Differential regulation of endoplasmic reticulum structure through VAP-Nir protein interaction.

The endoplasmic reticulum (ER) exhibits a characteristic tubular structure that is dynamically rearranged in response to specific physiological demands. However, the mechanisms by which the ER maintains its characteristic structure are largely unknown. Here we show that the integral ER-membrane protein VAP-B causes a striking rearrangement of the ER through interaction with the Nir2 and Nir3 proteins. We provide evidence that Nir (Nir1, Nir2, and Nir3)-VAP-B interactions are mediated through the conserved FFAT (two phenylalanines (FF) in acidic tract) motif present in Nir proteins. However, each interaction affects the structural integrity of the ER differently. Whereas the Nir2-VAP-B interaction induces the formation of stacked ER membrane arrays, the Nir3-VAP-B interaction leads to a gross remodeling of the ER and the bundling of thick microtubules along the altered ER membranes. In contrast, the Nir1-VAP-B interaction has no apparent effect on ER structure. We also show that the Nir2-VAP-B interaction attenuates protein export from the ER. These results demonstrate new mechanisms for the regulation of ER structure, all of which are mediated through interaction with an identical integral ER-membrane protein.

The endoplasmic reticulum (ER) 1 is an extensive network of membranes comprised of an array of interconnecting tubules and cisternae that emerges from the nuclear envelope (NE) and extends peripherally throughout the cell cytoplasm (1). It contains several structurally distinct domains, including the NE, the rough and smooth ER (rER and sER), and the regions that contact other organelles, such as the Golgi apparatus, the late endosomes, the lysosomes, mitochondria, peroxisomes, and the plasma membrane (2). The ER functions in diverse metabolic processes including lipid synthesis, carbohydrate metabolism, and the detoxification of drugs. It is responsible for the synthesis, translocation, glycosylation, folding, assembly, and processing of secretory and membrane proteins, and it functions in intracellular calcium storage and sequestering (3,4).
While the function of the ER in membrane trafficking, lipid biosynthesis, and calcium signaling have been extensively studied, the mechanism by which the ER maintains its characteristic structure in vivo remains largely unknown. Studies from yeast and mammalian cells have shown that the size and/or structure of the ER is extremely sensitive to certain cellular stress conditions, such as the unfolded protein response (UPR) or to the overexpression of a subset of ERresident membrane proteins (5,6). Overexpression of 3-hydroxymethyl-3-methylglutaryl coenzyme A (HMG-CoA) reductase, microsomal aldehyde dehydrogenase (msALDH), cytochrome P-450, and malfolded cytochrome P-450, causes the proliferation of ER membranes and stacking of the ER cisternae into organized structures known as crystalloid ER, sinusoidal ER, or karmellae (7)(8)(9)(10)(11)(12). The formation of these structures is easily visible and can be used as a quantitative method for studying ER membrane biogenesis (5). Nevertheless, the underlying mechanism of their formation is not completely understood. It has been proposed that proteins that regulate lipid metabolism, such as HMG-CoA reductase, induce the formation of these structures because of the extensive requirement of membrane biogenesis. However, subsequent studies suggested that the oligomerization state of the ER integral membrane proteins is crucial for their production (7,13,14).
The organization of the ER is also sensitive to drugs that induce microtubule depolymerization, such as colchicine or nocodazole. These drugs cause retraction of the ER tubules from the cell periphery and consequently, the formation of ER membrane aggregates around the nucleus. It has long been known that the ER uses the microtubules as a framework for extending and maintaining its reticular organization in animal cells (1,15). The interaction of the ER with microtubules is mediated by motor proteins, which interact directly with microtubules and concomitantly with the ER membranes via protein-protein interactions. Their interaction with the ER and their motor activity are required for the sliding of ER tubules along stationary microtubules and consequently, motility of the ER network. In contrast, the position of the ER within the cells and its motility by microtubule movement or tip-attachment mechanisms (3) require static association of the ER membranes with microtubules, and this is thought to be mediated by proteins such as p63 or cytoplasmic linker proteins (CLIPs). p63 is a type II integral ER membrane protein that binds directly to microtubules via its cytoplasmic domain and thereby links the * This work was supported by the Israel Science Foundation (Grant No. 1073/03), the Israel Cancer Research Foundation, and by the Harry and Jeanette Weinberg Fund for the Molecular Genetics of Cancer. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
VAP-B is also a type II integral membrane protein of ϳ31 kDa that has been previously localized to the ER and the pre-Golgi intermediates (19,20). It belongs to a highly conserved family of proteins, which are implicated in the regulation of neurotransmitter release, ER-Golgi and intra-Golgi transport, Glu4 (glucose transporter 4) trafficking, stabilization of presynaptic microtubules, and the expression of phospholipid biosynthetic genes (19,(21)(22)(23)(24)(25). These diverse functions have been demonstrated in different species and cell types, and are mediated by different members of this family. Nevertheless, the overall structures of VAP proteins are similar and consist of a large N-terminal region facing the cytoplasm and a hydrophobic C terminus that functions as a transmembrane domain (TMD) (20). The cytoplasmic region contains a conserved N-terminal domain of 100 amino acids, which shares high sequence similarity with the nematode major sperm protein (MSP). This 100 amino acid domain contains a highly conserved sequence of 16 amino acids. The central part of the cytoplasmic region contains a coiled-coil domain of ϳ40 amino acids, which is a common motif in many t-SNARE (soluble N-ethylmaleimide-sensitive factor attachment protein (SNAP) receptor) proteins (26).
The VAP proteins interact with several intracellular proteins and have the ability to interact with each other (27)(28)(29). Originally, the Aplysia ApVAP33 was isolated as a VAMP/synaptobrevin-interacting protein using the yeast two-hybrid screen (21). Subsequent studies demonstrated the interaction of the mammalian VAP-A with additional SNAREs, including syntaxin 1A, rbet1, rsec22, ␣SNAP, and NSF (29). VAP-A also interacts with the tight junction protein occludin (22), with microtubules (20,24), and with OSBP (oxysterol-binding protein) (28). In this study, we isolated VAP-B as an interacting protein with Nir2, using a pull-down experiment and mass spectrometry analysis. Similar to VAP-B, Nir2 also belongs to a highly conserved family of proteins, the Nir/rdgB, which are implicated in the regulation of membrane trafficking, phospholipid metabolism, and signaling (30,31). Here we show that the three Nir proteins: Nir1, Nir2, and Nir3, interact with VAP-B via their FFAT motif, and that Nir-VAP interactions differentially affect the organization of the ER.

EXPERIMENTAL PROCEDURES
Recombinant DNA and Antibodies-The cDNA of human VAP-B was isolated by PCR using a first-strand cDNA, which was synthesized from total RNA of HeLa cells (SuperScript II first strand synthesis system, Invitrogen), as a template and the following sense and antisense oligonucleutide primers; 5Ј-CCGGGATCCGCGAAGGTGGAGCA-3Ј and 5Ј-CGGGATATCCTACAAGGCAATCTTCCCAAT-3Ј, respectively. The amplified PCR product was subcloned into the pCAN-Myc1 and pGE-X4T1 expression vectors. The FFAT mutant of Nir2 was generated by replacing amino acids 349 -353 (EFFDA) with ALLAG using three sequential PCR steps with the following sense and antisense primers; 5Ј-GAGATCCTGGCCAACCGG-3Ј and 5Ј-GTGCCCAGCGAGGAGAGC-TTCCTCGGAGCTGTTCTC-3Ј for the first PCR reaction; 5Ј-GAAGCT-CTCCTCGCTGGGCACGAAGGCTTCTCGGAC-3Ј and 5Ј-GATGGCCT-GGAACTTCGGGGC-3Ј for the second PCR reaction. The two PCR products were mixed together and amplified by a third set of sense and antisense oligonucleotide primers; 5Ј-GAGATCCTGGCCAACCGG-3Ј and 5Ј-GATGGCCTGGAACTTCGGGGC-3Ј, respectively. The PCR product was digested with PstI and BglII and used to replace the corresponding fragment in Nir2 cDNA. The cDNAs of the three Nirs were subcloned into pRK5 mammalian expression vector downstream of the CMV promoter. Hemagglutinin (HA)-tag was fused to the Cterminal of Nir1, Nir2, and Nir3 coding sequences, essentially as de-scribed previously (30). Restriction enzyme analysis and DNA sequencing verified the DNA constructs. YFP-VSV-G construct was kindly provided by Koret Hirschberg (Tel-Aviv University, Israel). Antibody against Nir2 was raised in rabbits as described previously (32). Polyclonal antibody against VAP-B was raised in rabbits immunized with a recombinant GST-VAP-B fusion protein. The antiserum was first run through a GST-bound agarose column, to remove the anti-GST antibodies. The flow-through was then affinity-purified on a GST-VAP-B-bound agarose column. Monoclonal antibody against ␤-tubulin was purchased from Sigma. Monoclonal and polyclonal antibodies against HA and Myc were purchased from Santa Cruz Biotechnology, Inc. Antibody against PDI was purchased from ABR (Affinity BioReagents). Alexa-488 donkey anti-mouse and anti-rabbit IgG were purchased from Molecular Probes. Cy3-conjugated goat antirabbit or goat anti-mouse IgG were from Jackson ImmunoResearch Laboratories (West Grove, PA).
Cell Cultures, Transfections, and Indirect Immunofluorescence-HEK293 and HeLa cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum, penicillin (100 units/ml), and streptomycin (100 mg/ml). The cells were transfected by the calcium phosphate method as described previously (33). HeLa cells grown on glass coverslips were transfected as indicated, fixed in 4% paraformaldehyde in PBS for 20 min at room temperature, and immunostained essentially as described previously (33).
Cell Labeling, Extraction, Immunoprecipitations, and Pull-down Experiments-HeLa cells were washed with methionine-free DMEM and incubated in the same medium containing 10% dialyzed fetal bovine serum and 400 Ci/ml [ 35 S]methionine/cysteine (EXPRE 35 S 35 S, Dupont/PerkinElmer Life Sciences) for 6 h. The cells were harvested and lysed in lysis buffer containing: 20 mM Tris-HCl, pH 7.5, 1% Nonidet P-40, 150 mM NaCl, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, and 10 g/ml aprotinin. The lysate was centrifuged at 15,000 ϫ g for 15 min at 4°C, and the supernatant was used in either pull-down or immunoprecipitation experiments. Immunoprecipitations were performed essentially as described previously (34). GST fusion proteins were expressed in bacteria and purified by standard procedures (Amersham Biosciences). For protein identification by mass spectrometry, GST, or GST-Nir2-(205-424) were cross-linked to the agarose beads using dimethyl pimelimidate (Sigma). Following cross-linking termination and washing, the beads were incubated with HeLa cell lysate (from 10 7 cells) for 3 h at 4°C, extensively washed, boiled in SDS-sample buffer, and separated by SDS-PAGE. The gel was stained with Gel Code and destained by multiple washing steps with doubledistilled water. The protein bands were excised from the gel and subjected to protein identification by peptide mapping using MALDI-TOF mass spectrometry after in-gel trypsinization (Technion, Israel).
Chemical Cross-linking-HEK293 cells were transiently transfected with an expression vector encoding VAP-B-Myc. 24 h later, the cells were washed with PBS and incubated at 4°C for 60 min with 2 mM DSP (dithiobis(succinimidyl propionate); Pierce), a reversible chemical crosslinker. The cross-linking reaction was terminated by incubating the cells with 100 mM glycine for 15 min at 4°C. For reversing the crosslinking, the cells were incubated with 100 mM DTT for 30 min at 37°C. The cells were then washed in PBS and lysed in buffer containing 50 mM Tris-HCl, pH 7.5, 2 mM EDTA, 100 mM NaCl, 1% Triton X-100, 5 g/ml leupeptin, 5 g/ml aprotinin, and 1 mM phenylmethylsulfonyl fluoride. Insoluble materials were removed by centrifugation at 16,000 ϫ g for 20 min at 4°C. SDS sample buffer with or without 100 mM DTT was added, the samples were boiled for 5 min and separated by SDS-PAGE.
Sucrose Density Gradient Sedimentation-HEK293 cells expressing the indicated proteins were washed with PBS and lysed in buffer containing 50 mM Tris-HCl, pH 7.5, 50 mM KCl, 0.5% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, 5 g/ml leupeptin, and 5 g/ml aprotinin. Insoluble materials were pelleted by centrifugation at 16,000 ϫ g for 1 h at 4°C. Supernatants (0.5 ml) were layered onto 10 ml of continuous 5-30% (w/v) sucrose density gradients in a buffer containing 50 mM Tris-HCl, pH 7.5 and either 0.1% Triton X-100 or 1% SDS, as indicated. The gradients were centrifuged at 100,000 ϫ g for 18 h at 4°C, fractionated from the top to the bottom into 20 fractions of 0.5 ml, and analyzed by Western blotting. Protein markers including thyroglobulin (669 kDa), catalase (232 kDa), and bovine serum albumin (67 kDa) were separated and fractionated under the same conditions and detected by SDS-PAGE and Coomassie Blue staining.
VSV-G Transport Assay-HeLa cells grown on glass coverslips were transfected with an expression vector encoding the temperature-sensitive mutant of vesicular somatitis virus glycoprotein (VSV-G) fused to YFP (YFP-VSV-G), either alone or together with expression vectors encoding Nir2-HA, VAP-B-Myc, or Nir2-HA and VAP-B-Myc. Following 5 h, the cells were shifted to 40°C to accumulate YFP-VSV-G in the ER. The cells were then incubated at 32°C for different time periods in the presence of cycloheximide (100 g/ml) to permit the transport of VSV-G from the ER along the secretory pathway. The cells were fixed at the indicated time points, immunostained with antibody against Myc or HA, and analyzed by confocal microscopy.
Transmission Electron Microscopy-Hela cells grown in 35-mm Falcon dishes, were transfected as indicated, and fixed for 60 min in Karnovsky's fixative (3% paraformaldehyde, 2% glutaraldehyde, 5 mM CaCl 2 in 0.1 M cacodylate buffer, pH 7.4 containing 0.1 M sucrose). The cells were then washed and scraped. The cell pellet was embedded in agar noble (1.7%) and postfixed with 1% OsO 4 , 0.5% potassium dichromate, and 0.5% potassium haxacyanoferrate in 0.1 M cacodylate buffer. The cells were stained en bloc with 2% aqueous uranyl acetate, followed by ethanol dehydration. Sections were cut using a diamond knife (Diatome, Biel, Switzerland), and stained by 2% uranyl acetate in 50% ethanol and lead citrate. The samples were examined with transmission electron microscope FEI CM12 Eindhoven, Holland, at an accelerating voltage of 120 kV and recorded with a SIS Biocam CCD, 1024 ϫ 1024 pixel camera.

RESULTS
Isolation of VAP-B as a Nir2-interacting Protein-To gain a better insight into Nir2 cellular functions, we looked for Nir2-interacting proteins using coimmunoprecipitation assays. Control and Nir2-HA-transfected HeLa cells were metabolically labeled with [ 35 S]methionine, lysed, and immunoprecipitated with anti-HA antibody. Several proteins were detected in the Nir2 immunocomplex in a specific manner, including a prominent one of ϳ31 kDa (Fig. 1A). This protein was also found in the immunoprecipitates of several Nir2-truncated mutants, including a mutant that lacks the first 340 amino acids. However, it was not present in the immunocomplex of a truncated mutant lacking the first 440 amino acids of Nir2 (Fig. 1B), suggesting that the interaction of Nir2 with the 31 kDa protein is mediated through amino acids 341-439. These results were further confirmed by a pulldown experiment using a GST fusion protein consisting of amino acids 205-424 of Nir2 as an affinity column and [ 35 S]methionine-labeled HeLa cell lysate. The 31 kDa protein was recovered on the Nir2 affinity column in a specific manner (Fig. 1C). Therefore, this column was used in the subsequent large scale pull-down experiment as described under "Experimental Procedures," and the 31 kDa protein was identified by mass spectrometry. This analysis identified two peptides corresponding to amino acids 20 -31 and 201-216 of the human VAP-B protein. Since VAP-B protein consists 243 amino acids and apparently migrates as a ϳ31 kDa protein on SDS-PAGE (19), we assumed that it is the 31 kDa protein that interacts with Nir2. To explore this possibility, the human VAP-B cDNA was isolated and subcloned into a mammalian expression vector containing an N-terminal Myc tag. The Myc-tagged VAP-B protein was expressed in the expected size, and its interaction with Nir2 was determined by pull-down experiments using the GST-Nir2 fusion protein (amino acids 205-424) immobilized on glutathione beads as an affinity column. As shown in Fig. 1D, VAP-B interacts with the GST-Nir2 fusion protein but fails to interact with GST, suggesting that the 31 kDa protein is indeed VAP-B.

FIG. 1. Isolation and identification of a 31 kDa protein that interacts with Nir2. A, HeLa cells expressing wild-type
Nir2-HA and control non-transfected cells were metabolically labeled with [ 35 S]methionine, lysed, and subjected to immunoprecipitation with anti-HA antibody. The samples were separated by 10% SDS-PAGE, and then subjected to autoradiography. The 31 kDa protein that coimmunoprecipitated with Nir2-HA is marked by an arrow. B, HeLa cells that express either the wild-type Nir2-HA, the indicated Nir2-HA-truncated mutants, or control nontransfected cells were metabolically labeled, lysed, and immunoprecipitated as described above. The 31 kDa protein was not detected in the control or the immunocomplex of a mutant lacking the first 440 amino acids of Nir2. C, GST or GST-Nir2-(205-424) fusion protein immobilized on glutathione-agarose beads were incubated with [ 35 S]methionine-labeled HeLa cell lysate. The samples were extensively washed, resolved by SDS-PAGE, and subjected to autoradiography. D, HEK293 cells were transiently transfected with an expression vector encoding VAP-B-Myc. The cell lysate was incubated with GST or GST-Nir2-(205-424) bound to glutathioneagarose beads. Following washing, the samples were resolved by SDS-PAGE, transferred to nitrocellulose membrane, and immunoblotted with anti-Myc antibody.

Nirs Interact with VAP-B via Their Conserved FFAT Motif-
Recently, the yeast homolog of VAP-B, Scs2p was shown to interact with Opi1p, through a conserved EFFDAXE motif designated FFAT (35). This motif is also found in the Nir2 protein at amino acids 349 -355 and is highly conserved among the other Nir/rdgB family members. The presence of this motif within the region that mediates the interaction with the 31 kDa protein (amino acids 341 to 439; Fig. 1B) further supports our results and strongly suggests that the 31 kDa protein is indeed the VAP-B protein. To determine whether the interaction of Nir2 with VAP-B is mediated through its FFAT motif, the conserved EFFDA sequence was replaced by ALLAG. This mutagenesis completely abolished the interaction of Nir2 with VAP-B ( Fig. 2A), indicating that the interaction of Nir2 with VAP-B is mediated by its FFAT motif. To demonstrate the interaction between endogenous VAP-B and Nir2 proteins, we raised a polyclonal antibody against VAP-B as described under "Experimental Procedures." This antibody recognizes both the transfected and the endogenous VAP-B protein by immunopre-cipitation and immunoblotting (Fig. 2B). Moreover, Western analysis of endogenous VAP-B immunoprecipitated from HeLa cells, revealed the presence of endogenous Nir2 in the VAP-B immunocomplex. Similarly, VAP-B was detected in the immunocomplex of Nir2. These results suggest that Nir2 and VAP-B interact with each other in vivo (Fig. 2C).
Next, we assessed whether the other Nir proteins; Nir1 and Nir3, interact with VAP-B, as they also contain a FFAT motif (Fig. 2D). The results shown in Fig. 2E, clearly demonstrate the interaction of Nir1 and Nir3 with VAP-B, suggesting that the three Nir proteins interact with VAP-B via their FFAT motif.
The Interaction of VAP-B with Nir2 Affects the ER Structure-The VAP-B protein was previously localized to the ER and to pre-Golgi intermediates by biochemical and immunocytochemical methods (19). We have previously shown that Nir2 mainly localizes in the Golgi apparatus, but it is also found in the ER in interphase cells (36). To determine the localization of VAP-B in HeLa cells, we used the anti-VAP-B antibody in indirect immunofluorescence and confocal microscopy analyses. The VAP-B protein exhibits a typical ER reticular localization. Similar localization was observed with the transfected VAP-B-Myc protein, and double immunostaining with antibody against protein-disulfide isomerase (PDI), an ER marker, revealed their strong colocalization (Fig. 3A), suggesting that endogenous and transfected VAP-B proteins are localized in the ER.
To gain a better understanding on Nir-VAP-B interactions, we coexpressed them together in HeLa cells and examined their subcellular localization using indirect immunofluorescence and confocal microscopy analyses. Overexpression of wild-type Nir2 with VAP-B caused to production of large heterogeneous granular structures that were dispersed around the nucleus throughout the cytosol, in which Nir2 and VAP-B were strongly colocalized. These structures were not detected in cells that express Nir2 alone (data not shown), or coexpressed the FFAT mutant of Nir2 and VAP-B (Fig. 3B). These results suggest that the interaction of Nir2 with VAP-B, and not simply their overexpression, induces the formation of these un-usual structures. To determine the origin of these structures, we used several organelle-specific markers and analyzed their localization by confocal microscopy. As shown in Fig. 3B, the ER-resident protein PDI was strongly localized to these structures, whereas neither the Golgi markers, nor the endosomal or lysosomal markers were detected in these structures (data not shown). These results suggest that these structures were formed from the ER, and that Nir2-VAP-B interaction rearranges the normal structure of the ER. Furthermore, coexpression of Nir2 and VAP-B had no detectable effect on the cis, medial, or trans Golgi morphology, on COPI vesicles, on early or late endosomes, or on lysosomes, as determined by localization of their respective markers; p58, NAGT-I, sialyltransferase, ␤-COP, EEA1, rab7, and LAMP1 (data not shown). Thus, coexpression of Nir2 and VAP-B specifically modifies the ER structure.
To better characterize the effect of the Nir2-VAP-B interaction on ER structure, we used transmission electron microscopy (EM) analysis. The EM images shown in Fig. 4, demonstrate the remarkable reorganization of the ER, which is characterized by various structures consisting multiple membrane arrays that were organized in diverse forms. Some of them emerged from the NE, while others appeared in peripheral locations, consistent with our confocal microscopy analysis. Sometimes they were visualized as circular packed cisternae, or linear stacks of cisternae located far from the cell nucleus. Membrane wheels, or closed or partially opened loops were also observed (data not shown). Similar structures have been previously obtained by overexpression of several integral ER membrane proteins, including HMG-CoA (9), msALDH (7), and cytochrome b 5 (37), among others.

VAP-B Undergoes Oligomerization in Mammalian Cells-
The mechanism underlying the formation of stacked ER membrane arrays is not completely understood. However, several lines of evidence suggest that oligomerization of the membrane protein is critical to this process. Accordingly, interaction between the cytoplasmic domains of proteins on apposed ER membranes causes stacking of the ER cisternae (7,13,14,38). Other studies have suggested that oligomerization of the ER membrane protein is insufficient for the induction of alterations in membrane assembly; rather, proper folding of the TMD is the critical induction factor (10,37). We therefore assessed the oligomerization state of VAP-B in mammalian cells. VAP-B-Myc was expressed in HEK293 cells, and its oligomerization state was analyzed following treatment with DSP, a membrane-permeable reducible cross-linker. The results (Fig. 5A) indicated that VAP-B undergoes oligomerization in mammalian cells. A clear band of ϳ63 kDa, probably repre-senting a dimer, was detected, whereas a band with the expected migration for a tetramer, at ϳ120 kDa, was less abundant. Furthermore, deletion of the first 63 amino acids of VAP-B had no effect on its oligomerization, as demonstrated by the same experimental approach (data not shown). These results suggest that VAP-B undergoes self-oligomerization in mammalian cells, and that the TMD is probably critical for its oligomerization. We further confirmed this result by sucrose density gradient analysis (Fig. 5B). In this set of experiments, we expressed either VAP-B or Nir2 alone, or coexpressed VAP-B with wild-type Nir2 or its FFAT mutant in HEK293 cells. The cells were lysed as described under "Experimental Procedures" and layered onto a continuous sucrose density gradient in the presence or absence of 1% SDS. Following sedimentation, fractions were collected and analyzed by Western blotting. Protein markers, including thyroglobulin (669 kDa), catalase (232 kDa), and bovine serum albumin (67 kDa), were separated on a parallel sucrose gradient, fractionated under the same conditions, and detected by SDS-PAGE and Coomassie Blue staining. A logarithmic plot of the molecular mass of the marker proteins as a function of gradient fraction indicated that VAP-B sediments as expected for a dimer, consistent with the data shown in Fig. 5A. The peak intensity of the VAP-B protein under native conditions was found in fraction 5, which has an estimated molecular mass of ϳ63 kDa, while its peak under denaturing conditions (1% SDS) was shifted to fraction 3. Nir2 was detected in fractions 6 through 12, with a peak in fraction 10 corresponding to ϳ153 kDa. Coexpression of VAP-B and the FFAT mutant of Nir2 had no effect on the sedimentation profile of either VAP-B or Nir2, consistent with these proteins' inability to interact with one another. However, coexpression of wild-type Nir2 and VAP-B induced a dramatic shift in VAP-B sedimentation. The peaks of VAP-B and Nir2 intensities appeared in fraction 12, corresponding to ϳ220 kDa, and continued through to fraction 16 (ϳ440 kDa). Collectively, these results suggest that by itself, VAP-B is a dimer, and that its interaction with Nir2 creates complexes of various sizes that may represent different oligomerization forms. Accordingly, we suggest (Fig. 5C) that VAP-B undergoes dimerization in mammalian cells, probably mediated by the GXXXG motif present in its TMD (20,27). The binding of Nir2 to VAP-B induces a conformational change in the VAP-B cytoplasmic domain and facilitates its trans-oligomerization. This trans-oligomerization is mediated by a head-to-head interaction of VAP-B cytoplasmic domains on apposing ER membranes, which zips up the apposing membrane, yielding stacked ER cisternae. Alternatively, a headto-head interaction of Nir2 molecules could bridge VAP-B proteins on apposing ER membranes and induce their stacking. The sedimentation profile of Nir2 alone (Fig. 5B), which demonstrates the formation of larger Nir2 structures, may support this possibility. Both of these possibilities rely on head-to-head interactions, involving either VAP-B or Nir2 molecules. However, it could be that the binding of Nir2 to VAP-B modifies the folding of VAP-B TMD, which would then transmit a signal for the production of stacked membrane arrays (Fig. 5C, option 3).
Coexpression of VAP-B and Nir2 Attenuates Protein Export from the ER-Next, we assessed whether these structures have any effect on protein export from the ER to the Golgi apparatus and subsequently to the plasma membrane. For this purpose, HeLa cells were cotransfected with either VAP-B and a temperature sensitive mutant of vesicular stomatitis virus-glycoprotein (ts045 VSV-G) fused to YFP, or with Nir2, VAP-B, and YFP-VSV-G. At 40°C the VSV-G is synthesized and retained in the ER because of its misfolding (39). However, upon shifting the temperature to 32°C, the VSV-G protein folds and is transported to the Golgi apparatus and then to the cell surface within 20 -30 min and 60 -90 min, respectively. The results shown in Fig. 6, clearly demonstrate the accumulation of VSV-G in the granular structures at 40°C (time 0) in cells coexpressing Nir2 and VAP-B, consistent with the localization of PDI in these structures and their ER origin. However, 30 min after shifting the temperature to 32°C, VSV-G was mainly localized in the Golgi apparatus in the control non-transfected, Nir2-transfected (data not shown), or the VAP-B-transfected cells, and at 90 min was mainly at the plasma membrane. In contrast, in cells coexpressing Nir2 and VAP-B, VSV-G was retained in the granular structures even 1 h following shifting the temperature to 32°C, suggesting that these structures inhibit the export of VSV-G from the ER. The accumulation of VSV-G in these structures was not caused by a continuous synthesis of VSV-G protein, because the experiment was performed in the presence of cycloheximide.
Ectopic Expression of Nir3 and VAP-B Induces Microtubule Bundling along the ER Membranes-Since Nir1 and Nir3 also interact with VAP-B, we assumed that their coexpression with VAP-B would affect the ER structure in a similar manner to that of Nir2. We therefore coexpressed them with VAP-B and examined their localization by immunofluorescence and confocal microscopy analyses. Coexpression of Nir1 with VAP-B had no apparent effect on either Nir1 or VAP-B localizations (Fig.  7A). In contrast, coexpression of Nir3 and VAP-B strikingly changed the typical localization of VAP-B. It was visualized as thick bundles surrounding the nucleus that were extended peripherally throughout the cytosol, to which Nir3 was colocalized (Fig. 7B). To determine whether coexpression of Nir3 and VAP-B-Myc has any effect on the ER structure, HeLa cells that coexpress them were double immunostained with anti-PDI and anti-Myc antibodies and their localization was analyzed by confocal microscopy. The results shown in Fig. 7C demonstrate that PDI staining was strikingly different in cells that coexpress VAP-B and Nir3 as compared with the control non-transfected HeLa cells (Fig. 3A). In the cotransfected cells, PDI immunostaining appeared in tubular-like structures that were colocalized with VAP-B. Thus, in contrast to Nir1 and Nir2, coexpression of Nir3 with VAP-B modified the ER structure into a tubular pattern. A similar pattern has been previously obtained when the integral ER-membrane protein p63 was overexpressed in COS cells (16). Its overexpression caused to rearrangement of the ER and concomitantly bundling of microtubules along the altered ER membranes. This similarity led us to examine the morphology of the microtubules in cells that coexpress Nir3 and VAP-B compared with their morphology in cells that either express Nir3 or VAP-B alone. As shown in Fig.  7D, coexpression of Nir3 and VAP-B dramatically modified the organization of the microtubules; they were unusually thick and not organized, the microtubule organization center was not visible, and more importantly, they aligned with VAP-B or Nir3 immunostaining. These results suggest that the VAP-B-Nir3 interaction links the ER membranes to the microtubule network, and thereby modifies both the ER and microtubule organization. It is noteworthy that coexpression of Nir3 and VAP-B caused to partial dispersal of the Golgi (data not shown), consistent with the role of microtubules in positioning and maintenance of the Golgi apparatus (1).
The mouse and the Drosophila VAP-A proteins have been previously shown to directly interact with microtubules, yet, through an unidentified motif (20,24). Since the Drosophila VAP-A (DVAP-33A) was proposed to stabilize microtubules, we examined the effect of nocodazole treatment on the microtubule architecture in cells that overexpress VAP-B or coexpress VAP-B and Nir3. As shown in Fig. 7E, treatment with nocodazole for 3 h caused to complete depolymerization of microtubules in the VAP-B overexpressing or in the non-transfected control cells. However, in cells that coexpress VAP-B and Nir3, some microtubules were still detected following such long treatment with nocodazole, suggesting that Nir3-VAP-B interaction enhances the microtubule stability. These results are consistent with the proposed function of DVAP-33A in stabilizing microtubules (24). DISCUSSION Nir1, Nir2, and Nir3 belong to a highly conserved family of proteins, the Nir/rdgB, which have been implicated in regulation of phospholipid trafficking, metabolism, and signaling. Nir2 and Nir3 contain an N-terminal phosphatidylinositol (PI)transfer domain, followed by a short acidic region, six hydrophobic stretches, and a long highly conserved C-terminal domain (Fig. 2D). The N-terminal PI-transfer domain is not present in Nir1 or the zebrafish pl-RdgB (31,40), but all the other family members, including Drosophila retinal degeneration B (rdgB), have a functional PI-transfer domain that has the ability to transfer PI and phosphatidylcholine (PC) between membrane bilayers in vitro (41).
In this study, we show that the three Nir proteins interact with VAP-B through their conserved FFAT motif (Fig. 2), which is present within their acidic region. This motif consists of a conserved EFFDAXE sequence, which acts as an ERtargeting determinant by its direct interaction with VAP proteins (35). The FFAT motif has been identified in 17 distinct eukaryotic proteins, 14 of which are directly implicated in lipid binding or lipid sensing, including homologs of OSBP, homologs of Goodpasture's antigen-binding protein (GPBP), the Nir/rdgB proteins, and Opi1p, a transcriptional regulator of phospholipid synthesis in yeast. The FFAT motif in Opi1p mediates the interaction of Opi1p with Scs2p and thereby targets it to the ER. A similar targeting mechanism has been shown for the yeast homologs of OSBP, Osh1p, Osh2p, and Osh3p (35). In mammalian cells, OSBP also interacts with VAP-A protein (28), probably through its FFAT motif. However, in this particular case, the binding affinity of OSBP to VAP-A is largely dependent on its pleckstrin homology (PH) domain. A specific mutation in this domain, W174A, enhances its binding to VAP-A, and overexpression of this mutant in mammalian cells causes the production of ER inclusions, in which OSPB and VAP-A are colocalized (28). Very similar structures were obtained upon overexpression of Nir2 and VAP-B (Fig. 3B), suggesting that the interaction of FFATmotif-containing proteins with VAPs can induce the formation of these unusual ER structures. However, our results indicate that these structures are formed in a specific manner, which is largely dependent on the VAP-interacting protein, as neither Nir1 nor Nir3 induced their formation (Fig. 7), despite their interaction with VAP-B protein (Fig. 2E). Electron microscopy analysis revealed that these structures consist of multiple membrane arrays that are organized in diverse forms. The structures were found either emerging from the NE or peripherally throughout the cytosol (Fig. 4). Similar structures have been previously obtained in different cell types by overexpressing HMG-CoA reductase, an integral membrane protein that catalyzes the rate-limiting step in cholesterol biosynthesis (42). Detailed analysis using different mutants or chimeric proteins of HMG-CoA reductase suggests that its catalytic activity is not required for stimulation of membrane proliferation; rather, its membrane domain appears to be both necessary and sufficient for the induction of these structures (9,43). Subsequent studies using other transmembrane proteins led to the hypothesis that the proper folding of the transmembrane domain is critical for their formation, and that this folding is required for the transmission of a signal for membrane biogenesis and the production of stacked membrane arrays (5,37). In contrast, other studies proposed that homodimeric interactions between cytoplamic domains of ER-resident proteins consisting of TMDs is sufficient for generating these structures (14). Accordingly, headto-head interaction between the cytoplasmic domains of the integral membrane protein on apposing ER membranes zips the apposing membrane, thereby stacking the ER cisternae (7,13,14). This hypothesis prompted us to characterize the oligomerization state of VAP-B in mammalian cells (Fig. 5).
Our finding that VAP-B undergoes dimerization in mammalian cells (Fig. 5) is consistent with previous studies performed in yeast and in vitro (27)(28)(29). According to the yeast two-hybrid analysis (28), truncated VAP-A mutants lacking either amino acids 41-59, the TMD, or the last 84 amino acids (amino acids 1-160) interact with wild-type VAP-A but fail to interact with OSBP, suggesting that the C-terminal domain, including the TMD, is not required for VAP-A self-oligomerization. On the other hand, Weir et al. (29) showed that both the N-and C-terminal domains of VAP-A are required for its oligomerization, and Nishimura et al. (27) suggested that the TMD is critical for both homo-and hetero-oligomerization of VAP-A and VAP-B proteins in vitro. Indeed, the TMD of VAP-A/B contains a GXXXG motif, which has been identified in a number of TMDs and was proposed to induce strong self-assembly (44). Thus, it could be that several structural domains, including the TMD, mediate the oligomerization of VAP-A/B. We show here that VAP-B expression is insufficient to produce stacks of ER membrane arrays unless Nir2 is coexpressed (Fig.  3). We also show that VAP-B is found mainly as a dimer in mammalian cells, and that its interaction with Nir2 induces the formation of oligomers of various sizes (Fig. 5). These results suggest that the binding of Nir2 to VAP-B induces a conformational change in the VAP-B protein, which either facilitates the trans-oligomerization of VAP-B proteins on apposing ER membranes through interaction of their cytoplasmic domains, or induces specific folding of VAP-B TMD that transmits a signal for membrane assembly (Fig. 5C). Although the first possibility is consistent with the results of the yeast twohybrid interaction assays suggesting that the cytoplasmic region of VAP-A is involved in VAP oligomerization, at present we cannot exclude the second possibility. We also cannot exclude the possibility that a head-to-head interaction of Nir2 molecules bridges VAP-B proteins on apposing ER membranes, allowing them to stack together (Fig. 5C). While this is a reasonable possibility in light of the sedimentation profile of Nir2 (Fig. 5B), our preliminary studies indicate that overexpression of a truncated VAP-B protein lacking the highly conserved 16 amino acids within the N-terminal region induces the formation of stacked ER membrane arrays (data not shown). These results suggest that these structures are formed by conformational changes in the VAP-B protein.
In contrast to Nir2, coexpression of Nir1 with VAP-B had no effect on ER structure (Fig. 7A). Since Nir1 lacks the PI-transfer domain, we assumed that this domain is required for induction of the unusual morphology of the ER seen in Nir2/VAP-Bexpressing cells. However, deletion of most of the PI-transfer domain (amino acids 1-240), or even its flanking region (1-340), did not abolish the effect of Nir2-VAP-B interaction on the ER structure (data not shown). These results suggest that despite the high sequence similarity between the three Nir proteins, and their ability to interact with VAP-B via their FFAT motif (Fig. 2E), they distinctly affect the ER structure in the presence of VAP-B protein. Indeed, coexpression of Nir3 with VAP-B caused a striking rearrangement of the ER and concomitantly, bundling of microtubules along the altered ER membranes, to which Nir3 and VAP-B were colocalized (Fig. 7, C and D). Previously, it was shown that the Drosophila homolog of VAP-A, DVAP-33A, binds microtubules and is involved in stabilizing and directing microtubules during the budding of synaptic boutons. In DVAP-33A mutant flies, the presynaptic microtubule architecture is severely compromised, and DVAP-33A was therefore proposed to function as a bridge between microtubules and the presynaptic membrane (24). Our results support this hypothesis, and suggest that VAP-B bridges the ER membrane with the microtubule network. However, overexpression of VAP-B did not induce bundling of microtubules unless Nir3 was coexpressed. Thus, it could be that the binding of Nir3 to VAP-B induces a conformational change in the latter, which in this specific case enhances its binding to microtubules, or that Nir3 itself serves as a bridge between VAP-B and the microtubules.
Overexpression of many microtubule-associated proteins induces bundling of microtubules and stabilization of their structure (45). Indeed, we show that the microtubules in Nir3/VAP-B-expressing cells are more resistant to nocodazole treatment than the non-transfected control or VAP-B-transfected cells (Fig. 7E), suggesting that this interaction stabilizes the microtubules, and is consistent with the proposed function of DVAP-33A.
Although, our results were obtained through overexpression of VAP-B and the different Nir proteins, they imply that VAP proteins play essential roles in the regulation of ER structure. Thus, VAPs can be considered ER-receptors for many FFATcontaining proteins, which are differentially expressed in various tissue and cell types, and might be differentially regulated under specific physiological conditions. These receptors may undergo different conformational changes upon binding of their cognate ligands, and consequently, distinctly affect the ER structure. Thus, our results suggest that the interaction of VAPs with FFAT-containing proteins is not only required for targeting of FFAT-containing proteins to the ER, but is also involved in the regulation of ER organization and positioning.