Ascorbate peroxidase proximity labeling coupled with biochemical fractionation identifies promoters of endoplasmic reticulum–mitochondrial contacts

To maintain cellular homeostasis, subcellular organelles communicate with each other and form physical and functional networks through membrane contact sites coupled by protein tethers. In particular, endoplasmic reticulum (ER)–mitochondrial contacts (EMC) regulate diverse cellular activities such as metabolite exchange (Ca2+ and lipids), intracellular signaling, apoptosis, and autophagy. The significance of EMCs has been highlighted by reports indicating that EMC dysregulation is linked to neurodegenerative diseases. Therefore, obtaining a better understanding of the physical and functional components of EMCs should provide new insights into the pathogenesis of several neurodegenerative diseases. Here, we applied engineered ascorbate peroxidase (APEX) to map the proteome at EMCs in live HEK293 cells. APEX was targeted to the outer mitochondrial membrane, and proximity-labeled proteins were analyzed by stable isotope labeling with amino acids in culture (SILAC)-LC/MS-MS. We further refined the specificity of the proteins identified by combining biochemical subcellular fractionation to the protein isolation method. We identified 405 proteins with a 2.0-fold cutoff ratio (log base 2) in SILAC quantification from replicate experiments. We performed validation screening with a Split-Rluc8 complementation assay that identified reticulon 1A (RTN1A), an ER-shaping protein localized to EMCs as an EMC promoter. Proximity mapping augmented with biochemical fractionation and additional validation methods reported here could be useful to discover other components of EMCs, identify mitochondrial contacts with other organelles, and further unravel their communication.

Interorganelle communication has emerged as a critical physiological function required to maintain cellular homeostasis. The functional interplay between cellular organelles is mediated through the membrane contacts formed by protein tethers within the membranes of the two organelles holding them in close proximity (1)(2)(3)(4). Different forms of membrane contacts display a wide variety of functions depending on the organelles involved and the types of membrane contacts. For instance, the endoplasmic reticulum and mitochondrial contacts (EMC) 3 are the hub for Ca 2ϩ exchange, phospholipid exchange, autophagy, mitochondrial biogenesis, and ER stress response (5)(6)(7)(8)(9)(10). At ER-plasma membrane junctions, in addition to Ca 2ϩ and lipid transport, cell signaling regulation and ER shaping occurs (11)(12)(13). Bi-directional lipid trafficking and secretion of proteins are well-known to occur at the point of ER and Golgi juxtaposition (14 -16). In yeast, mitochondrial attachments to plasma membranes are important for mitochondrial fission and partitioning in cell division (17,18).
Membrane contacts are established by interactions between proteins presented on the outer surface of each organelle, allowing subdomains of membranes to appose each other within 10 -50 nm but without fusion (1,3,19). Likely due to their important roles in cellular homeostasis, defects in the regulation of these contacts have been linked to several human diseases. For example, overexpression of familial Alzheimer's disease-associated presenilin 2 mutations significantly increases EMCs and causes an imbalance of intracellular Ca 2ϩ homeostasis (20,21). Overexpression of hereditary spastic paraplegia-associated REEP1 mutations impairs the level of ER-mitochondrial contacts and results in neuritic degeneration in cortical neurons (22). Missense mutation (P56S) in the integral ER membrane protein vesicle-associated membrane protein-associated protein B (VAPB) has been linked to atypical amyotrophic lateral sclerosis (ALS8). P56S mutations impair ER-Golgi vesicle trafficking and perturb Ca 2ϩ handling by ER-mitochondrial contacts formed by VAPB and outer mitochondrial membrane proteins like protein-tyrosine phosphatase-interacting protein 51 (PTPIP51) (23)(24)(25). Another example is BAP31. This multifunctional protein forms an apoptosis platform with the fission 1 homologue (Fis1) at ER-mitochondrial contacts, controls the quality of newly synthesized proteins in the ER, and regulates T cell receptor signaling for T cell activation. Mutations in BAP31 have been linked to X-linked intellectual disabilities with severe phenotypes such as deafness, dystonia, and congenital microcephaly (26 -29). Finally, dysregulation of EMCs, resulting from mutations in various proteins participating in contact between these two organelles, has been extensively linked to neurodegenerative diseases (30 -34).
Understanding the cellular physiology and underlying mechanisms governed by interorganelle communication is crucial to gain new insights into potential therapeutic opportunities. Unfortunately, no well-defined screening assay existed in mammalian cells. Although genetic screening has been fruitful in identifying EMCs in yeast, both similarities and differences exist in mammalian cells. For instance, components of the ER-mitochondrial encounter structure (ERMES) complex in yeast were identified and characterized through a genetic screen in which lethal mutant strains were isolated by expressing synthetic ER-mitochondrial bridge protein (35). The ERMES complex was found to be composed of ER membrane protein (Mmm1), two outer mitochondrial membrane proteins (Mdm10 and Mdm34), and a cytosolic protein (Mdm12) (35)(36)(37). Unfortunately, none of the orthologues of the ERMES complex have been identified in mammalian cells. Herein, we have developed a systematic approach to identify protein components associated with EMCs in mammalian cells.
To identify proteins that coordinate EMCs, we have adopted the proximity-labeling technique using engineered ascorbate peroxidase (APEX). APEX-mediated labeling has proven to be a powerful tool for labeling proteins in live cells (38). It has been successfully used to map the proteome of the mitochondrial intermembrane, ER-PM junctions, and proteome of primary cilia (13, 39 -43). To capture proteins associated with EMCs, we fused a mitochondrial targeting sequence to APEX and expressed it in HEK293 cells. We utilized SILAC-based quantitative LC-MS/MS to identify proteins enriched following APEX labeling. We identified 405 proteins across two independent replicates, including reticulon 1a (RTN1A), an ER-shaping protein localized to EMC. We further validated that RTN1A functions in EMC using a live cell assay. Altogether, we have characterized a new function for RTN1A, and we have defined a method to identify proteins important in inter-organelle contacts. These data provide new insights into several human diseases and new potential targets for therapy.

Targeted expression of APEX to mitochondrial outer membrane (Mit-APEX)
To identify potential target proteins associated with ER-mitochondrial contacts, we used the experimental scheme depicted in Fig. 1A. Mit-APEX cannot label ER proteins when the respective membranes are separated by more than ϳ20 nm (41,44,45). At EMC sites, both ER and mitochondrial proteins will be labeled (Fig. 1A, contact). In contrast, when ER and mitochondrial members are not juxtaposed, the APEX system only labels mitochondrial proteins (Fig. 1A, no contact). We thus anticipate the proteins identified will include not only ER-mitochondrial tethering proteins but also ER and mitochondrial proteins that are not directly engaged in EMC but may still play a functional role.
pcDNA3.1-Mit-APEX was transfected in HEK293 cells, and protein expression was determined to be stable (Fig. 1B). A single band was detected by Western blotting when probed with an ␣-FLAG antibody indicating strong expression and no obvious degradation products. Next, we confirmed the mitochondrial localization of Mit-APEX. Mit-APEX expressed in HeLa cells showed clear localization to mitochondria as assayed by overlapping expression between FLAG (Mit-APEX) and TOM20 (a subunit of the outer mitochondrial membrane protein complex) (Fig. 1C). Finally, we examined the biotinylation activity of Mit-APEX. Proximity labeling was done essentially as described previously (41). Whole-cell lysates obtained from HEK293 cells expressing Mit-APEX were subjected to SDS-PAGE, and biotinylated proteins were probed with streptavidin-HRP. In the absence of Mit-APEX, no biotin labeling was detected (Fig. 1D, lanes 1 and 2). However, in the presence of Mit-APEX and the biotin substrate (but not in the absence of this substrate), strong biotinylation was present (Fig. 1D, lanes  3 and 4). Moreover, we compared our Mit-APEX targeted to the outer mitochondrial membrane with APEX expressed in the mitochondrial matrix (41); they showed distinctive biotinylation patterns indicating specificity to the Mit-APEX targeting (data not shown). Finally, we performed pilot proximity labeling to know whether Mit-APEX can label known ER-mitochondrial tethers, inositol 1,4,5Ј-triphosphate receptor II (IP3RII) calcium channel, vesicle-associated membrane protein-associated protein B (VAPB), and B cell receptor-associated protein 31 (BAP31). As predicted, Mit-APEX labeled those ER-mitochondrial tethering proteins, which is likely at EMCs. Taken together, these data indicate we successfully targeted an active and stable form of APEX to the outer mitochondrial membrane, and it is suitable for mapping proteins at EMCs.

Purification of proteins proximity-labeled by Mit-APEX after SILAC and biochemical subcellular fractionation
To facilitate the identification of proteins labeled by APEX at the ER-mitochondrial interface, we adopted a quantitative LC-MS/MS approach relying on SILAC (46 -48). HEK293 cells were equilibrated in light or heavy isotope-containing medium for a minimum of five passages, with the efficiency of isotope incorporation assessed by mass spectrometry (see Fig. 2A for experimental protocol). Excess proline was used as a supplement (200 mg/liter) to prevent the metabolic conversion of heavy arginine to heavy proline (49). Cell cultures equilibrated in SILAC media were either mock-transfected (SILAC-light) or transfected with pcDNA3.1-Mit-APEX (SILAC-heavy). An independent biological replicate with a reverse labeling strategy was also prepared. After labeling with heavy amino acids, proximity labeling by Mit-APEX was performed as described under "Experimental procedures," and then the microsomal fraction, containing ER, was isolated from cells by differential centrifugation. The biotinylated proteins recovered from the microsomal fractions by releasing them from streptavidin-magnetics RTN1A promotes ER-mitochondrial contacts beads, and this was followed by LC/MS-MS analysis ( Fig. 2A). Western blotting after biochemical fractionation validated that most proteins from the cytosol (tubulin) and nucleus (TCF7L2) were removed, whereas ER proteins (calnexin) and some mitochondrial proteins (ATP5A) were retained (Fig. 2B). We next compared biotinylation patterns in each fraction and found a nice separation of microsomal fractions and biotinylated proteins (Fig. 2B, streptavidin-HRP). As a final quality control step prior to mass spectrometry analysis, a separate protein purifi-cation was performed from control and experimental samples, and eluted proteins were subjected to SDS-PAGE/silver staining (Fig. 2C).

Proximity labeling by ascorbate peroxidase identified RTN1A as a promoter of ER-mitochondrial contacts
Mass spectrometry identified ϳ1,100 proteins that were common to experiment 1 (control cultured in light isotope, Mit-APEX in heavy isotope) and experiment 2 (control cul- Biotinylated proteins by Mit-APEX were detected using streptavidin-HRP. E, HEK293T cells were transfected with pcDNA-Mit-APEX, and pilot biotinylation was performed using whole-cell lysate. Biotinylated proteins were purified with streptavidin-magnetic beads, and the eluates were run on SDS-PAGE. Proteins blot was probed with antibodies as indicated.

RTN1A promotes ER-mitochondrial contacts
tured in heavy isotope, Mit-APEX in light isotope). To further identify significantly enriched proteins, we analyzed SILAC results based on fold enrichment ratios instead of comparing abundance among proteins because the absolute level of proximity-labeled proteins is dependent on the numbers of tyrosine residues accessible by biotin-phenol radicals. To determine the cutoff ratio, we normalized both replicates using quantile normalization method in R (RStudio Team (2015) RStudio: Integrated Development for R. RStudio, Inc., Boston). Cutoff ratios were determined by the fold change of IP3Rs (2.0 of log base 2), a known EMC tether. Applying a 2.0-fold cutoff ratio (log base 2), we identified 405 overlapping proteins in the two experiments (Table 1). A gene ontology (GO) analysis was performed using Functional Enrichment analysis tool, Funrich (50). For GO term cellular component (CC) definitions, the top 15 clusters (p value Ͻ 0.05) are reported (Fig. 3).
In the CC, the cytoplasmic (66.3%) cluster was omitted, and sub-terms were presented to show greater specificity of the results (Fig. 3). In total, 88 proteins were annotated to ER by Funrich (Table 1). Among the proteins, we identified inositol 1,4,5Ј-triphosphate receptor II (IP3RII) calcium channel, vesicle-associated membrane protein-associated protein B (VAPB), and B-cell receptor-associated protein 31 (BAP31), each well-known ER-mitochondrial tethers. The presence of the tethers in our screen was confirmed by Western blotting (Fig. 4B).

RTN1A promotes ER-mitochondrial contacts
The atlastin (ATL2) and reticulon (RTN1 and RTN3) ERshaping proteins were identified as interesting novel targets because of their known role in regulating ER structure, which is likely critical in forming EMCs. Furthermore, a study in yeast showed ER-shaping proteins facilitate lipid transfer from the ER to the mitochondria, which occurs at EMCs (51). When fold changes in SILAC are compared, RTN1A (2.7-fold increase) was higher than the known EMC tether IP3RII (2.0-fold in replicates 1 and 2) and VAPB (2.3-fold). For all of these reasons, we chose RTN1A as an important protein to study.
The coordinated action of ATLs and RTNs is important for forming and stabilizing the tubular ER structure (52)(53)(54)(55). However, little is known regarding their role in EMCs in mammals. To test whether ATLs and RTNs affect ER-mitochondrial contact, we used our Split Rluc8 complementation assay to test ATLs and RTNs (Fig. 4A). Notably, RTN1A created a 4.4-fold increase in ER-mitochondrial contacts (Fig. 4D), but no change was observed with ATLs (data not shown). RTN2B (2.5-fold) and RTN3B (1.3-fold) showed small but statistically significant fold increases in EMCs. Taken together, these results indicate RTN1A has a role in forming ER-mitochondrial contacts.
Interestingly, we noticed that RTN1A, which has a longer N terminus than the other RTN family members RTN1C, RTN2B, and RTN3B, showed higher Split-Rluc8 activity (ERmitochondrial contacts, Fig. 4, C and D). We postulated that the N terminus of RTN1A might have an important role in ER-mitochondrial contacts. To test this, we used the N-terminal (1-568 amino acids) deletion construct of RTN1A (RTN1A-⌬N568) and the N terminus of RTN1A (1-568 amino acids)-RTN3B fusion construct (RTN1A-N568-RTN3B). The N-terminal deletion construct of RTN1A showed minimal activity, whereas the RTN1A-RTN3B fusion construct increased EMCs (Fig. 4, C and E). Consistent with the Split-Rluc8 complement assay results, the N-terminal deletion of RTN1A significantly reduced the level of biotin labeling by Mit-APEX, but full-length RTN1A was enriched (Fig. 4F). These data indicate the N terminus of RTN1A is critical for this activity, but it does require the C terminus from an RTN family member.

Discussion
The cell's cytoplasm is tightly packed with organelles, macromolecules, and small molecules (56, 57). Communication

RTN1A promotes ER-mitochondrial contacts
between the organelles is highly orchestrated and essential for normal cellular homeostasis (1-3, 58). Indeed, a growing body of evidence indicates that a defect in inter-organelle communication may play a role in the pathogenesis of multiple human diseases (30,33,59,60). More specifically, abnormalities associated with EMCs have been reported to be linked to multiple neurodegenerative disorders (20,30,61). However, the pathological mechanisms underlying these disorders remain to be fully clarified. Identifying the full repertoire of proteins regulating ER-mitochondrial contacts and understanding molecular events occurring between ER and mitochondrial contacts are critical to developing potential therapeutic strategies.
In this study, we applied proximity labeling using engineered APEX to identify proteins that exist within ER-mitochondrial contacts. The short half-life (Ͻ1 ms) and small labeling radius (Ͻ20 nm) of the biotin-phenoxyl radical converted by APEX enable capture of proteins located in close proximity in live cells (13, 39 -43). To capture proteins associated with ER-mitochondrial contacts, we targeted APEX to the outer mitochondrial membrane in HEK293 cells, enriched the micro-somal fraction, and purified biotinylated proteins using streptavidin-magnetic beads. Although proximity labeling by APEX is highly specific, a combination of SILAC labeling and biochemical purification improved the specificity of targets identified by proximity labeling, particularly when APEX is exposed to the outside of membrane-enclosed cellular compartments. Indeed, inclusion of a fractionation step removed a large portion of cytosolic, nuclear, and mitochondrial biotinylated proteins (see Fig. 2B).
LC-MS/MS analysis identified 405 proteins overlapping from replicate experiments with a cutoff ratio 2.0 (log base 2). In total 88 proteins were annotated to the ER by Funrich ( Table 1). Identification of inositol 1,4,5Ј-triphosphate receptor II (IP3RII) calcium channel, VAPB, and B-cell receptor-associated protein 31 (BAP31), all well-known ERmitochondrial tethers, demonstrated the robustness of our approach (Table 1 and Fig. 4B). Unexpectedly, we did not identify mitofusin 2 (MFN2), which is known to participate in EMC. It is possible electron-rich residues such as tyrosine in MFN2 might not be accessible by phenoxyl radical. For

RTN1A promotes ER-mitochondrial contacts
instance, the electron-rich residues could be masked by interacting proteins with MFN2.
RTNs are a group of evolutionarily conserved proteins residing predominantly in the ER, despite not containing a known ER localization sequence (52,53). RTNs share highly conserved C-terminal reticulon homology domain, but their N-terminal domains are divergent in length and sequence. The N-terminal domain of RTN1A (isoform A, the longest form) might have a specific role in bridging ER-mitochondria in addition to formation of tubular ER given that RTN1C and RTN3B (shorter forms of RTNs among those tested) showed minimal effect on induction of ER-mitochondrial contacts (see Fig. 4, C and D).
Indeed, our data showed the N terminus (1-568 amino acids) of RTN1A has a functional role in coupling EMCs. This result suggests that RTN1A has several functions in its N terminus beyond its conserved role in ER shaping. Differential up-and down-regulation of RTNs have been linked to neurodegenerative diseases (e.g. Alzheimer's disease, amyotrophic lateral sclerosis, multiple sclerosis, as well as hereditary spastic paraplegia) (62)(63)(64). RTN1A has been implicated in the onset of several neurodegenerative diseases; however, what role it plays is unknown (62,65). Disease-linked mutations in RTN1A, specifically in its N terminus, have not been reported. RTN1A appears to be overexpressed in some chronic kidney diseases, CRluc-ER alone have no expected luciferase activity. However, when in proximity to each other they establish functional complementation, thus assembling an active luciferase that, in the presence of the Enduren live cell substrate, generates a luminescence product. The level of ER-mitochondria is measured as luminescence by catalytic conversion of Enduren substrate. B, Western blotting of biotin-phenol-labeled ER-mitochondrial tethering proteins. Proteins purified using streptavidin magnetic beads after biotin-phenol labeling by Mit-Apex were run on SDS-PAGE and probed with antibodies as indicated. C, scheme of reticulons used in this study, the numbers correspond to exons. D, reticulons (RTN1A, RTN2B, and RTN3B) increased Split Rluc8 activity, whereas RTN1C showed no increase. Empty control vector and Reep1 expression vector were used as negative and positive control, respectively. ***, p Ͻ 0.0001, and **, p Ͻ 0.001 (n ϭ 3). E, N-terminal (1-568 amino acids) deletion mutants of RTN1A (RTN1A-⌬N568) and the RTN1A N terminus (1-568 amino acids)-RTN3B fusion constructs (RTN1A-N568-RTN3B) were expressed in HEK293 cells, and the Split Rluc8 activity was compared between proteins indicated. F, RTN1A and RTN1A-⌬N568 were co-expressed with Mit-APEX in HEK293 cells, and biotin labeling and purification by streptavidin-beads were performed. The level of proximity labeling efficiency was compared by Western blotting.

RTN1A promotes ER-mitochondrial contacts
which is thought to increase ER stress (66). Because ERmitochondrial contacts are linked to ER stress, the unfolded protein response, and the formation of autophagosomes, we speculate that dysregulation of RTN1A expression might possibly be related to the onset of the disease. Although further studies are required to elucidate the exact function of RTN1 in EMCs, we suggest that dysregulation of EMCs caused by the ER-shaping protein RTN1A might underlie several neurodegenerative diseases.
The microsomal fraction we used in this study includes vesicle-like artifacts re-formed from pieces of the ER but also membrane vesicles formed from any membrane structure that was soluble after centrifugation. Because the artificial vesicles could be formed with pieces of membrane derived not only from ER but from mitochondria, Golgi, plasma membrane, nuclear envelope, etc., the microsomal fraction we obtained is heterogeneous. We adopted the terminology first used when the protocol was first published to avoid any confusion (67).
Interestingly, the CC analysis significantly clustered membrane-enclosed organelles, including endoplasmic reticulum, mitochondria, lysosomes, exosomes, and nuclei. These data suggest mitochondria make contacts with many other organelles, as expected, and the proteins on them can be mapped by Mit-APEX at the moment they come in close proximity.
CC analysis also identified microtubule components, which are likely due to the association of mitochondrial and microtubules for regulation of mitochondrial motility and fusion/fission. Interestingly, ribosome, ribonucleoprotein complex, and translation initiation complex were enriched, implying that protein translation activity is also closely related to the outer mitochondrial membrane. Gene set enrichment analysis using Funrich clustered the majority of proteins into membranebound compartments indicating that mitochondria establish contacts with many different organelles. Previous studies support these results. For example, ER-mitochondrial contacts have been most extensively studied and linked to Ca 2ϩ and lipid exchange, intracellular trafficking of ER and mitochondria, ER stress response, autophagy, mitochondrial biogenesis, and inflammasome formation (5-7, 10, 34, 68 -71). Peroxisomes and mitochondria functionally and physically interact to maintain lipid homeostasis through ␤-oxidation of fatty acids and scavenging reactive oxygen species (72,73). Mitochondria are also known to form stable contacts with Golgi and the nuclear envelope (74,75). Although little is known regarding mitochondria and plasma membrane contacts, the Num1-Mdm36 complex was discovered as a tethering complex in yeast (17,76). Interestingly, our screen identified stromal interaction molecule 1 (STIM1), which is involved in regulation of store-operated Ca 2ϩ entry at ER-PM junctions. This raises the question whether mitochondria might have a role in store-operated Ca 2ϩ entry (13).
In conclusion, we have successfully applied proximity labeling by Mit-APEX to discover candidate proteins localizing at the contact sites between mitochondria and other organelles. We have identified and demonstrated that RTN1A increases EMCs. Although we focused on one identified EMC protein, we realize there are many additional candidates derived from our assay, including those localized at the contact sites between mitochondria and other membrane-surfaced organelles such as the Golgi apparatus, peroxisome, and nucleus.

Cell culture, SILAC, and proximity labeling by Mit-APEX
HEK293T cells (American Type Culture Collection, Manassas, VA) were cultured in Dulbecco's modified Eagle's medium (Invitrogen), with 10% fetal bovine serum (Invitrogen) at 37°C with 5% CO 2 . For SILAC, HEK293T cells were metabolically labeled with Heavy L-arginine and L-lysine using the SILAC protein quantitation kits (Thermo Fisher Scientific, DMEM). The labeling efficiency was assessed by mass spectrometry before experiments were performed. Fully labeled HEK293T cells were cultured at 37°C with 5% CO 2 . For biotin labeling of proteins, cells (5 ϫ 10 6 /100-mm dish, 4ϫ) were plated and transfected with pcDNA3-Mit-APEX using PEI (ratio of DNA/ PEI ϭ 1:3, Polysciences, Warrington, PA). After 24 h in culture, the cells were incubated with biotin-phenol (500 M) for 30 min at 37°C and then treated with 1 mM H 2 O 2 for 1 min at room temperature. The biotinylation reaction was quenched by replacing medium with PBS containing 10 mM sodium azide, 10 mM sodium ascorbate, and 5 mM Trolox. After biotinylation, equal numbers of cells were mixed and subjected to biochemical fractionation to collect microsomes and reduce the number of nonspecifically labeled proteins. Microsomal fractionation was performed as described previously (67). In brief, HEK293 cell homogenate (supernatant) was centrifuged as follows: spun at 600 ϫ g for 5 min at 4°C (two times), at 7,000 ϫ g for 10 min at 4°C (two times), at 10,000 ϫ g at 4°C, and 100,000 ϫ g for 1 h at 4°C. Then the pellet was collected as the microsomal fraction. The isolated microsomal fraction was lysed in RIPA buffer (20 mM Hepes, pH 7.4, 150 mM NaCl, 5% glycerol, 1% Nonidet P-40, 0.5% deoxycholic acid, 0.1% SDS, 0.5 mM EDTA, 1.5 mM MgCl 2 , and protease inhibitors) and centrifuged at 13,000 rpm for 10 min at 4°C. The cleared supernatant was incubated with streptavidin magnetic beads (Thermo Fisher Scientific) for 1 h at 4°C. The beads were subsequently washed three times with RIPA buffer (5 min each time), RIPA ϩ 1 M NaCl, RIPA ϩ 2 M urea in 10 mM Tris-HCl, pH 8.0, and finally three times with TBS. The beads were dried and stored at Ϫ80°C before mass spectrometric analysis.

RTN1A promotes ER-mitochondrial contacts LC-MS/MS
Biotinylated targets of SILAC-encoded Mit-APEX pulldowns were digested directly on streptavidin beads, and resulting tryptic peptides were sequentially purified by reverse-phase and strong cation-exchange batch-mode chromatography. Peptides were separated on a 50-cm analytical column packed with 5-m monitor tC18 beads using a 60-min gradient from 99% solvent A (0.2 M acetic acid in water) to 35% solvent B (0.2 M acetic acid in acetonitrile). Eluted peptides were directly injected into a Q-Exactive HF mass spectrometer (Thermo Fisher Scientific, Waltham, MA) equipped with a Digital Pico-View electrospray source platform (New Objective, Woburn, MA) (77). The spectrometer was operated in data-dependent mode where the top 10 most abundant ions in each MS scan (resolution 240,000, isolation width ϭ 1.5 Da) were subjected to HCD fragmentation (30% normalized collision energy). Dynamic exclusion was enabled with a repeat count of 1 and exclusion duration of 15 s. electrospray ionization voltage was set to 3.8 kV. The lock-mass function was enabled during MS acquisition and set to recalibrate MS1 scans using the background ion (Si(CH 3 ) 2 O) 6 at m/z 445.12 Ϯ 5 ppm.
MS spectra were converted into a Mascot generic file format (.mgf) using multiplier scripts (78,79). Spectra were searched using Mascot (version 2.4) against three appended databases consisting of the following: (i) human protein sequences (downloaded from RefSeq on 07/11/2011); (ii) common laboratory contaminants; and (iii) a decoy database generated by reversing the sequences from these two databases. Precursor tolerance was set to 10 ppm and product ion tolerance to 0.02 Da. Search parameters included trypsin specificity, up to two missed cleavages, fixed carbamidomethylation (Cys, ϩ57 Da), and variable oxidation (Met, ϩ16 Da) with the SILAC quantitation method set to account for differentially labeled lysine (Lys, ϩ6) and arginine (Arg, ϩ10). Peptide spectral matches to the reverse database were used to limit the global false discovery rate to 1%. The intensity of each SILAC isotope for a given peptide feature was retrieved using custom multiplier scripts. The local noise intensity was used to impute isotopes missing from a peptide feature. Protein enrichment ratios were calculated by summing the SILAC intensities of their constituent peptides. Raw mass spectrometry data files have been archived at ftp://massive.ucsd.edu/MSV000081174.

Split Renilla luciferase complementation assay
For the split Renilla luciferase complementation assay (22), HEK293T cells (4 ϫ 10 5 /well) were plated in a 12-well culture plate the day before transfection. Expression constructs pcDNA3-Mit-NRluc91 and pcDNA3-CRluc92-ER were cotransfected with expression constructs encoding RTN1A, RTN2B, or RTN3B using PEI (1:4 DNA/PEI ratio, Polysciences). Empty control vector and Reep1 expression vector were used as negative and positive controls, respectively. The cells were re-split into a 96-well plate (4 ϫ 10 4 /well, Corning; white-wall plate, product no. 3610) 6 h post-transfection and further incubated for 18 h. Twenty four hours post-transfection, Enduren live cell substrate (Promega) was added to the culture media. The cells were further incubated in a CO 2 incu-bator for 2-3 h, and luminescence was measured by POLARstar Omega microplate reader (BMG LABTECH).

Immunofluorescent staining
After transfecting HeLa cells with pcDNA3.1-Mit-APEX and incubating for 24 h, the cells were fixed with 4% paraformaldehyde for 10 min at room temperature, rinsed with PBS, and permeabilized with 0.2% Triton X-100 in PBS for 5 min. After blocking in 10% goat serum in PBS, cells were incubated with primary antibodies (Sigma, mouse monoclonal anti-FLAG M2, 1:500; Santa Cruz Biotechnology, rabbit polyclonal anti-Tom20, 1:500) for 1 h at room temperature. After washing three times in PBS for 5 min each, secondary anti-mouse Alexa Fluor 488 or anti-rabbit Alexa Fluor 594 (Invitrogen) was incubated for 1 h at room temperature. Images were captured with Zeiss Zen Pro software using a Hamamatsu ORCA-Flash4.0 camera attached to a Zeiss Observer Z1 inverted microscope.

Western blotting and silver staining
For Western blotting, molecular weight markers (PageRuler plus prestained protein ladder, Thermo Fisher Scientific) were included with each gel. Whole-cell extract or fractionated protein samples in RIPA buffer were subjected to 4 -15% SDS-PAGE (Bio-Rad, Mini-PROTEAN precast gels) and blotted on a PVDF membrane (Bio-Rad, Trans-Blot Turbo TM mini PVDF transfer Packs) using Trans-Blot Turbo TM transfer system (Bio-Rad). The blotted membrane was soaked in 100% methanol for 30 s and dried completely. Primary antibody (Abcam, Rtn1, ab83049, IP3R-II (A-5), and BAP31 (D-6); Novus Biologicals, VAP-B antibody) was next applied for 1 h at room temperature, rinsed three times for 5 min in TBST (Trisbuffered saline with 0.5% Tween 20), and incubated with secondary antibody for 1 h at room temperature. Membranes were rinsed in TBST containing 1 M NaCl and TBST. Blots were developed using ECL kit (SuperSignal West Pico or Femto Chemiluminescent Substrate, Thermo Fisher Scientific) and scanned using ChemiDoc MP imaging system (Bio-Rad). For silver staining, protein eluates from streptavidin magnetic beads were separated in 4 -15% SDS-polyacrylamide gel and stained with SilverQuest TM silver staining kit (Thermo Fisher Scientific) according to the manufacturer's instruction.

Functional enrichment analysis
FunRich (Functional Enrichment analysis tool) was used for analyzing SILAC/LC-MSMS results.

Statistical analysis
All statistical analyses were done in Prism software using two-tailed unpaired Student's t test. All graphs are plotted as standard error of the mean, S.E.