Ceapins block the unfolded protein response sensor ATF6α by inducing a neomorphic inter-organelle tether

The unfolded protein response (UPR) detects and restores deficits in the endoplasmic reticulum (ER) protein folding capacity. Ceapins specifically inhibit the UPR sensor ATF6α, an ER-tethered transcription factor, by retaining it at the ER through an unknown mechanism. Our genome-wide CRISPR interference (CRISPRi) screen reveals that Ceapins function is completely dependent on the ABCD3 peroxisomal transporter. Proteomics studies establish that ABCD3 physically associates with ER-resident ATF6α in cells and in vitro in a Ceapin-dependent manner. Ceapins induce the neomorphic association of ER and peroxisomes by directly tethering the cytosolic domain of ATF6α to ABCD3’s transmembrane regions without inhibiting or depending on ABCD3 transporter activity. Thus, our studies reveal that Ceapins function by chemical-induced misdirection which explains their remarkable specificity and opens up new mechanistic routes for drug development and synthetic biology.


Introduction
The endoplasmic reticulum (ER) is the site of folding and assembly of secreted and transmembrane proteins. When ER homeostasis is disturbed, misfolded proteins accumulate and activate the unfolded protein response (UPR) (Walter and Ron, 2011). One of the ER-resident UPR sensors, ATF6a, is an ER-tethered transcription factor that is cytoprotective and necessary for cell survival when cells experience ER stress (Wu et al., 2007;Yamamoto et al., 2007). Under stress conditions, ATF6a traffics to the Golgi apparatus, where it undergoes intramembrane proteolysis, releasing a bZIP transcription factor domain that moves to the nucleus and activates transcription (Haze et al., 1999;Yoshida et al., 1998). The events leading to ATF6a activation and trafficking remain poorly understood, but require the Golgi-resident proteases S1P and S2P and general components involved in COPII trafficking (Nadanaka et al., 2004;Okada et al., 2003;Schindler and Schekman, 2009;Ye et al., 2000) that are not specific to ATF6a.
Using a cell-based high-throughput screen, we recently identified a series of selective small-molecule inhibitors of ATF6a signaling, termed Ceapins (from the Irish verb 'ceap' meaning 'to trap') . Ceapins act on the most upstream step of ATF6a activation by retaining ATF6a at the ER and excluding it from ER exit sites during ER stress. When this trafficking requirement is removed by collapsing the Golgi apparatus into the ER, making ATF6a accessible to S1P and S2P, ATF6a is still cleavable by the proteases in the presence of Ceapin. Upon Ceapin treatment, ATF6a rapidly and reversibly forms foci without requiring new protein synthesis . The molecular target(s) of Ceapins, let alone how Ceapins specifically inhibit ATF6a, especially in light of the fact that activation depends on components that are shared by other cellular process, have remained an enigma.
To identify the molecular target of Ceapin, we carried out an unbiased genome-wide screen and proteomic analysis. Our approaches converged on a single target, the peroxisomal transporter ABCD3. ATF6a and ABCD3 normally do not interact and, indeed, localize to different parts of the cell. Ceapins induce these novel physical associations between ATF6a and ABCD3 in cells and in vitro. Our results indicate that Ceapins achieve their remarkable specificity through an unprecedented mechanism of small molecule induced inter-organelle tethering.

ABCD3 KD desensitizes cells to Ceapin-A7
To decipher the molecular mechanism of Ceapins, we carried out a genome-wide CRISPR interference (CRISPRi) screen to identify genes whose knockdown (KD) resulted in reduced or enhanced sensitivity to the drug. To this end, we screened a genome-wide sgRNA library  in K562 cells that stably expressed dCas9-KRAB and an mCherry transcriptional reporter dependent on ATF6a activation ( Figure 1A). Treatment with tunicamycin (Tm), which blocks N-linked glycosylation, activates ATF6a signaling leading to a two-fold reporter induction that was completely dependent on ATF6a ( Figure 1A). As a positive control, knocking down MBTPS2, one of the Golgi proteases that processes ATF6a, also inhibited induction of the reporter ( To carry out our genome-wide screen, we transduced the K562 ATF6a reporter cell line and selected for sgRNA expressing cells. We then induced ER stress with Tm in the presence or absence of Ceapin-A7, a potent member of the Ceapin family, and sorted cells by FACS (fluorescence-activated cell sorting). We isolated populations with decreased or increased ATF6a signaling (bottom 30% and top 30% of the reporter signal distributions, respectively) and used next-generation sequencing to quantify frequencies of cells expressing each sgRNA in both pools to evaluate how expression of each individual sgRNA affects activation of the ATF6a reporter Sidrauski et al., 2015) ( Figure 1B).
As expected, KD of ATF6a or MBTPS2 (encoding S2P) inhibited reporter induction ( Figure 1C). Knocking down abundant ER quality control components such as HSPA5, induced ER stress and turned on the reporter independently of Ceapin treatment (labeled in red in Figure 1C, Figure 1figure supplement 1C-D). Ceapin independent genes localized to the diagonal because their knockdown changed the expression of the reporter to the same degree in both treatments (labeled in red in Figure 1-figure supplement 1C). Of particular interest were genes whose KD specifically made cells insensitive to Ceapin treatment allowing activation of the reporter by Tm in the presence of Ceapin (labeled in black in Figure 1-figure supplement 1C). Two genes, ABCD3 and PEX19, robustly retested among the more than twenty hits from the genetic screen we individually knocked down and tested in the ERSE reporter cell line.
ABCD3, which encodes a peroxisomal ABC transporter involved in long-chain fatty acid import into peroxisomes, desensitized cells to Ceapin treatment ( Figure 1C, Figure 1-figure supplement 1C-D). Additionally, PEX19, which is necessary for chaperoning and targeting ABCD3 to the peroxisome, also desensitized cells to Ceapin treatment ( Figure 1C, Figure 1-figure supplement 1C-D). We knocked down these candidates individually and performed ERSE-mCherry dose response assays using Tm. Retesting of these candidates revealed that ABCD3 and PEX19 KD cells remained

ABCD3 is required for Ceapin-induced ATF6a foci
Ceapin treatment induces rapid and reversible formation of ATF6a foci that are retained in the ER ( Figure 2A) . We next tested if ABCD3 was directly involved in the formation of these foci and would colocalize with ATF6a. Indeed, in Ceapintreated cells, ATF6a colocalized with ABCD3 as visualized by immunofluorescence (Figure 2A -B). This result was surprising because newly synthesized ABCD3 is inserted directly into the peroxisomal membrane using PEX19 as import receptor (Imanaka et al., 1996;Biermanns and Gärtner, 2001;Kashiwayama et al., 2007;Kashiwayama et al., 2005;Sacksteder et al., 2000). ABCD3 is not cotranslationally translocated into the ER, indicating there is not a pool of ABCD3 in the ER (Figure 2figure supplement 1) (Jan et al., 2014); indeed, it is commonly used as a reliable marker for peroxisomes (Uhlén et al., 2015). Since both ABCD3 and PEX19 scored as hits in our screen, it seemed plausible that Ceapin induces ATF6a colocalization with peroxisomal ABCD3. We next tested whether ATF6a also colocalized with other peroxisomal markers, peroxisomal membrane protein PEX14 and peroxisomal matrix protein Thiolase (a maker for mature import competent peroxisomes). In the absence of Ceapin, ATF6a and PEX14 or Thiolase did not colocalize ( Figure 2A,C). This result was consistent in PEX19 KD cells, where peroxisome biogenesis is affected and ABCD3 is no longer chaperoned and targeted to the peroxisome (Kashiwayama et al., 2007;Kashiwayama et al., 2005;Sacksteder et al., 2000), ATF6a no longer formed foci in the presence Figure 1 continued fluorescence of each subpopulation (Tm-treatment and Tm + Ceapin-treatment) were collected by FACS and processed to measure the frequencies of sgRNAs contained within each. (C) Volcano plot of gene-reporter phenotypes and p values from CRISPRi screen. Negative control sgRNA targeted genes (gray), Ceapin-independent genes (red), genes with growth phenotypes (blue), and Ceapin hits (black) are indicated. (*) denotes chromatin architecture and remodeling related genes that impact reporter transcription. The reporter phenotypes and p values for genes in CRISPRi screen are listed in Figure 1-source data 1. (D) K562 ERSE reporter cells with individual ABCD3 sgRNAs or control sgRNA (NegCtrl) were treated with Tm and increasing concentrations of Ceapin-A7 for 16 hr. Reporter fluorescence was measured by flow cytometry and median values were plotted (N = 3, ± SD). (E) K562 ERSE reporter ABCD3 and NegCtrl KD cells were treated with DMSO or Tm and reporter activation was measured as in (D). (F and G) qPCR analysis of ATF6a target genes HSPA5 and HSP90B1, respectively. HepG2 CRISPRi NegCrl and ABCD3 KD cell lines were treated with DMSO, thapsigargin (Tg) (100 nM), and Tg with Ceapin (6 mM). Tg blocks the ER calcium pump and induces ER stress. Data plotted are mRNA levels for HSPA5 and HSP90B1 normalized to GAPDH and then compared to unstressed NegCtrl cells ± standard deviation of duplicate technical replicates of two biological replicates. DOI: https://doi.org/10.7554/eLife.46595.002 The following source data and figure supplements are available for figure 1: After prolonged ER stress, ATF6a attenuates and forms foci that are reminiscent of Ceapin induced foci . We next asked whether Ceapin was acting on the normal mechanism of ATF6a attenuation by testing ABCD3 colocalization with stress attenuated ATF6a foci. To induce stress attenuated ATF6a foci, we treated U2OS cells expressing GFP-ATF6a with ER stress, Tm or Tg (thapsigargin, which inhibits the ER calcium pump) for 2 and 4 hr. In positive control cells treated with Ceapin, ATF6a colocalized with ABCD3 and PEX14. In stress induced cells, attenuated ATF6a foci did not colocalize with ABCD3 or PEX14 by immunofluorescence ( Figure 2D-E). Thus, Ceapin does not act on the ATF6a pathway by stabilizing the attenuated ATF6a state. The stress attenuated foci and Ceapin induced foci are distinct.

Ceapin treatment does not inhibit ABCD3 activity
Since Ceapin treatment inhibits ATF6a, we next tested whether Ceapin treatment also inhibits ABCD3. ABCD3 knockout mice and hepatocytes display defects in bile acid biosynthesis (Ferdinandusse et al., 2015). To test if Ceapin treatment affects ABCD3 activity, we measured bile acid levels in a liver cancer cell line (HepG2) after Ceapin treatment and ABCD3 KD. As expected, in ABCD3 KD cells, bile acid levels were decreased ( Figure 3). In control cells treated at the EC 50 and ten-times the EC 50 of Ceapin, bile acid levels were similar to cells treated with vehicle only ( Figure 3). Thus, Ceapin does not inhibit ABCD3 activity in cells.

Figure 2 continued
PEX14 from (A) with at least 30 cells imaged per condition. All cells imaged in ABCD3 KD (96% KD), including wildtype cells, were used in quantification. Statistical analysis used unpaired two-tailed t-tests, **** indicates p<0.0001. (D) U2-OS cells stably expressing GFP-ATF6a were treated either with vehicle (DMSO), Tg (100 nM), Tm (2 mg/ml), or Ceapin (6 mM) for 2 hr or 4 hr (shown) prior to fixation, co-staining with anti-ABCD3 and anti-PEX14, and fluorescent imaging. Stress attenuated GFP-ATF6a foci are indicated by arrowheads. Scale bar, 10 mm. (E) Quantification of correlation of GFP-ATF6a and ABCD3 within PEX14 sites. DOI: https://doi.org/10.7554/eLife.46595.006 The following figure supplements are available for figure 2: Ceapin-induced ATF6a-ABCD3 interaction does not require known ERperoxisome tethers The tight association between the ER and peroxisome is mediated by ER-peroxisome tethers, VAPA and VAPB on the ER and ACBD4 and ACBD5 on the peroxisomes (Costello et al., 2017a;Costello et al., 2017b;Hua et al., 2017). While the ER components are redundant, ACBD5 KD or overexpression alone decreases or increase ER-peroxisome contacts, respectively (Costello et al., 2017a;Hua et al., 2017). To determine whether proximity between the ER and peroxisomes induced by these tethers is required for Ceapin-induced foci formation, we knocked-down these known ER-peroxisome tethers. In tether KD cells treated with Ceapin, ATF6a foci still formed and ATF6a colocalized with ABCD3 ( Figure 4A-B). Additionally, tether KD cells were not resistant to Ceapin treatment ( Figure 4C), consistent with the results from our screen in which these components also did not score as hits.

Ceapin-induced interactions do not require ER localized ATF6a nor ABCD3 transporter activity
We next tested if ER membrane association of ATF6a is required for Ceapin induced foci. To this end, we knocked down endogenous ATF6a and FACS sorted for a narrow, low level of GFP expression for truncated variants of ATF6a containing its cytosolic regions without the transmembrane and ER-lumenal domains ( Figure 5A). We found that GFP-ATF6a(2-302), which was retained in the cytosol with a nuclear exit signal and was no longer associated with the ER, colocalized with ABCD3 and formed foci ( Figure 5A-B). Further truncations showed that only the first 89 amino acids of ATF6a were both necessary and sufficient for Ceapin-dependent foci formation and colocalization with ABCD3 and peroxisomes ( Figure 5A-B, Figure 5-figure supplement 1).
Since ABCD3 is a transporter, we then tested if ABCD3 catalytic activity was required for Ceapin action. Similarly to our ATF6a truncations, we also knocked down endogenous ABCD3 and FACS sorted for low level GFP expression of constructs with mutations of ABCD3 residues that mediate ATP binding (G478R) and hydrolysis (S572I) or a deletion of the entire catalytic domain (Roerig et al., 2001). There is one reported patient with a C terminal truncation of ABCD3 in which a reduced number of import competent peroxisomes are present (Ferdinandusse et al., 2015). Similarly, GFP-ABCD3DNBD cells, with a deletion of the entire catalytic domain, have reduced, enlarged peroxisomes ( Figure 5C, Figure 5-figure supplement 2). We also confirmed correct localization of the GFP-ABCD3 constructs to the peroxisome ( Figure 5-figure supplement 2). As a positive control, ABCD3 KD cells complemented with the full length ABCD3 construct were able to colocalize with and form ATF6a foci when treated with Ceapin ( Figure 5C-D). In our catalytic activity mutants, we found that ABCD3 ATP binding or hydrolysis was not required for Ceapin-induced foci formation ( Figure 5C-D). Although there are fewer larger peroxisomes in GFP-ABCD3DNBD cells, peroxisomal ABCD3 still induced foci formation and colocalized with ATF6a in the presence of Ceapin ( Figure 5C-D). These results indicate that Ceapin-induced interactions do not require ER localized ATF6a nor ABCD3 transporter activity.

Ceapin drives ATF6a-ABCD3 interaction in cells and in vitro
To identify components physically associating with ATF6a in the presence of Ceapin, we carried out native immunoprecipitation -mass spectrometric (IP-MS) analyses. We treated 3xFLAG-ATF6a HEK293 cells with Ceapin-A7 or an inactive analog, Ceapin-A5, in the presence of stress (Tg) and found that ABCD3 co-purified as the top hit with epitope-tagged ATF6a selectively in the presence of active Ceapin-A7 but not inactive Ceapin-A5 ( Figure 6A-B). The native reciprocal affinity purification with full-length GFP-ABCD3 cells confirmed these results ( Figure 6C). Furthermore, GFP-ABCD3DNBD, lacking the entire nucleotide binding domain, also physically associated with ATF6a in the presence of Ceapin ( Figure 6C).
We then tested if the minimal cytosolic domain of ATF6a, GFP-ATF6a(2-90), physically associated with peroxisomal ABCD3. We immunoprecipitated GFP-ATF6a(2-90) from detergent solubilized lysates and specifically enriched ABCD3 in the presence of active Ceapin-A7 but not inactive Ceapin-A5 ( Figure 6D). Thus, consistent with the above experiments where organelle tethering was not required, these results confirm that no other ER proteins are required for Ceapin-A7 induced ATF6a and ABCD3 physical association.    Finally, we tested whether purified ATF6a and ABCD3 were sufficient for Ceapin-induced tethering. In a binding assay with purified ATF6a(2-90) and ABCD3, our vehicle (DMSO) and inactive Ceapin-A5 controls did not induce ATF6a(2-90) and ABCD3 binding ( Figure 6E). In the presence of Ceapin-A7, however, the cytosolic domain of ATF6a(2-90) and ABCD3 associated in solution ( Figure 6E). Thus, Ceapin is directly responsible for tethering ABCD3 to ATF6a.

Discussion
Ceapins, named for their ability to trap ATF6a in the ER, act with exquisite selectivity; they do not affect signaling of ATF6a's close homolog ATF6b or SREBP (sterol response element binding protein) , which depend on broadly used vesicular trafficking ER-Golgi pathways and are activated by the same Golgi-resident proteases (Nadanaka et al., 2004;Okada et al., 2003;Schindler and Schekman, 2009;Ye et al., 2000). Here we discovered the basis of this specificity. Ceapins induce neomorphic inter-organelle junctions, forcing interactions between the cytosolic domain of ER-tethered ATF6a and the peroxisomal transmembrane protein ABCD3 to sequester ATF6a from its normal trafficking route (Figure 7), and do so without interfering with or depending on ABCD3's normal function. Since ABCD3 protein expression is ten-fold higher than ATF6 (Hein et al., 2015), it is likely ABCD3 is not saturated. Ceapin induced interaction of ABCD3 with the most N-terminal region of ATF6a also clarifies how ATF6a foci are excluded from COPII trafficking, while the transmembrane region of ATF6a remains accessible to protease cleavage. Mechanistically, Ceapins could act as molecular staples that physically bridge the respective proteins or bind to one or the other inducing allosteric changes that promote their association; but in either case, Ceapin is responsible for tethering ABCD3 to ATF6a.
Remarkably, in the absence of Ceapins, ATF6a and ABCD3 localize to different parts of the cell and are not known to interact physically or functionally. Indeed, an 89-amino acid fragment of ATF6a fused to GFP is sufficient to recruit GFP to peroxisomes, ruling out the need for endogenous inter-organellar tethers. This Ceapin-induced tethering enables an 'anchor away' strategy but one that uses an abundant, ubiquitously expressed endogenous acceptor protein. There has been increasing interest in small molecules that induce novel protein-protein interactions with therapeutic potential (de Waal et al., 2016;Han et al., 2017;Krö nke et al., 2014;Lu et al., 2014;Krö nke et al., 2015;Petzold et al., 2016;Uehara et al., 2017). Ceapins provide a novel example of such molecules and increase the repertoire to include the induction of inter-organellar connections, opening new mechanistic routes for drug development and synthetic biology by broadly enabling control of protein function through chemical-induced misdirection.
Understanding the mechanism of action of a chemical modulator of cellular stress and establishing that it is acting directly and specifically is critical for exploiting the utility of any stress modulators either as research or potential therapeutic agents. Our identification of the mechanism by which Ceapins achieve their remarkable specificity forms a foundation to explore the utility of ATF6a inhibition in the treatment of cancers, such as squamous carcinomas, in which ATF6a signaling protects dormant tumor cells from classical chemotherapies (Schewe and Aguirre-Ghiso, 2008). -FLAG affinity purification from 3xFLAG-ATF6a HEK293 cells treated with stress (100 nM Tg) and inactive Ceapin-A5 analog (6 mM) or active Ceapin-A7 (6 mM) with two replicates for each treatment condition. The proteins identified with affinity-purified FLAG-ATF6 treated with ER stress and Ceapin-A5 or Ceapin-A7 are listed in Figure 6-source data 1. SQSTM1 KD (*) was the top second hit in proteomics, however, SQSTM1 KD in the K562 ATF6 reporter cell Figure 6 continued on next page
Individual gene knockdowns were carried out by selecting sgRNA protospacers from compact hCRISPRi-v2 library and cloning into lentiviral plasmid pU6-sgRNA EF1a-puro-t2a-BFP (Addgene 60955) as previously described . Protospacer sequences used for individual knockdowns are listed in Table 1. The resulting sgRNA expression vectors were packaged into lentivirus by transfecting HEK293T with standard packaging vectors using TransIT-LTI Transfection Reagent (Mirus, MIR 2306). The viral supernatant was harvested 2-3 days after transfection and frozen prior to transduction into CRISPRi knockdown cell lines described above.
U2OS Flp-In cells were infected with UCOE-EF1a-dCas9-BFP-KRAB and FACS sorted for BFP expression. They were then stably transduced with sgRNA knockdown of endogenous ATF6a and GFP-ATF6a(2-90) construct, and FACS sorted for a narrow level of GFP expression. Parental cell lines and commercially available cell lines were authenticated by STR analysis and tested negative for mycoplasma contamination.

Genome-scale CRISPRi screen
Reporter screens were carried out using protocols similar to those previously described Gilbert et al., 2014;Sidrauski et al., 2015). The compact (five sgRNA/ gene) hCRISPRi-v2  sgRNA libraries were transduced into ERSE reporter cells at a MOI <1 (55% BFP+ cells). Cells were grown in spinner flasks for 2 days without selection, selected with 2 mg/ml puromycin for 2 days, and allowed to recover for 3 days. Cells were then split into two populations, which were treated for 16 hr with 6 mg/ml tunicamycin alone or 6 mg/ml tunicamycin and 3 mM Ceapin (EC90). Cells were then sorted based on reporter fluorescence using BD FACS Aria2. Cells with the highest (~30%) and lowest (~30%) mCherry expression were collected and frozen after collection. Approximately 20 million cells were collected per bin. Genomic DNA was isolated from frozen cells, and the sgRNA-encoded regions were enriched, amplified, and prepared for sequencing. Sequenced protospacer sequences were aligned and data were processed as described (Gilbert et al., 2014;Horlbeck et al., 2016) with custom Python scripts (available at https://github. com/mhorlbeck/ScreenProcessing). Reporter phenotypes for library sgRNAs were calculated as the log2 enrichment of sgRNA sequences identified within the high-expressing mCherry over the lowexpressing mCherry cells. Phenotypes for each transcription start site were then calculated as the average reporter phenotype of all five sgRNAs. Mann-Whitney test p-values were calculated by comparing all sgRNAs targeting a given TSS to the full set of negative control sgRNAs. For data presented in Figure 1B, screen hits are defined as those genes where the absolute value of a calculated reporter phenotype over the standard deviation of all evaluated phenotypes multiplied by the log10 of the Mann-Whitney p-value for given candidate is greater than 7. Growth screen data  was used to label genes with growth phenotype of at least À0.19. Ceapin independent genes are defined as genes that were hits in tunicamycin alone and tunicamycin with Ceapin treatment since their phenotype was independent of Ceapin treatment. Genes involved in chromatin remodeling and architecture have been previously described in UPR screens to act downstream and directly affect expression of the reporter (Jonikas et al., 2009). Chromatin related genes that impact reporter expression are labeled with ($) in Figure 1-figure supplement 1C-D.

Bile acid assay
HepG2 CRISPRi ABCD3 KD and NegCtrl cells were treated with DMSO or Ceapin at 600 nM or 6 mM for 24 hr. Cells were harvested in scrapping buffer (cold PBS with 10 mM MG132 and 1X protease inhibitor), spun down, resuspended in lysis buffer (50 mM Tris pH 7.4, 150 mM NaCl, 5 mM EDTA, 1X protease inhibitor, and 1% LMNG), and spun down at 10,000 x g for 10 min. The supernatant was used for bile acid assay (Cell Biolabs STA-631) as described by the manufacturer.

Quantitative RT-PCR
Cells were harvested and total RNA was isolated using the NucleoSpin RNA II (Macherey-Nagel) according to manufacturer's instructions. RNA was converted to cDNA using AMV reverse transcriptase under standard conditions with oligo dT and RNasin (Promega, Life Technologies). Quantitative PCR reactions were prepared with a 2x master mix according to the manufacturer's instructions (KAPA SYBR FAST qPCR Kit). Reactions were performed on a LightCycler thermal cycler (Roche). Primers used were against HSPA5 (forward, TGTGCAGCAGGACATCAAGT: reverse, AGTTCCAGCG TCTTTGGTTG) and HSP90B1 (forward, GGCCAGTTTGGTGTCGGTTT; reverse, CGTTCCCCGTCC TAGAGTGTT).

Immunofluorescence
Fluorescence confocal imaging was carried out as described in . 293 TREx, U2OS, HepG2, and HeLa cells were plated in 8-well ibiTreat mSlide (ibidi 80826) at 20-25,000 cells/well. In 3xFLAG-ATF6a imaging experiments ( Figure 4A-C, Figure 5A-D, Figure 5-figure supplement 2), 3xFLAG-ATF6a HEK293 CRISPRi cells were plated and induced with 50 nM doxycycline on the same day. On the following day, cells were treated with DMSO or 6 mM Ceapin for 30 min and then fixed with cold methanol or 4% PFA. For cold methanol fixation, media was removed, cold ethanol was added for 3 min at À20˚C, washed, and permeabilized with PHEM (60 mM PIPES, 25 mM HEPES, 10 mM EGTA, 2 mM MgCl 2 , pH 6.9) with 0.1% Triton X-100, and washed twice with PHEM. For PFA fixation, media was removed from slides, 4% PFA (EMS) was added for 10 min at room temperature, washed, permeabilized as above, and washed with PHEM. Slides were then treated with blocking buffer (5% goat serum (Jackson ImmunoResearch) in PHEM) for 1 hr at room temperature. Antibodies were diluted in blocking buffer and incubated with cells at 4˚C overnight. After three washes with PHEM, cells were incubated with secondary antibodies conjugated to Alexa 488, Alexa 568, and/or Alexa 633 (Invitrogen) for 1 hr at room temperature. Slides were imaged on a spinning disk confocal with Yokogawa CSUX A1 scan head, Andor iXon EMCCD camera and 100x ApoTIRF objective NA 1.49 (Nikon). Linear adjustments were made using ImageJ. Quantification of correlation between ATF6a with ABCD3, Thiolase, and/or PEX14 was calculated using CellProfiler 2.1.1. ABCD3, Thiolase, or PEX14 images were used to identify objects, a background threshold for ATF6a images was set to 1.2, and clumped objects were separated based on intensity. The resulting ABCD3, Thiolase, or PEX14 outlines were used as masks to count the ATF6a intensity within ABCD3, Thiolase, or PEX14. Data from CellProfiler was imported into GraphPad Prism version 6.0 for statistical analysis and plotting.
Nuclear translocation assay 3xFLAG-ATF6a HEK293 CRISPRi cells with ABCD3 KD and ABCD3 KD complemented with full length GFP-ABCD3 construct were plated in ibidi 96-well ibiTreat m-plate (ibidi 89626) and induced with 50 nM doxycycline on the same day. On the following day, cells were treated with DMSO or 100 nM Tg for 2 hr and then fixed with 4% PFA as described above. The plates were then treated with blocking buffer (5% goat serum (Jackson ImmunoResearch) in PHEM) for 1 hr at room temperature. Primary antibodies, mouse anti-FLAG M2 (Sigma F1804) and rat anti-GRP94 9G10 (abcam ab2791), were diluted in blocking buffer and incubated with cells at 4˚C overnight. After three washes with PHEM, cells were incubated with secondary antibodies conjugated to Alexa 568 and Alexa 633 (Invitrogen) and nuclear stain (DAPI, Molecular Probes D-1306, 5 mg/mL) for 1 hr at room temperature. Quantification ATF6a signal in ER and nucleus was calculated using CellProfiler 2.1.1 as described in . DAPI images were used to identify primary objects and clumped objects were distinguished based on fluorescence intensity. The GRP94 images were then used to generate secondary objects from primary objects using global Otsu two-class thresholding with weighted variance. The final ER mask was generated by subtracting the nuclear area from the ER area. Lastly, the FLAG-ATF6a images were used to calculate FLAG-ATF6a intensity in the nucleus and ER and determine the nucleus to ER ratio of each cell. Data from CellProfiler was exported as a MATLAB file for analysis and plotted on GraphPad Prism version 6.0.

Immunoprecipitation and immunoblot analysis
Cells were grown in 100 mm plates with two replicates for each treatment condition, treated with 50 nM doxycycline the following day, treated with 100 nM Tg and 6 mM Ceapin-A5 or 6 mM Ceapin-A7 for 30 min on the day of harvest, and harvested in scrapping buffer (cold PBS with 10 mM MG132 and 1X protease inhibitor). Ceapin A-7, inactive analog Ceapin A-5, or DMSO were kept in scrapping and lysis buffers throughout IP. Cells were lysed for 1 hr at 4˚C in lysis buffer (50 mM Tris pH 7.4, 150 mM NaCl, 5 mM EDTA, 1X protease inhibitor, and 1% LMNG (Anatrace NG322)). The lysates were cleared by centrifugation at 17,000 x g for 30 min. Dynabeads Protein-G (ThermoFisher) were bound with Sigma FLAG M2 antibody for 1 hr at 4˚C and crosslinked with 100 mM BS3 crosslinker for 30 min. 293 TREx 3XFLAG cell lysates were then incubated with these FLAG beads for 2 hr at 4˚C. IP beads were washed with wash buffer (lysis buffer without LMNG) and boiled and eluted in buffer containing 50 mM Tris pH 6.8, 300 mM NaCl, 2% SDS, and 10 mM EDTA. Protein samples were then precipitated, trypsin digested, labeled with tandem mass tags (TMT), and analyzed by liquid chromatography-mass spectrometry using Multidimensional Protein Identification Technology (MuD-PIT), as described previously (Mortenson et al., 2018;Plate et al., 2018). TMT intensities for proteins detected in each channel were normalized to the respective TMT intensity of ATF6a. TMT ratios for individual proteins were then calculated between Tg+Ceapin-A7/DMSO treatment or Tg +Ceapin-A5/DMSO treatment.
The reciprocal affinity purification with full-length GFP-ABCD3 or GFP-ABCD3DNBD cells was carried out by culturing, treating, and lysing cells as described above. 293 TREx 3XFLAG GFP-ABCD3 clarified cell lysate was then incubated with GFP-Trap_MA ChromoTek beads for 2 hr at 4˚C. IP beads were washed with wash buffer (lysis buffer without LMNG) and boiled in SDS sample buffer for 10 min.
Cells for in vitro incubation were lysed with lysis buffer containing LMNG (described above) and cleared by centrifugation at 17,000 x g for 30 min in the absence of any drug. Cleared supernatant was then incubated with Ceapin A-7 or inactive analog Ceapin A-5 for 30 min at room temperature, bound to GFP-Trap_MA ChromoTek beads for 1 hr at 4˚C, washed with wash buffer containing Ceapin A-7 or Ceapin A-5, and eluted by boiling in SDS sample buffer.
Samples were run on a precast 4-12% Bis-Tris polyacrylamide gel (Life Technologies) under denaturing conditions and transferred to nitrocellulose membrane. Antibodies described above for FLAG, GFP, and Pmp70 (SAB4200181) were used to detect proteins and blots were imaged for chemiluminescence detection using a ChemiDocTM XRS + Imaging System (Bio-Rad) ( Figure 6B and D-E) or LICOR system ( Figure 6C).

Generation of recombinant proteins
Human ATF6a(2-90) with an N-terminal 3XFLAG was cloned into pET16b-TEV-MBP-HIS 6X (Novagen) using Gibson assembly. The construct was expressed in in BL21-Gold(DE3) E. coli cells, grown to 0.6-0.8 OD 600 , and induced overnight at 16˚C with 0.25 mM IPTG (Gold Biotechnology). The cells were harvested and resuspended in buffer containing 50 mM HEPES pH 7, 150 mM NaCl, 10% glycerol, 2 mM TCEP, and complete EDTA-free protease inhibitor cocktail (Roche). After lysis by sonication, the lysate was clarified at 30,000 x g for 30 min at 4˚C. The clarified lysate was loaded onto a HisTrap HP 5 ml column, washed in binding buffer (50 mM HEPES, pH 7, 300 mM NaCl, 1 mM TCEP, 10% glycerol, and 25 mM imidazole), and eluted with a linear gradient of 25 mM to 1M imidazole in the same buffer. The ATF6a fractions eluted at 240 mM imidazole were collected and concentrated with an Amicon Ultra-15 concentrator (EMD Millipore) with a 30,000-dalton molecular weight cutoff. The ATF6a concentrated fraction was loaded onto a Mono Q HR16/10 column (GE Healthcare), washed in Buffer A (50 mM HEPES, pH 7.5, 100 mM NaCl, 10% glycerol, and 1 mM DTT) and eluted with a linear gradient of 100 mM to 1M NaCl in the same buffer. Fractions were collected, concentrated as above, and loaded onto a Superdex 200 10/300 GL column (GE Healthcare) equilibrated with buffer containing 30 mM HEPES, pH 7.5, 300 mM NaCl, 5% glycerol, and 1 mM DTT.
Expression and purification of human ABCD3: Full-length human ABCD3 isoform I was synthesized and cloned into modified pFastBac1 plasmid with a C-terminal -eGFP À8XHis-tag for baculoviral expression in Spodoptera frugiperda SF9 cells. Bacmid DNA was produced by transforming the recombinant pFastBac1 plasmid into E. coli DH10Bac strain. To express the protein, SF9 cells were infected with the bacmid made from recombinant pFastBac1 plasmid at multiplicity of infection (MOI) = 2 for 48 hr at 27˚C. The cells were harvested and resuspended in lysis buffer (50 mM Tris Cl, pH 7.5, 100 mM NaCl, 100 mM MgCl 2 , 10% glycerol) containing complete EDTA-free protease inhibitor cocktail (Roche), and lysed by sonication. The lysate was centrifuged at 186,010 x g for 2 hr to extract the membrane fraction. 3 g of the membrane was solubilized in 30 ml of lysis buffer containing 1% w/v lauryl maltose neopentyl glycol (LMNG) (Anatrace): 0.1% w/v cholesteryl hemisuccinate (CHS) (Anatrace) overnight. Solubilized membrane was clarified by centrifugation at 104,630 x g for 30 min with 5 mM imidazole added. A HiTrap TALON crude 1 ml column (GE Healthcare) was equilibrated with the lysis buffer containing 5 mM imidazole and solubilized membrane loaded onto the column. After binding the column was washed with 15 ml of 10 mM imidazole, 0.02% glyco-diosgenin (GDN) (Anatrace) in lysis buffer. The protein was eluted from the column with 10 ml of 150 mM imidazole, 0.02% GDN containing lysis buffer. The protein obtained was concentrated using Amicon Ultra-15 centrifugal filter units (MilliporeSigma) and size exclusion chromatography was done to further purify the protein in SEC buffer (20 mM HEPES, pH 7.5, 100 mM NaCl, 2 mM MgCl 2 , 2% glycerol and 0.02% GDN). advice on protein purification; Nico Stuurman and Vladislav Belyy for advice on fluorescence microscopy; John Christianson for advice on solubilization of intact membrane protein complexes; and members of the Walter and Weissman labs for helpful discussions. This research was supported by Collaborative Innovation Awards from the Howard Hughes Medical Institute (HHMI) and by NIH/ NIGMS New Innovator Award DP2 OD021007 (MK). Research of MG and RMS is supported by NIH GM111126 to RMS PW and JSW are Investigators of the HHMI.