Sterol oxidation mediates stress-responsive Vms1 translocation to mitochondria

Vms1 crystal structure shows the LRS binding a hydrophobic groove on the MTD surface

To further understand the intramolecular regulation of Vms1 localization to mitochondria, we visualized the Vms1 LRS-MTD interaction by determining a crystal structure of Vms1^LRS-ZnF-MTD (Figure 1A). Many constructs of S. cerevisiae Vms1 were prepared by expression and purification from E. coli and subjected to crystallization trials. The crystalized construct comprised residues 1-417, and lacked residues 38-69, which are poorly conserved and were predicted to form an unstructured loop (Figure 1A). Removal of this loop had no effect on Vms1 function in vivo (Figure S1). The structure was determined using single-wavelength anomalous diffraction from the selenomethionine-substituted protein, and was refined at a resolution of 2.7 Å to R_work/R_free values of 18.7/24.5% (Table 1).

The Vms1 model comprises the LRS (residues 13-35), portions of the ZnF domain including the Zn²⁺ ion and its coordinating residues (residues 74-80, 90-106), and the majority of the MTD (residues 188-197, 203-257, 267-286, 318-397) (Figure 1B). No other residues in the Vms1^LRS-ZnF-MTD construct are visible in the crystal structure, presumably because they are highly mobile or because they were excised by the low level of protease that we found to be necessary to grow crystals (Figure 1A). The Vms1^MTD adopts a fold in which a mixed β-sheet is flanked on both sides by α-helices (Figure 1B). Comparison with structures in the PDB by the DALI server (Holm and Rosenstrom, 2010) indicated that the MTD fold is most closely related to eukaryotic peptide chain release factor subunit 1 (3e1y-C), although the overlap is only on 117 pairs of Cα atoms and gives a large root mean square deviation (5.7Å). This raises the intriguing speculation that Vms1 might have an additional uncharacterized function as a release factor.

LRS and MTD mutations that disrupt the LRS-MTD interface alter mitochondrial localization

The LRS (residues 13-35) packs against the MTD through an interface, that includes LRS residues that are hydrophobic and conserved. The N-terminal helix (residues 14-18) is not highly conserved (Figure 1C). Additional conserved hydrophobic residues follow in an extended turn (residues 19-24) that are buried against the MTD (Figure 1C, Ile20, Phe21, Leu23). This is followed by a second helix (residues 25-30) that includes two conserved Leu residues that are partially buried. Lastly, residues 31-35 comprise a short loop containing highly conserved and buried Leu residues (Figure 1C, Leu31, Leu33). The only residue in the entire LRS that is conserved and not buried is Ser35. We previously identified the conserved Leu residues (Leu 23, 28, 31, 33) in the LRS as being important for the interaction with the MTD (Heo et al., 2013) (Figure 1C). Our structure revealed that these residues mediate multiple hydrophobic inter-domain contacts at the LRS-MTD interface (Figure 1D). Moreover, mutation of these LRS Leu residues led to greater mitochondrial localization of Vms1 (Heo et al., 2013), which suggests that displacement of the LRS may unmask a surface of Vms1 that is important for binding to a mitochondrial receptor and/or the mitochondrial membrane.

The MTD surface buried by the LRS is a large (1,205Ų) groove that is predominately comprised of hydrophobic residues that contact the conserved LRS hydrophobic residues (Figure 1D). The majority of these hydrophobic MTD residues reside in two strands (residues 190-194, 203-207), although Leu210, Leu223, and Leu392 also contribute (Figure 1E-F). A few hydrophilic residues, which are not conserved, also lie at the LRS/MTD interface (Asn227, His253, Arg255).

Guided by the structure, we further explored the importance of residues at the LRS-MTD interface and across the MTD surface for mitochondrial localization by determining the localization of Vms1-GFP in S. cerevisiae when 19 conserved MTD residues were substituted individually or in clusters. Localization was tested in the context of the Vms1^MTD-GFP construct because its constitutive and robust mitochondrial localization simplifies a quantitative comparison between mutants. We identified several mutants that displayed impaired mitochondrial localization of the Vms1^MTD-GFP fusion protein and others that did not (representative examples are shown in Figure 2A-B; summary shown in Figure 2D-E). The reduction in mitochondrial localization did not result from a reduction in protein abundance because the abundance of each mutant is similar to wild type (Figure 2C).

Building on the earlier mutational study of conserved LRS Leu residues (Heo et al., 2013), we verified that the MTD residues that are located at the LRS-MTD interface are also important for LRS-MTD binding in solution. This was done by mutating the MTD residues Phe193 or Ile204, which are buried against the LRS (Figure 2D-E). In both cases, co-expression of these mutated MTD-GFP constructs with a LRS-HA construct showed impaired interaction in co-immunoprecipitation experiments (Figure 2F). Interestingly, these mutants exhibited reduced mitochondrial localization (Figure 2A-B). This is in contrast to the analogous mutations of hydrophobic residues on the LRS side of the interface, which also showed reduced LRS-MTD interaction but displayed increased mitochondrial localization (Heo et al., 2013). These observations are consistent with the model that the LRS masks MTD residues that are important for mitochondrial localization, such that mutation of LRS residues that disrupt the MTD interaction uncovers a surface that promotes mitochondrial localization, whereas mutation of MTD residues that disrupt the LRS interaction might also disrupt a part of the MTD surface that mediates mitochondrial localization.

MTD surface residues close to the LRS are also important for mitochondrial localization

Our survey of residues across the MTD surface revealed a number of mutants that are not buried at the LRS interface yet displayed >10% reduced mitochondrial localization (Figure 2C-D). These residues were all from the region surrounding the Vms1^LRS interaction interface, whereas mutations elsewhere had little or no effect on Vms1 localization (Figure 2A-B). For example, Y190D and K194D/K196D, which are within 10Å of the LRS but are not at the LRS interface and display normal interaction with the Vms1^LRS, showed >10% reduced localization to mitochondria (Figure 2A-B, 2F). The observation that MTD residues both at and surrounding the LRS interface mediate MTD localization to mitochondria suggests that the localization of Vms1 to mitochondria involves remodeling of the structure to displace the LRS and form a mitochondrial binding surface formed by multiple MTD residues that are either surface exposed or buried at the LRS interface in the Vms1 crystal structure.

Vms1 binds mitochondria via a lipid species

To better understand how the stress-responsive translocation of Vms1 to mitochondria is regulated, we sought to identify the molecule(s) to which the Vms1^MTD binds on the mitochondrial outer membrane (MOM). We initially hypothesized that Vms1 mitochondrial binding is mediated by a protein and we screened for a reduction of Vms1^MTD localization in over 500 genetic mutants, each lacking a unique gene encoding a non-essential, nuclear-encoded, mitochondrial protein (Giaever et al., 2002). None of the mutant strains displayed a substantial reduction of Vms1^MTD localization to mitochondria (data not shown). Recognizing that Npl4 and other Cdc48 adaptor proteins bind ubiquitin to recruit Cdc48 to ubiquitylated substrates (Schuberth et al., 2004; Ye et al., 2003), we hypothesized that Vms1 might also bind (poly)ubiquitin at the mitochondrial surface. However, we observed that treating mitochondria with the Usp2 deubiquitylating enzyme had no effect on Vms1 interaction with mitochondria (data not shown).

To determine whether any mitochondrial surface protein is required for Vms1 binding, we treated purified mitochondria with proteinase K. After digestion, we observed efficient degradation of cytosol-exposed mitochondrial proteins (e.g. Tom22 and Fzo1), but very little cleavage of MOM-imbedded (e.g. Por1) and intra-mitochondrial proteins (e.g. Sdh1, Sdh2, and Mia40) suggesting that mitochondrial surface protein had been thoroughly digested without compromising mitochondrial integrity (Figure 3A). Surprisingly, we observed no significant difference in Vms1 affinity toward mitochondria after proteinase K treatment (Figure 3B-C). Although we cannot rule out the possibility that proteinase K resistant proteins or loops contribute to Vms1 binding, these results suggested that mitochondrial surface proteins are dispensable for Vms1 binding to mitochondria.

Because MOM protein appears to be dispensable for Vms1 binding, we hypothesized that Vms1 binds mitochondrial lipid. The majority of mitochondrial lipid is phospholipid, with much smaller quantities of sphingolipids and sterols principally comprising the remainder. Recent studies have described unique roles for both phospholipids (Chu et al., 2013;) and sphingolipids (Huang et al., 2012) in recruiting components of the mitophagy machinery to mitochondria. Due to their reactive phosphodiester linkage, phospholipids are susceptible to hydrolysis by mild base, whereas sphingolipids and sterols are not (Guan et al., 2010). We took advantage of this property to create liposomes containing total mitochondrial lipids or liposomes containing a lipid extract that lacked phospholipids. Vms1 bound liposomes containing total mitochondrial lipids, and this binding was enhanced with lipid extract that lacked phospholipids (Figure 3D). The increased level of binding seen upon hydrolysis and removal of the more abundant phospholipids may result from increased local concentration and liposome-incorporation of sterols/sphingolipids. These observations indicate that Vms1 binds a sterol or sphingolipid species.

To identify which specific lipid was responsible for Vms1 binding to mitochondria, we assayed for Vms1 binding to HPLC and TLC-fractionated, alkaline-resistant mitochondrial lipids. Sequential, orthogonal fractionation enabled the isolation of two lipid mixtures with highly enriched Vms1 binding activity (Figures 3E-G). We used TLC to compare lipids found in the two purified fractions that displayed the strongest binding activity towards Vms1, and identified a single lipid species that was specifically enriched in both fractions (Figure 3H). In parallel, analysis of these binding and adjacent non-binding fractions by LC/MS identified a single specific candidate species with a mass corresponding to the molecular formula, C₂₈H₄₄O₃.

Ergosterol peroxide is necessary and sufficient for Vms1 localization to mitochondria

The molecular formula, C₂₈H₄₄O₃, corresponds with hundreds of lipid species. To determine the molecular structure of the Vms1-binding lipid, we used silica gel column chromatography to purify a large quantity of the lipid, which we subjected to multiple structural analyses. As expected, the apparent binding activity towards Vms1 increased as C₂₈H₄₄O₃ was purified (Figures 4A-B), consistent with C₂₈H₄₄O₃ mediating Vms1 binding. We obtained a 1-H NMR spectrum of the isolated ligand (Figure 4C) and compared the peaks with published NMR spectra (Nowak et al., 2016). From this comparison, we determined that the peaks in the NMR spectra of our purified ligand were identical to those of ergosterol peroxide (EP), which has the same C₂₈H₄₄O₃ formula. Furthermore, our purified ligand and commercial EP possess identical retention factors by TLC (Figure 4D), identical retention times by LC/MS (Figure 4E), and equivalent MS/MS spectra at multiple collision energies (Figure 4F and Figure S2A-C). In combination, these data indicate that the C₂₈H₄₄O₃ lipid species that we isolated on the basis of its direct Vms1 binding activity is ergosterol peroxide.

EP is an oxidized sterol, structurally differing from ergosterol only by the presence of an endoperoxide. We found that Vms1 displayed comparable binding to nitrocellulose-spotted commercial EP and the purified C₂₈H₄₄O₃, while displaying no affinity towards ergosterol (Figure 4G). EP is generated when highly reactive singlet oxygen and the conjugated 5,7-diene of ergosterol spontaneously cyclize, in a [4+2] Diels-Alder type cycloaddition, to generate the 5,8-endoperoxide (Figure 4H). Ergosterol is found in many membranes throughout the cell. However, we found that EP levels were specifically enriched at mitochondria relative to other cellular membranes (Figures 4I, 4J, and S4D), which is consistent with the hypothesis that EP recruits Vms1 selectively to mitochondria.

Because EP is non-enzymatically produced from ergosterol and reactive oxygen species (ROS), we tested the necessity of EP for mitochondrial localization of Vms1 by eliminating these precursors. The importance of ROS was demonstrated by observing a significant reduction in Vms1^MTD localization to mitochondria in cells grown in anoxic conditions compared with cells grown in normoxia (Figures 5A-B), with no change in Vms1^MTD-GFP abundance (Figure S3A). Production of ergosterol, the other essential EP precursor, occurs through the mevalonate pathway, which is inhibited by the statin class of drugs. Statins (e.g. mevastatin) competitively inhibit HMG CoA reductase, the rate-controlling enzyme of the mevalonate pathway. We observed a significant reduction in Vms1^MTD localization to mitochondria upon mevastatin treatment (Figures 5C-D), with no effect on the steady state abundance of Vms1^MTD-GFP (Figure S3B). Thus, the EP precursors, ergosterol and oxygen/ROS, are both necessary for normal Vms1^MTD localization to mitochondria.

Although sterol synthesis is essential for cell survival, some steps in the ergosterol biosynthetic pathway are not essential in S. cerevisiae. Loss of those steps leads to the synthesis of ergosterol-like species that are capable of fulfilling the essential functions of ergosterol (Bocking et al., 2000), but may be differentially capable of mediating Vms1 mitochondrial localization. To test this possibility, we measured Vms1^MTD-GFP localization to mitochondria (Figures 5E-F) and mitochondrial ergosterol and EP levels (Figures 5G-H) in strains lacking non-essential ergosterol biosynthetic genes. The specific “ergosterol” or “EP” species in these mutant strains are likely unique to each strain and are therefore indicated as ergosterol* and EP*. Vms1^MTD localization to mitochondria was most affected in erg2Δ, erg3Δ, erg5Δ, and erg24Δ mutants. Erg2 and Erg3 catalyze the formation of the 7,8 and 5,6-alkenes, respectively (Figures 5I), which comprise the 5,7-diene required for EP formation (Veen and Lang, 2005), while the desaturase Erg5 and reductase Erg24 catalyze reactions near the 5,7-diene (Figures 5I) (Veen and Lang, 2005). In contrast, Vms1^MTD localization and mitochondrial EP* levels were only mildly affected in the absence of the reductase Erg4, methyltransferase Erg6, and glucosyltransferase Atg26. None of these mutants had a significant effect on Vms1^MTD-GFP abundance (Figure S3C). These data therefore indicate that Vms1^MTD localization to mitochondria is dependent on ergosterol biosynthesis, and particularly on the availability of the 5,7-diene that becomes oxidized to form EP.

To determine if EP is sufficient to promote Vms1 binding to lipid membranes, we tested the impact on Vms1 binding of introducing EP or ergosterol into liposomes prepared from mitochondrial lipids. We used mitochondrial lipids obtained from erg2Δ cells because this strain exhibited decreased Vms1^MTD localization to mitochondria and reduced levels of EP without grossly affecting the abundance of ergosterol-like sterols (Figures 5G-H). Exogenous addition of EP to erg2Δ mitochondrial lipids restored Vms1 binding, while the addition of ergosterol had no effect (Figures 5J-K). Taken together, these observations indicate that EP is necessary and sufficient for Vms1 binding to mitochondria and mitochondria-derived lipids.

Stress stimulates mitochondrial ergosterol peroxide abundance

Having determined that EP binds Vms1 and is necessary for Vms1 localization to mitochondria, we next sought to determine if EP serves as a stress signal to recruit Vms1 specifically to damaged mitochondria. This idea is supported by our previous observation that Vms1 preferentially localizes to mitochondria in vivo that have been selectively damaged by Killer Red-generated ROS (Heo et al., 2013). Because mitochondria in the same cell that were not damaged by laser-activation of the Killer Red did not show localization of Vms1, we concluded that the stress signal that promotes Vms1 translocation is confined to damaged mitochondria. To further test this hypothesis, we evaluated the in vitro interaction between Vms1 and mitochondria independently isolated from both stressed and unstressed cells. As predicted from the model that stressed mitochondria display enhanced Vms1 binding, we observed that mitochondria isolated from cells stressed with either rapamycin, which among other things induces endogenous ROS, (Kissova et al., 2006) or H₂O₂, have an increased affinity towards Vms1 in vitro compared with mitochondria from unstressed cells. Conversely, Vms1 purified from stressed (rapamycin or H₂O₂) and unstressed cells exhibited identical mitochondrial binding in vitro (Figures 6A-B). This suggests that the stress-induced changes that promote Vms1 localization are localized to mitochondria.

We further demonstrated that the stress-induced change that recruits Vms1 is displayed within the mitochondrial lipid fraction by using an in vitro liposome-floatation binding assay. Consistent with previous results, we observed that Vms1 exhibits enhanced affinity towards liposomes prepared from mitochondrial lipids isolated from stressed cells (Figures 6C-D). The simplest explanation of these data is that elevated ROS causes elevated EP abundance, which recruits Vms1 to mitochondria. This model was further supported by our finding of elevated EP levels in mitochondrial lipids isolated from cells that had been treated with either rapamycin or H₂O₂ (Figures 6E-F).

Vms1^LRS and EP-containing membranes compete for Vms1^MTD binding

To test the hypothesis that the Vms1^LRS and EP-containing (EP⁺) membranes compete for binding with the Vms1^MTD, we used the LRS mutant, “L4A”, in which four of the LRS buried leucines are changed to alanine, that we previously showed disrupted the Vms1^LRS/MTD interaction and promoted constitutive localization to mitochondria (Heo et al., 2013) (Figure 7A). In floatation assays, Vms1^L4A bound EP⁺ liposomes substantially more strongly than Vms1^WT, particularly as the amount of liposomes in the assay was decreased (Figures 7B-C), thereby validating that disruption of the Vms1^LRS/MTD interaction enhances Vms1 binding to EP⁺ liposomes.

Given that disrupting the Vms1^LRS/MTD interaction promoted EP⁺ membrane binding, we postulated that restricting Vms1^LRS/MTD dissociation would reduce membrane binding. To test this idea, we engineered a Vms1 protein in which the Vms1^LRS and Vms1^MTD can be tethered via chemical crosslinking and are thereby stabilized against dissociation. To crosslink Vms1 with the sulfhydryl-specific bismaleimidohexane (BMH), we introduced a single Cys in the Vms1^LRS and a single Cys in the Vms1^MTD (Vms1^Cys-Cys-HA) such that the two Cys residues would be in close proximity (Figure 7D). C387, the only reactive Cys residue in the Vms1^LRS/MTD construct, was mutated to Ala in the Vms1^Cys-Cys constructs. All Vms1^Cys-Cys mutants were fully functional in vivo (data not shown). To monitor crosslinking, we introduced a PreScission protease site in a loop N-terminal to the Vms1^MTD, such that migration on SDS-PAGE could distinguish the crosslinked species (Figure 7E). Importantly, protease cleavage of purified Vms1 did not alter migration on gel filtration chromatography, which indicates that the two cleaved Vms1 fragments maintained a stable association under the solution conditions used and that the structure is not grossly distorted. Cleaved Vms1^WT or Vms1^Cys-Cys was incubated with DMSO or BMH crosslinker and analyzed by SDS-PAGE. Crosslinking was incomplete, but an identifiable crosslinked species appeared only in the Vms1^Cys-Cys protein subjected to BMH (Figure 7E). Incomplete crosslinking enabled us to compare crosslinked and non-crosslinked protein within a single floatation assay. Three separate Vms1^Cys-Cys proteins were probed for their ability to bind EP⁺ liposomes and all three were found to exhibit a reduced ratio of crosslinked:non-crosslinked Vms1 in liposome-bound fractions compared to input (Figure 7E-F). This indicates that crosslinked Vms1^Cys-Cys, in which the LRS-MTD interaction is stabilized, binds EP⁺ membranes less efficiently than a cleaved version of the same protein in which the LRS-MTD interaction is not stabilized. These data suggest that dissociation of the Vms1^LRS/MTD interaction is required for efficient Vms1 binding to EP⁺ membranes.

Next, we wanted to determine the fate of the LRS when the MTD binds an EP⁺ lipid membrane. Based on our previous results, we predicted that the LRS would dissociate from the MTD upon lipid binding. To test this model we designed a GFP-Vms1-HA construct that contained a PreScission protease site immediately C-terminal to the LRS (Figure 7G) and followed the binding of the GFP-LRS and MTD-HA in a floatation assay. Cleavage of this Vms1 construct with PreScission protease did not alter migration on gel filtration chromatography (Figure 7H-I), indicating that the Vms1^LRS/MTD interaction remained intact as observed previously. We quantified binding of the full-length construct (GFP-Vms1-HA) and the two halves of the cleaved construct (GFP-LRS/MTD-HA) to EP⁺ liposomes in a floatation assay (Figure 7J). The difference in binding affinity was quantified as the ratio of GFP-Vms1:GFP-LRS or Vms1-HA:MTD-HA in the bound and unbound fractions (Figure 7K). GFP-Vms1 bound more strongly than GFP-LRS, while MTD-HA bound more strongly than Vms1-HA, thereby indicating that LRS association diminishes MTD membrane binding activity. Taken together, these three observations strongly support a model wherein the LRS and EP⁺ lipid membranes compete for binding to the MTD.

Finally, we wanted to determine the effect of mutating the MTD on EP⁺ liposome binding. Based on our model, we predicted that mutations that disrupt the LRS/MTD interface would promote membrane interaction, while mutants that maintain the LRS/MTD interaction but reduce mitochondrial localization will reduce membrane interaction. To test this model, we purified the mutants described in Figure 2 in the context of His₁₂-Vms1^1-417, and quantified binding to EP⁺ liposomes (Figure S4). K194D/K196D, a mutant with reduced mitochondrial localization that maintains LRS interaction, had significantly reduced binding to liposomes (Figure S4). The other mutants maintained or increased membrane interaction, which supports the model that disrupting the LRS-MTD interaction increases EP⁺ membrane binding in vitro, regardless of how the mutant behaves in vivo.

Yeast strains, media, and growth conditions

Saccharomyces cerevisiae strain JRY2509 (MATa his3 leu2 ura3 met15) was used as the WT strain. Deletion mutant strains were taken from the Saccharomyces Genome Deletion Project (Giaever et al., 2002). Yeast were transformed by the lithium acetate method, and grown at 30°C in SD med ium (0.67% yeast nitrogen base, 2% glucose) with amino acids unless otherwise indicated (Sherman, 1991). To test the effect of stress on Vms1-mitochondria co-migration, WT cells transformed with empty vector were grown to mid-log phase in SD-Ura medium and treated with one of the following prior to harvesting: 200 ng/mL rapamycin (LC Laboratory #r-5000) for 2.5 hours or 3 mM H₂O₂ (Millipore #386790) for 90 minutes.

Plasmids

A plasmid (pJR132011A) expressing C-terminally GFP-tagged Vms1^MTD was generated by PCR amplifying the VMS1 coding region from nucleotide 544 to 1251 and the VMS1 promoter from 335 nucleotides upstream to the VMS1 start codon from genomic DNA and ligating them in frame into a pRS416-based vector containing a GFP tag (pJR1415A). Mitochondria were identified with a plasmid (pJR9509B) encoding RFP targeted to mitochondria as described previously (Heo et al., 2010). Details of expression constructs are given in key resources table. All mutations and deletions were introduced via sewing PCR.

Protein Expression and Purification

The Vms1^{1-417, Δ38-69} construct was expressed in BL21(DE3) codon+(RIL) E. coli cells (Stratagene #230245) in autoinduction media ZYP-5052 (Studier, 2005). Cultures were grown in baffled flasks at 37°C to an optical density 600 (OD600) of ~0.5 and transferred to 19°C for ~18 hours. Cultures were harvested via centrifugation and pellets were frozen at −80 °C. Pellets were resuspended in ice cold lysis buffer (20 mM Tris pH 8.0, 300 mM NaCl, 5% glycerol, 1 mM DTT) supplemented with protease inhibitors (aprotinin, leupeptin, pepstatin A, and PMSF) (Sigma). Resupended cells were treated with 100 ug/ml lysozyme (Gold Biotechnology #L-040-100) and DNAse (Gold Biotechnology #9003-98-9) for 30 minutes and then subjected to sonication. Sonicated lysates were clarified via centrifugation.

Recombinant Vms1 was purified via three chromatographic steps at 4°C. Purification was monitored by SDS-PAGE. Clarified lysate was incubated with Glutathione sepharose resin (EMD Biosciences #70541-4) for ~3 hours, washed with 3 × 10 column volumes (CV) of lysis buffer, and incubated with 10 CV of low salt lysis buffer (100 mM NaCl) containing TEV protease (0.005 mg/ml) overnight. Protein cleaved off of the GST resin was loaded on a Hi-Trap Q HP column (GE Healthcare #17-1154-01), washed with 10 cv of low salt lysis buffer, and eluted with a 0-500mM NaCl gradient over 15 cv. Q fractions containing Vms1 were concentrated and run on a Superdex 200 (GE Healthcare) gel filtration column equilibrated in low salt lysis buffer. For purification of the selenomethionine-substituted protein, 3 mM DTT was used instead of 1 mM DTT.

For the His₁₀-Vms1^Cys-Cys proteins, constructs were expressed and lysed as described above, except that DTT was not included in the lysis buffer. Clarified lysates were added to Ni-NTA resin (Qiagen #30250) for 1 hour, washed with 10 CV of lysis buffer, 10 cv of lysis buffer with 40 mM imidazole, and eluted with lysis buffer with 250 mM imidazole. Eluted protein was dialyzed into lysis buffer to remove the imidazole, concentrated, and run on a Superdex 200 as described above.

For the His₁₂ tagged proteins, constructs were transformed into JRY1734 (pep4::HIS3 prb1::LEU2 bar1::HISG lys2::GAL1/10-GAL4) and grown in synthetic media lacking Uracil with 3% glycerol and 2% ethanol. When the OD 600 reached ~0.5, 0.5% galactose was added to the cultures and incubated for 6 hours. Cells were harvested, washed with sterile H2O, and flash frozen. Cells were lysed using a pulverizer (SPEX SamplePrep 6870) and the lysed powder was thoroughly resuspended in the above lysis buffer. The resuspended lysate was clarified via centrifugation, and purified as described for Vms1^Cys-Cys.

Mitochondria purification

Cells were exposed to the indicated conditions and harvested in mid-log phase. Mitochondria were then purified by sucrose gradient ultracentrifugation as previously described (Hao et al., 2009).

Lipid isolation and alkaline hydrolysis

Lipids were isolated from mitochondria, post-mitochondrial supernatant (PMS), or whole cell extract (WCE) using a modified Folch extraction. Pelleted mitochondria (1-5mg as measured by protein concentration) were resuspended in 500 μL of PBS (Gibco). 500 μL of methanol (Sigma) and 1 mL of chloroform (Fischer Scientific) were added to 500 μL of WCE, PMS, or resuspended mitochondria and vortexed at high speed for 10 minutes. The resulting mixture was centrifuged for two minutes at 1000×g. The lower organic phase was transferred to a new tube and the remaining aqueous phase was extracted with chloroform two additional times. The pooled organic extracts were washed two times with water, and the washed organic extract evaporated by SpeedVac (Savant) to yield a lipid film. This lipid film was analyzed as a total lipid fraction or subjected to NaOH hydrolysis as previously described (Guan et al., 2010) to create an alkaline-resistant lipid fraction enriched for sterols, and sphingolipids.

Proteinase K treatment

Purified mitochondria were resuspended in SEM buffer (250 mM sucrose, 1 mM EDTA, 10 mM MOPS pH 7.2) to a concentration of 5 mg/mL. 1 μL of Proteinase K (New England BioLabs #P8107S) was added for every 50 μg of mitochondria, the mixture was quickly mixed, and then incubated on ice for 30 minutes. 1 μL of 25 mg/mL PMSF (in isopropanol) was added for every 50 μg of mitochondria, and the resulting mixture was centrifuged at 3000×g for 5 minutes after which the supernatant was discarded. The pellet was washed with 250 μL SEM buffer, centrifuged at 3000×g for 5 minutes, and the supernatant discarded. To test the efficiency of protein cleavage at the MOM, untreated and proteinase K-treated mitochondria were subjected to SDS-PAGE and Western blot for Fzo1 (Dr. Janet Shaw), Tom22 (Dr. Kostas Tokatlidis), Por1 (Abcam #110326), Mia40 (Dr. Kostas Tokatlidis), Sdh1 (21^st Century Biochemicals #Pr1852a), and Sdh2 (21^st Century Biochemicals #Pr1633).

Mitochondrial co-migration assay

100μg of purified mitochondria were incubated with 2.5 μg of purified recombinant His₁₀-Vms1^1-417-HA in 300 μL of SEM buffer for 1 hour at 4°C. 30 μL was taken as an input control and the remaining mixture was loaded on a step-sucrose gradient (60%, 32%, 23%, 15% sucrose in MOPS buffer) and mitochondria-bound Vms1 was separated from unbound Vms1 by high speed ultracentrifugation as previously described (Meisinger et al., 2000). After centrifugation, fractions were collected, TCA precipitated, and subjected to SDS-PAGE and Western blot with α-His₆ (Clontech #631212) and α-Por1 (Abcam #110326) antibodies.

LC/MS analysis

Samples were resuspended in 25 μL of isopropyl alcohol (IPA):water:acetonitrile (ACN) (72:16:12) in 10 mM ammonium formate and 0.1% formic acid, centrifuged for 5 min at 15,000 g, and the supernatant transferred to a LC/MS vial with insert. Lipid extracts were separated on an Acquity UPLC Charged Surface Hybrid C18 1.7 μm 2.1 × 100 mm column maintained at 60°C connected to an Agilent HiP 1290 Sampler, Agilent 1290 Infinity pump, equipped with an Agilent 1290 Flex Cube and Agilent 6520 Accurate Mass Quadrupole time-of-flight dual electrospray ionization mass spectrometer. The source gas temperature was set to 350 °C, with a gas flow of 11.1 L/min and a nebuliz er pressure of 24 psig. Capillary voltage was 5000 V, fragmentor voltage 250 V, skimmer voltage 74.4 V, and Octopole RF peak voltage 750 V. Reference masses (m/z 121.0509 and 922.0098) were infused with nebulizer pressure at 2 psig and acquired with the scan range between m/z 100 – 1700 in positive mode. Mobile phase A comprised ACN:water (60:40 v/v) in 10 mM ammonium formate and 0.1% formic acid, and mobile phase B comprised IPA:water (90:10 v/v) in 10 mM ammonium formate and 0.1% formic acid. The chromatography gradient started at 15% mobile phase B, increased to 30% B over 4 min, increased to 52% B from 4-5 min, increased to 82% B from 5-22 min, increased to 95% B from 22-23 min, and then increased to 99% B from 23-27 min. From 27-38 min it was held at 99%B, then decreased to 15% B from 38-38.2 min, and was held at that level from 38.2-44 min. The flow rate was 0.3 mL/min throughout. The injection volume was 20 μL for fractionation experiments and 3μL for analytical experiments. For fractionation experiments, we used an in-line splitter, to divert post-column flow at a 10:1 ratio to collect 1 min fractions (0.3 mL per fraction), with the 1 part used for mass spectrometry analysis and the other 10 parts for liposome formation. Tandem mass spectrometry was conducted with the same LC gradient and source conditions, but without fraction collection, in the negative mode using collision energies of 10 eV, 20 eV and 40 eV.

Thin-layer chromatography (TLC) separation, staining, and quantification

Lipids were dissolved in 1:1 chloroform:methanol and spotted on to TLC plates (EMD Millipore #1.05554.0001). The TLC was resolved in a 40:1 chloroform:methanol solvent system. Lipids were extracted from TLC plates by vortexing the strip of TLC plate in a 1:1 chloroform:methanol mixture for 10 minutes. The TLC strip was removed and the mixture was centrifuged for two minutes at high speed to pellet the loose silica. The supernatant was transferred to a fresh tube and the resulting organic solution was evaporated by SpeedVac (Savant) to produce a lipid film. TLC plates were stained with an orcinol solution. 200 mg of orcinol (Sigma) were dissolved in 11.4 mL of H₂SO₄ and brought up to 100 mL with water. Resolved TLC plates were submerged briefly in the orcinol solution, left to air dry for five minutes, and then heated on a hot plate at ~95°C until bands appeared. TLC has been repeatedly used to quantify abundance (Khedr and Sheha, 2003). For EP quantification, we ran a dilution series of commercial EP and quantified each standard spot on the TLC using ImageJ. Using these values, we generated standard curves with R² values ≥ 0.97, suggesting that dilutions of the EP standards can be accurately quantified using TLC. We then quantified our unknown samples at a dilution where the EP band was in the linear range of our standard curve.

Silica chromatography

Starting at the base of the column, a classic preparative chromatography column (with solvent reservoir) was prepared with a glass wool plug, washed sand (Fisher Scientific), silica gel (VWR International #SX0143U-1) at a ratio of 50:1 silica mass:sample mass that had been suspended in hexane (Sigma), and another layer of washed sand on top. The hexane was drained from the column and the lipid mixture, resuspended in hexane, was loaded slowly on to the column and allowed to migrate through the sand into the silica gel layer. Lipids were eluted with the following solvent system: 1 column volume hexane (Sigma), 2 column volumes dichloromethane (Fisher Scientific), 2 column volumes of 1:1 dichloromethane:chloroform (Fischer Scientific), and then chloroform until the desired C₂₈H₄₄O₃ had completely eluted as monitored by TLC.

NMR

1-H NMR spectra were acquired with CDCl₃ (deutered chloroform) as the solvent on a Varian Inova 500 MHz NMR spectrometer in the University of Utah NMR Core Facility.

Lipid blot

1 μL of a lipid solution, containing lipids purified from yeast, EP (Chemfaces #CFN98035), or ergosterol (Fischer Scientific #AC11781), in 100% chloroform was spotted on to nitrocellulose membrane (Fischer Scientific #10600002). The membrane was blocked with 3% fatty-acid free BSA (Sigma #A6003) in TBS (25 mM Tris, 150 mM NaCl, pH 7.5) for one hour at room temperature. The blocking solution was replaced with a 1 μg/mL solution of His₁₂-Vms1^1-417-HA in TBS with 3% fatty-acid free BSA and incubated overnight at 4°C. The membrane was then subjected to standard Western blot techniques with 1° and 2° antibody incubation in TBS with 3% fatty-acid free BSA to visualize protein-lipid binding.

Liposome preparation

A control lipid solution was prepared by mixing 0.82 molar equivalents of DOPC (Avanti #850375P) and 0.18 molar equivalents of cholesterol (Avanti #700000P) in a glass vial. Additional lipids were added to the control liposome solution where indicated. Solvent was evaporated by gentle vortexing under a steady stream of argon gas to make a lipid film around the walls of the vial. These films were dried under vacuum for one hour at −52°C. The lipid film was then resolubilized in hexane (Sigma) and subsequently evaporated under a gentle stream of argon while vortexing, followed by a second round of vacuum-drying for three to four hours at −52°C. Lipid films were then hydrated with TBS to produce a 1mg lipid/mL buffer mixture. This mixture was rotated overnight at 4°C and extruded through 1.0 μM membranes (Fischer Scientific #05-71-5120) to produce a semi-homogenous liposome population. Aliquots were stored at −80°C. Liposomes prepared from total mitochondrial lipids were prepared without the addition of DOPC and cholesterol. Total mitochondrial lipids isolated from the indicated strains and additional lipids, if indicated, were mixed and dissolved in chloroform. Liposomes were prepared from this lipid solution as described above.

Liposome floatation assay

100 μL of liposomes were incubated with 2.5 μg of purified Vms1 protein at 4°C for one hour. 10 μL of the mixture was taken as an input control and mixed 1:1 with 4× laemmeli buffer (40% glycerol, 250 mM Tris-HCl pH 6.8, 8%SDS, 0.04% bromophenol blue, 0.1M DTT). 300 μL of 2M sucrose in TBS was added to the liposome-protein mixture and mixed well. A step-sucrose gradient was then created by gently loading 300 μL of 1M sucrose in TBS on top of the liposome mixture, followed by 300 μL of 0.5 M sucrose in TBS, and finally 75 μL of TBS. This was spun for 30 minutes in a Beckman Optima-Max table-top ultracentrifuge at 55,000 rpm allowing the liposomes to float to the top of the gradient, thereby separating unbound protein from liposome-bound protein (Koirala et al., 2013). 150 μL was taken from the top (bound Vms1) and bottom (unbound Vms1) of the sucrose gradient and analyzed by SDS-PAGE and Western blot with α-His₆ antibody (Clontech #631212).

Microscopy

The WT (JRY2509) or mutant strains were transformed with a plasmid expressing Vms1^MTD-GFP (or point mutants) under the native VMS1 promoter and a plasmid expressing mitochondria-targeted RFP. The cells were grown to mid-log phase and then imaged using a Zeiss Axioplan 2 Imaging microscope (Carl Zeiss). To test the effect of statin-treatment on Vms1^MTD localization, WT cells transformed with a plasmid expressing Vms1^MTD-GFP under the native VMS1 promoter and a plasmid expressing mitochondria-targeted RFP were inoculated, back-diluted, and grown in media containing vehicle or 500 μM mevastatin (Santa Cruz Biotech #SC-200853A) for 24 hours prior to imaging. To test the dependence of Vms1 lipid receptor production on the presence of oxygen, WT (JRY2509) cells were transformed with a plasmid expressing Vms1^MTD-GFP under the native VMS1 promoter and a plasmid expressing mitochondria-targeted RFP. After reaching mid-log phase, these cells were grown for 6 hours either in anoxia or normoxia with supplemented ergosterol (20 μg/mL) and fatty acids (5 mg/mL Tween 80) (Reiner et al., 2006). After the 6-hour treatment cells were imaged as described above.

Quantitation methods

All Western blots were quantitated using ImageJ 1.50i software. Microscopy was quantified as the mean GFP intensity (Vms1^MTD-GFP) in pixels containing RFP signal (mitochondria) using Cell Profiler (Carpenter et al., 2006). All statistics are derived from the comparison of the mean ± standard error of the mean in each experimental condition to the WT or control sample utilizing an unpaired t-test, unless specifically indicated otherwise. * p≤0.05, ** p≤0.01, ***p≤0.001.

Site-specific crosslinking

Purified Vms1^Cys-Cys proteins were incubated with PreScission protease (0.01 mg/ml) overnight and run on SDS-PAGE to monitor cleavage. Cleaved protein was desalted into 20 mM Tris pH 8.0, 300 mM NaCl, 5% glycerol to remove any reducing agent. 1 ul of either DMSO or 2 mM bismaleimidohexane (BMH, ThermoFisher #22330) were added 100 ul of 0.5 mg/ml protein and incubated on ice for 2 hours. Crosslinked was quenched by adding 3 mM DTT to the reaction.

Crystallization and Structure Determination

Gel filtration fractions containing Vms1 were concentrated to ~10 mg/ml for crystallization trials. All crystals were grown in sitting drops of 1:1 protein:reservoir at 13°C. Vms1 was initially crystallized against a reservoir of 20% PEG MME 2000, 100 mM Tris pH 8.5, and 0.2 M TMAO (Index screen F2, Hampton Research #HR2-144). Optimized crystals were obtained by screening around the initial reservoir conditions (22-24% PEG was ideal) combined with streak seeding after ~1 day.

Crystals were briefly immersed in reservoir solution with 20% glycerol and plunged into liquid nitrogen prior to data collection. Diffraction data were collected on SSRL beamline 11-1 at a wavelength of 0.97925 Å (selenomethionine inflection) with a Pilatus 6M detector, and were processed with HKL2000 (Otwinowski and Minor, 1997). The structure was determined by the SAD method and refined using PHENIX (Adams et al., 2010). The model was built using COOT (Emsley et al., 2010). Figures of molecular structures were made in Chimera (Pettersen et al., 2004). SBGrid was used throughout the structure determination (Morin et al., 2013). See table 1 for crystallography statistics.

Co-Immunoprecipitations

p^VMS1-VMS1^1-182-HIS₆-HA was co-expressed with p^VMS1-Vms1^182-417-GFP in JRY764 (vms1Δ). ~50 OD were harvested and resuspended in lP buffer (20 mM Tris pH 7.4, 50 mM NaCl, 0.2% TritonX-100), vortexed, clarified via centrifugation, and added to anti-HA magnetic beads (Thermo scientific #88836). After 4 hours of incubation, beads were pelleted at 2K G and washed 4× with 1ml of IP buffer. Proteins were eluted with 50 ul of 2× laemmli buffer (20% glycerol, 125 mM Tris-HCl pH 6.8, 4% SDS, 0.02% bromophenol blue)