RESEARCH

◥ zation, we combined biochemical and structural RESEARCH ARTICLE SUMMARY approaches to define the function and mecha- nism of the P5A-ATPase. STRUCTURAL BIOLOGY RESULTS: P-type ATPases form a large class of The endoplasmic reticulum P5A-ATPase is a active transporters that are present in all kingdoms of life and predominantly trans- transmembrane helix dislocase port ions or lipids across cellular membranes. The P5A-ATPase belongs to a eukaryotic- Michael J. McKenna*, Sue Im Sim*, Alban Ordureau, Lianjie Wei, J. Wade Harper, specific subfamily of P-type ATPases with un- Sichen Shao†, Eunyong Park† known substrate specificity. We reconstituted membrane protein insertion into organelles in a cell-free system and used site-specific cross- INTRODUCTION: Eukaryotic cells contain geted transmembrane proteins from the ER linking to reveal that the P5A-ATPase interacts membrane-bound organelles with distinct membrane are incompletely understood. directly with the TM of a mitochondrial tail– identities and functionalities that depend anchored protein. Human cells lacking ATP13A1 on protein composition. Correct localization RATIONALE: As a model to study membrane pro- showed mislocalization of mitochondrial tail– of proteins is thus critical for organelle func- tein localization, we focused on tail–anchored anchored proteins to the ER and secretory tion and cellular homeostasis. The endoplas- proteins, which contain a single C-terminal TM pathway. In in vitro assays, newly synthesized mic reticulum (ER) and mitochondrial outer that is necessary and largely sufficient for or- mitochondrial tail–anchored proteins aber- Downloaded from membrane are the primary destinations for ganelle localization. We reasoned that factors rantly accumulated in ER vesicles lacking newly synthesized proteins with hydrophobic that mediate mitochondrial tail–anchored pro- P5A-ATPase activity. This accumulation was transmembrane segments (TMs). Membrane tein localization would interact directly with due to the impaired extraction of misinserted protein localization requires not only high- nascent proteins. We used an unbiased, site- mitochondrial proteins from ER membranes fidelity protein targeting but also quality con- specific cross-linking and mass spectrome- lacking ATP13A1. Cryo–electron microscopy trol mechanisms that selectively remove try approach to identify such protein TMs. structures of Saccharomyces cerevisiae Spf1 mislocalized proteins. At the mitochondrial This approach revealed that the ER-resident in different conformations at 3.3 to 3.7 Å res- http://science.sciencemag.org/ outer membrane, the ATP-dependent motor orphan P-type pump P5A-ATPase (Spf1 in yeast; olutions revealed that the P5A-ATPase has an protein Msp1/ATAD1 removes some mislocal- ATP13A1 in humans) interacted directly with a atypically large substrate-binding pocket ized transmembrane proteins. By contrast, al- mitochondrial tail–anchored protein. Because compared with other P-type ATPases with though protein targeting to the ER is well genetic studies have linked the P5A-ATPase to known structures. The pocket alternately opens studied, the mechanisms that remove mistar- mitochondrial tail–anchored protein mislocali- toward the ER lumen and cytosol while remain- ing accessible to the lipid bilayer through a lateral opening. Trapping putative substrates for structure determination revealed an addi- A C tional membrane-spanning density at the lat-

Cytosol Mitochondria eral opening, which resembles an a-helical TM. on January 4, 2021 Together with proteomics of wild-type and P5A-ATPase knock-out cells, our results indi- Mitochondrial TM cate that the P5A-ATPase can dislocate mode- rately hydrophobic TMs with short hydrophilic lumenal domains that misinsert into the ER.

Mistargetingg CONCLUSION: Our findings define the func- tion of the P5A-ATPase as a dislocase of TMs N at the ER membrane. This assignment estab- B A lishes polypeptides as P-type ATPase transport P substrates in addition to ions and lipids. Ac- TM tive dislocation of misinserted proteins from the ER by the P5A-ATPase also represents a previously unknown cellular safeguarding and quality control mechanism that helps main- tain ER and mitochondrial homeostasis, pos- WT P5A-ATPase P5A-ATPase ATP Inward-open sibly explaining the pleiotropic phenotypes Mitochondrial TM KO Outward-open hydrolysis (E1 form) linked to P5A-ATPase dysfunction.▪ Mitochondria ER lumen (E2 form)

The list of author affiliations is available in the full article online. P5A-ATPase dislocates mistargeted TMs from the ER. (A) Diagram of a eukaryotic cell showing the *These authors contributed equally to this work. nucleus (blue), ER (pale green), and mitochondria (pale purple). (B) Immunofluorescence images showing †Corresponding author. E-mail: [email protected] mislocalization of a mitochondrial tail–anchored protein containing the mitochondrial TM OMP25 (green) (S.S.); [email protected] (E.P.) Cite this article as M. J. McKenna et al., Science 369, in P5A-ATPase knock-out cells. A mitochondrial marker (TOM20) is shown in purple. (C) Model for eabc5809 (2020). DOI: 10.1126/science.abc5809 P5A-ATPase–mediated removal of mistargeted TMs from the ER membrane based on cryo–electron microscopy structures showing different conformations of the yeast P5A-ATPase (Spf1; surface READ THE FULL ARTICLE AT representations) and the position of a substrate TM (green ribbon) bound to the outward-open form. https://doi.org/10.1126/science.abc5809

McKenna et al., Science 369, 1583 (2020) 25 September 2020 1of1 RESEARCH

◥ protein containing the mitochondrial OMP25 RESEARCH ARTICLE TM during reconstituted targeting reactions (Fig.1A).Weincorporatedtheultraviolet STRUCTURAL BIOLOGY (UV) light–activated cross-linker p-benzoyl-L- phenylalanine (Bpa) into the TM [referred to The endoplasmic reticulum P5A-ATPase is a as TA(Bpa)] and purified recombinant, FLAG- tagged TA(Bpa) in complex with calmodulin transmembrane helix dislocase (CaM), a calcium-dependent TM chaperone (fig. S1A) (22, 23). We confirmed that FLAG- Michael J. McKenna1*, Sue Im Sim2*, Alban Ordureau1, Lianjie Wei1†, J. Wade Harper1, TA(Bpa) released from CaM by EGTA formed Sichen Shao1‡, Eunyong Park2,3‡ UV-dependent cross-links to the TM chaper- one SGTA (24)(Fig.1Bandfig.S1C).Whenwe Organelle identity depends on protein composition. How mistargeted proteins are selectively recognized released FLAG-TA(Bpa) in the presence of and removed from organelles is incompletely understood. Here, we found that the orphan P5A–adenosine crude yeast membranes that could recon- triphosphatase (ATPase) transporter ATP13A1 (Spf1 in yeast) directly interacted with the transmembrane stitute TM insertion (fig. S1B), we observed segment (TM) of mitochondrial tail–anchored proteins. P5A-ATPase activity mediated the extraction of cross-links to various proteins. Because photo- mistargeted proteins from the endoplasmic reticulum (ER). Cryo–electron microscopy structures of cross-linking generates covalent adducts, we Saccharomyces cerevisiae Spf1 revealed a large, membrane-accessible substrate-binding pocket that expected this approach to enrich for otherwise alternately faced the ER lumen and cytosol and an endogenous substrate resembling an a-helical TM. transient interactions made by substrates un- Our results indicate that the P5A-ATPase could dislocate misinserted hydrophobic helices flanked by short dergoing targeting or QC. Indeed, tandem Downloaded from basic segments from the ER. TM dislocation by the P5A-ATPase establishes an additional class of P-type mass tag mass spectrometry (TMT-MS) anal- ATPase substrates and may correct mistakes in protein targeting or topogenesis. ysis of UV-dependent membrane interactors of FLAG-TA(Bpa) identified known mitochon- drial protein receptors and QC factors (Fig. 1C, ore than one-third of the proteome in changes driven by phosphorylation and dephos- fig. S1D, and data S1) (25–27). Unexpectedly, eukaryotes consists of membrane and phorylation of a conserved aspartate (5–7). the ER-resident P5A-ATPase Spf1 was a promi- secretory proteins with diverse hydro- Among the five subfamilies categorized by nent UV-dependent interactor of FLAG-TA(Bpa) http://science.sciencemag.org/ M phobic sequences that must insert sequence similarity, the functions of P1- to (Fig. 1, C and D). correctly into the appropriate cellular P3-ATPases as cation transporters and P4- Although Spf1 has been linked to mitochon- membrane. Protein mistargeting causes or- ATPases as lipid flippases are well defined. drial TA protein mislocalization (4, 11), a direct ganelle dysfunction and cellular and organis- In addition, ATP13A2, one of four mamma- interaction has not been reported. To confirm mal stress (1, 2), underscoring the importance lian P5B-ATPases, was recently proposed to the interaction between the P5A-ATPase and TA of quality control (QC) mechanisms that re- be a lysosomal polyamine transporter (8). proteins, we translated radiolabeled TA(Bpa) move mislocalized proteins. Defined by a single Otherwise, little is known about the substrates in vitro and performed site-specific cross- C-terminal transmembrane segment (TM), tail- and biological functions of P5-ATPases. linking with crude yeast membranes or human anchored (TA) proteins pose a distinct sorting Humans and yeast each have a single P5A- ER-derived rough microsomes (RMs) contain-

challenge because they rely exclusively on post- ATPase, ATP13A1 and Spf1, respectively, which ing FLAG-tagged Spf1 or ATP13A1, respective- on January 4, 2021 translational mechanisms to localize to the likely perform a conserved task based on func- ly. UV-dependent TA protein cross-links to endoplasmic reticulum (ER), mitochondria, or tional complementation (9, 10). In addition to each P5A-ATPase were detected after denatur- peroxisomes. Although TA protein targeting to disrupting TA protein localization and turn- ing immunoprecipitations (Fig. 1E and fig. S2). the ER is well studied (3), how cells remove mis- over (4, 11), mutating SPF1 causes a wide spec- Cross-linking was specific to Bpa positioned targeted TA proteins from the ER and properly trum of phenotypes, including induction of within the TM, but not the cytosolic domain, localize non-ER TA proteins is poorly understood. the ER unfolded protein response, defects in and was enhanced by flanking C-terminal se- A yeast genetic screen for mutants that dis- lipid and sterol homeostasis, and dysregulated quences that contain basic residues (Fig. 1E), a rupt mitochondrial TA protein localization protein N-glycosylation, topogenesis, and turn- common feature of mitochondrial TA pro- identified only SPF1, the deletion of which over (10, 12–20). It has been speculated that teins. Thus, the P5A-ATPase interacts directly results in mitochondrial TA protein accumu- these pleiotropic phenotypes may be explained with TMs. lation at the ER (4). SPF1 encodes the ER- bytheP5A-ATPasefunctioningasamanga- resident orphan P5A–adenosine triphosphatase nese or calcium transporter or as a lipid flip- ATP13A1 mediates ATP-dependent (ATPase), a conserved subtype of the eukaryotic- pase (13–15, 18, 21). However, direct evidence TM dislocation specific P5 subfamily of P-type ATPases. P-type supporting any of these functions is lacking, To investigate whether P5A-ATPase activity ATPases are a major class of active transporters and the identity of physiological P5A-ATPase affected TA protein localization to the ER, found in all organisms that pump substrates substrates remains elusive. Here, using cell-free we performed in vitro insertion reactions of across membranes through conformational reconstitutions, high-resolution cryo–electron radiolabeled TA protein with RMs isolated microscopy (cryo-EM), and proteomics, we from wild-type (WT) or ATP13A1 knock-out found that the P5A-ATPase could dislocate (KO) cells without or with reexpression of 1Department of Cell Biology, Blavatnik Institute, Harvard Medical School, Boston, MA 02115, USA. 2Department of Molecular and moderately hydrophobic terminal transmem- ATP13A1 (fig. S3A). TA protein containing the Cell Biology, University of California, Berkeley, CA 94720, USA. brane helices from the ER membrane, a mech- OMP25 TM preferentially accumulated in RMs 3 California Institute for Quantitative Biosciences, University of anism required to ensure correct mitochondrial lacking WT ATP13A1 activity (fig. S3, B to D). California, Berkeley, CA 94720, USA. *These authors contributed equally to this work. TA protein localization. Two possibilities may explain this result: de- †Present address: Biochemistry, Biophysics, and Structural leting ATP13A1 may enhance TM insertion (4) Biology Program, Division of Biology and Biomedical Sciences, The P5A-ATPase interact directly with TMs or it may impair TM extraction. To distinguish Washington University in St. Louis, St. Louis, MO 63110, USA. ‡Corresponding author. E-mail: [email protected] We used an unbiased strategy to identify fac- between these scenarios, we uncoupled OMP25 (S.S.); [email protected] (E.P.) tors that interacted directly with a model TA TM insertion and extraction by reisolating RMs

McKenna et al., Science 369, eabc5809 (2020) 25 September 2020 1of14 RESEARCH | RESEARCH ARTICLE

Fig. 1. The P5A-ATPase directly A C FLAG-TA(Bpa)-CaM interacts with the TM of TA Msp1 proteins. (A) Scheme to identify mito. TM 8 Tom22 Mim1 direct interactors of a mito- photo-XL 6 chondrial TA protein. Recombinant Tom70 FLAG-tagged TA protein with + crude membranes

( P -value) Spf1 or SGTA the OMP25 TM containing the 10 4 Ubx2 Tom20

UV-reactive cross-linker (photo-XL) Log

+ EGTA − Bpa was purified in complex with 2 CaM [FLAG-TA(Bpa)-CaM]. EGTA UV-induced XL dissociates FLAG-TA(Bpa) from 0 CaM, permitting TM insertion into −1012345678 B Log ratio (+/− UV) organelles in a crude membrane + SGTA + memb. 2 fraction or association with purified UV: -++-+ + D E SGTA. Exposure to UV light induces EGTA: --+ - - + cyto. Bpa TM Bpa UV: − + site-specific covalent cross-links. 135 - CTD: WT R-E (B) Anti-FLAG blots of reactions as 100 - tot. - Spf1- UV:-+-++ in (A). Unmodified FLAG-TA(Bpa) 75 - ** GFP 58 - - x Spf1-FLAG and FLAG-TA(Bpa) cross-links x SGTA/** IPs 46 - - Spf1- of oligomers (x TA) and to SGTA - x TA x TA GFP Downloaded from 32 - elu. (x SGTA) or membrane-specific - FLAG- - TA(Bpa) -** x TA tot. proteins (orange asterisks) 25 - TA(Bpa) trunc. TA are indicated. (C) Six replicates 17 - of reactions as in (A) without or 11 - - FLAG- TM CTD TA(Bpa) with UV irradiation were affinity purified and analyzed by TMT-MS. http://science.sciencemag.org/ Volcano plot shows relation of the P value and log2 enrichment of proteins by UV irradiation. (D) Five percent input (tot.) of reactions as in (A) with organelles containing Spf1-GFP were purified by FLAG-TA(Bpa) (elu.) and immunoblotted for GFP to show UV-dependent Spf1 interaction. (E) Autoradiography of total (tot.) and denaturing anti-FLAG immunoprecipitations (IPs; top) of photo-cross-linking reactions of IVT-radiolabeled untagged TA(Bpa) with organelles containing Spf1-FLAG. OMP25 TM and C-terminal domain (CTD) sequence is below. Bpa is positioned in the cytosolic (cyto.) domain (lanes 1 and 2) or the OMP25 TM (lanes 3 to 5). R-E, TA(Bpa) in which the arginines in the OMP25 CTD are mutated to glutamates; trunc. TA, truncated translation products that do not incorporate Bpa.

after short insertion reactions for extraction in ATP13A1 KO cells (Fig. 2D and figs. S5 and can extract TMs from the outer mitochondrial assays with an energy-regenerating system S6), consistent with spf1D yeast phenotypes membrane (26, 28, 29). Knocking down ATAD1

and excess SGTA to sequester dislocated TMs (4, 11). Localization of TA proteins with ER- in ATP13A1 KO cells significantly restored mito- on January 4, 2021 (Fig. 2A) (28). WT RMs displayed higher ex- targeted TMs was not affected by ATP13A1 KO chondrial localization of both BAK1 and OMP25 traction activity than KO RMs based on the (fig. S6). Reexpressing WT ATP13A1, but not (fig. S8, A to C). These results support a model amount of TA protein dislocated into the sol- D533A ATP13A1, in KO cells restored local- in which ATAD1 and ATP13A1 each facilitates uble fraction and bound to SGTA (Fig. 2, B ization and expression levels of both OMP25 TM removal from mitochondria and the ER, and C, and fig. S4A). TM removal required and BAK1 (Fig. 2, D and E, and fig. S5). Thus, respectively, to contribute to TA protein local- ATP (fig. S4B), and the impaired dislocation ATP13A1 is required for proper mitochondrial ization at steady state (fig. S8D). Impairing activity of KO RMs was rescued by WT ATP13A1, TA protein localization. ATP13A1 would result in the ER acting as a but not D533A ATP13A1, a catalytically inac- To confirm that mitochondrial TA protein “sink” that accumulates misinserted TMs be- tive variant in which the conserved phospho- mislocalization in ATP13A1 KO cells was due cause of the disruption of a major mechanism rylation site required for transport is mutated. to impaired TM removal from the ER rather for removing them. In the absence of ATP13A1, Thus, ATP13A1 mediates ATP-dependent re- than altered insertion efficiencies, we assayed misinserted TA proteins appear to undergo moval of a mitochondrial TM from the ER. the targeting of in vitro–translated (IVT) OMP25 alternative substrate-specific processes that to organelles of semipermeabilized cells over may include vesicular trafficking and lyso- ATP13A1 minimizes TA protein mislocalization short time periods to minimize downstream somal or ER-associated degradation (11, 30). We next investigated how ATP13A1 affects TA processes (fig. S7A). OMP25 localized similarly protein localization in human cells. In WT to mitochondria in both WT and KO semi- Features of ATP13A1-dependent proteins cells, FLAG-tagged mitochondrial TA protein permeabilized cells (fig. S7, B and C), indicat- The observation that reexpressing WT ATP13A1 reporters containing either the OMP25 or ingthatATP13A1wasnotrequiredforinitial in KO cells restored mitochondrial TA protein BAK1 TM (referred to as OMP25 or BAK1, re- TA protein targeting. Nonetheless, at steady reporter levels suggested that some endoge- spectively) localized to mitochondria as ex- state, mitochondrial TA proteins were not nous proteins that require ATP13A1 function pected(Fig.2,DandE,andfig.S5,AandB). only mislocalized, but were also apparently could be identified by quantitative proteomics. Knocking out ATP13A1 resulted in substrate- depleted from mitochondria in ATP13A1 KO To investigate this, we analyzed lysates of WT, specific changes in TA protein levels: OMP25 cells (Fig. 2D and figs. S5A and S7C). Because ATP13A1 KO, and rescue cells using TMT-MS levels decreased and BAK1 levels increased targeting to mitochondria was not affected, (fig. S9A and data S2). We identified 34 pro- (fig. S5, C and D). Both mitochondrial TA pro- we reasoned that this observation may have teins with decreased abundance in ATP13A1 teins displayed increased mislocalization to resulted from other mechanisms. In particu- KO cells, the abundance of which was specif- the ER and other secretory pathway organelles lar,theAAA-ATPaseMsp1(ATAD1inhumans) ically rescued by acutely reexpressing WT

McKenna et al., Science 369, eabc5809 (2020) 25 September 2020 2of14 RESEARCH | RESEARCH ARTICLE

ACB radiolabeled TA (IVT) + RMs: WT KO KO + WTKO + D533A WT KO KO + WTKO + D533AWT KO KO + WTKO + D533A RMs insertion tot. - TA

reisolated RMs + energy, - x SGTA spin XL (tot.) SGTA

+ energy + SGTA - TA spin sup. soluble stain - TA TA-SGTA (SGTA) (sup.) tot. sup.

D FLAG-β-OMP25 TOM20 Merge Merge HA

KO + WT Downloaded from WT HA-ATP13A1

ATP13A1 KO + D533A

KO HA-ATP13A1 http://science.sciencemag.org/

E **** **** 1

F CTSA mito. TA TM MAVS 5 TMED2 N-term. mito. TM LRP8 NUCB1 CHST10 SS (sol.) CCDC134 RMDN3 SS (other) GGH DSEL POMGNT2 on January 4, 2021 FLAG- β -OMP25 4 N-term. type II TM DNAJB9 st

MCC (coloc. w/TOM20) MCC (coloc. TGFB1 1 TM type II SELENOM TGFB2 0 GLB1 DNASE1L1 WT KO KO + KO + OMP25 MAOB PRSS23 WT D533A DNAJB11 3 LAMA1 SUMF1 G EFNA5 1 ( P -value) LTF TMED3 **** **** 10 PDGFD FBLN5 PGAP1 TWSG1 SDF2 -Log 2 TRIM13 TF

1 Endogenous OMP25 MCC (coloc. w/TOM20) MCC (coloc. 0 0 WT KO KO + KO + −1 −0.5 0 0.5 1 1.5 2 WT D533A Log2 ratio (KO + WT / KO + D533A)

Fig. 2. ATP13A1 mediates TM removal from the ER. (A) Scheme to (left) WT, ATP13A1 KO, or (right) KO Flp-In T-REx HeLa cells transfected with reconstitute TA protein dislocation from ER-derived RMs. IVT-radiolabeled TA WT or D533A HA-tagged ATP13A1 (yellow arrowheads, transfected cells; red protein is incubated with RMs from WT, ATP13A1 KO, or KO Flp-In T-REx 293 arrowheads, untransfected cells; also see fig. S5E). Scale bar, 10 mm. Panels cells stably reexpressing WT or D533A ATP13A1. RMs containing inserted TA below show high-magnification views of boxed regions; intensity of FLAG- protein (tot.) are reisolated and incubated with an energy-regenerating system b-OMP25 is enhanced in ii, iv, and v. (E) Mean ± SEM and individual MCC values and the cytosolic TM chaperone SGTA. TA proteins dislocated from RMs are of FLAG-b-OMP25 colocalization with TOM20. ****P < 0.0001. (F) Volcano separated by centrifugation (sup.). (B) SDS-PAGE and autoradiography of tot. plot showing relation of the P value and log2 enrichment of proteins in ATP13A1 and extracted (sup.) TA protein containing the OMP25 TM with the indicated KO Flp-In T-REx HeLa cells reexpressing WT versus D533A FLAG-ATP13A1. RMs. Equal recovery of SGTA (Coomassie) and TA protein cross-links to SGTA Significantly enriched proteins are classified on the basis of expected cellular (x SGTA) are indicated. (C) As in (B), after cross-linking with 250 mM BMH. localization and TM topology. mito., mitochondrial; N-term., N-terminal; SS, signal (D) IF of endogenous TOM20 (purple; mitochondrial marker) and FLAG-tagged sequence; sol., soluble protein after SS cleavage. (G) Mean ± SEM and individual TA protein containing the OMP25 TM (FLAG-b-OMP25; green) stably expressed in MCC of endogenous OMP25 colocalizationwithTOM20intheindicatedcells.

McKenna et al., Science 369, eabc5809 (2020) 25 September 2020 3of14 RESEARCH | RESEARCH ARTICLE

ATP13A1, but not D533A ATP13A1 (Fig. 2F and in ER protein maturation. Although highly di- cerevisiae using single-particle cryo-EM (Fig. 3, figs.S9BandS10,AtoD).Exceptforone,allof verse, signal sequences and type II TMs are A and B, and table S1). We attached a green these candidates were mitochondria- or ER- generally less hydrophobic than TMs of the fluorescent protein (GFP) tag to the 3′ end of targeted proteins. opposite topology and often have positively the chromosomal SPF1 gene and extracted Endogenous OMP25 and two additional charged N-termini (32). The N-termini of can- endogenous Spf1 from yeast membranes with mitochondrial TA proteins, MAVS and MAOB, didate signal sequences also appear enriched a dodecyl maltoside (DDM)/cholesteryl hemi- were among these candidates, consistent with in prolines (fig. S10, B and G). These biophys- succinate (CHS) mixture. Spf1 purified using the decrease in OMP25 reporter levels in ical features resemble mitochondrial TA pro- anti-GFP nanobody affinity resin eluted as a ATP13A1 KO cells that was rescued by WT teins and raise the possibility that ATP13A1 single peak in size-exclusion chromatography ATP13A1 (fig. S5E). Another identified mito- may mediate the removal of these terminal (fig. S11, A and B). Cryo-EM images of apo Spf1 chondrialprotein,RMDN3,hasasingleN- hydrophobic helices if they are inserted in purified without any nucleotide or phosphate terminal TM that inserts in the Nlumen/Ccyto the wrong orientation (Nlumen/Ccyto) (fig. S10, analogs displayed well-dispersed particles (fig. orientation. These candidates all have a TM B and H). Together, our data support a QC S11C). Two-dimensional (2D) image classifica- flanked by a short, positively charged lumenal function for P5A-ATPases in removing mis- tion was used to remove empty micelles and segment (fig. S10A). Immunofluorescence (IF) inserted terminal hydrophobic helices from some poor-quality particles (fig. S11, D and verified mislocalization of endogenous OMP25 the ER. E), resulting in 216,994 particles that showed and RMDN3 in ATP13A1 KO cells, which was 2D projections of Spf1 from various angles specifically rescued by reexpressing WT ATP13A1 Cryo-EM structure of apo Spf1 with clear protein features (fig. S11E). Ab initio (Fig. 2G and fig. S10, E and F). The P5A-ATPase is an orphan transporter, and reconstruction and 3D classification indicated The remaining ATP13A1-dependent proteins our results suggesting a role in TM dislocation that most Spf1 molecules were in a single Downloaded from contain an N-terminal ER-targeting signal se- were puzzling because P-type ATPases are gen- conformation (fig. S11D). Approximately 60% quence or type II TM, which should adopt an erally assumed to transport cations or lipids. of particles from the 2D classification were Ncyto/Clumen topology (fig. S10B) (31). Proteins To understand the substrate selectivity of this used to reconstruct the final apo map at 3.5-Å without an N-terminal hydrophobic helix were family of transporters, we determined the struc- resolution (Fig. 3A, movie S1, and fig. S11, D to not enriched, arguing against a general defect ture of the P5A-ATPase Spf1 from Saccharomyces G). Many side chains were clearly visible, http://science.sciencemag.org/

A N-domain C P-domain N A-domain

NH2 Arm A P NTD COOH

ArAArm Cytosol

Cytosol 4b

on January 4, 2021

TMs ab123 56 78910

ER 90° 4a membrane

ER lumen ER lumen NTD T-domain S-domain TMs 1 & 2 TMs 3-10

B D TM2 TM4b TM3 TM5 TM3

Cytosol TM4a TM5

90° TM1 TM1

90° TM8 TM2 TM4a TM6 TM6 ER lumen Front TM3

Fig. 3. Structure of apo Spf1 in an inward-open conformation. (A and B) 3.5-Å-resolution cryo-EM reconstruction (A) and atomic model (B) of the apo Spf1. Note the V-shaped opening of the substrate-binding pocket between TMs 1 and 2 (brown) and TMs 3 to 10 (tan). Dotted lines show the approximate boundaries of the ER membrane. (C) Domain architecture of Spf1. Note that the helix of TM4 (outlined in green) is unraveled halfway by a conserved PP(E/D)LP motif. (D) Views into the substrate-binding pocket. Left is a front view. For clarity, only TMs 1 to 6 are shown. The solvent-accessible volume (~3500 Å3) is shown in semitransparent gray. Right panel is a view from the cytosol (TM8 is also shown). Side chains lining the cavity are shown in a stick representation.

McKenna et al., Science 369, eabc5809 (2020) 25 September 2020 4of14 RESEARCH | RESEARCH ARTICLE enabling de novo building of an atomic model (33–37). In P4-ATPase lipid flippases, the equiv- motif for dephosphorylation, to access the (fig. S12). alent pocket, which is also small and binds to phosphorylation site through an ~24° inward The apo Spf1 structure showed all core do- aphospholipidheadgroup,isexposedonlyto rotation of the A domain and NTD toward the mains of P-type ATPases, namely the cytosolic thelipidphase(38–40). The unusual size and N-P interface. Because TM1 and TM2 are linked A (actuator), P (phosphorylation), and N (nu- topology of the substrate pocket suggests that to the A domain, this motion substantially re- cleotide binding) domains and the T (trans- the P5A-ATPase has fundamentally different structures the substrate-binding pocket. TM2 port) domain formed by TMs 1 to 6 (Fig. 3, A to substrate specificities than other P-type ATPases and TM6, which contact each other on the ER C,andfig.S13).Inaddition,asiscommonin with known structures. The structure did not lumenal side in a V-shaped arrangement in P2- to P5-type ATPases, Spf1 contains an S reveal any sites in the pocket of Spf1 that the E1 states, spread out into a parallel ar- (support) domain (TMs 7 to 10) that tightly could specifically coordinate a metal ion, argu- rangement, and the cytosolic segment of TM4 packs against TM5 and TM6. The structure ing against the idea that the P5A-ATPase (denoted TM4b) moves in between TM2 and also showed two features specific to the P5A- transports cations (10, 18). TM6 toward the lipid phase by a tilting motion ATPase:anN-terminaldomain(NTD)preced- (Fig. 4D and movie S3). Consequently, the ing the A domain and an “arm”-like elongated Structure of Spf1 in an inward-open V-shaped substrate pocket trans- domain protruding between the P domain outward-open conformation forms into an outward-open, inverted U–shaped and TM5. The NTD contains a seven-stranded P-type ATPases undergo a series of conforma- pocket(Fig.4,CandD,andfig.S15B).During b-barrel in the cytosol, which is tightly bound tional changes with ATP binding, phosphoryl- this E1 to E2 transition, the lateral opening to to the A domain, and two additional TMs ation of a conserved aspartate (D487 in Spf1, the lipid phase is maintained but shifts from (TMa and TMb), which appear to be only D533 in ATP13A1), and subsequent dephos- the cytosolic to the lumenal leaflet of the ER loosely associated with TMs 1 and 2 through phorylation in a scheme known as the Post- membrane (movie S3). The surface of the pock- Downloaded from lipid and/or detergent molecules (fig. S14A). Albers cycle (i.e., E1 → E1-ATP → E1P-ADP → et maintains a negative electrostatic potential The arm domain extends from the P domain E1P → E2P → E2-Pi → E1; fig. S17A). Trans- (Fig. 4C), suggesting that substrates likely con- with a conserved a-helix and positions its dis- itions between the E1 and E2 states result in tain positive charges in addition to a hydro- tal portion upon the detergent micelle (Fig. 3A large conformational changes in the T domain phobic component, which would interact with and fig. S14B). An atomic model for this region that mediate substrate transport. Because the the acyl chains of lipids at the lateral opening. could not be built because of poor local reso- apo structure of Spf1 represents only the E1 These features are consistent with the TMs of lution, but the micelle around the contact was form, we sought to elucidate different confor- mitochondrial TA proteins and the terminal http://science.sciencemag.org/ noticeably pulled up toward the arm domain mational states by adding nucleotide or phos- hydrophobic helices of ATP13A1-dependent (fig. S14B), suggesting that the arm domain phate analogs to purified Spf1. To this end, we proteins identified in our proteomics analyses might affect local membrane structure. obtained structures of Spf1 in complex with (Fig. 2F). b,g-methyleneadenosine 5′-triphosphate (AMP- Apo Spf1 has a large inward-open − − Visualization of a putative Spf1 substrate PCP), AlF4 , and BeF3 , at overall resolutions substrate-binding pocket of 3.7, 3.4, and 3.3 Å, respectively (fig. S18). Finally, we sought to visualize substrate-engaged In all P-type ATPases, the substrate-binding AMP-PCP, a nonhydrolyzable ATP analog, locks Spf1. Although we did not observe substrate site is formed mainly by TM2, TM4, and TM6 the enzyme in an ATP-bound state, and the density in the structures obtained with Spf1 − − oftheTdomainatthepointwheretheTM4 phosphate analogs AlF4 and BeF3 form a purified as described above, we reasoned that

helix is disrupted [in Spf1, at the PP(E/D)LP stable complex with the side-chain carboxylate endogenous substrates maycopurifywithSpf1 on January 4, 2021 motif between TM4a and TM4b]. Apo Spf1 of D487 to mimic a phosphorylated aspartate trapped in the outward-open E2P conforma- also shows a prominent pocket for substrate (37). All three structures show clear EM den- tion. Accordingly, we repeated the extraction − binding in this site, but with an unusually sity for these inhibitors bound at the phos- and purification of Spf1 with BeF3 included in large V-shaped cavity compared with other phorylation site (fig. S19, A to C). On the basis all buffers. The resulting 3.3-Å-resolution struc- P-type ATPases (Fig. 3D and fig. S15). The pock- of the arrangement of the A, P, and N domains ture (fig. S18, J to L, P) was essentially identical et is blocked on the ER lumenal side and open and comparison with other P-type ATPase to the E2P structure but with a prominent toward the cytosol. Thus, the apo structure rep- structures (37, 39), we assigned the AMP-PCP, difference: an additional elongated density − − resents an inward-open conformation of the AlF4 , and BeF3 structures as the E1-ATP, of a putative substrate traversing the lateral transporter. The inner surface is lined with a E1P, and E2P states, respectively. The AMP- opening of the substrate pocket (Fig. 4E and − mixture of hydrophilic and hydrophobic side PCP– or AlF4 -bound Spf1 structures were fig. S20A). chains (fig. S15, A and C) and exhibits negative essentially identical to the apo structure (root Theputativesubstratedensity spans almost electrostatic potential (fig. S16A) that is con- mean square deviations of 0.56 and 0.48 Å, the entire membrane and closely resembles tributed collectively by D215, E232, D434, E446, respectively; fig. S19, D and E). Only a slight an a-helical TM that can accommodate a 20- and E450. These acidic amino acids are highly (7°) inward rotation was seen with the N do- residue-long polyalanine model (Fig. 4, E and conserved in P5A-ATPases across species but main of AMP-PCP–bound Spf1 as AMP-PCP F). The large cylindrical shape is incompatible not in the closest subfamily of P5B-ATPases bridges the interface of the N and P domains. with detergents, lipids, and sterols (fig. S20B). (fig. S16B). The T domain maintained the inward-open In the cytosolic leaflet of the membrane, the One noticeable feature of the putative conformation in both the AMP-PCP– and density is exposed to the lipid phase and may − substrate-binding pocket is that it is open AlF4 -bound structures. interact with hydrophobic amino acids of TM2 − both to the cytosol and laterally toward the By contrast, the BeF3 -bound E2P structure (F223, M227, M231), TM4b (M449, M453), and lipid phase in the cytosolic leaflet of the ER showed a substantially different conformation the conserved PP(E/D)LP motif (P445) be- membrane between TM2 and TM6 (Fig. 3D (Fig. 4, A and B; movie S2; and fig. S19, F and tween TM4a and TM4b. In the lumenal leaf- and fig. S15A). In P2-ATPase cation pumps, G). Similar to other P-type ATPases, the N do- let, the density is directed to the inverted such as the sarcoplasmic/endoplasmic reticu- main of Spf1 undergoes a large (~50°) outward U–shaped pocket. This density likely rep- lum Ca2+-ATPase (SERCA) and Na+,K+-ATPase, rotation with respect to the P domain during resents an average of multiple different co- the T domain forms a small ion-binding cav- the E1P to E2P transition. This allows the A purified substrates, possibly explaining the ity that is shielded away from the membrane domain, which contains a conserved (S/T)GES lower resolution compared with Spf1. Altogether,

McKenna et al., Science 369, eabc5809 (2020) 25 September 2020 5of14 RESEARCH | RESEARCH ARTICLE

ABCN-domain

A-domain P-domain

F NTD View from cytosol TM3 TM5 TM1 Cytosolsol TM4b ER membrane

ER lumenumen TMs 1 & 2 TM2 TMs 3-10 −5 +5 (kT/e) TM6 DETM5 View from front

TM4b TM4b Cyto. TM4b TM3 M453 Downloaded from Cytosol M231 M449

TM2 L1033 M227 TM2 90° P445

F223 ER membrane ER

TM1 TM6 http://science.sciencemag.org/ TM6 TM6 Lumen TM4a ER lumen Y220 TM4a TM4a TM3

Fig. 4. Structure of Spf1 in an outward-open conformation. (A and B)3.3-Å- the outward-open (BeF3-bound; colored) states. Left is a front view and right resolution cryo-EM reconstruction(A)andatomicmodel(B)oftheBeF3-bound is a side view. The outward-open substrate-binding pocket is indicated by Spf1.Shownarethefrontviews.Wenotethat, unlike the inward-open structure, an orange dotted line. (E) Density of copurified putative substrate (magenta TMa and TMb are flexible in this conformation. TMa and TMb were modeled mesh; 5-Å low-pass filtered) fitted with a polyalanine helix model (also see as polyalanine helices on the basis of the apo structure. (C) Heatmap of surface fig. S20A). (F) As in (E) but with amino acid side chains of Spf1 facing the electrostatics. The substrate-binding cavity is outlined with a dashed line. substrate density shown in a ball-and-stick representation. Top is a cytosolic on January 4, 2021 (D) Conformational changes of TMs 1 to 6 from the inward-open (apo; gray) to view and bottom is a front view. our structural and functional data indicate P-type ATPases characterized to date. We pears similar to Msp1 at the outer mitochon- that the P5A-ATPase dislocate misinserted propose that the P5A-ATPase uses this speci- dria membrane; although structurally unrelated, TMs from the ER membrane. alized pocket to extract mistargeted or mis- both use ATP as an energy source to dislocate inserted TMs with a short flanking lumenal TMs, possibly providing opportunities for re- Discussion hydrophilicsegmentfromtheER(Fig.5).In insertion into the correct membrane. This idea Our results assign transmembrane helices as this working model, we hypothesize that the is consistent with emerging observations that the substrates of the ER-resident P5A-ATPase, substrate TM largely remains in the lipid ER and mitochondrial factors cooperate to de- thereby defining an additional P-type ATPase phase, whereas the lumenal segment is flipped grade aberrant TM-containing proteins (11, 30) function of transporting polypeptides out of across the membrane by the outward- to and to ensure high-fidelity mitochondrial pro- lipid bilayers. Until recently, P-type ATPases inward-open transition of the large cavity tein import, as in the ‘ER-SURF’ pathway (41). were thought to be exclusively cation or lipid in the T domain. Moderate hydrophobicity The P5A-ATPase may also dislocate ER-targeted pumps. Together with the report that the characteristic of signal sequences and mito- N-terminal hydrophobic helices such as signal P5B-ATPase ATP13A2 transports polyamines chondrial TMs may lower the energetic bar- sequences that erroneously insert in the Nlumen/ (8), our study reveals that P-type ATPases rier for dislocation. Positive charges in the Ccyto orientation, although the exact basis of mediate transport of unexpectedly diverse sub- lumenal segment may be the preferred clients, these substrates’ dependence on the P5A-ATPase strates. Although the T domains of all P-type whereas large lumenal segments are likely ex- remains unclear. Biophysical features of P5A- ATPases display a similar overall arrange- cluded because of the limited space in the ATPase–dependent clients resemble substrates ment of TMs 1 to 6, our structures demonstrate pocket. oftheERmembraneproteincomplex(EMC), that they can form fundamentally different TM dislocation by the P5A-ATPase also re- which inserts moderately hydrophobic TA pro- substrate-binding sites. The large substrate- veals a previously unidentified protein QC and teins and terminal TMs in the Nlumen/Ccyto bindingpocketofSpf1,withanopeningthat safeguarding mechanism at the ER. Active orientation (23, 42). EMC and the P5A-ATPase not only alternately faces the cytosol and the transport by the P5A-ATPase is required to may thus have mechanistically opposing ac- ER lumen but also remains laterally accessi- dislocate mitochondrial TMs that mistarget to tivities at the ER. Disruption of EMC also ble from the membrane, is distinctive among the ER. P5A-ATPase function at the ER ap- causes a wide range of ER-related dysfunctions

McKenna et al., Science 369, eabc5809 (2020) 25 September 2020 6of14 RESEARCH | RESEARCH ARTICLE

ADP ATP immunoblotting, horseradish peroxidase– conjugated anti-mouse (Jackson ImmunoResearch E1P E2P E2-Pi E1 Laboratories 115-035-003; RRID:AB_10015289) Pi Substrate N and anti-rabbit (Jackson ImmunoResearch TM helix NNLaboratories 111-035-003; RRID:AB_2313567) A A A P P P antibodies were used at 1:5000. For IF, Alexa Cytosol Fluor 488–conjugated goat anti-mouse IgG (Jackson ImmunoResearch Laboratories 115-545-003; RRID:AB_2338840), Alexa Fluor 564–conjugated goat anti-rabbit IgG second- ER lumen (+) ary antibodies (Jackson ImmunoResearch Inward-facing Outward-facing Inward-facing Laboratories 111-585-003; RRID:AB_2338059), Fig. 5. Proposed model for TM dislocation by the P5A-ATPase. The P5A-ATPase likely alternates between and DyLight 680-conjugated anti-hemagglutinin two conformations, inward-open E1 and outward-open E2 states. During the E1P-to-E2P transition, the N (HA) antibody (Thermo Fisher Scientific 26183- domain rotates (red) and the A domain (yellow) moves closer to the N-P interface, causing conversion of the D680; RRID:AB_2533054) are commercially inward-open substrate-binding pocket of the transmembrane domain (brown and tan) to the outward-open available. conformation. In this model, a substrate TM helix (magenta) with a short, preferentially positively charged Yeast strains lumenal segment would bind to the outward-open pocket and the E2P-to-E1 transition would flip the TM by a switch from the outward-open to inward-open conformation. For simplicity, some intermediate steps BY4741 and Spf1-GFP (e.g., Fig. 1D) yeast strains used for membrane preparations and cross- of the Post-Albers cycle are not shown. Downloaded from linking analyses (Fig. 1 and fig. S1) were obtained from the Finley laboratory (Harvard (17, 43), similar to the pleiotropic phenotypes 132 amino acids of ATP13A1 not present in Medical School). To tag endogenous Spf1 with associated with P5A-ATPase mutants. Thus, the cDNA construct and the D533A muta- a 3xFLAG tag (e.g., Fig. 1E), a DNA fragment cells appear to spread the burden of manag- tions was introduced by Gibson assembly. To containing coding sequence for a linker ing very diverse TMs by using an arsenal of modulate induced expression levels (fig. S9A), (GSGGRIPGLINM), 3xFLAG, and a hygrom- different machineries, each with varying sub- the Promega-series CMVd1 truncation was ycin resistance marker flanked by 50 to 60 bp http://science.sciencemag.org/ strate selectivity, that collaborate to fine-tune generated in pcDNA5/FRT/TO expression vec- homologous to the 3′ regions of SFP1, was TM localization and topology. tors using conventional molecular biology generated by polymerase chain reaction (PCR) techniques. ATP13A1 coding sequences were using the pFA-6a-3xFLAG-Hyg plasmid (a gift Materials and Methods transferred from pcDNA5 to pcDNA3.1-based from P. Feng, Harvard Medical School) as a Plasmids and antibodies vectors behind 3xFLAG and 3xHA tags for template and a pair of primers (forward: 5′- To express recombinant TA protein in com- transient transfection (e.g., Fig. 2D) by restrict- caaa cc atc aga cat ttc tgt gca aca ggt caa gat plex with CaM (fig. S1A), the open reading ion enzyme digestion and T4 ligation. pOG44 tgc ctc taa agg atc cgg agg acG GAT CCC CGG frame of a model TA protein containing the (Thermo Fisher Scientific V600520) and pX459 GTT AAT TAA-3′; reverse: 5′- ggg taa tat aag cytosolic domain of SEC61b and OMP25 TM (48) are as described previously. Guide RNA tat ata aat aca aaa agg ggt act aca taa aag att

as described in (44) was appended after a 3C sequences (GAG CAC CGT CCC ATA CCC GG, taG AAT TCG AGC TCG TTT AAA CTG-3′; on January 4, 2021 protease cleavage site to the coding sequence ACG CGC TCA CTG TCC TCT CG) were cloned lowercase and uppercase, specific to the of pGEX-CaM (22). N-terminal FLAG tags and into pX459 for generating KO cells using chromosomal and template sequences, respec- amber codons were introduced using con- Gibson assembly. tively). DNA was introduced to BY4741 by a ventional molecular biology techniques. pGEX- Homemade rabbit polyclonal antibodies standard lithium acetate transformation pro- SGTA (24), pEVOL-Bpa for expressing amber against TRAPa [1:5000 for immunoblotting tocol before selection with 100 mg/ml hygro- codon suppression components in bacteria (45), (IB); 1:300 for IF], GFP (1:5000 for IB), and mycin in yeast extract, peptone, and dextrose and plasmids for expressing recombinant Bpa- 3F4 (1:350 for IP) were gifts from the Hegde (YPD) agar. Successful FLAG tagging and pro- RS and for suppressor tRNA transcription (46) laboratory. Antibodies against TOM20 (Abcam tein size was confirmed by immunoblotting have been described previously. ab186735) (1:5000 for IB, 1:300 for IF), TOM20 and Sanger sequencing. SP64-based plasmids used to generate in vitro (Santa Cruz Biotechnology sc-17764; RRID: To enable affinity purification of endoge- transcription templates were as previously de- AB_628381; 1:50 for IF), ATP13A1 (Protein- nous Spf1 from S. cerevisiae, a cleavable GFP scribed (47). TM substitutions and introduc- tech 16244-1-AP; RRID:AB_2290293; 1:5000 tag was introduced to the 3′ end of the SPF1- tion of amber stop codons for Bpa incorporation for IB), FLAG M2 (Sigma-Aldrich F1804; coding sequence of the WT strain BY4741. (e.g., Fig. 1E) were accomplished using con- RRID:AB_262044; 1:5000 for IB; 1:500 for First, a DNA fragment containing a coding ventional Gibson assembly and Phusion muta- IF), CNX (Enzo Life Sciences ADI-SPA-865; sequence for GFP and a nourseothricin re- genesis approaches. Model TA protein-coding RRID:AB_10618434; 1:5000 for IB), Por1 sistance marker, which is flanked by ~60-bp sequences were transferred to pcDNA5/FRT/ (Abcam ab110326; RRID:AB_10865182; 1:2500 sequences homologous to the 3′ regions of SFP1, TO vectors for stable incorporation into Flp-In for IB), ATAD1 (Abcam ab94583; RRID:AB_ was generated by PCR using the pSK-B399 T-REx HeLa cells (a gift from B. Raught) be- 10672816; 1:1000 for IB), DNAJB11 (Protein- plasmid (a gift from S. Klinge, Rockefeller Uni- hind a doxycycline-inducible cytomegalovirus tech 15484-1-AP; RRID:AB_2094400; 1:2000 versity) as a template and a pair of primers (CMV)promoter(e.g.,Fig.2,DandE).TMse- for IB), MAVS (Cell Signaling Technology (forward: 5′-c ttc atg gac gac aaa cca tca gac att quences were exchanged using Gibson assembly. 3993; RRID:AB_823565; 1:1000 for IB); OMP25 tct gtg caa cag gtc aag att gcc tct aaa GGG GCA ATP13A1 cDNA (Harvard plasmID no. (Abcam ab224217; 1:50 for IF); OMP25 (Pro- ACT GGT GGT AG-3′; reverse: 5′-a ctt aga agc HsCD00082872) was cloned into pcDNA5/ teintech 15666-1-AP; RRID:AB_2201149; 1:1000 tgt ttt aca aat aat aca gca ctt tca taa actt aac aat FRT/TO behind an N-terminal 3xFLAG tag for IB); RMDN3 (Thermo Fisher Scientific acacccCGGCGTTAGTATCGAATCG-3′; by restriction enzyme digestion and Gibson PA5-52680; RRID:AB_2646594; 1:1000 for IB; lowercase and uppercase, specific to the chro- assembly. Sequence encoding the N-terminal 1:100 for IF) are commercially available. For mosomal and template sequences, respectively).

McKenna et al., Science 369, eabc5809 (2020) 25 September 2020 7of14 RESEARCH | RESEARCH ARTICLE

The DNA segment also contains a 31-amino- lysed in ice-cold homogenization buffer acid, incubated on ice for 15 min, and cen- acid-long protease-cleavable linker [GAT- [10 mM Tris-HCl (pH 7.4), 0.6 M sorbitol, trifuged at 21,000g for 1 min. The pellet was GGSTAGGATTASGTG ENLYFQG TASGGG; 1mMEDTA,1mMphenylmethylsulfonyl washed twice with ice-cold acetone and re- underlined, a Tobacco etch virus (TEV) pro- fluoride (PMSF), 0.2% bovine serum albumin suspended in protein sample buffer. tease cleavage site] between the last codon (BSA)] using a prechilled glass-Teflon Dounce of SPF1 and the first codon of GFP. DNA was homogenizer. One volume of homogenization Generation of KO and stable inducible cell lines then introduced to BY4741 by a standard buffer was added and the mixture was cen- All cell lines were maintained in Dulbecco’s lithium acetate transformation protocol. Trans- trifuged at 1500g for 5 min. The supernatant modified Eagle’s medium (DMEM) with high formants were isolated from a YPD agar me- was centrifuged at 4000g for 5 min and again glucose, GlutaMAX, and sodium pyruvate dium (1% yeast extract, 2% peptone, 2% glucose, at 12,000g for 15 min. The pellet from the (Thermo Fisher 10569) with 10% fetal bovine and 2% Bacto Agar) containing 100 mg/ml 12,000g spin was resuspended in SEM buf- serum (FBS) at 37°C and 5% CO2. Parent nourseothricin. Correct chromosomal inte- fer [20 mM MOPS-KOH (pH 7.2), 250 mM HEK293T, Flp-In T-REx 293, and HeLa T-REx gration was confirmed by PCR and Sanger sucrose] and layered on top of a step gradient cell lines were authenticated by short tandem sequencing. of 60%, 32%, 23%, and 15% sucrose in 10 mM repeat analysis. ATP13A1 KO cell lines were MOPS-KOH (pH 7.2). Crude membranes were generatedbytransfectingFlp-InT-RExHeLa Protein purification for biochemical assays centrifuged at 134,000g for 1 hour in a SW (Fig. 2D and figs. S5 to S9) or Flp-In T-REx 293 To purify TA(Bpa) in complex with CaM, 40Ti rotor, and mitochondria-enriched mem- (Fig. 2, A to C, and figs. S3 and S4) cells with BL21(DE3) cells were cotransformed with branes were collected from the 32%/60% pX459 plasmids containing target guide RNAs pGEX-CaM-FLAG-OMP25(amb) and pEVOL- sucrose boundary. The membranes were ad- using Lipofectamine 2000 or TransIT293 (Mirus Bpa and grown in Luria broth under appro- justed to 20 to 25 mg/ml protein as measured Bio), respectively, according to the manufac- Downloaded from priate antibiotic selection to an optical density after solubilization in 1% sodium dodecyl sul- turer’s instructions. After 24 hours, transfected 600 nm (OD600)of0.6.Thecultureswerethen fate (SDS), aliquoted, and frozen in liquid nit- cells were selected with 2 mg/ml puromycin for supplemented with 1 mM Bpa (Bachem 4017646) rogen. ER-enriched membranes were obtained 48 hours. Single clones were isolated and KOs and induced with 1% arabinose and 0.2 mM by centrifuging the post-12,000g supernatant validated by immunoblotting and amplicon isopropyl-b-D-thiogalactopyranoside for 6 hours at 234,787g for45mininatype45Tirotor.The sequencing. at 25°C. Cells were harvested and resuspended pellet was resuspended in RM buffer [50 mM To establish stable doxycycline-inducible in ice-cold lysis buffer [1× phosphate-buffered Hepes (pH 7.5), 100 mM KOAc, 2.5 mM Mg cell lines expressing either FLAG-tagged tail- http://science.sciencemag.org/ saline (PBS), 0.5 mM CaCl2, 1 mM dithiothrei- (OAc)2, 14% glycerol, 1 mM DTT, 1× PIC], ad- anchored proteins or FLAG-tagged ATP13A1, tol (DTT), 1× protease inhibitor cocktail (PIC); justed to ~50 mg/ml protein, aliquoted, and WT or ATP13A1 KO Flp-In T-REx HeLa or 293 Sigma-Aldrich 5056489001] and lysed by pass- flash-frozen in liquid nitrogen. Note that all cells were cotransfected with a 1:1 ratio of ing through a microfluidizer two times. Soluble membrane fractions showed cross-linking to pOG44 and pcDNA5/FRT/TO vector contain- material after centrifugation at 34,541g for Spf1 caused by residual ER contamination in ing the desired insert using Lipofectamine 20 min in a SS34 rotor (Sorvall) was applied the mitochondria-enriched membrane fractions. 2000/Lipofectamine 3000 or TransIT293 (Mirus to 1 ml of packed glutathione Sepharose (GE Bio), respectively. After 24 hours, cells were Life Sciences 17075601) resin equilibrated in Photo-cross-linking of recombinant TA with selected with 10 mg/ml blasticidin and 150 to lysis buffer per liter of culture. Columns were yeast membranes 300 mg/ml hygromycin for at least 2 weeks,

washed with 10 column volumes of lysis buf- Reactions were performed in import buffer and expression was validated by induction on January 4, 2021 fer and eluted with 50 mM Tris (pH 8), 25 mM [50 mM Hepes (pH 7.5), 100 mM KOAc, 2.5 mM with 10 ng/ml doxycycline for 24 to 48 hours reduced glutathione, and 0.5 mM CaCl2.Peak Mg(OAc)2, 0.5 mM CaCl2] containing 7.5 to and immunoblotting. elution fractions were pooled and dialyzed 20 mg/ml FLAG-TA(Bpa) in complex with CaM against 50 mM Hepes (pH 7.4), 150 mM KOAc, and 5 mg/ml mitochondria-enriched mem- Isolation of RMs 2mMMg(OAc)2, 10% glycerol, 1 mM DTT, and branes. FLAG-TA(Bpa) was released from CaM RMswereisolatedfromWT,ATP13A1KO,and 0.5 mM CaCl2 for 2 hours before the addition by addition of 2 mM EGTA, and the samples ATP13A1 KO Flp-In T-REx 293 cells stably re- of 1:100 w/w GST-3C protease and dialysis were transferred to a prechilled 96-well plate expressing FLAG-tagged WT or D533A ATP13A1 against fresh buffer for 12 to 16 hours. Dia- ~10cmunderaUVPB-100serieslampfor essentially as described previously (42). Cells lyzed material was passed over glutathione 10 min. Ten volumes of IP buffer [50 mM were cultured for at least two passages in Sepharose column to subtract cleaved GST Hepes (pH 7.5), 100 mM KOAc, 2.5 mM Mg antibiotic-free medium. Reexpression of ap- and GST-3C before being aliquoted and frozen (OAc)2, 1% Triton X-100, 1× PIC] were added proximately endogenous levels of FLAG- in liquid nitrogen. SGTA, GST-3C protease, to each well, and the samples were incubated ATP13A1 in the rescue lines (fig. S3A) was and Bpa-RS were expressed in BL21(DE3) on ice for 10 min to solubilize membranes and achieved by trace levels of tetracycline present Escherichia coli and purified as described centrifuged at 21,000g for 5 min. The super- intheFBSintheculturemedium,andnoad- previously (24, 46). natant was added to M2 anti-FLAG affinity ditional doxycycline was added. Cells were resin (Sigma-Aldrich A2220) and mixed at 4°C harvested and washed twice in cold PBS and Preparation of yeast membranes for 2 hours. M2 FLAG beads were washed centrifuged at 500g for 5 min. The cell pellet Yeast strains were grown in YPG to an OD600 twice with IP buffer, twice with IP buffer + was resuspended in three volumes of 10 mM of 2, harvested, washed in distilled water, re- 0.5 M NaCl, and another two times with IP Hepes (pH 7.4), 250 mM sucrose, 2 mM MgCl2, suspended in 100 mM Tris-H2SO4 (pH 9.4) buffer. Elutions were performed with 0.15 mg/ and1×PIC,andlysedby25to30passages with 10 mM DTT, and incubated at 30°C for ml 3x FLAG peptide in IP buffer at 30°C for through a 27-gauge needle. Lysates were cen- 20 min. Cells were then washed in zymolyase 20 min. Samples for TMT-MS analysis were trifuged twice at 3800g for 30 min, and the buffer [20 mM potassium phosphate (pH 7.4), treated with 10 mM DTT at 55°C for 30 min, supernatant was centrifuged at 75,000g for 1.2 M sorbitol], resuspended in zymolyase followed by 50 mM chloroacetamide at 25°C 1 hour at 4°C in a TLA-55 or TLA-100.3 rotor. buffer containing 3 mg of zymolyase per gram in the dark for 20 min. For immunoblotting, The RM pellet was resuspended in 10 mM of cells and incubated at 30°C for 30 min. After elutions were precipitated by the addition of Hepes (pH 7.4), 250 mM sucrose, 1 mM MgCl2, another wash in zymolyase buffer, cells were one-fifth the elution volume of trichloroacetic 0.5 mM DTT, adjusted and normalized to an

McKenna et al., Science 369, eabc5809 (2020) 25 September 2020 8of14 RESEARCH | RESEARCH ARTICLE

absorbance at 280 nm (A280)valueof30to60 15 min. Samples were centrifuged at 131,440g For imaging, all cells were washed twice in as measured after solubilization in 1% SDS, and for15mininaTLA100rotorandthepellets PBS and fixed in 4% formaldehyde in PBS for used directly for insertion and extraction assays. resuspended in protein sample buffer. 15 min. Cells were washed once in PBS, per- Insertion reactions of TA proteins with meabilized with 0.1% Triton X-100 in PBS for In vitro translation, insertion, extraction, and hRMs were at 32°C for 20 min (Fig. 2B, tot.) 5 min, and then incubated at room temper- photo-cross-linking or 45 min (fig. S3, B to D). For photo-cross- ature with blocking solution (10% FBS in For most in vitro transcription reactions, lin- linking (e.g., Fig. 1E), reactions were diluted PBS) for 1 hour, followed by primary anti- ear DNA templates were generated by PCR of 5-fold with PSB, UV irradiated, and immuno- bodies in 10% FBS/PBS 1 hour. After three SP64-based vectors using primers annealing precipitated as described above. Cotransla- washes in PBS, cells were incubated with slightly upstream of the SP6 promoter and tional insertion reactions (fig. S10H) were 1:500 fluorophore-conjugated secondary anti- ~200 bp downstream of the open reading subjected to denaturing IPs with a3F4 anti- bodies in 10% FBS/PBS for 1 hour and washed frame. Linear DNA templates for transcrib- bodies and protein A resin (CaptivA). Carbon- three times in PBS before imaging. For ing different C-terminal domain sequences ate extractions (fig. S3D) were performed by ATP13A1 transfection rescue experiments, (Fig. 1E) were generated using reverse primers the addition of 100 reaction volumes of cold cells were incubated with 1:100 DyLight containing the desired C-terminal sequence 100 mM NaCO3 (pH11.5),incubationonicefor 680-conjugated anti-HA antibody (Thermo and 16 bp downstream of the TGA stop codon 25 min, followed by centrifugation at 250,000g Fisher Scientific 26183-D680) in 10% FBS/ as follows: TGA GAA TTC CTA ATC ATG TC. for30mininaTLA100.3rotor.Pelletswere PBS for 1 hour at room temperature or over- DNA templates for testing type II constructs directly resuspended in protein sample buffer night at 4°C and washed three times with PBS. (fig. S10H) were generated by PCR directly for SDS-PAGE and autoradiography analysis. Next, 13 mm coverslips were placed onto from gBlocks (IDT) containing the SP6 pro- For analysis of insertion efficiency (fig. S3, B SlowFade Gold Antifade Mountant (Thermo Downloaded from moter, the desired protein-coding sequence, and C) and extraction reactions (Fig. 2, B and Fisher Scientific S36936) and sealed using nail and 16 bp downstream of the stop codon as C,andfig.S4),RMswerepelletedbycentrif- polish. Cells in 24- or 48-well plates were above. In vitro transcription reactions were ugationoverfourreactionvolumesofa20% imaged directly in PBS. performed with SP6 polymerase as described sucrose cushion in PSB at 186,000g for 20 min All images were collected with a Yokagawa previously (47)for1hourat37°Candused at 4°C in a TLA-55 rotor and resuspended in CSU-X1 spinning-disk confocal microscope with directly for in vitro translation. Suppressor half the initial reaction volume of PSB. Ex- Spectral Applied Research Aurora Borealis tRNA (e.g., Fig. 1E) was transcribed from traction reactions contained ~20 mg/ml RMs modification on a Nikon Ti motorized inverted http://science.sciencemag.org/ BstN1-lineared plasmid with T7 polymerase (estimated on the basis of total protein con- microscope equipped with Plan Apo 100×/1.4 at 37°C for 1 hour and isolated by phenol: centration), 1 mg/ml SGTA, and either an numerical aperture oil-immersion objective. chloroform extraction. Translation reactions energy-regenerating system (1 mM ATP, 1 mM Fluorescence was excited with solid-state lasers were performed as described previously (47) GTP, 1.2 mM creatine phosphate, 20 mg/ml at 488 nm (100 mW), 561 nm (100 mW), and at 32°C for 20 to 30 min. For photo-cross- creatine kinase) or 1 mM ATP, 1 mM AMP- 642 nm (100 mW) and collected using ET525/ linking, translation reactions were supple- PCP, or 0.1 U/ml apyrase as indicated, and 50m, ET620/60m, or ET700/75m emission fil- mented with an amber suppression system incubated at 32°C for 30 min. Samples were ters, respectively. Images were acquired with a (46) composed of 5 mM in vitro–transcribed directly analyzed before or after centrifuga- Hamamatsu ORCA-ER cooled CCD camera. suppressor tRNA, 0.5 mM Bpa, and 0.25 mM tion over a 20% sucrose cushion at 186,000g Z-series optical sections were collected with

Bpa-RS. Reactions for targeting assays (fig. S7) for 20 min. Chemical cross-linking (Fig. 2C) was a step size of 0.5 mm and single sections were on January 4, 2021 included 40 mM cold methionine in place of with 250 mM BMH (Thermo Fisher Scientific selected from approximately the center of 35S-methionine. Translation reactions of TA 22330)at4°Cfor1houranddirectlyquenched the cell. Brightness and contrast were adjusted proteins were stopped by the addition of 50 mg/ with protein sample buffer. identically for compared image sets using Fiji ml RNaseA before the addition of yeast mem- software. branes (final concentration of 2 to 5 mg/ml) IF assays Background subtraction of IF images was (e.g., Fig. 1E) or 1:10 v/v RMs (e.g., Fig. 2, A to Flp-In T-REx HeLa cells were cultured either performed using the robust Otsu threshold- C). Translation reactions of type II substrates on 13-mm coverslips or in glass-bottomed 24- ing method (49). A region of interest (ROI) (fig. S10H) included human cell RMs (hRMs). or 48-well plates (Mattek). For assays with cell was drawn around each cell in the FLAG Insertion reactions with yeast membranes lines harboring inducible substrates (Fig. 2D channel, and colocalization was assessed in were at 25°C for 20 min. For UV-activated and figs. S5 to S8) and analysis of endogenous 3D using the ImageJ Coloc2 plugin with a cross-linking, insertion reactions were diluted substrates (Fig. 2G and fig. S10, E and F), TA point spread function of 2 and 25 random- 10-fold in PSB [50 mM Hepes (pH 7.5), 100 mM protein or Flag-ATP13A1expressionwasin- izations. Manders’ colocalization coefficients KOAc, 2.5 mM Mg(OAc)2]andplaced~10cm duced with 100 ng/ml doxycycline for 24 hours. (MCCs) (50) (e.g., Fig. 2E) indicate the contri- under a UVP B-100 series lamp for 10 min in a Forrescueassayswithstablereporters,cellsin bution of FLAG signal above 0 that overlaps prechilled ice block. SDS was added to a final a 24-well glass-bottomed plate were tran- with mitochondria normalized to the total concentration of 1% and samples were heated siently transfected with 200 ng of pcDNA3 FLAG signal within the ROI. Pearson’scorre- to 65°C for 5 min. Ten volumes of IP buffer containing HA-tagged WT or D533A ATP13A1 lation coefficients (PCCs) are shown for cor- and 10 mlofpackedM2anti-FLAGresinwere using Lipofectamine 3000 according to the relation of FLAG signal with the signal of the added, and samples were incubated at 4°C manufacturer’sinstructions48hoursbefore ER marker TRAPa (fig. S6) because of the in- for 1 hour. M2 beads were washed four times analysis. For targeting assays (fig. S7), Flp-In fluence of unreliable thresholding of the dif- in IP buffer before the addition of protein T-REx HeLa cells in a 48-well glass-bottomed fuse ER signal on MCC values. Nonetheless, sample buffer. For carbonate extractions dish were washed twice with PSB and incu- both MCC1 and MCC2 values with various (fig. S1B), membranes from 4 mlofinsertion bated with 200 mg/ml digitonin in PSB for thresholding techniques support the trends reactions were pelleted by centrifugation over 10 min at 4°C. The cells were washed once shown in fig. S6, where SEC61b is unaffected, a 15% sucrose cushion in PSB at 10,000 g for with PSB before being incubated with 25 ml but mitochondrial TA protein colocalization 10 min at 4°C, resuspended in 150 ml 100 mM of in vitro translation reaction mixture of with the ER marker increases in ATP13A1 NaCO3 (pH 11.3), and incubated on ice for FLAG-b-OMP25 at 32°C for 15 min. KO cells. For rescue experiments by transient

McKenna et al., Science 369, eabc5809 (2020) 25 September 2020 9of14 RESEARCH | RESEARCH ARTICLE transfection (Fig. 2, D and E, and fig. S5, A proline content within the first 10 amino acids by ultracentrifugation (Beckman Type 45 Ti and B), HA staining was used to identify trans- (fig. S10G) was analyzed using the Biostrings rotor) at 120,000g for 1.5 hours and resus- fected cells; nontransfected cells were not R package and R scripts. RNA extraction and pended in buffer containing 50 mM Tris analyzed. The numbers of cells and fields of quantitative real-time reverse transcription– (pH 7.5), 200 mM NaCl, 1 mM EDTA, 1 mM views (in parentheses) analyzed were as fol- PCR analyses (fig. S10D) were performed by DTT, 10% glycerol, 2 mM BeF3,10mMMgCl2, lows: Fig. 2E, OMP25 WT, n=14 cells (4 fields Syd Labs, Inc. from frozen cell pellets using 5 mg/ml aprotinin, 5 mg/ml leupeptin, 1 mg/ml of view), KO, n=14 (3), KO + WT, n=17 (5), the following primers: MAVS forward, CAT pepstatin A, and 2 mM PMSF. The membranes KO + D533A, n=16 (4); fig. S5, BAK1 WT, n= GCT GGG CAG GTC AGT TA; MAVS reverse, were then solubilized with 1% DDM and 0.2% 18 (5), KO, n=18 (6), KO + WT, n=35 (10), TCA AAG CTA CCC TGG AAC GG; OMP25 CHS for 2.5 hours. After clarification by ultra- KO + D533A, n=22 (9); fig. S6, SEC61b WT, forward, ACC CTC CAC CCC TGC TAT TC; centrifugation, the lysate was incubated with n=10 (3), SEC61b KO, n=9(3),BAK1WT,n= OMP25 reverse, GGT TTG GGG TGG GTA TCA anti-GFP nanobody beads for 3 hours. The 8(3),BAK1KO,n=12 (3), OMP25 WT, n=9 GTC; ATP13A1 forward, GTT CGT CGG CTT beads were then washed with ~30 column (3), OMP25 KO, n=7(3);fig.S8C,BAK1WT CAT TGT GG; ATP13A1 reverse, TGG GAC GCA volumes of buffer W additionally containing cont., n=16 (3), WT KD, n=15 (5), KO cont., TTCTGGATCTC;SRP14forward,ATTGGT 10 mM MgCl2 and 2 mM BeF3. After eluting n=25 (5), KO KD, n=47 (10); fig. S8C, CTG TGC TGC ATG GT; and SRP14 reverse, with the TEV protease, purified Spf1 was in- OMP25 WT cont., n=11 (3), WT KD, n=15 TGG GTT GGG AAA GGA ATA AAG AG. SRP14 jected into a Superose 6 column equilibrated (3), KO cont., n=14 (4), KO KD, n=44 (10); and ATP13A1 were used as normalization and with the same wash buffer. Peak fractions were Fig. 2G, endogenous OMP25 WT, n=32 (10), positive controls. pooled, concentrated to ~5 mg/ml, and imme- KO, n=32 (10), KO + WT, n=19 (8), KO + diately used for cryo-EM grid preparation. D533A, n=32 (10); and fig. S10F, endoge- Purification of S. cerevisiae Spf1 for Downloaded from nous RMDN3 WT, n=30 (7), KO, n=25 (7), cryo-EM analysis Cryo-EM grid preparation and data acquisition KO + WT, n=23 (10), KO + D533A, n=25 (8). Yeast cells, in which endogenous Spf1 is tagged To prepare cryo-EM grids, gold Quantifoil To compare the level of IVT or stably ex- with C-terminal TEV-GFP, were grown in YPD R 1.2/1.3 Holey Carbon Grids were treated pressed tail-anchored protein localized to medium (1% yeast extract, 2% peptone, and with glow discharge for 35 s using a PELCO mitochondria (fig. S7C), FLAG signal within 2% glucose) in shaker flasks at 30°C to an easiGlow system (0.39 mBar air, 25 mA). mitochondrial ROIs was quantified from five OD600 of 2 to 4. Cells were harvested by cen- Then, 3 ml of the Spf1 sample were applied to (WT + IVT), four (KO + IVT; WT steady state), trifugation, frozen in liquid nitrogen, and agrid.Thegridwasblottedfor2to3swith http://science.sciencemag.org/ or three (KO steady state) different fields of stored at −80°C until use. Cells were lysed by grade 1 Whatman filter paper and plunge- view. Results were analyzed and plotted using cryomilling and resuspended in buffer con- frozen in liquid nitrogen–cooled liquid eth- GraphPad Prism software, and statistical sig- taining 50 mM Tris (pH 7.5), 200 mM NaCl, ane using Vitrobot Mark IV (FEI) operated at nificance was tested with unpaired two-tailed 1mMEDTA,1mMDTT,10%glycerol,5mg/ml 4°C and 100% humidity. Student’s t tests. aprotinin, 5 mg/ml leupeptin, 1 mg/ml pepsta- The apo and AlF4 datasets were collected tinA,and2mMPMSF.Allsubsequentsteps onaTalosArcticaelectronmicroscope(FEI) Miscellaneous biochemistry and analyses were performed at 4°C. To purify apo Spf1, 1% operated at an acceleration voltage of 200 kV, SDS-PAGE was with 8%, 10%, or 12% Tris- n-dodecyl-b-D-maltopyranoside (DDM; Anatrace) and the other datasets were collected on a tricine polyacrylamide gels with the Bio-Rad and 0.2% cholesteryl hemisuccinate (CHS; Titan Krios electron microscope (FEI) oper-

mini-gel electrophoresis system. Transfers to Anatrace) were directly added to the cell ly- ated at an acceleration voltage of 300 kV and on January 4, 2021 0.2 mm nitrocellulose membranes and immuno- sate. After 3 hours of solubilization, the lysate equipped with a Gatan GIF Quantum LS blotting were performed using conventional was clarified by ultracentrifugation (Beckman image filter (a slit width of 20 eV) (table S1). techniques in 5% milk or BSA in PBS + 0.1% Type 45 Ti rotor) at 120,000g for 1 hour, and Dose-fractionated images (~50 to 60 electrons Tween and imaged on a BioRad ChemiDoc the supernatant was mixed with Sepharose per Å2 over 42 frames) were recorded on a K3 Touch or by exposure to XAR film. Transfers beads conjugated with anti-GFP nanobody for direct electron detector (Gatan) operated in in 25 mM potassium phosphate (fig. S1C) (51) 2.5 hours. The beads were washed with ~30 the superresolution mode using SerialEM soft- were at 15 V for 12 hours at 4°C. For auto- column volumes of a wash buffer (buffer W) ware (54). The physical pixel size was 1.14 Å for radiography, gels were dried on filter paper containing 20 mM Tris (pH 7.5), 100 mM NaCl, apo, 0.90 Å for AlF4,and1.19Åfortheother before phosphor imaging or exposure to MR 1 mM EDTA, 1 mM DTT, 0.04% DDM, and datasets. Target defocus values were from −1.0 film. Quantification of autoradiography sig- 0.008% CHS. Spf1 was eluted by incubating to −2.4 mm. nal was done using phosphor imaging and the beads with ~10 mg/ml TEV protease (in ImageQuant software (GE Healthcare) or by buffer W) for ~14 hours. The eluate was con- Cryo-EM structure determination scanning film and with ImageJ software (NIH). centrated and injected into a Superose 6 In- The single-particle analysis procedures are de- Native size fractionations on 200-ml5to25% crease column (GE Life Sciences) equilibrated scribedinfigs.S11DandS17,BandC,andtable sucrose gradients (fig. S1A) were performed by with buffer W. Peak fractions were pooled and S1. First, tile-based motion correction of movies centrifugation at 55,000 rpm for 2 hours 25 min concentrated to ~5 mg/ml before cryo-EM grid (using 7 by 5 tiles), contrast transfer function in a TLS-55 rotor followed by manually collect- preparation. For Spf1 bound to AMP-PCP, AlF4, (CTF) estimation, and automatic particle pick- ing 11 fractions as described previously (24, 52). and BeF3, the running buffer for size-exclusion ing were performed using Warp (55). During Classification of candidates identified by chromatography additionally contained 10 mM this process, the corrected movies were 2× quantitative proteomics was initially analyzed MgCl2, and the purified Spf1 was incubated frame binned by averaging each two frames, by Gene Ontology (GO) enrichment terms and with 2 mM of AMP-PCP, AlF4, or BeF3 for resulting in a total of 21 frames per movie. then by manually curating specific annota- 1 hour at 4°C before grid preparation. Micrographs that were not suitable for image tions in UniProt. Signal sequences were re- For the substrate-bound structure, the puri- analysis (mostly those containing crystalline trieved from UniProt or, for unclearly annotated fication procedure was modified as follows. ice) were removed by manual inspection. All candidates, predicted using SignalP-5.0. Ex- After cryomilling, unbroken cells and large subsequent image processing was performed perimentally validated eukaryotic signal se- debris were removed by centrifugation at using cryoSPARC v2 (56) with the particles quences were extracted from SPdb (53), and 4000g for 10 min. Membranes were pelleted and2×-binnedmoviesimportedfromWarp.

McKenna et al., Science 369, eabc5809 (2020) 25 September 2020 10 of 14 RESEARCH | RESEARCH ARTICLE

Except for the substrate-bound Spf1 dataset, they were invisible or poorly resolved in the over a 90-min run into 96 fractions by high pH the particles were polished by per-particle mo- densitymaps:Nto3,45to54(aloopbetween reverse-phase HPLC (Agilent Technologies tion correction on the 2×-binned movies (57). TMa and TMb), 646 to 652 (a loop in the N LC1260) through an aeris peptide xb-c18 col- The dimensions of extracted particle images domain), 853 to 957 (the arm domain), and umn (Phenomenex; 250 mm × 3.6 mm) with were 256 pixels (320 pixels for the AlF4 data- 1212 to C. In addition, the segments 562 to 571 mobile phase A containing 5% acetonitrile and set). The particle images were then subjected (a part of the N domain), 842 to 852 (the arm 10 mM NH4HCO3 in liquid chromatography to reference-free 2D classification. Most par- domain), and 958 to 962 (the connection from (LC)–MS grade H2O, and mobile phase B con- ticles discarded by the 2D classification were the arm domain to the P domain) were mod- taining 90% acetonitrile and 10 mM NH4HCO3 empty detergent micelle particles. For the AlF4 eled with alanine because we could not un- in LC-MS grade H2O (both pH 8.0). The 96 re- and BeF3 datasets, the first 2D classification ambiguously register amino acids. Model sulting fractions were then pooled in a noncon- was performed before the particle polishing refinement was done in real space using tinuous manner into 24 fractions [as outlined and another 2D classification after the polishing. Phenix 1.16 (59), with the refinement resolution in supplementary figure 5 of (67)]. Fractions These particles were used for generation of limit set to the overall resolution of the map. were vacuum centrifuged to near dryness. Each three(forapo,AMP-PCP,andAlF4)orfive(for MolProbity (60) and EMRinger (61) in the consolidated fraction was desalted by StageTip, BeF3- and substrate-bound Spf1) initial models Phenix package was used for structural valid- dried again by vacuum centrifugation, and re- (ab initio reconstruction), followed by one or ation (table S1). Protein electrostatics were cal- constituted in 5% acetonitrile (ACN), 1% formic two rounds of 3D classification (heterogeneous culated using the Adaptive Poisson-Boltzmann acid (FA) for LC-MS/MS processing. refinement). For the apo, AMP-PCP, and AlF4 Solver (62, 63) with default parameters built in datasets, only one class (containing ~55 to 80% PyMOL v2.3 (with monovalent ion concentra- LC-MS/MS particles from the 2D classification) showed tions of 0.15 M each). Cavities were detected by For UV-cross-linking samples (Fig. 1C), MS data Downloaded from features of Spf1, which was used for 3D refine- Hollow (64). UCSF Chimera (65), Chimera X were collected using an Orbitrap Fusion Lumos ment. In the case of the BeF3 dataset (without (66), and PyMOL (Schrödinger) were used to mass spectrometer (Thermo Fisher Scientific) substrate), a minor class of the apo-like inward- prepare the structural figures in the paper. coupled to a Proxeon EASY-nLC1200 LC pump open structure (23% particles) in addition to (Thermo Fisher Scientific). Peptides were sep- the major class of the outward-open structure Sample preparation and digestion for MS arated on a 100-mm-inner-diameter micro- (57% particles) was found in the 3D classifica- Protein extracts were subjected to disulfide capillary column packed in-house with ~35 cm tion. In the case of the substrate-bound struc- bond reduction with 5 mM tris(2-carboxyethyl) of Accucore150 resin (2.6 mm, 150 Å, Thermo http://science.sciencemag.org/ ture, three out of five classes (total 87% phosphine (TCEP; room temperature, 10 min) Fisher Scientific) with a gradient consisting particles) showed equally clear Spf1 features and alkylation with 25 mM chloroacetamide of 5 to 22% (0 to 100 min), 22 to 28% (100 to with the outward-open conformation, whereas (room temperature, 20 min) followed by 110 min) (ACN, 0.1% FA) over a total 120-min the two other minor classes showed little or methanol-chloroform precipitation (Fig. 2F run at ~500 nl/min. Each analysis used the no protein densities. All three good classes and fig. S9) or trichloroacetic acid precipi- Multi-Notch MS3-based TMT method (68)to also showed a clear additional TM helix-like tation (Fig. 1C), before protease digestion. reduce ion interference compared with MS2 density of a putative endogenous substrate. Samples were resuspended in 100 mM EPPS quantification, combined with newly imple- Because substrate occupancy is likely partial, (pH 8.5) containing 0.1% RapiGest and di- mented Real Time Search analysis software we subjected particles to a second round of 3D gested at 37°C for 2 hours with LysC protease (69, 70). The scan sequence began with an 1

classification using two reference models, one at a 200:1 protein-to-protease ratio. Trypsin MS spectrum (Orbitrap analysis; resolution on January 4, 2021 of the three substrate-bound classes contain- was then added at a 100:1 protein-to-protease 120,000 at 200 Th; mass range 400 to 1400 m/z; ing the substrate density and a class without ratio, and the reaction was incubated for automatic gain control (AGC) target 1 × 106; a substrate (from the outward-open BeF3-bound 6 hours at 37°C. TMT labeling of each sample maximum injection time 240 ms). Precursors structure without a substrate). This resulted in was performed by adding indicated amount of for MS2 analysis were selected using a 3 s two classes with a stronger (class A) and weaker the 20 ng/mlstockofTMTorTMTproreagent TopSpeed method. MS2 analysis consisted of (class B) substrate density, and particles from along with acetonitrile to achieve a final ace- collision-induced dissociation [quadrupole ion class A were used for 3D refinement. Particles tonitrile concentration of ~30% (v/v). Typical- trap analysis; rapid scan rate; AGC 2.5×104; from selected classes were refined to the final ly, 5 mlofTMTproreagentwasaddedfor50mg isolation window 0.7 Th; normalized collision maps by CTF refinement and nonuniform re- of protein input (Fig. 2F and fig. S9) and 3 mlof energy (NCE) 35; maximum injection time finement. The overall resolution was esti- TMT reagent for affinity purification samples 60 ms]. Monoisotopic peak assignment was mated on the basis of gold-standard Fourier (Fig. 1C). After incubation at room temper- used, and previously interrogated precursors shell correlation of independently refined ature for 1 hour, labeling efficiency of a small were excluded using a dynamic window (120 s ± half maps and the 0.143 cutoff criterion. Local aliquot was tested, and the reaction was then 7 ppm). After acquisition of each MS2 spec- resolution was estimated in cryoSPARC using quenched with hydroxylamine to a final con- trum, a synchronous-precursor-selection (SPS) default parameters (figs. S12A and S18, M to centration of 0.5% (v/v) for 15 min. The TMT- API-MS3 scan was collected on the top 10 most P). Unless indicated otherwise, the maps shown labeled samples were pooled together at a 1:1 intense ions b or y ions matched by the online in figures were sharpened with B factors es- ratio.Thesamplewasvacuumcentrifugedto search algorithm in the associated MS2 spec- timated in the nonuniform refinement and near dryness, resuspended in 5% formic acid trum (69, 70). MS3 precursors were fragmented low-pass filtered at their resolution. for 15 min, centrifuged at 10,000g for 5 min at by high-energy collision-induced dissociation The initial atomic model was built de novo room temperature, and subjected to C18 solid- (HCD) and analyzed using the Orbitrap (NCE into the sharpened, low-pass–filtered com- phase extraction (Sep-Pak, Waters). 65; AGC 2.5×105; maximum injection time bined map of the apo Spf1 using Coot (58). For whole-proteome samples (Fig. 2F and 200 ms, resolution was 50,000 at 200 Th). Pre- Models for the other structures were built fig. S9), dried TMT-labeled sample was resus- viously identified peptides corresponding to using Coot after rigid-body fitting of indi- pended in 100 mlof10mMNH4HCO3 (pH 8.0) human SEC61B where included in a targeted vidual domains (or groups of TMs) of the apo andfractionatedusingbasic pH reverse-phase mass list and given priority for MS2 and SPS-MS3. Spf1 structure into corresponding maps. The high-performance liquid chromatography (HPLC) For whole-cell proteome analysis (Fig. 2F following segments were not modeled because (67). Briefly, samples were offline fractionated and fig. S9), MS data were collected using an

McKenna et al., Science 369, eabc5809 (2020) 25 September 2020 11 of 14 RESEARCH | RESEARCH ARTICLE

Orbitrap Fusion Lumos mass spectrometer set to seven amino acids. TMT tags on lysine and collapsed to a 1% peptide FDR, and then combined with a high-field asymmetric wave- residues and peptide N termini (+229.163 Da) collapsed further to a final protein-level FDR form ion mobility spectrometry (FAIMS) Pro and carbamidomethylation of cysteine resi- of 1%. Moreover, protein assembly was guided interface, coupled to a Proxeon EASY-nLC1200 dues (+57.021 Da) were set as static modifica- by principles of parsimony to produce the LC pump (Thermo Fisher Scientific). Peptides tions, and oxidation of methionine residues smallest set of proteins necessary to account were separated on a 100-mm-inner-diameter (+15.995 Da) was set as a variable modifica- for all observed peptides. Proteins were quan- microcapillary column packed in house with tion. For TMT-based reporter ion quantita- tified by summing reporter ion counts across ~35 cm of Accucore150 resin (2.6 mm, 150 Å, tion, we extracted (integration tolerance of all matching PSMs using in-house software, as Thermo Fisher Scientific) with a gradient con- 0.003 Da) the summed signal-to-noise ratio described previously (73). PSMs with poor sisting of 5 to 15% (0 to 70 min), 15 to 23% (70 (S:N) for each TMT channel and found the quality, MS3 spectra with more than 6 TMTpro to 85min) (ACN, 0.1% FA) over a total 95-min closest matching centroid to the expected mass reporter ion channels missing, or isolation spe- run at ~500 nl/min. For analysis, we loaded of the TMT reporter ion. For protein-level cificity <0.6, or with TMT reporter summed one-third of each fraction onto the column. To comparisons, peptide-spectrum matches (PSMs) S:Ns <200 or with no MS3 spectra were excluded reduce ion interference compared with MS2 were identified, quantified, and collapsed to a from quantification. quantification, each analysis used the Multi- 1% peptide FDR, and then collapsed further to Protein quantification values were exported Notch MS3-based TMT method (68), combined a final protein-level FDR of 1%. Moreover, pro- for further analysis in Microsoft Excel and with newly implemented Real Time Search tein assembly was guided by principles of par- Perseus (74), and statistical test and param- analysis software (69, 70) and the FAIMS Pro simony to produce the smallest set of proteins eters used are indicated in the corresponding Interface. The scan sequence began with an necessary to account for all observed peptides. datasets (data S1 and S2). Briefly, Welch’s t test MS1 spectrum (Orbitrap analysis; resolution Proteins were quantified by summing reporter analysis was performed to compare two data- Downloaded from 120,000 at 200 Th; mass range 400 to 1600 m/z; ion counts across all matching PSMs for spe- sets using the s0 parameter (in essence, a min- AGC target 8 × 105; maximum injection time cific peptides only. PSMs with poor quality, imal fold change cutoff) and correction for 50 ms). Precursors for MS2 analysis were sel- MS3 spectra with TMT reporter summed S: multiple comparisons was achieved by the ected using a cycle type of 1.25 s/CV method Ns that were less than 10 per channel, isola- permutation-based FDR method, both func- [FAIMS CV = –40/–60/–80 previously opti- tion specificity ≤0.2, or had less than 70% of tions that are built in in Perseus software. For mized for TMT multiplexed samples (71)]. MS2 correctly selected SPS ions were excluded from whole-cell proteome (Fig. 2F and fig. S9) anal- analysis consisted of collision-induced dissoci- quantification. ysis, each reporter ion channel was summed http://science.sciencemag.org/ ation (quadrupole ion trap analysis; rapid scan For whole-proteome analysis (Fig. 2F and across all quantified proteins and normalized rate; AGC 1.0 × 104; isolation window 0.7 Th; fig. S9), spectra were converted to mzXML assuming equal protein loading of all 16 sam- NCE 35; maximum injection time 35 ms). using a modified version of ReAdW.exe. Data- ples. For affinity-purified samples (Fig. 1C), nor- Monoisotopic peak assignment and a precursor base searching included all canonical entries malization based on SEC61B (total TMT sum fit filter were used (70% for a fit window of fromthehumanReferenceProteomeUniProt S:N) was performed. 0.7 Th), and previously interrogated precur- database (SwissProt 2019-01), as well as an in- sors were excluded using a dynamic window house curated list of contaminants. This data- (120 s ± 10 ppm). After acquisition of each base was concatenated with one composed of REFERENCES AND NOTES MS2 spectrum, an SPS API-MS3 scan was col- all protein sequences in the reversed order. 1. E. A. Costa, K. Subramanian, J. Nunnari, J. S. Weissman, Defining the physiological role of SRP in protein-targeting

lected on the top 10 most intense ions b or y Trypsin was used as the digestion enzyme, on January 4, 2021 efficiency and specificity. Science 359, 689–692 (2018). ions matched by the online search algorithm two missed cleavages were allowed, and the doi: 10.1126/science.aar3607; pmid: 29348368 in the associated MS2 spectrum (69, 70). MS3 minimal peptide length was set to seven ami- 2. M. Gamerdinger, M. A. Hanebuth, T. Frickey, E. Deuerling, The principle of antagonism ensures protein targeting specificity precursors were fragmented by HCD and an- no acids. Searches were performed using a – 5 at the endoplasmic reticulum. Science 348, 201 207 (2015). alyzed using the Orbitrap (NCE 45; AGC 2.5×10 ; 20-ppm precursor ion tolerance for total doi: 10.1126/science.aaa5335; pmid: 25859040 maximum injection time 200 ms, resolution protein-level analysis. The recommended 3. U. S. Chio, H. Cho, S. O. Shan, Mechanisms of tail-anchored was 50,000 at 200 Th). product ion parameters for ion trap ms/ms membrane protein targeting and insertion. Annu. Rev. Cell Dev. Biol. 33, 417–438 (2017). doi: 10.1146/ were used (1.0005 tolerance, 0.4 offset (mono Proteomics data analysis annurev-cellbio-100616-060839; pmid: 28992441 masses), theoretical_fragment_ions = 1). TMTpro 4. K. Krumpe et al., Ergosterol content specifies targeting of Mass spectra were processed using a Comet- tags on lysine residues and peptide N termini tail-anchored proteins to mitochondrial outer membranes. Mol. Biol. Cell 23, 3927–3935 (2012). doi: 10.1091/ based (v2018.01 rev.2) in-house software pipe- (+304.2071 Da) and carbamidomethylation of mbc.e11-12-0994; pmid: 22918956 line (72, 73)orSequest-HT(forFig.1C)using cysteine residues (+57.021 Da) were set as sta- 5. M. Dyla, M. Kjaergaard, H. Poulsen, P. Nissen, Structure and Proteome Discoverer (v2.3.0.420; Thermo Fish- tic modifications, whereas oxidation of methi- mechanism of P-type ATPase ion pumps. Annu. Rev. Biochem. 89, 583–603 (2019). doi: 10.1146/annurev-biochem-010611- er Scientific). onine residues (+15.995 Da) was set as a 112801; pmid: 31874046 The identification of UV-cross-linked proteins variable modification. PSMs were adjusted 6. J. P. Andersen et al., P4-ATPases as phospholipid flippases: (Fig. 1C) was performed using the SEQUEST-HT to a 1% FDR, and PSM filtering was performed Structure, function, and enigmas. Front. Physiol. 7, 275 engine (Proteome Discoverer v2.3.0.420; Thermo using a linear discriminant analysis, as de- (2016). doi: 10.3389/fphys.2016.00275; pmid: 27458383 7. D. M. Sørensen, M. J. Buch-Pedersen, M. G. Palmgren, Fisher Scientific) against the UniProt S. cerevisiae scribed previously (73), while considering the Structural divergence between the two subgroups of (S288c) Reference Proteome (07/2017), as well following parameters: Comet Log Expect and P5 ATPases. Biochim. Biophys. Acta 1797, 846–855 (2010). as an in-house curated list of contaminants Diff Seq. Delta Log Expect, missed cleavages, doi: 10.1016/j.bbabio.2010.04.010; pmid: 20416272 8. S. van Veen et al., ATP13A2 deficiency disrupts lysosomal and the human SEC61B sequence, using the peptide length, charge state, and precursor polyamine export. Nature 578, 419–424 (2020). doi: 10.1038/ following parameters (after recalibration): a mass accuracy. For TMT-based reporter ion s41586-020-1968-7; pmid: 31996848 tolerance level of 10 ppm for MS1 and 0.7 Da quantitation, we extracted the summed S:N 9. D. M. Sørensen et al., Parkinson disease related 2 ATP13A2 evolved early in animal evolution. PLOS ONE for MS and false discovery rate (FDR) of the ratio for each TMT channel and found the 13 , e0193228 (2018). doi: 10.1371/journal.pone.0193228; Percolator decoy database search was set to closest matching centroid to the expected pmid: 29505581 1%. Trypsin and LysC were used as the di- mass of the TMT reporter ion (integration 10. S. R. Cronin, R. Rao, R. Y. Hampton, Cod1p/Spf1p is a P-type ATPase involved in ER function and Ca2+ homeostasis. gestion enzymes, two missed cleavages were tolerance of 0.003 Da). For protein-level com- J. Cell Biol. 157, 1017–1028 (2002). doi: 10.1083/ allowed, and the minimal peptide length was parisons, PSMs were identified, quantified, jcb.200203052; pmid: 12058017

McKenna et al., Science 369, eabc5809 (2020) 25 September 2020 12 of 14 RESEARCH | RESEARCH ARTICLE

11. V. Dederer et al., Cooperation of mitochondrial and ER factors 32. R. S. Hegde, H. D. Bernstein, The surprising complexity of 55. D. Tegunov, P. Cramer, Real-time cryo-electron microscopy in quality control of tail-anchored proteins. eLife 8, e45506 signal sequences. Trends Biochem. Sci. 31, 563–571 (2006). data preprocessing with Warp. Nat. Methods 16, (2019). doi: 10.7554/eLife.45506; pmid: 31172943 doi: 10.1016/j.tibs.2006.08.004; pmid: 16919958 1146–1152 (2019). doi: 10.1038/s41592-019-0580-y; 12. C. Suzuki, Y. I. Shimma, P-type ATPase spf1 mutants show 33. C. Toyoshima, M. Nakasako, H. Nomura, H. Ogawa, Crystal pmid: 31591575 a novel resistance mechanism for the killer toxin SMKT. structure of the calcium pump of sarcoplasmic reticulum at 56. A. Punjani, J. L. Rubinstein, D. J. Fleet, M. A. Brubaker, Mol. Microbiol. 32, 813–823 (1999). doi: 10.1046/ 2.6 A resolution. Nature 405, 647–655 (2000). doi: 10.1038/ cryoSPARC: Algorithms for rapid unsupervised cryo-EM j.1365-2958.1999.01400.x; pmid: 10361284 35015017; pmid: 10864315 structure determination. Nat. Methods 14, 290–296 (2017). 13. D. M. Sørensen et al., The P5A ATPase Spf1p is stimulated 34. T. L. Sørensen, J. V. Møller, P. Nissen, Phosphoryl transfer and doi: 10.1038/nmeth.4169; pmid: 28165473 by phosphatidylinositol 4-phosphate and influences cellular calcium ion occlusion in the calcium pump. Science 304, 57. J. L. Rubinstein, M. A. Brubaker, Alignment of cryo-EM movies sterol homeostasis. Mol. Biol. Cell 30, 1069–1084 (2019). 1672–1675 (2004). doi: 10.1126/science.1099366; of individual particles by optimization of image translations. doi: 10.1091/mbc.E18-06-0365; pmid: 30785834 pmid: 15192230 J. Struct. Biol. 192, 188–195 (2015). doi: 10.1016/ 14. S. R. Cronin, A. Khoury, D. K. Ferry, R. Y. Hampton, Regulation 35. J. P. Morth et al., Crystal structure of the sodium-potassium j.jsb.2015.08.007; pmid: 26296328 of HMG-CoA reductase degradation requires the P-type pump. Nature 450, 1043–1049 (2007). doi: 10.1038/ 58. P. Emsley, B. Lohkamp, W. G. Scott, K. Cowtan, Features ATPase Cod1p/Spf1p. J. Cell Biol. 148, 915–924 (2000). nature06419; pmid: 18075585 and development of Coot. Acta Crystallogr. D Biol. Crystallogr. doi: 10.1083/jcb.148.5.915; pmid: 10704442 36. K. Abe, K. Irie, H. Nakanishi, H. Suzuki, Y. Fujiyoshi, 66, 486–501 (2010). doi: 10.1107/S0907444910007493; 15. S. Vashist, C. G. Frank, C. A. Jakob, D. T. Ng, Two distinctly Crystal structures of the gastric proton pump. Nature pmid: 20383002 localized p-type ATPases collaborate to maintain organelle 556,214–218 (2018). doi: 10.1038/s41586-018-0003-8; 59. P. V. Afonine et al., Real-space refinement in PHENIX for homeostasis required for glycoprotein processing and pmid: 29618813 cryo-EM and crystallography. Acta Crystallogr. D Struct. quality control. Mol. Biol. Cell 13, 3955–3966 (2002). 37. M. Bublitz, H. Poulsen, J. P. Morth, P. Nissen, In and out of Biol. 74, 531–544 (2018). doi: 10.1107/S2059798318006551; doi: 10.1091/mbc.02-06-0090; pmid: 12429838 the cation pumps: P-type ATPase structure revisited. pmid: 29872004 16. A. Ando, C. Suzuki, Cooperative function of the CHD5-like Curr. Opin. Struct. Biol. 20, 431–439 (2010). doi: 10.1016/ 60. V. B. Chen et al., MolProbity: All-atom structure validation for protein Mdm39p with a P-type ATPase Spf1p in the j.sbi.2010.06.007; pmid: 20634056 macromolecular crystallography. Acta Crystallogr. D Biol. maintenance of ER homeostasis in Saccharomyces cerevisiae. 38. M. Timcenko et al., Structure and autoregulation of a Crystallogr. 66,12–21 (2010). doi: 10.1107/ Mol. Genet. Genomics 273, 497–506 (2005). doi: 10.1007/ P4-ATPase lipid flippase. Nature 571, 366–370 (2019). S0907444909042073; pmid: 20057044 s00438-005-1153-6; pmid: 15909163 doi: 10.1038/s41586-019-1344-7; pmid: 31243363 61. B. A. Barad et al., EMRinger: Side chain-directed model

17. M. C. Jonikas et al., Comprehensive characterization of genes 39. M. Hiraizumi, K. Yamashita, T. Nishizawa, O. Nureki, Cryo-EM and map validation for 3D cryo-electron microscopy. Downloaded from required for protein folding in the endoplasmic reticulum. structures capture the transport cycle of the P4-ATPase Nat. Methods 12, 943–946 (2015). doi: 10.1038/nmeth.3541; Science 323, 1693–1697 (2009). doi: 10.1126/science.1167983; flippase. Science 365,1149–1155 (2019). doi: 10.1126/science. pmid: 26280328 pmid: 19325107 aay3353; pmid: 31416931 62. N. A. Baker, D. Sept, S. Joseph, M. J. Holst, J. A. McCammon, 18. Y. Cohen et al., The yeast p5 type ATPase, spf1, regulates 40. L. Bai et al., Autoinhibition and activation mechanisms of the Electrostatics of nanosystems: Application to microtubules and manganese transport into the endoplasmic reticulum. eukaryotic lipid flippase Drs2p-Cdc50p. Nat. Commun. 10,4142 the ribosome. Proc. Natl. Acad. Sci. U.S.A. 98, 10037–10041 PLOS ONE 8, e85519 (2013). doi: 10.1371/ (2019). doi: 10.1038/s41467-019-12191-9; pmid: 31515475 (2001). doi: 10.1073/pnas.181342398; pmid: 11517324 journal.pone.0085519; pmid: 24392018 41. K. G. Hansen et al., An ER surface retrieval pathway safeguards 63. T. J. Dolinsky, J. E. Nielsen, J. A. McCammon, N. A. Baker,

19. C. Suzuki, Immunochemical and mutational analyses of the import of mitochondrial membrane proteins in yeast. PDB2PQR: An automated pipeline for the setup of http://science.sciencemag.org/ P-type ATPase Spf1p involved in the yeast secretory Science 361, 1118–1122 (2018). doi: 10.1126/science.aar8174; Poisson-Boltzmann electrostatics calculations. pathway. Biosci. Biotechnol. Biochem. 65, 2405–2411 (2001). pmid: 30213914 Nucleic Acids Res. 32 (Web Server), W665-7 (2004). doi: 10.1271/bbb.65.2405; pmid: 11791712 42. P. J. Chitwood, S. Juszkiewicz, A. Guna, S. Shao, R. S. Hegde, doi: 10.1093/nar/gkh381; pmid: 15215472 20. D. J. Tipper, C. A. Harley, Yeast genes controlling responses EMC is required to initiate accurate membrane protein 64. B. K. Ho, F. Gruswitz, HOLLOW: Generating accurate to topogenic signals in a model transmembrane protein. topogenesis. Cell 175, 1507–1519.e16 (2018). doi: 10.1016/ representations of channel and interior surfaces in molecular Mol. Biol. Cell 13, 1158–1174 (2002). doi: 10.1091/ j.cell.2018.10.009; pmid: 30415835 structures. BMC Struct. Biol. 8, 49 (2008). doi: 10.1186/ mbc.01-10-0488; pmid: 11950929 43. N. Volkmar, J. C. Christianson, Squaring the EMC: How 1472-6807-8-49; pmid: 19014592 21. D. M. Sørensen, H. W. Holen, T. Holemans, P. Vangheluwe, promoting membrane protein biogenesis impacts 65. E. F. Pettersen et al., UCSF Chimera: A visualization system for M. G. Palmgren, Towards defining the substrate of orphan cellular functions and organismal homeostasis. J. Cell Sci. exploratory research and analysis. J. Comput. Chem. 25, P5A-ATPases. Biochim. Biophys. Acta 1850, 524–535 (2015). 133, jcs243519 (2020). doi: 10.1242/jcs.243519; 1605–1612 (2004). doi: 10.1002/jcc.20084; pmid: 15264254 doi: 10.1016/j.bbagen.2014.05.008; pmid: 24836520 pmid: 32332093 66. T. D. Goddard et al., UCSF ChimeraX: Meeting modern 22. S. Shao, R. S. Hegde, A calmodulin-dependent translocation 44. E. Itakura et al., Ubiquilins chaperone and triage mitochondrial challenges in visualization and analysis. Protein Sci. 27,14–25 pathway for small secretory proteins. Cell 147, 1576–1588 membrane proteins for degradation. Mol. Cell 63,21–33 (2018). doi: 10.1002/pro.3235; pmid: 28710774

(2011). doi: 10.1016/j.cell.2011.11.048; pmid: 22196732 (2016). doi: 10.1016/j.molcel.2016.05.020; pmid: 27345149 67. J. A. Paulo et al., Quantitative mass spectrometry-based on January 4, 2021 23. A. Guna, N. Volkmar, J. C. Christianson, R. S. Hegde, The ER 45. J. W. Chin, A. B. Martin, D. S. King, L. Wang, P. G. Schultz, multiplexing compares the abundance of 5000 S. cerevisiae membrane protein complex is a transmembrane domain Addition of a photocrosslinking amino acid to the genetic proteins across 10 carbon sources. J. Proteomics insertase. Science 359, 470–473 (2018). doi: 10.1126/science. code of Escherichia coli. Proc. Natl. Acad. Sci. U.S.A. 99, 148,85–93 (2016). doi: 10.1016/j.jprot.2016.07.005; aao3099; pmid: 29242231 11020–11024 (2002). doi: 10.1073/pnas.172226299; pmid: 27432472 24. S. Shao, M. C. Rodrigo-Brenni, M. H. Kivlen, R. S. Hegde, pmid: 12154230 68. G. C. McAlister et al., MultiNotch MS3 enables accurate, Mechanistic basis for a molecular triage reaction. 46. Z. Lin et al., TTC5 mediates autoregulation of tubulin via mRNA sensitive, and multiplexed detection of differential expression Science 355, 298–302 (2017). doi: 10.1126/science.aah6130; degradation. Science 367, 100–104 (2020). doi: 10.1126/ across cancer cell line proteomes. Anal. Chem. 86, 7150–7158 pmid: 28104892 science.aaz4352; pmid: 31727855 (2014). doi: 10.1021/ac502040v; pmid: 24927332 25. N. Wiedemann, N. Pfanner, Mitochondrial machineries for 47. Q. Feng, S. Shao, In vitro reconstitution of translational arrest 69. B. K. Erickson et al., Active instrument engagement combined protein import and assembly. Annu. Rev. Biochem. 86, 685–714 pathways. Methods 137,20–36 (2018). doi: 10.1016/ with a real-time database search for improved performance of (2017). doi: 10.1146/annurev-biochem-060815-014352; j.ymeth.2017.12.018; pmid: 29277545 sample multiplexing workflows. J. Proteome Res. 18, – pmid: 28301740 48. F. A. Ran et al., Genome engineering using the CRISPR-Cas9 1299 1306 (2019). doi: 10.1021/acs.jproteome.8b00899; 26. Y. C. Chen et al., Msp1/ATAD1 maintains mitochondrial system. Nat. Protoc. 8, 2281–2308 (2013). doi: 10.1038/ pmid: 30658528 function by facilitating the degradation of mislocalized tail- nprot.2013.143; pmid: 24157548 70. D. K. Schweppe et al., Full-featured, real-time database anchored proteins. EMBO J. 33, 1548–1564 (2014). 49. N. Otsu, A threshold selection method from gray-level searching platform enables fast and accurate multiplexed – doi: 10.15252/embj.201487943; pmid: 24843043 histograms. IEEE Trans. Syst. Man Cybern. 9,62–66 (1979). quantitative proteomics. J. Proteome Res. 19, 2026 2034 27. C. U. Mårtensson et al., Mitochondrial protein translocation- doi: 10.1109/TSMC.1979.4310076 (2020). doi: 10.1021/acs.jproteome.9b00860; pmid: 32126768 associated degradation. Nature 569, 679–683 (2019). 50. E. M. M. Manders, F. J. Verbeek, J. A. Aten, Measurement 71. D. K. Schweppe et al., Characterization and optimization of doi: 10.1038/s41586-019-1227-y; pmid: 31118508 of co-localization of objects in dual-colour confocal images. multiplexed quantitative analyses using high-field asymmetric- 28. M. L. Wohlever, A. Mateja, P. T. McGilvray, K. J. Day, J. Microsc. 169, 375–382 (1993). doi: 10.1111/ waveform ion mobility mass spectrometry. Anal. Chem. 91, – R. J. Keenan, Msp1 is a membrane protein dislocase for j.1365-2818.1993.tb03313.x 4010 4016 (2019). doi: 10.1021/acs.analchem.8b05399; tail-anchored proteins. Mol. Cell 67, 194–202.e6 (2017). 51. M. T. Hincke, Conditions for improved adsorption of calmodulin pmid: 30672687 doi: 10.1016/j.molcel.2017.06.019; pmid: 28712723 to nitrocellulose: Detection by 45Ca binding. Electrophoresis 72. J. K. Eng, T. A. Jahan, M. R. Hoopmann, Comet: An open- 29. V. Okreglak, P. Walter, The conserved AAA-ATPase Msp1 9, 303–306 (1988). doi: 10.1002/elps.1150090704; source MS/MS sequence database search tool. Proteomics 13, – confers organelle specificity to tail-anchored proteins. pmid: 3234368 22 24 (2013). doi: 10.1002/pmic.201200439; pmid: 23148064 Proc. Natl. Acad. Sci. U.S.A. 111, 8019–8024 (2014). 52. M. C. J. Yip et al., Mechanism for recycling tRNAs on stalled 73. E. L. Huttlin et al., A tissue-specific atlas of mouse protein – doi: 10.1073/pnas.1405755111; pmid: 24821790 ribosomes. Nat. Struct. Mol. Biol. 26, 343–349 (2019). phosphorylation and expression. Cell 143, 1174 1189 (2010). 30. S. Matsumoto et al., Msp1 clears mistargeted proteins by doi: 10.1038/s41594-019-0211-4; pmid: 31011209 doi: 10.1016/j.cell.2010.12.001; pmid: 21183079 facilitating their transfer from mitochondria to the ER. Mol. Cell 53. K. H. Choo, T. W. Tan, S. Ranganathan, SPdb: A signal peptide 74. S. Tyanova et al., The Perseus computational platform for 76, 191–205.e10 (2019). doi: 10.1016/j.molcel.2019.07.006; database. BMC Bioinformatics 6, 249 (2005). doi: 10.1186/ comprehensive analysis of (prote)omics data. Nat. Methods 13, – pmid: 31445887 1471-2105-6-249; pmid: 16221310 731 740 (2016). doi: 10.1038/nmeth.3901; pmid: 27348712 31. E. Park, T. A. Rapoport, Mechanisms of Sec61/SecY-mediated 54. D. N. Mastronarde, Automated electron microscope protein translocation across membranes. Annu. Rev. Biophys. tomography using robust prediction of specimen movements. ACKNOWLEDGMENTS 41,21–40 (2012). doi: 10.1146/annurev-biophys-050511- J. Struct. Biol. 152,36–51 (2005). doi: 10.1016/ We thank D. Toso and J. Remis for electron microscope operational 102312; pmid: 22224601 j.jsb.2005.07.007; pmid: 16182563 support; the Nikon Imaging Center at HMS for training and

McKenna et al., Science 369, eabc5809 (2020) 25 September 2020 13 of 14 RESEARCH | RESEARCH ARTICLE

support; C. Yapp at the Image and Data Analysis (IDAC) core at collected and analyzed cryo-EM data; E.P. built atomic models bound Spf1. Renewable yeast strains, cell lines, and biochemical HMS and M. Hoyer for help with colocalization analysis; M. C. J. Yip of Spf1; M.J.M. and S.S. performed biochemical experiments; reagents are available by contacting E.P. or S.S. for help generating reagents; the R. S. Hegde laboratory for M.J.M., L.W., and S.S. generated cell lines; M.J.M. performed all IF mammalian Bpa suppression constructs; S. Klinge for the yeast experiments; A.O. performed all TMT-MS analysis under the SUPPLEMENTARY MATERIALS integration plasmid; S. Elsasser, K. Y. S. Hung, and the D. Finley supervision of J.W.H.; and S.S. and E.P. supervised the project and science.sciencemag.org/content/369/6511/eabc5809/suppl/DC1 laboratory for advice and yeast strains; J. A. Paulo for proteomics wrote the paper with input from all authors. Competing interests: Figs. S1 to S20 support; T. A. Rapoport and R. S. Hegde for critical reading of J.W.H. is a founder and advisory board member for Caraway Table S1 the manuscript; and Shao and Park laboratory members for Therapeutics, an adviser for X-Chem Inc., a reviewing editor for Movies S1 to S3 discussions. Funding: This work was supported by the American Elife, and an associate editor for Science Advances. Data and Data S1 to S2 Heart Association (grant no. 19POST34400009 to M.J.M.), the materials availability: EM maps and models are available through MDAR Reproducibility Checklist CRSP of HMS Department of Cell Biology (L.W.), NIH grant nos. EM Data Bank (EMDB) and the Protein Data Bank (PDB) under R37NS083524 and RO1NS110395 (J.W.H), NIH grant no. DP2GM137415 the following accession codes: EMD-22260 and PDB ID 6XMP for apo View/request a protocol for this paper from Bio-protocol. (S.S.), the Smith Family Foundation (S.S.), a Packard fellowship Spf1; EMD-22261 and 6XMQ for AMP-PCP–bound Spf1; EMD-22262

(S.S.), and the Vallee Scholars Program (S.S., E.P.). Author and 6XMS for AlF4-bound Spf1; EMD-22263 and 6XMT for BeF3-bound 1 May 2020; accepted 28 July 2020 contributions: S.I.S. prepared cryo-EM samples; S.I.S. and E.P. Spf1; and EMD-22264 and 6XMU for BeF3/endogenous substrate- 10.1126/science.abc5809 Downloaded from http://science.sciencemag.org/

on January 4, 2021

McKenna et al., Science 369, eabc5809 (2020) 25 September 2020 14 of 14 The endoplasmic reticulum P5A-ATPase is a transmembrane helix dislocase Michael J. McKenna, Sue Im Sim, Alban Ordureau, Lianjie Wei, J. Wade Harper, Sichen Shao and Eunyong Park

Science 369 (6511), eabc5809. DOI: 10.1126/science.abc5809

Dissecting membrane dislocation Mistargeting and misinsertion of membrane proteins causes proteostasis stress and dysfunction of intracellular organelles, which can lead to disease. McKenna et al. found that a conserved orphan P-type adenosine triphosphatase (ATPase) transporter removes misinserted transmembrane segments from the endoplasmic reticulum (ER). Functional

reconstitutions and cryo−electron microscopy structures show how this ATPase selectively extracts mitochondrial Downloaded from proteins that are mistargeted to the ER and transmembrane segments that are inserted in the wrong orientation. This work identifies polypeptides as a new class of P-type ATPase substrates and defines a new protein quality-control mechanism at the ER. Science, this issue p. eabc5809 http://science.sciencemag.org/

ARTICLE TOOLS http://science.sciencemag.org/content/369/6511/eabc5809

SUPPLEMENTARY http://science.sciencemag.org/content/suppl/2020/09/23/369.6511.eabc5809.DC1 MATERIALS

REFERENCES This article cites 74 articles, 19 of which you can access for free on January 4, 2021 http://science.sciencemag.org/content/369/6511/eabc5809#BIBL

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Use of this article is subject to the Terms of Service

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. The title Science is a registered trademark of AAAS. Copyright © 2020 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works