were examined in the absence of extracellular Ca 2ϩ , to preclude the contribution of SOCE to the Ca 2ϩ signal, a significant drop in cytosolic Ca 2ϩ level was observed in WT but not in RTN4-KO MEFs (Fig. 6). Thus, these results indicate that the altered Ca 2ϩ dynamics observed in RTN4-KO MEFs are due to defective SOCE.
Defective SOCE Renders RTN4-KO Cells Resistant to Ca 2ϩ Overload Induced Apoptosis-Next, we examined the physiological importance of this alteration in SOCE. Previous studies have shown that cells can utilize SOCE to promote their own demise. Prolonged cellular Ca 2ϩ overload, sustained by SOCE, leads to mitochondrial dysfunction, release of pro-apoptotic factors, activation of caspases and Ca 2ϩ -dependent degrada-tion enzymes, and subsequently cell death (reviewed in Rev. 40). Under such conditions, reducing SOCE activity can protect cells from apoptosis (41)(42)(43)(44)(45). To examine whether the reduction in SOCE observed in RTN4-KO MEFs hindered their capacity to undergo apoptosis, WT, and RTN4-KO MEFs were treated with TG or staurosporine (STS, 0.1 M), agents that trigger cell death via sustained elevation of cytosolic Ca 2ϩ levels (46 -49). As expected, TG treatment of WT cells for 24 h resulted in robust activation of caspases-3 and -8 and PARP cleavage. These hallmark features of apoptosis were virtually absent in RTN4-KO MEFs (Fig. 7A). Similar results were obtained in response to 10 h exposure to STS (Fig. 7A). These results were further confirmed using FACS analysis of annexin-V/PI co-labeled cells (Fig. 7B). Moreover, lentiviral reintroduction of RTN4b, but not empty vector, restored apoptosis of RTN4-KO MEFs to levels observed in WT cells (Fig. 7C). Overall, these results highlight the importance of RTN4b-regulated ER morphogenesis in SOCE and cellular homeostasis and strengthen the biological relevance of our findings.
Defective SOCE in RTN4-KO Cells Results from Altered STIM1 Distribution-Next, we sought to examine the SOCE defect in RTN4-KO MEFs at the molecular level. In non-excitable cells, SOCE is largely mediated by STIM1 and Orai1. In addition to these essential components, TRPC1 (transient receptor potential channel-1) and SERCA have also been implicated in regulating SOCE (reviewed in Ref. 50). Immunoblotting for these proteins revealed their levels to be unaffected by the loss of RTN4b (Fig. 8A). Furthermore, we detected no changes in the mobility of STIM1 on SDS-PAGE implying that the phosphorylation state, reported to regulate STIM1 function (31,51), was also unaltered.
To investigate STIM1 localization, mCherry-STIM1 was coexpressed with GFP-Sec61b in WT and RTN4-KO MEFs. Common to previous publications (8, 28, 31, 52, 53), confocal imaging of WT cells revealed an interlaced, tubular distribution   MARCH 28, 2014 • VOLUME 289 • NUMBER 13 JOURNAL OF BIOLOGICAL CHEMISTRY 9387 of STIM1 (Fig. 8B, top panels). In contrast, because of the ER morphological changes in RTN4-KO cells, STIM1 distributed predominantly to ER sheets (Fig. 8B, bottom panels). These findings have significant implications for SOCE since the recruitment of STIM1 to ER-PM junctions is diffusion-limited and spatially restricted to a 2-m perimeter (53), and diffusion to a particular target becomes less efficient and slower as dimensionality is increased (7). Therefore, the probability of STIM1 clusters reaching ER-PM junctions is reduced in the ER sheet-enriched network of RTN4-KO MEFs compared with the predominantly tubular ER network of WT MEFs. This was corroborated by TIRFM assessment of STIM1 dynamics. When low levels of mCherry-STIM1 were transfected into WT and RTN4-KO MEFs, and cells were challenged with TG to deplete internal Ca 2ϩ stores (in the absence of extracellular Ca 2ϩ ), a 2-fold reduction in STIM1 clustering was observed in RTN4-KO compared with WT MEFs (Fig. 8, C and D). Consequently, these results suggest that the diminished SOCE in RTN4-KO cells could in part be due to altered STIM1 diffusion dynamics resulting from ER remodeling and STIM1 mislocalization.

RTN4 Is Important for Store-operated Calcium Entry
Endogenous STIM1 Function Is Impaired in RTN4-KO Cells-To avoid potential pitfalls or artifacts associated with overexpression of tagged STIM1 molecules, we used a variety of methods to directly or indirectly investigate the function of endogenous STIM1. Direct visualization of endogenous STIM1 remains a challenge, potentially due to low protein levels and the compact nature of STIM1 under resting conditions. On that account, when distribution of endogenous STIM1 was examined by confocal microscopy, contrary to the obvious ER-localized staining observed with STIM1 overexpression, a weak diffuse staining pattern was observed under resting conditions in both WT and RTN4-KO cells (Fig. 9A, top panels). Despite the limited resolution, TG induced clustering of endogenous STIM1 was readily observed in WT MEFs (Fig. 9A, bottom left  panel). Reflective of previous results, endogenous STIM1 clustering occurred to a lesser extent in RTN4-KO MEFs (Fig. 9A,  bottom right panel). These changes were also mirrored by quantitative analysis of the size of STIM1 clusters (Fig. 9B).
To strengthen these findings, endogenous STIM1 function was also investigated indirectly via FRET and TRIFM using measurements collected with Orai1 overexpression. Although baseline FRET, between overexpressed Orai1-CFP and Orai1-YFP, was expected to be high (ϳ8% in both WT and RTN4-KO cells) due to the dimeric state of Orai1 under resting conditions (54,55), a further increase should arise from endogenous STIM1-driven clustering of Orai1 dimers. This occurred in both cell lines albeit to a significantly lesser extent in RTN4-KO compared with WT cells (Fig. 9C). This impairment of endogenous STIM1 function, reflected by reduced Orai1 clustering, was also apparent when visualized by TIRFM. Frame differencing of images collected before and after TG treatment in WT and RTN4-KO MEFs overexpressing low levels of Orai1-mCherry revealed a substantial difference in endogenous STIM1-induced Orai1 clustering (Fig. 9D). Because STIM1 and subsequent Orai1 clustering become apparent only upon reaching ER-PM junctions and coupling to one another, respectively, these results suggest of defective targeted diffusion of endogenous STIM1 following internal Ca 2ϩ stores depletion in RTN4-KO MEFs. Finally, we would anticipate that bypassing this ER-associated defect should rescue defective SOCE. Indeed transfection of a soluble Orai1-activating domain (14) (CAD-mCherry), but not its truncated/inactive form (D7-mCherry), rescued defective TG-induced SOCE in RTN4-KO cells ( Fig  10). Collectively, these data imply that RTN4b supports SOCE by promoting correct STIM1 localization and efficient diffusion on ER tubules to ER-PM junctions and coupling to Orai1 upon activation.
Loss of ER Tubulation Is Responsible for Defective SOCE-To confirm that loss of ER tubulation is responsible for the diminished SOCE activity in RTN4-KO MEFs, we conducted rescue experiments using co-expression of mCherry-KDEL to highlight ER sheets, in conjunction with either full-length or various mutants of RTN4b tagged with GFP. Full-length RTN4b and the truncation mutant RHD, missing the N-terminal-half of the protein, contain the curvature-inducing RHD. The TM2mut is a modified form of RTN4b retaining proper membrane topology and cellular localization but incapable of inducing membrane curvature (35). This inactivation is achieved through lengthening of the second hairpin of the RHD to penetrate deeper and span both leaflets of the ER membrane. The structures of these various constructs are depicted graphically in Fig  11A. Reconstitution with both full-length GFP-RTN4b and GFP-RHD restored ER tubulation and rescued the ER morphology defects in RTN4-KO cells, whereas GFP did not ( Fig 11B). As expected, the expression of GFP-TM2mut in RTN4-KO MEFs failed to restore ER tubulation (Fig 11B, right panels).
Finally, we examined whether this rescue of ER tubulation also extends to STIM1 distribution and SOCE rescue. For these experiments, lentiviral delivery of RTN4b constructs was used to achieve expressions levels close to physiological and avoid ER stress responses commonly associated with transient overexpression of ER proteins. Endogenous STIM1 distribution was estimated from results obtained with overexpression of low levels of STIM1-YFP. When STIM1-YFP was overexpressed in RTN4-KO MEFs stably expressing untagged RTN4b or RHD (Fig 12A, second and third panels, respectively), a reticulated tubular distribution was observed, similar to that seen in WT MEFs (Fig. 8B, top panels). In contrast, in RTN4-KO MEFs stably expressing RFP or untagged TM2mut (Fig 12A, first and  last panels, respectively), the distribution of STIM1-YFP was sheet-enriched and resembled that observed in RTN4-KO MEFs (Fig. 8B, bottom panels). Furthermore, when SOCE was assessed in RTN4-KO MEFs stably expressing full-length RTN4b or RHD (Fig 12B), Ca 2ϩ influx was completely restored to levels observed in WT MEFs (Fig. 5D). This rescue was not observed in cells expressing TM2mut or RFP (Fig 12B). Hence, the loss of ER tubulation is the major factor responsible for the reduction in SOCE in cells lacking RTN4b. Collectively, these data support a critical role for RTNs in Ca 2ϩ homeostasis and demonstrate for the first time that RTN4b, via its ER tubulating actions, is a positive regulator of SOCE.

DISCUSSION
Although it is presumed that ER structure is critical for SOCE, there is little direct experimental evidence. The present study was carried out to directly assess this hypothesis. Cell lines exhibiting opposing ER morphologies were generated via genetic manipulation of the ER-shaping protein RTN4. The loss of RTN4a/b, and lack of compensation from other RTNs, resulted in remodeling of the ER from a predominantly tubular to a sheet-enriched network. This change in ER structure did not lead to a generalized dysfunction of the ER, as indicated by unaltered triglyceride synthesis, lipid droplet formation and unfolded protein response. The loss of ER tubulation did however suppress SOCE. The associated resistance of RTN4-KO cells to Ca 2ϩ overload-induced cell death emphasized the physiological importance of the diminished SOCE. When investigated at the molecular level, the suppression of SOCE stemmed from STIM1 mislocalization and alteration of its membrane diffusion dynamics due to altered ER structure. We postulate that the transition from a predominantly tubular to a sheet- enriched ER increases the dimensionality for STIM1 diffusion, which in turn significantly reduces the encounter probability of STIM1 molecules and targeting to ER-PM junctions, resulting in diminished STIM1-Orai1 coupling and SOCE. These findings represent the first demonstration of a specific role for RTNs in Ca 2ϩ homeostasis and reveal novel insights into how ER morphology may regulate intracellular signaling.
RTN4-KO Cells Exhibit Altered ER Morphology-Characterization of MEFs showed that primary as well as immortalized cells express abundant levels of RTN4 isoforms. Among these, we detected the RTN4a and b isoforms, with the RTN4b isoform being predominant. In yeast, the two rtn paralogs, rtn1 and rtn2, generate each a single isoform. A triple deletion of both yeast paralogs and an unrelated curvature-inducing protein Yop1 was necessary to observe a loss of ER tubulation and enrichment of ER sheets (25). A different study reported similar changes in ER morphology with the single deletion of rtn1 (27).  In mammalian cells, four rtn paralogs can give rise to at least 24 different isoforms. In the human osteosarcoma cell line U2OS, only upon knockdown of isoforms encoded by rtn1, rtn3, and rtn4 together, were the investigators able to observe morphological changes in the ER (56). Given these observations, we were surprised to find altered ER morphology in RTN4-KO cells despite the expression of other RTNs in these cells. Although these cells lack both RTNa and RTN4b, the reintroduction of RTN4b was sufficient to rescue the observed ER and SOCE phenotypes. This suggests that RTN4b is the major isoform that maintains the tubular ER in MEFs. As to why this isoform is preferred over the others, it is unlikely that it is in direct relation to the basal function of curvature induction as this function is redundant among all RTNs, and the ER morphology defect can be rescued by the reintroduction of a truncated mutant expressing only the common RHD domain of RTN4. Therefore, the answer probably lies within the variable N-terminal region, whose function remains elusive and will require further studies to decipher its role in the functionality of these proteins.
The Structure-Function Relationship of the ER-The evolutionary conservation of the characteristic shapes of organelles suggests an intimate connection between structure and function. Based on this assumption, an alteration to the structure of an organelle could be expected to affect certain functions. This was recently substantiated in an elegant study in mitochondria where it was demonstrated that alteration of the structure of mitochondrial cristae, by depletion of the mitochondria-shaping protein OPA1, destabilizes structure-sensitive protein complexes and consequently leads to diminished mitochondrial respiratory function (57). Similar to this study, we genetically ablated RTN4 to directly assess the impact of the associated structural changes on ER functions. The fact that the ER is a multifunctional organelle that is extremely dynamic and adopts various forms depending on cellular demand is indicative of a strong correlation between structure and function. Therefore, disruption of the predominantly tubular ER network of WT MEFs and conversion to a sheet-enriched network as seen in RTN4-KO MEFs could be expected to severely impact ER functions. However, this was not the case since lipid biosynthesis, assessed via triglyceride synthesis and lipid droplet formation, as well as the UPR signaling were unaltered by structural changes sustained by the organelle. This outcome is in line with a previous study (58) that showed that membrane expansion in response to ER stress occurs through generation of ER sheets, requires UPR signaling, and is driven by lipid biosynthesis. Conversion of ER sheets to tubules via RTN overexpression did not affect these functions nor did it alter the ability of the cells to cope with ER stress. Collectively, these results indicate that certain ER functions are extremely permissive with regards to ER morphology. However, other ER functions are less permissive to structural changes. When the relationship between ER structure and SOCE was examined we found evidence that ER shape determines the assembly of STIM1-Orai1 complexes and hence SOCE efficiency. We arrived at this conclusion using various techniques (TIRFM, FRET, Ca 2ϩ influx) that assessed, either directly or indirectly, the function of STIM1 in WT and RTN4-KO MEFs. We initially observed that the alterations of ER structure associated with the depletion of RTN4b resulted in attenuated SOCE. Using overexpression of low levels of tagged STIM1 and TIRFM, we found a significant reduction of STIM1 cluster formation in RTN4-KO compared with WT MEFs. Because STIM1 levels were comparable in WT and RTN4-KO cells and bypassing the ER using a soluble form of STIM1 rescued SOCE, such an outcome could be reflective of altered STIM1 diffusion dynamics. To form visible puncta, STIM1 molecules must reach ER-PM junctions (53). Under such circumstances, it appears that a lower amount of STIM1 molecules are reaching these sites in RTN4-KO MEFs. These TIRF data are supported by direct detection of endogenous STIM1 using immunofluorescence before and after ER Ca 2ϩ depletion, where fewer STIM1 clusters were detected in RTN4-KO relative to WT MEFs. Endogenous STIM1 dysfunction in RTN4-KO MEFs was also observed using Orai1-Orai-1 FRET and Orai1 cluster formation by TIRFM. These findings suggest that the extent of ER tubulation is an important determinant of the efficiency of STIM1 translocation to ER-PM junctions. The mode of translocation of STIM1 molecules to ER-PM junctions is through passive diffusion on ER membranes and limited to a 2-m perimeter (53). Additionally, the encounter probability of a molecule diffusing to a target decreases as dimensionality is increased (7). Hence, whereas in WT MEFs STIM1 can efficiently diffuse to ER-PM junctions on the tubular configuration of the ER network and readily couple to Orai1 resulting in robust Ca 2ϩ influx, in RTN4-KO MEFs the diminished STIM1-Orai1 encounter probability, due to the loss of ER tubulation and the enrichment of ER sheets, appears to be the key mechanism responsible for SOCE dysfunction. Therefore, regulation of the extent of ER tubulation via the modulation of RTNs levels appears to be an important part of how cells regulate ER-PM communication and by extension SOCE. These findings may have broad implications for understanding the dynamic nature of the ER and its relationship to cellular physiology. For instance, alterations in ER morphology, observed during cell division, could be a contributing factor, along with STIM1 phosphorylation (31), Orai1 internalization (59) and microtubule rearrangement (29), to SOCE inhibition observed during that stage of the cell cycle.
Our data provide insights into the structure-function correlation in the ER and demonstrate a new level of regulation of ER functions through structural remodeling of ER elements. Here we demonstrated in a simple cell system that the loss of RTN4b results in ER morphological changes that selectively disrupt certain structure-sensitive ER processes such as SOCE and highlight the importance of RTN4b in Ca 2ϩ homeostasis. Additionally, these findings also raise the possibility that the involvement of RTN4b in SOCE may explain in part certain of the phenotypes observed in RTN4-KO mice (36, 60 -62), especially since processes such as inflammatory responses and neural growth are heavily dependent upon Ca 2ϩ signaling (reviewed in Refs. 63,64).