Myristoylated methionine sulfoxide reductase A is a late endosomal protein

Methionine residues in proteins provide antioxidant defense by reacting with oxidizing species, which oxidize methionine to methionine sulfoxide. Reduction of the sulfoxide back to methionine is catalyzed by methionine sulfoxide reductases, essential for protection against oxidative stress. The nonmyristoylated form of methionine sulfoxide reductase A (MSRA) is present in mitochondria, whereas the myristoylated form has been previously reported to be cytosolic. Despite the importance of MSRA in antioxidant defense, its in vivo binding partners and substrates have not been identified. Starting with a protein array, and followed by immunoprecipitation experiments, colocalization studies, and subcellular fractionation, we identified the late endosomal protein, StAR-related lipid transfer domain–containing 3 (STARD3), as a binding partner of myristoylated MSRA, but not of nonmyristoylated MSRA. STARD3 is known to have both membrane-binding and cytosolic domains that are important in STARD3-mediated transport of cholesterol from the endoplasmic reticulum to the endosome. We found that the STARD3 cytosolic domain localizes MSRA to the late endosome. We propose that the previous conclusion that myristoylated MSRA is strictly a cytosolic protein is artifactual and likely due to vigorous overexpression of MSRA. We conclude that myristoylated MSRA is a late endosomal protein that may play a role in lipid metabolism or may protect endosomal proteins from oxidative damage.

Methionine residues in proteins provide antioxidant defense by reacting with oxidizing species, which oxidize methionine to methionine sulfoxide. Reduction of the sulfoxide back to methionine is catalyzed by methionine sulfoxide reductases, essential for protection against oxidative stress. The nonmyristoylated form of methionine sulfoxide reductase A (MSRA) is present in mitochondria, whereas the myristoylated form has been previously reported to be cytosolic. Despite the importance of MSRA in antioxidant defense, its in vivo binding partners and substrates have not been identified. Starting with a protein array, and followed by immunoprecipitation experiments, colocalization studies, and subcellular fractionation, we identified the late endosomal protein, StAR-related lipid transfer domaincontaining 3 (STARD3), as a binding partner of myristoylated MSRA, but not of nonmyristoylated MSRA. STARD3 is known to have both membrane-binding and cytosolic domains that are important in STARD3-mediated transport of cholesterol from the endoplasmic reticulum to the endosome. We found that the STARD3 cytosolic domain localizes MSRA to the late endosome. We propose that the previous conclusion that myristoylated MSRA is strictly a cytosolic protein is artifactual and likely due to vigorous overexpression of MSRA. We conclude that myristoylated MSRA is a late endosomal protein that may play a role in lipid metabolism or may protect endosomal proteins from oxidative damage.
Until relatively recently, the functions of methionine in proteins have been poorly defined, except of course, for its role in protein initiation. Biochemistry texts often treat it as a generic hydrophobic amino acid, but investigations from many laboratories have changed this view (1)(2)(3)(4)(5)(6)(7)(8). Methionine in proteins functions in antioxidant defense, protein structure, and redox sensing and regulation.
Methionine residues provide antioxidant defense through reversible oxidation and reduction. Methionine is rather readily oxidized to methionine sulfoxide (9, 10) by reactive oxygen and nitrogen species. Oxidation of methionine creates a chiral center at the sulfur so that the methionine sulfoxide produced is a mixture of the S and R epimers (11). Reduction of the sulfoxide back to methionine is catalyzed by the methionine sulfoxide reductases (MSR), 2 enzymes found in almost all organisms from microbes to humans (12). Recycling by the reductases allows the methionine residue to react again with oxidizing species, creating a system with catalytic efficiency in scavenging reactive species.
MSRA acts on the S epimer, and MSRB acts on the R epimer. The A class was described some years ago and has been characterized in considerable detail, especially by Weissbach, Brot, and colleagues (12). The B class of reductases, some of which are selenoproteins in higher animals, was discovered more recently and has also been studied intensively (13,14). Mammals have 3 isoforms of the B class, and one of the A class. There is only one gene for MSRA, but it has two initiation sites (15). Initiation at the first site generates a protein with a mitochondrial targeting sequence. The protein produced by initiation at the second site lacks the mitochondrial targeting sequence and is myristoylated. In monkey kidney and HEK293 cells transfected with the full-length gene, the myristoylated form has been localized to the cytoplasm (ϳ75%) and the nonmyristoylated form to mitochondria (ϳ25%) (15,16). The role of myristoylation is ill-defined, but the myristoylated form protects the heart against ischemia-reperfusion injury, whereas a cytosolic, nonmyristoylated form does not (17). Myristoylation increases the catalytic efficiency of MSRA in vitro by improving its interaction with substrates (18,19), perhaps by stabilizing an encounter complex (20).
The in vivo importance of the reductases is well-established, particularly for MSRA. Knocking out the enzyme caused increased susceptibility to oxidative stress in mice (21), yeast (22), and bacteria (23)(24)(25)(26). Conversely, overexpressing MSRA conferred increased resistance to oxidative stress in Drosophila (27), Saccharomyces (28), Arabidopsis (29), PC-12 cells (30), and human T cells (28). Helicobacter pylori, a causative agent of gastric ulcers and carcinoma, requires MSRA for protection against oxidative stress, and MSRA appears to act through reduction of methionine sulfoxide in the bacterial catalase (26,31). Despite the importance of the reductase in defense against oxidative stresses, the mechanisms of its action and even its in vivo binding partners and substrates are essentially unknown.
Work by several investigators has established the surprising ability of cells to increase the methionine content of their pro- teins when exposed to oxidative stress (7,32). The mechanism was elucidated by Kim and colleagues (8) who showed that ERK1/2 phosphorylates methionyl tRNA synthetase in cells experiencing oxidative stress. The phosphorylation renders the synthetase promiscuous, due to increased affinity for non-methionine tRNAs, thus increasing the methionine content of proteins during oxidative stress. Cyclic oxidation and reduction of methionine residues, like any reversible covalent modification, has the potential to function as a regulatory switch. This function has been established for the cytosolic MSRB1. The oxidoreductase MICAL mediates thestereospecificoxidationofMet-44inactin,leadingtodepolymerization of actin (6). The cytosolic MSRB1 reduces methionine sulfoxide 44 (MetO-44) back to Met, restoring its ability to polymerize (4,5). MSRA is a bifunctional enzyme, capable of stereospecifically oxidizing Met and reducing methionine sulfoxide residues, although no in vivo substrates have yet been identified (33).
Many proteins with methionine sulfoxide residues have been shown to be substrates of MSRA in vitro, although thus far no in vivo substrates nor binding partners have been confidently identified. We previously attempted to find interacting proteins in vivo using a variety of techniques, but none led to confirmed substrates. The methods used included immunoprecipitation, affinity purification, chemical cross-linking, yeast two-hybridization, HPLC-MS (34), and in vivo labeling of endogenous proteins with photoactivatable methionine and leucine (35). We now report that late endosomes are the primary subcellular localization site of myristoylated MSRA and that the previously reported cytosolic localization is an artifact of too robust overexpression in transfected cells. Myristoylated MSRA is anchored to the late endosome by interaction with STARD3, an interaction initially detected with a protein array of human proteins. STARD3 is anchored in the membrane of late endosomes and functions to transfer lipids between organelles, notably cholesterol from the endoplasmic reticulum into late endosomes (36 -38). The interaction of STARD3 and myristoylated MSRA clarifies the role of myristoylation of MSRA and broadens the possible in vivo functions of MSRA.

A protein array identifies STARD3 as a candidate binding partner of MSRA
We utilized a microprotein array to identify candidate binding partners of MSRA. The arrays are printed on nitrocellulose, are the same dimensions as cDNA arrays used for transcriptomics, and are scanned in the same instruments as cDNA arrays. The protein array we used has 9,483 human proteins, produced in baculovirus with glutathione S-transferase tags (39). Cys-74 3 is required for catalytic activity, and thus, C74S lacks both reductase and oxidase activities. We utilized it to facilitate capture of substrates that would bind and not be readily released. We incubated the arrays with biotinylated human myristoylated C74S MSRA and interrogated the array with streptavidin tagged with Alexa Fluor 647. Several candidate proteins emerged, although all but one had relatively low confidence scores ( Table 1). The one high confidence candidate was STARD3, a late lysosomal protein located in the lysosomal membrane with a cytosolic domain (40).

Myristoylated MSRA binds to STARD3 in vivo
To test whether MSRA actually interacts with STARD3 within the cell, we co-transfected HEK293 cells with FLAGtagged Stard3 and WT or C74S myristoylated MsrA. A homogenate was then immunoprecipitated with anti-FLAG antibody. Fig. 1A shows that MSRA was co-precipitated with STARD3. Interaction occurred with both the active site mutant and the WT MSRA. Although STARD3 can be detected in nontransfected HEK293 cells, MSRA cannot (15). By transfecting the HEK293 cells only with MSRA, we examined whether it would interact with STARD3 present at normal intracellular concentration, and we again found that both the active-site mutant and the WT enzyme interacted with STARD3 ( Fig. 1C and D).
We then tested whether myristoylation affects the interaction of MSRA and STARD3. Gly-24 is the myristoylated residue in MSRA, so that G24A MSRA is not myristoylated. The G24A MSRA did not interact with STARD3 ( Fig. 2A). STARD3 has a well-defined cholesterol-binding site (40) that might accommodate a myristoyl moiety. This raises the question of whether STARD3 interacts specifically with myristoylated MSRA or generally with myristoylated proteins. We investigated whether ARF1 (ADP-ribosylation factor 1) interacted with STARD3. ARF1 is a GTPase that is important in protein trafficking and in vesicle budding and uncoating within the Golgi (41). We chose ARF1 because it is myristoylated and its myristoylated-binding site shares similarities with that of MSRA (18). Coexpressing ARF1 and STARD3 in HEK293 cells revealed that STARD3 did not interact with either myristoylated nor nonmyristoylated ARF1 (Fig. 2C). We conclude that STARD3 does not bind nonspecifically to myristoylated proteins. Rather, it interacts specifically with myristoylated MSRA.

Methionine sulfoxide reductase A is an endosomal protein Myristoylated MSRA colocalizes with STARD3 at late endosomes
STARD3 is localized to late endosomes (36). An N-terminal domain (MENTAL) has 4 transmembrane helices that embed in the late endosomal membrane (42). MENTAL is followed by the START domain that is cytosolic and contains the lipidbinding domain. We expressed FLAG-tagged STARD3 in COS7 cells and confirmed that it colocalized with the late endosomal marker LAMP1 but not the early endosomal marker EEA1 (Fig. 3, A and B). Expression of myristoylated MSRA revealed that it colocalized with the late endosomal marker LAMP1 (Fig. 3C) and with STARD3 (Fig. 3D). The active site mutant also colocalized with STARD3 ( Fig. 3E). Myristoylated MSRA did not colocalize with the mitochondrial marker DsRed-Mito (Fig. 4). Because of the small size of endosomes, it was difficult to determine whether MSRA was in the interior of the endosome or on its membrane. Treatment of cells with ammonium chloride causes enlargement of endosomes, allowing demonstration that myristoylated MSRA localizes to the late endosomal membrane, whereas nonmyristoylated MSRA does not (Fig. 5).
We then performed subcellular fractionation of liver from WT and MSRA knockout mice, using a sucrose gradient. STARD3 and the late endosomal marker, RAB7, were prominent in the first fraction, as expected for late endosomes (Fig. 6). Myristoylated MSRA was also most prominent in the fraction, confirming the presence of both STARD3 and myristoylated MSRA in nontransgenic mouse liver.
We next investigated the effect of reducing the endogenous level of STARD3 in COS7 cells by treatment with an siRNA against STARD3. Fig. 7 shows that knocking down the level of STARD3 reduced the amount of myristoylated MSRA localized to the late endosome. This provides additional support for the conclusion that MSRA and STARD3 interact directly and not simply through mutual interaction with a third protein. Fig. 7D shows that both myristoylated MSRA and STARD3 levels decrease by about half in response to the siRNA. Fig. 7E presents the colocalization of the two proteins. Without siRNA, colocalization is quite high, with a Pearson's correlation coefficient of 0.8. With siRNA, there is a modest, but statistically significant, decrease in colocalization, with a correlation coefficient of 0.6. Although a change in the level of STARD3 might alter the expression of genetically encoded MSRA, one would not expect the synthesis of the transfected GFP-myristoylated MSRA to be affected by a change in STARD3, although it was ( Fig. 7). We suggest that the decreased level of the GFP-myristoylated MSRA is due to increased degradation of the fraction not bound to late endosomes.

STARD3 features affecting interaction with myristoylated MSRA
The crystal structure of the cytosolic START domain of STARD3 was solved by Tsujishita and Hurley (40) who showed that cholesterol binds in a hydrophobic tunnel formed by a U-shaped ␤-barrel and helix ␣4, which forms a roof over the barrel. They pointed out that Met-307 and Asn-311 are conserved, uncharged residues that influence binding specificity; the double mutant, M307R/N311D, could not bind cholesterol. The N-terminal domain that forms helix ␣1 does not participate in forming the lipid-binding site but could be the docking site for MSRA. Helix ␣4 forms a roof over the lipid-binding site and might regulate access to that site. We investigated whether these features affected the interaction of MSRA with STARD3. First, we compared the full-length STARD3 and its cytosolic

Methionine sulfoxide reductase A is an endosomal protein
START domain (residues 231-444 of STARD3) that was utilized in the crystal structural studies (40,44). We found no difference in their interaction with myristoylated MSRA, consistent with the immunohistochemical localization of binding to the cytosolic-facing membrane of late endosomes (Fig. 8A). All other studies were then performed with the cytosolic START domain. The mutant that could not bind lipid, M307R/ N311D, was also impaired in binding MSRA. Deletion of the N-terminal helix ␣1 greatly increased binding, whereas deletion of helix ␣4 modestly increased binding (Fig. 8, B and C). Thus, neither of those helices is required for MSRA binding. Met-307 is within the binding pocket for lipids, and a recent high-resolution crystal structure shows the sulfur of Met-307 is just 3.8 Å from Trp-310, forming a bond with the aromatic ring (3,45). We therefore also evaluated the effect of mutating just Met-307 and found that the interaction of STARD3 and MSRA was greatly reduced (Fig. 8, D and E).

Oxidized STARD3 is a substrate for MSRA
In addition to Met-307, STARD3 has one other methionine residue, Met-427. To determine whether methionine sulfoxide in STARD3 is a substrate for MSRA, we prepared oxidized STARD3 by treating it with hypochlorite to oxidize methionine to methionine sulfoxide. The oxidation conditions yielded proteins with a low fractional oxidation of the methionine residues, as might be expected occur in vivo. We then incubated the oxidatively modified protein with MSRA and quantitated the -tailed t test. C, myristoylated or nonmyristoylated ARF1, tagged with HA at its C terminus, was overexpressed with FLAG-tagged STARD3 in HEK293 cells. Cell lysates were immunoprecipitated with anti-FLAG antibody and detected by anti-HA antibody. The asterisk marks the heavy chain of IgG. This experiment was performed 3 times on separate days, none of which detected an interaction between ARF1 and STARD3. fraction of methionine sulfoxide at positions 307 and 427 that were reduced. methionine sulfoxide 307 was a substrate for MSRA (Fig. 9A), whereas the effect on methionine sulfoxide 427 did not reach statistical significance (Fig. 9B).

Reconciliation with earlier reports
As noted in the Introduction, Kim and Gladyshev (15,16) and we previously reported that the nonmitochondrial form of MSRA is cytosolic. That conclusion was based on studies of monkey kidney cells (CV-1 or COS) transfected with the MSRA gene. Neither paper reported the amount of DNA used for transfection, but a review of our laboratory notebooks found that it was 1 g. Given the results above that demonstrate a late endosomal localization, we wondered whether the amount of DNA used in the earlier studies may have been excessive, saturating the binding sites on late endosomes and leading to accumulation of the excess MSRA in the cytosol. We examined that possibility by transfecting COS7 cells with varying amounts of DNA (Fig. 10). With 0.125 or 0.25 g of DNA, MSRA colocalized to late endosomes. Although that colocalization was maintained with 0.50 and 0.75 g of DNA, the cytosol was now flooded with MSRA. We conclude that the earlier finding that myristoylated MSRA is cytosolic was artifactual. It is a late endosomal protein.

Discussion
The major finding reported here is that the nonmitochondrial, myristoylated form of mammalian MSRA binds to STARD3 and is thus located on the cytosolic face of late endosomes, although we do not rule out the possibility that some myristoylated MSRA is in the cytoplasm or other subcellular locations. The interaction between STARD3 and MSRA was demonstrated by 4 independent methods: 1) binding of MSRA to STARD3 on a protein array; 2) localization of GFP-tagged MSRA to late endosomes by immunofluorescence in intact cells; 3) coimmunoprecipitation of STARD3 and MSRA from cell lysates; and 4) subcellular fractionation of cell lysates.
MSRA is known to be myristoylated (15). Although other myristoylated proteins are cytosolic, in general they translocate to membranes under specific conditions (46). We previously considered the possibility that oxidative stress would trigger translocation to a membrane, but application of various oxidative stressors failed to alter its distribution (24). We noted that, "the role of myristoylation remains to be elucidated." The experimental data presented in this study establish that a role of myristoylation is to mediate the binding of MSRA to the cytosolic domain of STARD3.
STARD3 has been known for years to bind cholesterol as well as carotenoids and be a late endosomal protein (36,40,44). Although implicated in cholesterol and lipid metabolism, its detailed function was unclear until recently, when Alpy and colleagues (37) established that it mediates the transfer of cholesterol from the endoplasmic reticulum to the endosome. Why MSRA is bound to STARD3 is a matter of speculation at this point, although one can reasonably suggest that MSRA plays a role in lipid/cholesterol metabolism.
In apolipoprotein E-deficient mice, feeding a high-fat "Western diet" causes atherosclerosis and hepatic steatosis. Hepatic overexpression of MSRA in those mice reduced the plasma very low density lipoprotein/low density lipoprotein ratio and reduced the hepatic steatosis and aortic atherosclerosis (47). In humans, a G to A polymorphism in MSRA is associated with an increased risk of cardiovascular disease (48,49). In a genomewide association study, an A to G intron variant was associated with hypertension, a known risk factor for cardiovascular disease (50).
One possibility is that MSRA provides an oxidative defense for other components in the cholesterol-metabolic pathways. Stocker and colleagues (51,52) have shown that potentially toxic lipid hydroperoxides are reduced by oxidation of methionine in apolipoproteins to methionine sulfoxide, and MSRA can reduce the sulfoxide in apolipoprotein A-I (53). We showed that MSRA can reduce oxidatively modified Met-307 back to methionine. Met-307 is situated near the distal end of the lipidbinding site and likely binds to STARD3's lipid cargo (40). Met-307 in STARD3 may be particularly susceptible to methionine oxidation by lipid hydroperoxides. The anchoring of MSRA would greatly enhance the efficiency of its action on STARD3. Such a mechanism can explain why myristoylated MSRA protected the mouse heart from ischemia-reperfusion injury, whereas nonmyristoylated, cytosolic MSRA did not (15).

DNA constructs
Myristoylated and nonmyristoylated human MSRA (residues 23-235) was amplified by PCR and digested with restriction enzymes BamHI and NotI and then inserted into a pCR3.1 vector. Nonmyristoylable MSRA was produced by mutating Gly-24 to Ala by PCR. An inactive form of MSRA was produced by mutating Cys-74 to Ser with the QuikChange site-directed mutagenesis kit (Stratagene). Human MSRA with a C-terminal GFP tag, pCMV6-AC-GFP-MSRA was purchased from Origene (accession number NM_012331, catalogue number RG 208916) and amplified by PCR to generate constructs for myristoylated MSRA and nonmyristoylated MSRA (G24A). A human STARD3 cDNA clone (NM_006804) in a pCMV6-Myc-FLAG vector was purchased from OriGene (RC206802). To generate truncated forms of START domain, START was amplified by PCR and inserted into a pCMV-Myc-FLAG vector using SgfI and MluI restriction enzymes. The Met-307 and M307R/N311D mutants of the START domain were generated with the QuikChange kit. pDsRed2-Mito was purchased from Clontech (number 6975-1). mRFP-RAB7 was a gift from Ari

Methionine sulfoxide reductase A is an endosomal protein
Helenius (Addgene plasmid number 14436). Both the WT and G2A ARF1 plasmids, pXS-ARF1-HA, and Escherichia coli cells expressing N-myristoyltransferase were kindly provided by Dr. Julie G. Donaldson (National Institutes of Health, NHLBI).

Production and purification of recombinant human His-STARD3(216 -445)
The cytosolic form of STARD3(216 -445) with an N-terminal His tag and tobacco etch virus protease-cleavage site was amplified by PCR and digested with restriction enzymes NdeI and BamHI and then inserted into a pET17b vector. BL21 cells were transformed with this His-Stard3(216 -445) construct, grown at 37°C in 2-liter flasks, and induced with 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside. After a 4-h induction, cells were collected and lysed (50 mM Tris-Cl, pH 8.0, 5 mM imidazole, 100 mM NaCl, 0.1 mM EDTA, 1 mM PMSF with 1ϫ protease inhibitor mixture I).

Rabbit mAb specific for myristoylated MSRA
Epitomics (now Abcam) created rabbit hybridomas by immunization against recombinant mouse myristoylated MSRA. The myristoylated and nonmyristoylated mouse MSRA were used to select for a hybridoma with high specificity for the myristoylated form, assayed by Western blotting. Among 53 clones tested, only one was specific for the myristoylated form. The titer of IgG in the supernatant of this clone was very low, leading us to concentrate and purify the IgG. Culture medium from a 1-liter culture (150 mm dishes) was clarified by centrifugation. The supernatant was concentrated to 100 ml and then precipitated with 50% ammonium sulfate. The precipitate was purified by one of two methods. In one, it was dissolved in and dialyzed against 10 mM Tris-HCl, pH 7.5, and purified by protein G-agarose chromatography. In the second, it was dissolved in and dialyzed against 50 mM sodium phosphate, 100 mM NaCl, pH 7.2, and purified by gel filtration chromatography. The purified anti-myristoylated MSRA was stored in aliquots at Ϫ80°C until use. The specificity of the antibody is documented in Fig. 6B.  . Subcellular fractionation demonstrates that myristoylated MSRA is enriched in the late endosome fraction. A, myristoylated MSRA is most prominent in fraction 1 from wildtype (WT) mouse liver and is not present in fractions prepared from the MSRA knockout liver (KO). Fraction 1 also contains most of the intact late endosomes. Each fraction was evaluated by immunoblot with these marker proteins: RAB7 (late endosome), mtHSP70 (mitochondria), myristoylated MSRA, total MSRA, and STARD3. The asterisk (*) indicates a nonspecific band. The molecular size markers of protein are indicated in kDa. The specificity of the mAb for the myristoylated form of MSRA is shown in panel B. 100 ng of purified, recombinant human myristoylated or nonmyristoylated MSRA were subjected to SDS-PAGE followed by immunoblotting by either the anti-myristoylated MSRA antibody or the general anti-MSRA antibody. As indicated under "Experimental procedures," the concentration of protein was determined spectrophotometrically. However, to assure that equal amounts of the two proteins were loaded for immunoblotting, we also subjected 1 g of each protein SDS-PAGE followed by staining with Coomassie Brilliant Blue. C, the anti-myristoylated MSRA and general anti-MSRA antibodies are specific for MSRA. This was demonstrated by immunoblotting homogenates of liver and kidney from WT and MSRA knockout mice. The asterisk (*) marks nonspecific bands detected by the general anti-MSRA antibody. Tubulin served as a loading control.  KH 2 PO 4 , and 56 mM Na 2 HPO 4 )) with 5% glycerol, 0.05% Triton X-100, 1% BSA, 5 mM MgCl 2 , and 0.5 mM dithiothreitol (DTT) and incubated with streptavidin-conjugated Alexa 647 (Invit-rogen, number S32357) in the dark at room temperature for 1 h. Protein arrays (ProtoArray Human Protein Microarray version 5.0, Invitrogen) were incubated with blocking buffer (PBS with 0.1% Tween 20 and 1% BSA) for 1 h at 4 o C and then incubated with the biotin-strepavidin complex for 2 h at 4 o C. Arrays were then washed 3 times with washing buffer II (PBS with 0.05% Triton X-100, 1% BSA, 5 mM MgCl 2 , and 0.5 mM DTT). Arrays were scanned with a GenePix 4200AL and analyzed with Invitrogen Prospector software using the ORF browser database.

Cell culture and immunoprecipitation assays
Mammalian cells were cultured in Dulbecco's modified Eagle's medium (Invitrogen 11965092) including 10% fetal calf serum and 1% penicillin and streptomycin at 37 ºC in a humidified atmosphere of 5% CO 2 and 95% air. Cells were transfected using calcium phosphate (Invitrogen, number K278001) for HEK293 cells and Lipofectamine 3000 (Invitrogen, number L3000008) for COS7 and HeLa cells. To deplete STARD3,

Methionine sulfoxide reductase A is an endosomal protein
COS7 and HeLa cells were transfected with siRNA against STARD3 and nontargeting control for 48 h using Lipofectamine RNAi Max (Invitrogen, number 13778150). For immunoprecipitation assays, HEK293 cells were cotransfected with pCR3.1 human MsrA and pCMV Stard3-FLAG on 100-mm plates for 24 h and then lysed with 25 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% Nonidet P-40, 5% glycerol,1 mM PMSF, and 1ϫ protease inhibitor. Lysates were cleared by centrifugation for 20 min at 20,800 ϫ g at 4 o C and then incubated with 3 g of mouse anti-FLAG antibody (Origene TA50011) for 4 h at 4 o C. Protein-antibody complexes were precipitated with protein G beads for 2 h at 4 o C. Immunoprecipitated proteins were separated by electrophoresis on 10 -20% Tris glycine gels (Invitrogen, number XP10205) and transferred to a nitrocellulose membrane (Bio-Rad, number 1704158). Membranes were probed with the indicated antibodies and quantitatively detected by an Odyssey IR scanner (Li-Cor Biosciences).

Subcellular fractionation
To prepare a fraction enriched in late endosomes, subcellular fractionation was performed using a modification of a published procedure (56). 100 mg of liver from WT or MSRA knockout mice was washed with cold PBS. Tissues were gently minced and homogenized on ice with 1 ml of homogenization buffer (250 mM sucrose, 1 mM EDTA, 1 mM PMSF) using a Dounce homogenizer (Wheaton 357542, loose-fitting). After centrifugation at 1,000 ϫ g for 10 min, the supernatant was transferred to a new tube and made 25% in sucrose with Opti-Prep gradient media (Cosmo Bio USA) in a total volume of 1.5 ml. 2 ml of 45% Opti-Prep was placed in an ultracentrifuge tube for an SW41Ti rotor (Beckman Coulter, 14 ϫ 89 mm). Then 4 ml of 35% OptiPrep was layered on, followed by 3 ml of 25% Opti-Prep. Then the 1.5-ml sample was layered on the top. The tubes were centrifuged for 1 h at 100,000 ϫ g after which 1-ml fractions were collected and placed in microcentrifuge tubes. These were centrifuged at 18,000 ϫ g for 30 min. The pellet of each fraction was resuspended with 2ϫ sample loading buffer. Following electrophoresis and blotting, the membranes were probed with various antibodies. With this method, late endosomes will be enriched in fractions 1 and 2 (56).

Oxidative modification of STARD3
To generate a low fractional oxidation of the methionine residues in STARD3, 4 M His-STARD3(216 -244) was incubated with 500 M sodium hypochlorite (Sigma, number 239305) in 20 mM Tris, pH 7.4, with 150 mM sodium chloride for 5 min at 25°C. Residual hypochlorite was scavenged by the addition of 10 mM DTT after the 5-min incubation. The sample was then divided, with half receiving no additional treatment, whereas the other half was incubated with 4 M myristoylated MSRA for 30 min at 37°C. Incubation longer than 30 min did not increase reduction of the STARD3.
Alkylation and tryptic digestion was performed as described (57). HPLC-MS was performed on an Agilent 1200 series HPLC system equipped with an autosampler set to 4°C and a column compartment set to 30°C. Separations were performed on a Zorbax 300Å StableBond C18 MicroBore column (1.0 ϫ 50 mm, 3.5 m particle size, Agilent catalogue number 865630-902). The initial solvent was water, 0.05% TFA and elution was by a gradient with acetonitrile, 0.05% TFA developed at 1%/min with a flow rate of 20 l/min. The effluent from the column was mixed in a tee with 20 l/min just prior to the electrospray needle to displace the bound TFA and generate internal standards for the mass spectra (43,58). Electrospray MS was performed on an Agilent model 6520 accurate mass quadropole-TOF instrument. The effluent-positive electrospray ionization spectra were obtained in the mass range of 100 -2500 m/z. The drying gas temperature was 350°C with a flow rate of 10 liters/ min and a nebulizer pressure of 2 bar. The voltages were capillary 3500 V, fragmentor 235 V, skimmer 65 V, and octopole 1 750 V. MS/MS fragmentation used a collision energy of 30 with a data collection range of 20 -2000 m/z.
Mass spectra were analyzed using Agilent software, Mass-Hunter version B.05. Extracted ion chromatograms were generated for the methionine and methionine sulfoxide tryptic peptides containing the 2 methionine residues in STARD3, Met-307 and Met-427. We confirmed that the selected peaks were the desired peptides by MS-MS sequencing. The areas of the peaks in the extraction ion chromatogram were used to calculate the percentage of each peptide that was oxidized. The m/z values used to extract chromatograms were 790.4 and 806.4 for reduced and oxidized Met-327 (Z ϭ 1), and 765.6 and 771.0 for reduced and oxidized Met-427 (Z ϭ 3).

Immunofluorescence and confocal microscopy
Cells were seeded on poly-L-lysine-coated coverslips in a 24-well plate and then transfected with various constructs. Cells were fixed with 4% (w/v) paraformaldehyde for 10 min at room temperature and permeabilized with blocking buffer