Post-translational Membrane Insertion of Tail-anchored Transmembrane EF-hand Ca2+ Sensor Calneurons Requires the TRC40/Asna1 Protein Chaperone*

Background: Calneurons are EF-hand Ca2+ sensors that regulate Golgi-to-plasma membrane trafficking. Results: Calneurons are tail-anchored proteins that interact with TRC40/Asna1 with their minimal Golgi-targeting sequence, which assists their membrane insertion. Conclusion: Calneurons are post-translationally inserted into the ER as monomers, where they dimerize and are transported to the Golgi and the plasma membrane. Significance: This is the first complete mapping of the calneuron biogenesis pathway. Calneuron-1 and -2 are neuronal EF-hand-type calcium sensor proteins that are prominently targeted to trans-Golgi network membranes and impose a calcium threshold at the Golgi for phosphatidylinositol 4-OH kinase IIIβ activation and the regulated local synthesis of phospholipids that are crucial for TGN-to-plasma membrane trafficking. /In this study, we show that calneurons are nonclassical type II tail-anchored proteins that are post-translationally inserted into the endoplasmic reticulum membrane via an association of a 23-amino acid-long transmembrane domain (TMD) with the TRC40/Asna1 chaperone complex. Following trafficking to the Golgi, calneurons are probably retained in the TGN because of the length of the TMD and phosphatidylinositol 4-phosphate lipid binding. Both calneurons rapidly self-associate in vitro and in vivo via their TMD and EF-hand containing the N terminus. Although dimerization and potentially multimerization precludes TRC40/Asna1 binding and thereby membrane insertion, we found no evidence for a cytosolic pool of calneurons and could demonstrate that self-association of calneurons is restricted to membrane-inserted protein. The dimerization properties and the fact that they, unlike every other EF-hand calmodulin-like Ca2+ sensor, are always associated with membranes of the secretory pathway, including vesicles and plasma membrane, suggests a high degree of spatial segregation for physiological target interactions.

ment (6). Interestingly, membrane localization can be controlled by a Ca 2ϩ -myristoyl switch. In the case of Hippocalcin, binding of Ca 2ϩ induces a conformational change and exposure of a hydrophobic myristoyl tail and subsequent translocation of the protein to the plasma membrane (PM) and the Golgi complex (7). Association with these particular membrane compartments is controlled by a direct interaction of Hippocalcin with phosphatidylinositol 4-phosphate (PI(4)P) and phosphatidylinositol 4,5-bisphosphate (PI(4,5)P 2 ) (6). Phosphoinositides contribute to the unique identity of organelles, and binding of PI(4)P, specific for the trans-Golgi network (TGN), and PI(4,5)P 2 at the plasma membrane might explain at least in part the targeting of Hippocalcin and NCS-1 to these membranes (6).
The Golgi by itself is a Ca 2ϩ store that contains release and sequestration apparatus, and several studies have shown that Ca 2ϩ regulates the passage of proteins along the secretory pathway as well as the exit of vesicles from the TGN (8,9). The local synthesis of PI(4)P and PI(4,5)P 2 is crucial for TGN to PM trafficking, and the activity of phosphatidylinositol 4-OH kinase III ␤ (PI-4KIII␤) at the Golgi membrane is the first mandatory step in this process (10). Interestingly, the enzymatic activity of PI-4KIII␤ is regulated by an interaction with NCS-1 and calneurons. At low Ca 2ϩ levels, PI-4KIII␤ is preferentially associated with calneurons, whereas high Ca 2ϩ levels favor binding of NCS-1, and in sharp contrast to the activating role of NCS-1, calneurons strongly inhibit PI-4KIII␤ activity with markedly attenuated PI-4KIII␤ activity at low to medium Ca 2ϩ levels (11). It was therefore suggested that calneurons operate as a filter that suppresses PI-4KIII␤ activity at submaximal amplitudes of Golgi Ca 2ϩ transients and thereby provides a tonic inhibition that is only released under conditions of sustained Ca 2ϩ release in secretory cells (11,12).
Calneurons are highly abundant at the Golgi apparatus in neurons, and their Golgi association is much more prominent than those of other calcium sensor proteins like Caldendrin and NCS-1 (11,13). Structurally, calneurons possess four EF-hand motifs out of which EF-hands three and four are nonfunctional in the sense that they do not bind Ca 2ϩ (Fig. 1A). Although they are efficiently localized to the TGN, calneurons do not contain an N-myristoylation motif that could provide them with a lipid membrane anchor. Thus, the mechanism by which calneurons can be localized to the TGN is unclear. Recently, it was suggested that they are transmembrane proteins and that the membrane localization of calneurons might be provided by the C-terminal hydrophobic region that serves as the transmembrane domain (TMD), with the N terminus oriented toward the cytosol (Fig. 1A) (14). In this study, we aimed to identify the mechanisms that target calneurons to the TGN.

EXPERIMENTAL PROCEDURES
Information about the plasmids and antibodies used in this study as well as a detailed description of subcellular fractionation of HeLa cells and quantification for the Golgi localization of calneuron-1 constructs can be found in the supplemental material.
Cell Culture, Transient Transfection, and Immunostaining-HEK-293T, COS-7, and HeLa cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 2 mM L-glutamine, 100 units/ml penicillin/streptomycin, and 5% (v/v) heat-inactivated fetal bovine serum (FBS) (Invitrogen). For immunofluorescence studies and live imaging experiments, COS-7 cells were transfected with Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. 24 h after transfection, the cells were fixed with 4% paraformaldehyde and processed as described previously (15). Except for the GFP, live staining COS-7 cells were preincubated with anti-GFP rabbit antibody (1:200) diluted in DMEM growth media for 1 h at 37°C and then fixed, washed with PBS, and processed for immunostaining. For the bimolecular fluorescence complementation assay, COS-7 cells were co-transfected with equal amounts of the fusion constructs containing the N-and C-terminal halves of split YFP variant Venus. 40 h after transfection, cells were fixed and processed for immunostaining. HEK-293T cells for BRET experiments were transfected with the corresponding fusion protein cDNAs by polyethyleneimine (Sigma) as described previously (16).
Liposome Sedimentation Assays-Liposome sedimentation assays were done using liposomes obtained from bovine brain fraction (Folch fraction 1, Sigma) and purified His 6 SUMO-calneuron-1 proteins as described previously (18). All steps were carried out in the presence of either 100 M CaCl 2 or 2 mM EDTA.
Size Exclusion Chromatography-Equal amounts of protein samples incubated with either 500 M Ca 2ϩ or 500 M EDTA were loaded on a Superdex TM 200 HR 10/30 column (GE Healthcare) and pre-equilibrated with a buffer containing 50 mM Tris, pH 7.5, 100 mM KCl, and either 1 mM Ca 2ϩ or 1 mM EDTA, respectively. The separation was performed at room temperature using a flow rate of 0.5 ml/min and monitored at 280 nm. Gel filtration standards from Bio-Rad were used for both calibration and determination of void volume.
Dynamic Light Scattering (DLS)-DLS experiments were performed in the presence of either 500 M Ca 2ϩ or 500 M EDTA on a Flexible Correlator Photocor-FC (Photocor Instruments. Inc.) with a 632.8 nm laser at 25°C. The software, Alango DynaLS version 2.0, provided with the instrument was used to analyze the data thus obtained and calculate the mean hydrodynamic radius using the Stokes-Einstein equation, Dt ϭ k B T/6RH.
Co-immunoprecipitation (Co-IP) and Pulldown Assay-Heterologous co-IP was performed with extracts from double transfected COS-7 cells using the MACS TM GFP isolation kit (Miltenyi Biotec GmbH, Germany) according to the manufacturer's protocol. Eluted samples were checked on SDS-PAGE/ Western blot using anti-calneuron-1 rabbit (1:300, Protein-Tech) antibody. For pulldown assay experiments, COS-7 cells were transfected with different EGFP-calneuron-1 constructs. GST-TRC40/Asna1 or GST-coupled Sepharose beads were incubated with COS-7 cell extracts overnight at 4°C. Experiments were carried out as described previously (11). The pulldown assay with MBP-calneuron-1 coupled beads was performed in the presence of either 2 mM EDTA or 100 M Ca 2ϩ . Detection on the immunoblot was done using anti-GFP mouse antibody (1:5000, Covance).
Protein-Lipid Overlay (PLO) Assay-For the PLO assay, COS-7 cells were transfected with untagged calneuron-1, and a cell extract (extraction buffer: 1ϫ TBS buffer containing 0.25% of Tween 20 and a protease inhibitor mixture, Roche Applied Science) was prepared 48 h after transfection. The lipid strips containing different lipids (Lipid Strips and PIP Strips from Echelon Biosciences (Mobitec, Goettingen, Germany) were incubated with a COS-7 cell extract at 4°C overnight and then developed with an anti-calneuron-1 rabbit antibody.
Protein-Protein Overlay Assay-Different recombinant His 6 -SUMO-calneuron-1 proteins or His 6 -SUMO control were diluted in 1ϫ Native Sample Buffer also containing 0.5% SDS and 0.2% ␤-mercaptoethanol, subjected to SDS-PAGE (5 g/lane), transferred onto the nitrocellulose membrane, blocked in 5% nonfat milk in TBS-T, and incubated at 4°C overnight with recombinant GST-TRC40/Asna1 (10 g/ml). After extensive washing, the membrane was incubated with anti-Asna1 mouse antibody following standard protocols.
BRET Assays-BRET assays were performed in HEK-293T cells transiently co-transfected with a constant amount of cDNA encoding for the protein fused to Rluc and increasing amounts of cDNA corresponding to the protein fused to YFP exactly as described previously (19). Specifications of particular experiments can be found in the supplement material.
Laser Scanning Microscopy, FRAP Experiments, and Image Analysis-All fluorescence images were obtained on a TCS SP5 II confocal laser scanning microscope (Leica, Germany) using a 63ϫ oil objective and zoom factors in the range of 1-4ϫ. For FRAP experiments, COS-7 cells were co-transfected with pEGFP-C1-calneuron-1 and pDsRed-Monomeric-Golgi or pGFPC1-Sec61␤. 24 h later DMEM was replaced by KD buffer, and cells were placed under a confocal laser scanning microscope. A 568 and 488 nm laser line was used to monitor the DsRed channel and a 488 nm laser line for both imaging of EGFP and photobleaching. A 63ϫ oil objective and 4ϫ confocal zoom were used. Single plane images were recorded every 20 s for the ER marker and every minute for the calneuron-1 at the Golgi complex. Photobleaching was performed using the FRAP Wizard mode with 488 nm laser at maximum efficiency. Regions of interest of comparable size were taken for each experiment. For the evaluation of % of FRAP, the initial fluores-cence was taken as 100%, and data were graphically plotted. Images were analyzed using ImageJ software (National Institutes of Health).
Brefeldin A Assay-COS-7 cells were co-transfected with untagged calneuron-1, GFP-Sec61␤, and DsRed-Monomeric-Golgi for 24 h. Inhibition of ER-to-Golgi trafficking was induced by prolonged (3 h) incubation with brefeldin A (Cell Signaling, New England Biolabs) at a concentration of 100 ng/ml. Control cells were treated with DMSO. Thereafter, cells were fixed and stained with an anti-calneuron-1 rabbit antibody.
VSV-G Trafficking Assay-COS-7 cells were grown on coverslips and co-transfected with a VSV-G-GFP (ts045) expression plasmid and pcDNA3.1 encoding untagged calneuron-1. The transfected cells were transferred to 39.5°C 4 h post-transfection and incubated overnight. Accumulation of VSV-G-GFP at the Golgi was induced by shifting the temperature to 20°C for 2 h. Then the Golgi block was removed, and Golgi-to-PM trafficking was allowed by incubating the coverslips at 32°C for 20 min, followed by formaldehyde fixation and immunostaining.
Statistical Analysis-Statistical analysis was performed with the SPSS Statistics software (IBM, Ehningen, Germany). Oneway analysis of variance was used to compare individual groups.

Calneurons Are Tail-anchored (TA) Proteins with a 23-Amino Acid-long TMD Responsible for Their Golgi Localization-
We and others have demonstrated previously that calneuron-1 and -2 tightly associate with Golgi membranes and are particularly abundant at the TGN, cytosolic vesicular structures, and presumably endosomal compartments (11,13), and overexpressed calneuron-1 appears to traffic to the plasma membrane (20). We therefore reasoned that if calneurons are transmembrane proteins in the secretory pathway, a C-terminal EGFP tag of a calneuron fusion protein should be detectable at the surface of the plasma membrane of COS-7 cells after overexpression in living cells (Fig. 1A). Incubation of COS-7 cells with a GFP antibody indeed showed membrane labeling only in the case of calneuron-1-EGFP-transfected cells (Fig. 1B), although we could not see any GFP antibody staining with an N-terminally tagged EGFP-calneuron-1 construct (Fig. 1B). A 17-amino acid (aa) region (residues 193-209 for calneuron-1 and 189 -205 for calneuron-2) at the C terminus of both calneurons has high scores for transmembrane ␣-helices, as predicted by "The TMpred algorithm." In fact, mCherry fused to a 17-aa TMD of calneuron-2 has been shown to localize at the TGN like the full-length protein, whereas deletion constructs lacking the TMD are no longer associated with Golgi membranes and show a diffuse cytosolic distribution (13). We generated EGFP constructs containing the same 17 aa from the proposed TMD of calneuron-2 and also observed a perinuclear Golgi accumulation of EGFP fluorescence in some of the transfected cells.
However, many cells exhibited a diffuse cytosolic or ER localization that was not visible to this extent with a full-length construct (data not shown). The width of the membrane bilayer increases from the ER to the Golgi and the PM, with the lipid composition favoring TMDs of 15-17 residues in the ER and TMDs of more than 20 at the TGN and endosomes. Moreover, longer TMDs are essential for proteins to exit the Golgi complex and to be transported to the PM (21). A re-analysis of calneuron-1 and -2 sequences with HMMTOP 2.0 (22) and ProtScale software indicated the presence of a 23-aa stretch that could serve as a TMD. Both calneuron-1 and -2 contain highly hydrophobic TMDs with calculated hydrophobicity scores of 48.1 and 50.7 for the 17-aa TMDs and 48.8 and 52.7 for the 23-aa TMDs, respectively. We next asked whether the TGN localization is more prominent with this longer TMD and designed a number of EGFP fusion constructs of calneuron-1 (Fig. 2B). COS-7 cells were transfected with EGFP-calneuron-1 full-length (aa 1-219), EGFP-calneuron-1_⌬C (aa 1-190), EGFP-calneuron-1_CT (aa 191-219), EGFP-calneuron-1_23 aa (aa 192-214), EGFP-calneuron-1_17 aa (aa 192-208), or a EGFP control plasmid, fixed 24 h later, and stained with the Golgi markers GM130 and TGN38 ( Fig. 2A). A quantitative analysis of TGN localization revealed that the 23-aa stretch predicted to set up the TMD was sufficient for proper TGN recruitment and is as efficient as full-length calneuron-1 in TGN targeting (Fig. 2, A and C). In contrast, the construct with the shorter TMD EGFP-calneuron-1_17aa (aa 192-208) exhibited much less efficient TGN targeting (Fig. 2C).
The idea that calneurons represent true TA proteins is also based on digitonin permeabilization experiments and subsequent protein extraction (14). We repeated this experiment and investigated whether there was any loss in EGFP-calneuron-1 fluorescence after digitonin permeabilization in different cellular compartments. To obtain a readout for permeabilization efficiency, COS-7 cells were co-transfected with a construct expressing soluble red fluorescent protein (supplemental Fig.  S1A). Instantly after addition of 40 M digitonin to KD buffer, red fluorescent protein fluorescence was lost, first in the cytosol and then with a short delay in the nucleus (supplemental Fig.  S1, A and B). We then measured the intensity changes for EGFP-calneuron-1 at the TGN where most of the protein was localized but also in the cytosol, in the nucleus, and in the extracellular buffer. For each of the regions of interest, the average of the base-line fluorescence was taken as 100%. As expected, and in accord with previous data (14), we observed no change in EGFP-calneuron-1 fluorescence at the Golgi, but cytosolic and nuclear regions of interest values were reduced (supplemental Fig. S1, A and B). However, the relative fluorescence intensity of EGFP-calneuron-1 in the cytosol and the nucleus as compared with the TGN was much lower and could be due to overexpression (supplemental Fig. S1, A and B). Moreover, when we transfected COS-7 cells with either EGFP-calneuron-1 or a deletion construct lacking the TMD region (EGFP-calneuron-1_⌬C) and then treated the cells with digitonin for different periods of time, EGFP-calneuron-1_⌬C could be detected in the buffer 30 s after treatment (supplemental Fig. S1C), whereas a weak band for the full-length EGFP-calneuron-1 band was first visible after 60 s. The TGN marker Syntaxin 6 remained com- The green circle represents the EGFP tag fused to the N or C terminus of the protein. In live staining experiments, the fluorophore (magenta circle)-tagged antibody can bind only to the extracellularly exposed antigen. B, COS-7 cells transfected with EGFP-calneuron-1 or calneuron-1-EGFP were subjected to live staining with the rabbit anti-GFP antibody for 1 h at 37°C. Only the cells transfected with calneuron-1-EGFP show immunoreactivity with the GFP antibody, indicating that the tag is extracellular. Scale bar, in B is 10 m.
pletely associated with the membrane fraction (supplemental Fig. S1C, upper panel) and did not diffuse out of the cells after permeabilization. In contrast, soluble ␤-actin, used as a loading control, exited the cells after 30 s (supplemental Fig. S1C).
In conjunction with previous findings (14), this suggests that calneurons are the first EF-hand proteins belonging to the generic group of TA proteins. TA proteins contain a C-terminal TMD that serves also as a signal peptide. The targeting information is encoded in the TMD region alone and is based on its hydrophobicity, lipid binding, and length (23). We next checked whether calneurons associate with lipids by performing lipid pulldown experiments with bacterially produced His 6 -SUMO fusion constructs of calneuron-1 and the lipids from a bovine Folch fraction I. Interestingly, full-length calneuron-1 showed only weak binding to the lipids in a Ca 2ϩ -dependent manner, although most of the protein remained in the supernatant fraction (Fig. 2D). Calneuron-1_⌬C, a construct lacking the TMD, also exhibited only weak binding that was Ca 2ϩindependent (Fig. 2D). In contrast, the calneuron TMD, including the 5 aa of the intraluminal part, was quantitatively pulled down with the lipids after ultracentrifugation at 140,000 ϫ g (Fig. 2D). These data suggest that the hydrophobic C terminus of calneuron-1 might be masked in full-length calneuron-1 by intramolecular or intermolecular interactions and could be exposed by Ca 2ϩ -induced conformational changes. To address the question with which the particular membrane lipid calneuron-1 associates, we performed a PLO assay. We found that untagged calneuron-1, expressed in COS-7 cells (Fig. 2E), associates strongly with lipids that are enriched in the TGN and the plasma membrane, namely PI(4)P, PI(4,5)P 2 , and PI(3,4,5)P 3 (Fig. 2F). Interestingly, and different from Hippocalcin (24), which interacts equally well with PI(3)P, PI(4)P, and PI(5)P, calneuron-1 did not show clear binding to other PIPs.
Calneurons Are Inserted in ER Membranes and Then Accumulate in the Golgi-No instances of direct Golgi insertion of newly synthesized TA proteins have been reported, and we therefore reasoned that calneurons are inserted into ER membranes following translation and then transported toward the TGN via the classical membrane trafficking pathway. To test this hypothesis, we first double transfected COS-7 cells with different combinations of calneuron-1 and ER or Golgi markers. As expected, EGFP-calneuron-1 fluorescence showed a high degree of overlap with the fluorescence of a DsRed-monomeric Golgi fusion construct of the Golgi resident enzyme 1,4galactosyltransferase (supplemental Fig. S2A). Of note, in many transfected cells, calneuron-1 was also highlighting the nuclear envelope, a structure that is in continuity with ER membranes. A more detailed analysis after co-expression of TagRFP-calneuron-1 and the ER marker GFPC1-Sec61␤ revealed a partial overlap of fluorescence for both constructs (supplemental Fig.  S2). Triple transfection with untagged calneuron-1, DsRedmonomeric Golgi, and GFPC1-Sec61␤ again indicated that although most of calneuron-1 staining is restricted to the Golgi, a low intensity signal is also found at the ER (Fig. 3A).
To confirm that Golgi-localized calneuron-1 is inserted first in the ER, we disrupted ER-to-Golgi protein transport by prolonged incubation with brefeldin A. COS-7 cells, triple transfected with untagged calneuron-1, GFP-Sec61␤, and DsRedmonomeric Golgi showed a clear redistribution of calneuron-1 staining toward the ER. A similar effect was seen for Golgi resident 1,4-galactosyltransferase expressed as DsRed-Monomeric-Golgi construct (Fig. 3B). To visualize ER-to-Golgi transport of calneuron-1 with an independent method, we implemented FRAP experiments with COS-7 cells transfected with EGFPcalneuron-1 and DsRed-monomeric Golgi (Fig. 3C). We obtained base-line fluorescence recordings from cells where calneuron-1 localized to the Golgi as well as to Golgi marker negative reticular structures resembling ER in the cytosol. After acquiring images of the initial fluorescence, the Golgi fraction of EGFP-calneuron-1 was photobleached, and fluorescence recovery was monitored over 20 min with 1 frame/min ( Fig. 3C; supplemental Fig. 3 movie). An average reduction of 18% from the initial fluorescence was calculated for the Golgi area after photobleaching (Fig. 3D). Fluorescence recovery in these experiments correlated with the reduction of EGFP-calneuron-1 fluorescence at the ER. No significant decrease was measured in the nucleus where soluble calneuron was diffusely distributed. ER and the Golgi complex are not completely spatially segregated, and both organelles can be found in the perinuclear somatic area. To exclude the possibility that only lateral diffusion of calneuron-1 within the ER was measured, we compared FRAP dynamics for EGFP-calneuron-1 and GFP1-Sec61␤. Previous studies on GFP-tagged lumenal and membrane ER markers subjected to FRAP revealed that these proteins rapidly diffuse throughout the ER, with no regional restriction for diffusional exchange of these molecules (25). We observed similar fast kinetics for GFPC1-Sec61␤, and the recovery rate was considerably faster than the Golgi FRAP of EGFP-calneuron-1 (Fig. 3E). Because calneuron-1 is also associated with the PM (Fig. 1, A and B), the next question was how calneuron-1 is transported to the PM. After passing through the Golgi complex, secretory cargo was packaged into post-Golgi transport FIGURE 2. Calneuron-1 is a tail-anchored protein that localizes to TGN membranes. A, most of EGFP-calneuron-1 fluorescence is localized to the Golgi area defined by TGN 38 staining, and the 23 amino acids at the C terminus of calneuron-1 (Caln1) are sufficient for its TGN localization. B, schematic representation of EGFP-calneuron-1 constructs used to map the minimal region required for efficient Golgi localization. C, quantification of TGN targeting for different EGFP-calneuron-1 deletion constructs. Box plot reflects the distribution of EGFP fluorescence between the Golgi and total cell area. *, group comparison between EGFP and calneuron-1 constructs; E, group comparison between full-length calneuron-1 and its deletion constructs. *, p Ͻ 0.05; ***, p Ͻ 0.001. D, full-length His 6 -SUMO-calneuron-1 shows Ca 2ϩ -modulated interaction with lipids of the Folch fraction 1 from bovine brain in sedimentation assays. The C terminus of calneuron-1 (23 aa of the TMD ϩ 5 intraluminal aa) is efficiently pulled down with lipids in a Ca 2ϩ -independent manner. S stands for supernatant and P for pellet fraction. E, untagged calneuron-1 overexpressed in COS-7 cells and detected by an anti-calneuron-1 rabbit antibody. No endogenous calneuron-1 or additional unspecific bands were detected. 20 g/lane of total protein was loaded. F, for the protein-lipid overlay assay, an extract of COS-7 cells, expressing untagged calneuron-1, was incubated with two different lipid strips and developed with the same anti-calneuron-1 rabbit antibody. Note the binding to the negatively charged phosphoinositides PI (weak), PI(4)P, PI(4,5)P 2 , and PI(3,4,5)P 3 . TG, triglyceride; DG, diacylglycerol; PA, phosphatidic acid; PS, phosphatidylserine; PE, phosphatidylethanolamine; PC, phosphatidylcholine; PG, phosphatidylglycerol; CL, cardiolipin; SM, sphingomyelin; LPA, lysophosphatidic acid; LPC, lysophosphocholine; S-1-P, sphingosine 1-phosphate. OCTOBER  intermediates, which translocate to the PM. We utilized a temperature-sensitive mutant of VSV-G for the Golgi-to-PM trafficking assay. We found that calneuron-1 immunofluorescence shows overlap with GFP-VSV-G at the Golgi complex and at the post-Golgi intermediates 20 min after removing the Golgi block (Fig. 3F). These data indicate that fraction of calneuron-1 protein is transported through the same secretory pathway as VSV-G to the PM.

Calneurons Are Targets of the TRC40/Asna1 Chaperone
Calneurons Are Novel Targets of TRC40/Asna1 Chaperone-Classically, newly synthesized type 1 transmembrane proteins are co-translationally inserted into membranes via interaction with signal recognition particle (SRP) and Sec61 translocon machinery (26,27). In TA proteins, the hydrophobic TMD is not exposed to the cytosol until the polypeptide is terminated and released from the ribosome. Therefore, the newly synthesized protein will not have a chance to interact with SRP, and insertion into the ER must occur posttranslationally (28). TA proteins can be either spontaneously inserted into ER membranes or with the assistance of the Hsp40/Hsc70 or TRC40/Asna1 chaperone, and the hydrophobicity of the TMD usually defines which mechanism will be used (29). More hydrophobic TMDs need assistance for insertion, although less hydrophobic TMDs have been shown to be capable of "spontaneous insertion" that occurs without the help of cytosolic or membrane-associated factors (28). Conversely, TMDs with a net hydrophobicity of ϳ35 display an obligatory requirement for Hsp40/Hsc70. However, none of the TA proteins with a net hydrophobicity above 40 are solely dependent upon the Hsp40/Hsc70 route, and they instead utilize TRC40/Asna1 (29). Because of the calculated hydrophobicity scores of the 23-aa TMD of calneuron-1 and -2 of 48.8 and 52.7, respectively, the spontaneous insertion into the lipid bilayers is very unlikely, and the most likely scenario is the interaction of the nascent polypeptide with TRC40/Asna1, a dimeric cytosolic ATPase that recognizes the TMD of TA proteins (30).  To check if calneuron-1 might be a new substrate of TRC40/ Asna1 or whether it is also capable of interaction with the SPR/ translocon complex, we first performed heterologous co-immunoprecipitation from COS-7 cells co-transfected with untagged full-length calneuron-1 and YFP-TRC40/Asna1, and GFP1-Sec61␤ is one of the subunits of translocon complex or a EGFP control using a mouse GFP antibody coupled to magnetic beads (Fig. 4A). Calneuron-1 efficiently co-immunoprecipitated with YFP-TRC40/Asna1, and only a very weak band was seen after elution of immunoprecipitated GFP1-Sec61␤ (Fig.  4A). We next expressed GST-TRC40/Asna1 and His 6 -SUMOcalneuron-1 either alone or in combination in E. coli. When both proteins were co-expressed and then purified by glutathione-Sepharose, His 6 -SUMO-calneuron-1 co-purified with GST-TRC40/Asna1 (Fig. 4B, left and middle panel). GST-TRC40/Asna1 co-purification with His 6 -SUMO-calneuron-1 was also observed when affinity chromatography was performed with the Probond resin binding His 6 tag (Fig. 4B, right  panel). These data suggest that calneuron-1 and TRC40/ASNA form a complex, that led to their co-purification.
To examine whether indeed the predicted TMD of calneuron-1 is capable of interacting with GST-TRC40/Asna1, we performed GST pulldown assays with EGFP-tagged full-length calneuron-1, calneuron-1_⌬C, calneuron-1_CT, calneuronlong TMD (23 aa), and calneuron-short TMD (17 aa) (Fig. 4C). Full-length calneuron-1 showed modest binding typical for a transient interaction, whereas stronger binding was observed for the C-terminal calneuron-1_CT construct and the long TMD (23 aa) fragment (Fig. 4C). Surprisingly, we could not pull down with GST-TRC40/Asna1 the short calneuron-1 TMD fragment (17aa, Fig. 4C). A possible explanation might come from the fact that in analogy with its yeast homologue Get3, TRC40/Asna1 forms a dimer that provides the hydrophobic groove that adopts the TMD and protects it from the other interaction. The optimal length of the TMD is in the range of about 20 aa (31). But longer sequences, for instance the 25-aa TMD of RAMP4, also can be inserted via TRC40/Asna1 (32).
We next applied an in situ PLA to provide further evidence that the interaction will occur in vivo (33). A strong PLA signal was detected in COS-7 cells co-transfected with EYFP-TRC40/ Asna1 and untagged calneuron-1 (Fig. 5A, regarding antibody characterization see supplemental Fig. S3). We then analyzed the mean gray values of the PLA signal, selecting cells showing EYFP-TRC40/Asna1 fluorescence or EGFP fluorescence. Invariably, the PLA signal in cells co-transfected with EYFP-TRC40/Asna1 and calneuron-1 was much stronger than in cor- responding controls (Fig. 5B). We next utilized BRET methodology in living cells to further demonstrate a physical interaction between EYFP-TRC40/Asna1 and Rluc-calneuron-1 in vivo (Fig. 5C) (16). BRET measurements were performed in HEK-293T cells transiently co-transfected with a constant amount of Rluc-calneuron-1 and increasing amounts of EYFP-TRC40/Asna1. The BRET signal was saturated at higher concentrations of TRC40/Asna1 indicating a specific interaction between these two proteins (Fig. 5C). From the saturation curve, a BRET max of 37 Ϯ 1 and a BRET 50 of 24 Ϯ 4 was calculated. These values are in the range of modest to low affinity binding that is expected for proteins that undergo a transient interaction with TRC40/Asna1 before their membrane insertion. No BRET signal was detected in a negative control where CaM-Rluc and EYFP-calneuron-1 were used as a pair (Fig. 5D). Taken together, these results suggest calneuron-1 is a new target of the TRC40/Asna1 chaperone in vivo.
Calneurons Operate as Dimers or Potentially as Oligomers in the Membrane-Without the assistance of chaperones, calneurons as many other proteins with hydrophobic TMDs might readily interact with each other to form cytosolic dimers and oligomers, which are in many cases prone to aggregation (28). During our efforts to purify calneurons, we realized that both calneuron-1 and -2 formed soluble dimers and higher order oligomers when produced in bacteria. Interestingly, these dimers could be easily dissociated under denaturating conditions. Dimer formation has been shown previously for certain calcium sensors, and in fact some neuronal calcium sensor proteins like DREAM, for example, need dimerization for their cellular function (34 -36). We therefore decided to investigate in more detail the requisites and conditions of dimer formation. We first purified MBP-calneuron-1 and performed a pulldown assay with EGFP-calneuron-1 and EGFP-calneuron-2 expressed in COS-7 cells in the presence and absence of Ca 2ϩ . Both proteins formed homo-and heterodimers with each other regardless of the presence of Ca 2ϩ in the buffer (Fig. 6A) but not with NCS-1, another Ca 2ϩ sensor protein from the CaM superfamily. MBP and EGFP controls did not show any binding (Fig.  6A).
To check whether dimerization occurred only between the hydrophobic TMDs, we purified bacterially expressed His 6 -SU-MO-calneuron-1_⌬C and performed size exclusion chromatography using a Superdex 200 gel filtration column. Peak positions of both the apo-and Ca 2ϩ -bound form of His 6 -SUMOcalneuron-1_⌬C overlapped with each other on the elution profile (Fig. 6B). The molecular masses corresponding to the peaks and calculated from the protein standards run under the same buffer conditions (supplemental Fig. S5, A and B) are 88.6 and 78.7 kDa for apo-and Ca 2ϩ -bound His 6 -SUMO-calneuron-1_⌬C, respectively (Fig. 6B). These molecular masses are inbetween the theoretical molecular size of a dimer (70.2 kDa) and a trimer (105.3 kDa). Interestingly, we observed a reduction of intensity and broadening of the peak in the presence of Ca 2ϩ . Because the area under the curves for both the apo-and Ca 2ϩbound proteins are the same (0.030 and 0.0295 ml, respectively), the reduction in the peak intensity cannot be caused by loading unequal amounts of proteins on the column. The peak broadening indicates that either a heterogeneous population of FIGURE 6. Calneuron-1 and -2 form dimers and potentially oligomers in vitro. A, pulldown assays for recombinant MBP-calneuron-1 or MBP control and EGFP-calneuron-1, -2, or EGFP only. Both calneurons form Ca 2ϩ -independent homodimers and heterodimers with each other but not with a GFP control. IN, input. B, His 6 -SUMO-calneuron-1_⌬C lacking the hydrophobic TMD domain forms a dimer as shown by a gel filtration experiment. Addition of Ca 2ϩ leads to peak broadening, indicating that a heterogeneous population of the protein exists in the presence of Ca 2ϩ or that a conformational change upon Ca 2ϩ binding alters its mobility through the column. For column calibration see supplemental Fig. S5. C, dynamic light scattering data indicate the presence of a heterogeneous population of trimers and oligomers of recombinant His 6 -SUMO-calneuron-1_⌬C irrespective of the presence of Ca 2ϩ . D, GST-TRC40/Asna1 binds only the monomeric form of fulllength calneuron-1 or its C terminus in protein-protein overlay assay. Left, Coomassie Blue stain of SDS-PAGE with different His 6 -SUMO-calneuron-1 constructs run at semi-denaturing conditions. Right, immunoblot with anti-Asna1 mouse antibody used to detect the interaction.
Calneurons Are Targets of the TRC40/Asna1 Chaperone the protein exists in the presence of Ca 2ϩ or that a conformational change upon Ca 2ϩ binding alters its mobility on the column (Fig. 7B). To address this issue with an independent method, we performed DLS experiments. The observed mean hydrodynamic radii (and mean polydispersity) were found to be 5.607 nm (0.311) and 7.978 nm (0.387) for the apo-and Ca 2ϩbound forms, respectively (Fig. 6C). These values predict that a heterogeneous population of trimers and higher order oligomers exists in both the presence and absence of Ca 2ϩ . Thus, with both methods we found that the majority of calneuron-1 does not exist as a monomer but instead as a dimer or a higher order oligomer.
Taken together, these results suggest that calneuron-1 (and probably calneuron-2 as well) has two dimerization interfaces, a hydrophobic TMD and the cytosolic EF-hand containing part of the protein. For the latter, Ca 2ϩ could play a role for the on/off rate of dimerization. At this point, we cannot conclude that in solution only the N-terminal part of the protein would contribute to the dimer because the TMD region of the fulllength His 6 -SUMO-calneuron-1 appears to be inaccessible for the interaction with lipids from a bovine brain Folch fraction I (Fig. 2D). Consequently, the obvious question arises whether the dimer of calneuron-1 can still bind the TRC40/Asna1 chaperone and thereby has the potential for insertion into the ER membrane. To address this question, we performed proteinprotein overlay assays with recombinant GST-TRC40/Asna1 and different calneuron-1 fusion proteins. His 6 -SUMO-calneuron-1, His 6 -SUMO-calneuron-1_⌬C, His 6 -SUMO-calneuron-1_CT, and a His 6 -SUMO control protein were subjected to SDS-PAGE at low SDS and ␤-mercaptoethanol concentrations. Under these semi-native conditions, calneuron-1 migrates as two bands corresponding to the dimer (70 kDa for the full length) and the monomer (35 kDa for the full length, Fig. 6D, left panel). After developing the blot with an Asna1 mouse antibody, specific bands could be observed only at the size of the monomer of both His 6 -SUMO-calneuron-1 and His 6 -SUMOcalneuron-1_CT. Interestingly, a stronger interaction was again found for the C-terminal fragment. No binding to His 6 -SUMOcalneuron-1_⌬C or His 6 -SUMO control was detected (Fig. 6D,  right panel). In summary, this suggests that calneuron-1 can only interact with TRC40/Asna1 as a monomer. Of note, a considerable amount of GFP-calneuron-1, extracted from COS-7 cells, migrates as a dimer under native and semi-native conditions (supplemental Fig. S4, A and B). The monomer/dimerization distribution might account for the relatively low propor-tion of GFP-calneuron-1 that is capable of GST-TRC40/Asna1 binding in the pulldown assay (Fig. 4C).
To learn whether an endogenous calneuron dimer might also exist in vivo, the extract from HeLa cells (containing 1% of Triton X-100) was subjected to native PAGE. Immunoblot with anti-calneuron-2 antibody indicated the presence of bands corresponding to the size of monomeric, dimeric, and higher order oligomeric calneuron-2 (Fig. 7A). The bands corresponding to dimers and oligomers disappeared when semi-native or nonnative sample dye was used to solubilize the sample (Fig. 7A). To obtain independent proof for in vivo dimer and to learn about the orientation of the dimer, we employed BRET technology in living cells using different combinations of N-and C-terminally Rluc-and EYFP-targeted calneuron-1 fusion constructs. If both proteins dimerize in a parallel head-to-head (N-terminal to N-terminal) orientation, then only protein pairs with the Rluc and EYFP at the same terminus will exhibit energy transfer. BRET experiments revealed that indeed only a combination of Rluc-calneuron-1 and EYFP-calneuron-1 showed a positive and saturable BRET signal that was not affected much by increasing intracellular Ca 2ϩ levels (Fig. 7B, left panel). From the BRET saturation curves, a BRET max of 169 Ϯ 8 or 164 Ϯ 15 milliBRET units (mBU) and a BRET 50 of 24 Ϯ 6 or 28 Ϯ 7 were calculated in the absence or in the presence of ionomycin, respectively. Importantly, only a very low and linear BRET for the Rluc-calneuron-1 and calneuron-1-EYFP pair was detected, suggesting that only a parallel dimer forms in vivo (Fig. 7B, right  panel).
During transfection studies in COS-7 cells, we always saw a low intensity calneuron-1 signal in the cytosol and the nucleus (supplemental Fig. S1). We therefore wondered whether a cytosolic calneuron-1 pool might exist as a parallel dimer, where the TMD is not accessible for the interaction with TRC40/Asna1 and subsequent membrane insertion. HeLa cells express considerable amounts of calneuron-2 (14,37). Using HeLa cell lysates, we performed ultracentrifugation experiments to obtain a cytosolic and a membrane fraction, which includes microsomes. Subsequent immunoblotting revealed that calneuron-2 is exclusively associated with the membrane fraction (Fig. 7C). Similarly, Syntaxin 6, a classical transmembrane Golgi SNARE protein and also a TA protein (38), was not detectable in the cytosolic fraction even after long exposure times, but instead all protein was pelleted with membranes (Fig. 7C). It is therefore likely that calneurons are predominantly membrane proteins and that a considerable cytosolic pool only exists after A, native PAGE indicates the presence of higher order bands (dimers and multimers) for endogenous calneuron-2 extracted from HeLa cells. These bands disappear when sample is solubilized in semi-native Sample Buffer or solubilized and heated up to 100°C to 5 min in non-native Sample Buffer. B, calciumindependent BRET is detected only for the parallel dimer (Rluc and EYFP tags are at the N terminus) of calneuron-1; no energy transfer was measured in the absence or presence of Ca 2ϩ for the antiparallel dimer (Rluc from the N terminus and EYFP from the C terminus of calneuron-1 and vice versa) (n ϭ 8 for each experiment). mBU, milliBRET. C, endogenous calneuron-2 exists only in the membrane-bound form in HeLa cells. Following differential ultracentrifugation of HeLa cell homogenate, 20 g of protein from each fraction were loaded on SDS-PAGE and analyzed by immunoblotting. Immunoreactivity for calneuron-2 is seen in membrane-containing fraction P2 but not in the SN2 cytosolic fraction obtained after 2 h of centrifugation at 400,000 ϫ g. The band corresponding to TGN marker Syntaxin-6 is present only in homogenate and P2 fraction. Equal loading of fractions was confirmed by the comparison of ␤-Actin bands in all the fractions. H, homogenate. D, large population of membrane-bound calneuron-1 (Caln1) is visible as a parallel dimer in fluorescence complementation assay. COS-7 cells were co-transfected with calneuron-1 constructs fused at the N or C terminus to the N-or C-terminal split-YFP (Venus) tags. Complementation is observed only in the case of the parallel dimer. COS-7 cells were also stained with the Golgi marker GM130 and nuclear marker DAPI. E, higher magnification images for the combinations of split YFP-tagged calneuron-1 forming a parallel dimer with the tag reconstituted either in cytosol or inside of the Golgi/ER lumen. Note the high overlap of the YFP and GM130 fluorescence. A weak YFP signal was also observed in the ER (nuclear envelope), vesicular structures, and the plasma membrane. overexpression when TRC40/Asna1 chaperone is oversaturated with the excess of newly produced protein.
Because we observed a strong dimerization with BRET in vivo, but could not find an endogenous cytosolic calneuron-2 pool, we wondered what would be the subcellular localization of the protein dimer. Therefore, we employed a fluorescence complementation assay where calneuron-1 was fused at the N or C terminus to the N-or C-terminal part of the split YFP variant Venus. This method allows for the visualization of a direct protein-protein interaction in vivo (39,40), because the N-and C-terminal parts of Venus can only assemble a functional fluorophore when they are brought in close proximity. In agreement with the BRET experiments, we could only detect parallel dimers formed in both cases when calneuron-1 was fused N-or C-terminally with the N-and C-terminal halves of Venus (Fig. 7, D and E). The fluorescence was detected again only at the Golgi complex, and the distribution pattern was very similar to those of untagged or EGFP-tagged calneuron-1 (Fig.  7, D and E).

DISCUSSION
In previous work, we have demonstrated that both endogenous and overexpressed calneuron-1 and -2 are localized at the TGN in pyramidal neurons as well as in secretory cells (11). The localization of calneurons tightly correlates with their function as "calcium threshold filters" for activation of PI-4KIII␤, production of PI(4)P, and regulation of vesicular trafficking from the TGN to the plasma membrane. Shih et al. (20) have recently shown that a fraction of calneuron-1 resides at the PM where calneuron-1 is involved in inhibition of N-type Ca 2ϩ channels. Interestingly, the presence of the hydrophobic C-terminal fragment is required to induce this inhibition (20). In this study, we addressed two questions important for the understanding of calneuron biosynthesis and function at the TGN and PM. Why are these proteins enriched at the TGN and which route of post-translation insertion do they utilize?
It was previously suggested that calneurons are transmembrane proteins, which is an uncommon feature for CaM-like EF-hand Ca 2ϩ sensors (14). We could confirm these findings and found further evidence that they are indeed TA proteins. However, in accordance with the requirement for TRC40/ Asna1 binding for membrane insertion and TGN membrane thickness, the TMD region and minimal TGN targeting sequence appears to be longer than those published previously (13,14). The joint feature of the heterogeneous group of TA proteins is that they harbor a TMD at their C terminus and undergo post-translation insertion into different membrane organelles, including the outer membrane of bacteria, membranes of mitochondria, and chloroplasts, peroxisomes, and ER membranes (41). Once TA proteins are integrated into the ER membrane, they can also be sorted to other membranes within the secretory pathway (42). TA proteins are always oriented in the membrane with the larger N-terminal region facing the cytosol. This region is usually important for the biological function of the protein (23). In the case of calneurons, the EF-hand domains are exposed to the cytosol where they are involved in Ca 2ϩ -dependent regulation of vesicular trafficking at the TGN (11,14) and inhibition of N-type Ca 2ϩ channels (20) at the PM.
Trafficking from the TGN to PM appears to follow the route of VSV-G. In turn, this provides a unique mechanism for a highly restricted localization of these Ca 2ϩ sensor proteins, and calneuron-2 indeed appears to be exclusively situated at HeLa cell membranes.
In addition, we found that the 23-aa-long TMD of calneuron-1 is necessary and sufficient for TGN localization of the protein. The previously suggested 17-aa shorter TMD region (13) co-localized less efficiently with the Golgi marker GM130, and we observed a more diffuse distribution that is commonly observed for ER proteins or proteins present in the cytosol. It has been shown previously that information decoded in the length and the hydrophobicity of the TMD defines the destination of TA proteins (20,32,42,43). The reason why calneuron-1_CT and calneuron-1_23aa are targeted to the TGN more efficiently than the shorter TMD probably relates to the properties of the TGN membrane bilayers. Membranes of different intracellular compartments have different lipid compositions that result in increasing thickness and decreasing fluidity of the bilayers from the ER-to-Golgi complex and the PM (44). Glycerophospholipids and sphingolipids are initially synthesized in the ER and then transferred between different compartments (45,46). Lippincott-Schwartz and co-workers (46) suggested the so-called "rapid partitioning model." Based upon the cellular glycerophospholipids and sphingolipid compositions of ER, Golgi, and plasma membranes, they propose a steady state cisto-trans gradient in sphingolipid/glycerophospholipid ratio across the Golgi stack (46). Several lines of evidence suggest that TGN membranes are thicker than ER membranes (46,47), and a number of proteins depend on features of their TMD for correct sorting between the ER and the Golgi apparatus (46,48). TA proteins frequently require short and moderately hydrophobic TMDs to remain resident at the ER, where they are excluded from the more ordered bilayer regions that get transported to the Golgi (49). For example, Ceppi et al. (44) could demonstrate that the wild type form of the ER resident TA protein cytochrome b 5 with a 17-aa TMD could be found at the PM when five extra nonpolar amino acids were fused to extend the transmembranal helix. Apart from the longer TMD, another reason for the localization of calneurons to the post-ER components of the secretory trafficking pathway is probably the interaction of calneuron-1 with phospholipids enriched in TGN, endosomes, and the PM. We found specific binding to PI(4)P-phosphoinositide, a lipid characteristic for the TGN (50). PI(4)P is a lipid product of TGN localized PI-4KIII␤ (10) that is required to recruit several effectors to promote the budding and fission of Golgi-derived transport vesicles (10). Moreover, PI(4)P is a precursor of PI(4,5)P 2 , which in turn is enriched at the PM; and in a PLO assay, we also found binding of calneuron-1 to this phosphoinositide.
TA proteins are first inserted into the ER membranes and, later on, depending upon their features, distributed to other compartments of the secretory pathway. With blocking of ERto-Golgi protein transport by applying brefeldin A and with independent FRAP experiments, we could show that EGFPcalneuron-1 is indeed transported from the ER to the Golgi. Moreover, we show that calneuron-1 strongly co-localizes with VSV-G post-Golgi clusters 20 min after removing the Golgi block, suggesting that fraction of the protein follows the route of VSV-G trafficking from ER-to-plasma membranes. The highly hydrophobic TA core of calneurons as expected requires assistance of chaperones for the integration into the ER membrane (23). Calneuron-1 interacts with the TRC40/Asna1 complex in vitro and in vivo, and this complex only binds the monomeric form of calneuron-1. Of note, we also observed a weak binding between calneuron-1 and Sec61␤, one of the subunits of Sec61 protein translocation channel associated with SRP complex. Interestingly, Sec61␤ is actually a TA protein by itself and it is inserted into the ER membranes via TRC40/Asna1 (30). SRP-dependent delivery of TA proteins to the ER has been reported previously, and this route has been shown to act as a complementary pathway for TA proteins such as Syb2 and Sec61␤ (51). Therefore, the possibility that calneuron-1 in the absence of TRC40/Asna1 might also utilize the SRP-dependent insertion mechanism cannot be excluded.
Finally, we found that calneurons dimerize and potentially multimerize in vitro and in vivo. We found no evidence that a parallel dimer might give rise to a cytosolic pool that is not accessible for TRC40/Asna1. Instead, the dimer appears to be present at the TGN and potentially in later steps of the secretory pathway. Interestingly, we found that the N-terminal domain lacking the TMD is capable of dimerization by itself. At present, it is therefore unclear how the membranous dimer is formed, but it has been shown that dimerization of the CaMlike part can play an important role for the function of EF-hand Ca 2ϩ sensors (36). A calneuron dimer or multimer might also regulate the retention of the protein at the TGN membrane as has been shown for other membrane proteins (52,53). It is also conceivable that self-association of calneurons can influence membrane curvature, which has been shown for many other proteins with a large cytosolic domain and a membrane spanning region (18,54,55).