The majority of ER‐targeted tail‐anchored (TA) proteins are inserted into membranes by the Guided Entry of Tail‐anchored protein (GET) system. Disruption of this system causes a subset of TA proteins to mislocalize to mitochondria. We show that the AAA+ ATPase Msp1 limits the accumulation of mislocalized TA proteins on mitochondria. Deletion of MSP1 causes the Pex15 and Gos1 TA proteins to accumulate on mitochondria when the GET system is impaired. Likely as a result of failing to extract mislocalized TA proteins, yeast with combined mutation of the MSP1 gene and the GET system exhibit strong synergistic growth defects and severe mitochondrial damage, including loss of mitochondrial DNA and protein and aberrant mitochondrial morphology. Like yeast Msp1, human ATAD1 limits the mitochondrial mislocalization of PEX26 and GOS28, orthologs of Pex15 and Gos1, respectively. GOS28 protein level is also increased in ATAD1−/− mouse tissues. Therefore, we propose that yeast Msp1 and mammalian ATAD1 are conserved members of the mitochondrial protein quality control system that might promote the extraction and degradation of mislocalized TA proteins to maintain mitochondrial integrity.
See also: RS Hedge (in press)
Tail‐anchored proteins are often mistargeted to mitochondria, thereby affecting their function. The conserved AAA+ ATPase Msp1/ATAD1 acts in a quality control system that prevents this aberrant accumulation.
The Msp1/ATAD1 proteins are conserved participants in mitochondrial protein quality control.
Yeast Msp1 is required to maintain mitochondrial integrity upon loss of the GET system.
Mammalian ATAD1 is required to maintain normal mitochondrial structure and function.
Msp1 interacts with the Pex15 and Gos1 tail‐anchored proteins that mislocalize to mitochondria.
Msp1 and ATAD1 limit the mitochondrial mislocalization of tail‐anchored proteins by facilitating their degradation.
Tail‐anchored (TA) proteins are a distinct subset of membrane proteins, which are uniquely characterized by a single transmembrane domain (TMD) at the C‐terminus with the majority of the protein extending into the cytoplasm. They play critical roles in a variety of cellular processes including intracellular trafficking (e.g., all SNARE proteins), protein translocation and maturation (e.g., Sec61β, Tom5, Tom6, Tom7, and Tom22), apoptosis (e.g., Bcl2 family), organelle ultrastructure (e.g., Fis1), and metabolism (e.g., CPT1, Cyb5) (Wattenberg & Lithgow, 2001; Borgese et al, 2003). The majority of TA proteins are targeted to two distinct membrane systems: the mitochondrial outer membrane and the ER membrane from where TA proteins can be subsequently sorted to the nuclear envelope, Golgi complex, peroxisome, vacuole/lysosome, and plasma membrane, probably by vesicle‐mediated trafficking (Kutay et al, 1995).
The topology of TA proteins precludes them from being targeted by the canonical co‐translational signal recognition particle pathway that is used by the vast majority of membrane proteins (Kutay et al, 1995; Steel et al, 2002; Yabal et al, 2003; Brambillasca et al, 2005). Instead, a post‐translational targeting system, GET (Guided Entry of Tail‐anchored proteins), or TRC (TMD Recognition Complex) in Saccharomyces cerevisiae (yeast) or mammals, respectively, chaperones newly synthesized TA proteins and mediates their insertion into the ER membrane (Stefanovic & Hegde, 2007; Schuldiner et al, 2008; Hegde & Keenan, 2011; Denic, 2012). The mechanism whereby TA proteins are targeted and inserted into mitochondria has not yet been determined. In yeast, the GET system starts with capture of the nascent TA protein by the pre‐targeting subcomplex composed of Get4, Get5, and Sgt2 (Jonikas et al, 2009; Wang et al, 2010). Via interaction with Get4, Get3 binds the cargo protein complex and binds the TA protein by directly associating with the TMD (Chartron et al, 2010; Wang et al, 2010). Through docking of Get3 onto the ER membrane receptor, a Get1/Get2 heterodimer, the TA protein is then released and properly inserted (Schuldiner et al, 2008; Wang et al, 2010; Mariappan et al, 2011). However, the fidelity of protein targeting is not perfect, and when it is impaired by mutation (e.g., of the GET system), TA proteins mislocalize to the cytosol or mitochondria (e.g., peroxisomal Pex15 and ER Ubc6) (Schuldiner et al, 2008; Jonikas et al, 2009). To this end, a protein quality control system has evolved to clear TA protein aggregates from the cytosol (Hessa et al, 2011). We hypothesized that a parallel system evolved to clear TA proteins that have mislocalized to the mitochondria and constituted a threat to mitochondrial function.
The mitochondrion is a complex organelle that performs functions that are fundamental to many aspects of cell biology. Therefore, failure or dysregulation of this organelle is associated with many forms of human disease. This includes cancer (Kroemer & Pouyssegur, 2008), diabetes (Patti & Corvera, 2010), neurodegenerative disease (Lessing & Bonini, 2009), and others (Lopez‐Armada et al, 2013; Lopez‐Otin et al, 2013). This critical importance necessitates that cells employ a multi‐tiered quality control system to maintain the integrity of mitochondria (Rugarli & Langer, 2012). One line of defense consists of intra‐mitochondrial proteases that enable the degradation of misfolded or damaged proteins. Among them, two AAA+ (ATPase Associated with various cellular Activities) proteases, m‐AAA and i‐AAA, enforce protein quality in the matrix and inter‐membrane space compartments, respectively (Gerdes et al, 2012; Janska et al, 2013). They are characterized by a AAA+ domain that, in most cases, oligomerizes to form a channel to unfold protein substrates using the energy derived from ATP hydrolysis. Unfolded polypeptides are subsequently degraded by an associated proteolytic domain (Sauer & Baker, 2011). Mutations affecting these proteases cause a plethora of phenotypes in yeast, including respiratory deficiency, loss of mitochondrial DNA (mtDNA), altered mitochondrial morphology, and decreased chronological lifespan (Thorsness et al, 1993; Campbell et al, 1994; Janska et al, 2013). Cells employ a related quality control system on the mitochondrial outer membrane, which enables the extraction and proteasomal degradation of proteins exposed to the cytosol (Taylor & Rutter, 2011). This system appears to also utilize the AAA+ ATPase, Cdc48 in yeast or p97 in mammals, for protein extraction prior to engagement of the proteasome (Heo et al, 2010; Tanaka et al, 2010; Xu et al, 2011). When mitochondria depolarize, cells initiate the autophagic degradation of mitochondria or mitophagy as a second line of defense (Youle & Narendra, 2011; Ashrafi & Schwarz, 2013). In animals, neurons seem to be particularly sensitive to the presence of damaged mitochondria. An increasing number of neurodegenerative diseases, including spastic paraplegia, spinocerebellar ataxia, Parkinson's disease, Alzheimer's disease, and peripheral neuropathies, are linked with defective mitochondrial quality control systems (Rugarli & Langer, 2012).
As a result of our ongoing project to functionally annotate the mitochondrial proteome (Hao et al, 2009; Heo et al, 2010; Bricker et al, 2012; Chen et al, 2012), we here describe the previously unknown function of Ygr028 (or Msp1) in yeast and ATAD1 (also known as Thorase) in mammals, which are mitochondrial members of the AAA+ family of proteins (Zhang et al, 2011). Our data suggest that they are components of the mitochondrial protein quality control system and play a crucial role in degrading TA proteins that escape from the GET system to mislocalize to mitochondria. Depletion of Msp1 in yeast or ATAD1 in human cells and mice leads to ectopic accumulation of a subset of TA proteins that mislocalize to mitochondria. We also showed that Msp1 and ATAD1 physically interact with these same mislocalized TA proteins. Finally, loss of this system causes severe mitochondria damage including respiratory deficiency, loss of mitochondria, and altered mitochondrial morphology. As a result, we hypothesize that Msp1 and ATAD1 are novel mitochondrial protein quality control components that might extract and facilitate the degradation of mislocalized TA proteins from the mitochondria.
Msp1 is an evolutionarily conserved mitochondrial outer membrane protein
Protein sequence alignment and domain prediction of yeast Msp1, fly CG5395, mouse ATAD1, and human ATAD1 suggested that they all contain a transmembrane domain near the N‐terminus and a highly conserved AAA+ domain toward the C‐terminus (Fig 1A). Because yeast Msp1 was previously annotated as a mitochondrial protein (Nakai et al, 1993), we first confirmed its mitochondrial localization. We generated a yeast strain co‐expressing a fully functional Msp1‐GFP fusion protein from the native MSP1 promoter and mitochondrial RFP (Mito‐RFP) (see Supplementary Fig S2A; the construct fully complements the growth phenotype of the get3Δ msp1Δ mutant strain, which will be discussed in the next section and Fig 2). As shown in Fig 1B, we observed a nearly complete overlap of RFP and GFP, demonstrating mitochondrial localization of Msp1‐GFP. We also observed extra‐mitochondrial GFP puncta (marked by arrows), however, which we showed to be peroxisomes based on co‐localization with an RFP protein fused with a peroxisomal targeting sequence (SKL) (Supplementary Fig S1A).
To determine in which mitochondrial compartment Msp1 resides, we performed a biochemical fractionation experiment using a strain expressing a fully functional Msp1‐His6‐HA3 fusion protein from the native MSP1 promoter (see Supplementary Fig S2A). Msp1 was absent from the post‐mitochondria supernatant, but present in purified mitochondria. When we subjected intact, hypotonically swollen or Triton X‐100 lysed mitochondria to Proteinase K digestion, Msp1 was completely degraded in all three situations, like the mitochondrial outer membrane protein Fzo1 (Fig 1C). Furthermore, similar to Om45 and Por1, both integral mitochondrial membrane proteins, neither high salt nor alkaline carbonate extracted Msp1 from mitochondria, suggesting that it is an integral mitochondrial outer membrane protein (Fig 1D).
To confirm that the mitochondrial localization of this protein family is conserved in higher eukaryotes, we established a number of human cell lines that stably express human ATAD1 (hATAD1) as a C‐terminal GFP fusion. As shown in Fig 1E and Supplementary Fig S1B, the GFP signal overlapped with the mitochondrial fluorescent dye, Mitotracker Red. Consistent with yeast Msp1, we also observed that hATAD1‐GFP localized to extra‐mitochondrial puncta, which were verified to be peroxisomes by co‐localization with a peroxisomal RFP marker (Supplementary Fig S1C). We also observed some cytoplasmic localization of ATAD1 (Supplementary Fig S1B and C). Taken together, the aforementioned data indicate that yeast Msp1 and human ATAD1 localize to multiple cellular compartments. Specifically, yeast Msp1 resides on the mitochondrial outer membrane, probably anchored via the N‐terminal TMD, with the majority of the protein including the AAA+ domain exposed to cytosol. While Msp1 and hATAD1 might have important peroxisomal (or cytoplasmic for ATAD1) functions, this study will primarily focus on their roles in mitochondria.
The msp1Δ mutant exhibits synthetic growth and mitochondrial defects in combination with GET mutants
We generated yeast mutants lacking MSP1, which did not show any significant growth phenotype in regular laboratory growth conditions (Fig 2A). Two independent synthetic genetic array experiments, however, both suggested that an msp1Δ mutant exhibits a strong negative genetic interaction with get2Δ or get3Δ mutants (lacking critical components of the GET system), indicating that the growth of the double‐mutant strain is more severely compromised than predicted from the phenotypes of the single knockouts (Costanzo et al, 2010; Hoppins et al, 2011). To verify these data, we generated get2Δ msp1Δ and get3Δ msp1Δ double‐deletion strains and found that both double mutants showed more severe growth phenotypes than the single mutants when grown on glucose at 30°C, which was exacerbated at 37°C (Fig 2A, top panel). Both double mutants completely failed to grow on glycerol medium, which necessitates mitochondrial respiration for growth, whereas all of the single mutants exhibited wild‐type growth (Fig 2A, top panel). These growth phenotypes are fully rescued by plasmid expression of the native or C‐terminally tagged Msp1 protein, but not by expression of the N‐terminally tagged Msp1 protein (which presumably fails to localize to mitochondria) or Msp1E193Q, which is predicted to lack ATPase activity (Supplementary Fig S2A and B). In fact, expression of the Msp1E193Q causes dominant‐negative growth impairment when expressed in a get3Δ mutant (Supplementary Fig S2B). Given the functions of Get2 and Get3, it is reasonable to speculate that MSP1 might have a similar genetic relationship with other GET genes including GET1, GET4, GET5, and SGT2. As expected, the synthetic growth phenotype of the get1Δ msp1Δ mutant is identical to that of the get2Δ msp1Δ and get3Δ msp1Δ double mutants (Fig 2A, top panel). The get4Δ msp1Δ, get5Δ msp1Δ, and msp1Δ sgt2Δ mutants exhibited modest synthetic growth defects, particularly on glycerol medium (Fig 2A, bottom panel). This is consistent with the published observations that Get1, Get2, and Get3 are more central players in the GET system, while Get4, Get5, and Sgt2 play a more auxiliary role (Jonikas et al, 2009).
The complete failure of get1Δ msp1Δ, get2Δ msp1Δ, and get3Δ msp1Δ mutants to grow on glycerol medium suggested that mitochondrial function might be impaired. To explore a potential mitochondrial phenotype further, we extracted whole‐cell lysates from the wild‐type (WT) and single and double mutants and immunoblotted for proteins that localize to different mitochondrial compartments (Fig 2B). We observed that the get1Δ msp1Δ, get2Δ msp1Δ, and get3Δ msp1Δ mutants were completely devoid of the mtDNA‐encoded proteins, Cox2 and Cox3, and were severely depleted of all nuclear‐encoded mitochondrial proteins examined, Atp2, Sdh1, Sdh2, Por1, and Om45 (Lane 4, 6, and 8). The other three double‐mutant strains exhibited a partial depletion of some proteins, particularly those encoded by mtDNA (Lane 10, 12 and 14). The respiratory growth defect of the get1Δ msp1Δ, get2Δ msp1Δ, and get3Δ msp1Δ strains was also irreversible as transformation of an Msp1‐expressing plasmid failed to restore the growth phenotype (Supplementary Fig S2C). We suspected that these double mutants might have lost their mitochondrial DNA and become rho0. To directly test this, we employed a classic cross‐complementation test and found that the strains lacking MSP1 and one of the GET genes failed to complement the growth phenotype of a rho0 tester strain on glycerol medium (Fig 2C). These data demonstrate that the get1Δ msp1Δ, get2Δ msp1Δ, and get3Δ msp1Δ strains lack functional mtDNA. While mtDNA depletion could explain the loss of mtDNA‐encoded proteins, it causes a much less severe depletion of nuclear‐encoded mitochondrial proteins (Supplementary Fig S2D).
Finally, we directly visualized mitochondria by expressing Mito‐RFP in the WT and single‐ and double‐mutant strains (Fig 2D). In addition to the expected observation that the mitochondrial RFP signal in the get1Δ msp1Δ, get2Δ msp1Δ, and get3Δ msp1Δ strains was much weaker than WT or the single mutants (Supplementary Fig S2F), we also found that they exhibited severely altered mitochondrial morphology. Again, the severity of the morphological changes cannot be explained by the lack of mtDNA, which has only modest effects on mitochondrial morphology (Supplementary Fig S2E). The get4Δ msp1Δ, get5Δ msp1Δ, and msp1Δ sgt2Δ strains did not have overt morphological changes, but did have some areas of apparent atypical mitochondrial swelling (marked by arrows, Fig 2D and Supplementary Fig S2G). In summary, we have found that the combined impairment of Msp1 and the GET system causes severe mitochondrial damage, including depletion of resident proteins, loss of the mitochondrial genome, and aberrant morphology.
Depletion of mammalian ATAD1 causes mitochondrial damage
Because a large fraction of human ATAD1 localizes to mitochondria, we wanted to assess whether this mitochondrial function is also conserved in higher eukaryotes. We tested whether depletion of ATAD1 compromises mitochondrial function. First, we performed immunoblots on lysates from brain, which was previously shown to highly express ATAD1, of the WT and ATAD1−/− mice (Fig 3A and B) (Zhang et al, 2011). We observed a significant decrease of many mitochondrial proteins in the ATAD1−/− brain of both female and male mice. The diminished mitochondrial content is further supported by the decreased basal and mitochondrial respiratory capacity of the mouse embryonic fibroblasts (MEFs) lacking ATAD1 (Fig 3C). Second, we also directly visualized mitochondrial morphology of ATAD1−/− MEFs. While WT MEFs exhibit tubular mitochondria, we observed that the mitochondria of ATAD1−/− MEFs were severely fragmented (Fig 3D). The same mitochondrial fragmentation was observed in ATAD1 knockdown HeLa cells (Supplementary Fig S3). Therefore, we conclude that mammalian ATAD1, similar to yeast Msp1, plays an important role in mitochondrial function.
As a complementary test of the conservation of the mitochondrial function of the Msp1 protein family, we expressed human ATAD1 and Drosophila melanogaster CG5395 under control of the yeast ADH1 promoter and found that they either partially or fully rescued the growth phenotype of the get3Δ msp1Δ mutant, respectively (Fig 3E). The glycerol growth defect was also particularly suppressed, suggesting that the mitochondrial role of this protein family spans across the eukaryotic kingdom.
Msp1/hATAD1 is required to minimize the level of the TA protein Pex15/PEX26 mislocalized to mitochondria
Potential mechanisms that might underlie the severe mitochondrial defects caused by combined loss of Msp1 and the GET system were not clear. One potential connection, however, was the previous observation that disruption of the GET system causes some TA proteins to aggregate in the cytoplasm and others to mislocalize to mitochondria (Schuldiner et al, 2008; Jonikas et al, 2009). Perhaps, loss of Msp1 causes mitochondria to be particularly sensitive to the damaging effects of TA protein mislocalization to mitochondria found upon impairment of the GET system. Considering that Msp1 is part of the AAA+ protein family that, in most cases, extracts, disassembles, or unfolds protein substrates, we hypothesized that Msp1 might extract TA proteins that mislocalize to mitochondria upon loss of the GET system. We tested this hypothesis for Pex15, a peroxisomal TA protein that mislocalizes to mitochondria upon disruption of the GET system (Schuldiner et al, 2008; Jonikas et al, 2009). Surprisingly, msp1Δ single mutants displayed significant mitochondrial localization of a GFP‐Pex15 fusion, nearly identical to previous observations with the get3Δ single mutant (Fig 4A) (Schuldiner et al, 2008; Jonikas et al, 2009). Much of the Pex15 protein continues to be targeted to peroxisomes in the single‐mutant strains (Supplementary Fig S4A). In the get3Δ msp1Δ mutant, we observed more complete co‐localization of the GFP and RFP signals, which can only be seen with greatly extended exposure time due to the diminished mitochondrial content in this mutant (Fig 4A; see also Fig 2B and D). We also performed the converse experiment by overexpressing Msp1 in the get3Δ mutant expressing GFP‐Pex15 and observed complete depletion of mislocalized Pex15 from the mitochondria (Fig 4B).
We tested the possibility that Msp1 binds and perhaps extracts mislocalized Pex15. Given that the wild‐type Msp1 protein may associate with substrates transiently, we performed co‐immunoprecipitation experiments using yeast strains expressing a ‘substrate trap’ E193Q mutant (asterisk in Fig 1A), which is predicted to bind substrates but fail to release them efficiently (Weibezahn et al, 2003; Hanson & Whiteheart, 2005). Consistent with this notion, exogenous expression of Msp1E193Q was unable to rescue the growth defect of the get3Δ msp1Δ mutant; moreover, it induced a dominant‐negative phenotype in the get3Δ single mutant (Supplementary Fig S2B). As shown in Fig 4C, we observed that when Pex15 was overexpressed, it stably interacted with Msp1E193 particularly in the get4Δ mutant background. Note that we consistently observed higher accumulation of Msp1E193Q in spite of it being expressed under the endogenous MSP1 promoter. This possibly also contributes to the apparent interaction affinity relative to wild‐type Msp1. It is also noteworthy that Pex15 exhibits elevated accumulation in the get4Δ mutant expressing the dominant‐negative Msp1E193Q, possibly related to mislocalization and stabilization on mitochondria (Fig 4C–input panel).
If the severe phenotype of the get3Δ msp1Δ double mutant is due to accumulation of mislocalized TA proteins in the mitochondrial outer membrane, overexpression of proteins like Pex15 should exacerbate this phenotype. To test this, we compared growth upon overexpression of GFP‐Pex15 from the GAL1 promoter. While the get3Δ msp1Δ mutant exhibited impaired growth on its own, the GAL1::PEX15 construct caused a complete loss of growth on galactose medium (Fig 4D). One possible mechanism whereby Msp1 might maintain a low level of mislocalized TA proteins is to extract them and facilitate their degradation. We tested the half‐life of GFP‐Pex15 utilizing the stringent glucose repression of new transcription from the GAL1 promoter. We first noticed that the steady‐state level of Pex15 was increased in the msp1Δ mutant compared to WT (t = 0 min, Supplementary Fig S4B). Furthermore, the time‐course of Pex15 degradation was significantly slower in the msp1Δ mutant (Supplementary Fig S4B and C) most likely due to impaired extraction and degradation of mitochondrial Pex15. To more rigorously test whether Msp1 limits mitochondrial Pex15 by facilitating its degradation, we measured the half‐life of Pex15 in the get3Δ mutant expressing either the wild‐type Msp1 or mutant Msp1E193Q (Fig 4E, Supplementary Fig S4D and E). Therefore, in both cases, Pex15 is strongly mislocalized to mitochondria (t = 0 min, Supplementary Fig S4D). Pex15 is significantly more stable in the strain expressing Msp1E193Q (Fig 4E and Supplementary Fig S4E). When we visualized the localization of GFP‐Pex15 throughout the time‐course, we observed that mitochondrial Pex15 was degraded faster in the get3Δ mutant expressing the wild‐type Msp1 compared to the Msp1E193Q, which still retained significant mitochondrial Pex15 after 180 min (Supplementary Fig S4D). Taken together, these results suggest that Msp1 is required to minimize the abundance of the Pex15 TA protein on mitochondria, most likely by promoting its extraction and degradation.
We also tested whether hATAD1 functions similarly to protect mammalian mitochondria from the mislocalization of PEX26, the human ortholog of Pex15. We established a human hepatocellular carcinoma (HepG2) cell line stably expressing GFP‐PEX26 and repeatedly observed very modest mitochondrial GFP localization in a minority of cells (Fig 4F‐top; marked by arrows). Knockdown of ATAD1 with three distinct siRNAs (Fig 4G), however, greatly increased the overlap of GFP‐PEX26 and Mitotracker Red (Fig 4F). We also observed that the overall intensity of the mitochondrial GFP signal was significantly enhanced, consistent with increased steady‐state level of PEX26 in the whole‐cell lysates from multiple cell lines (Fig 4F and Supplementary Fig S4F). Taken together, both yeast Msp1 and human ATAD1 protein are important to minimize the mitochondrial mislocalization of the TA proteins, Pex15, and PEX26, respectively.
Msp1/hATAD1 physically interacts with and is required to prevent TA protein Gos1/GOS28 misaccumulation on mitochondria
To identify bona fide substrates of Msp1, we performed a two‐step purification using a yeast strain expressing Msp1‐His6/HA3 tag from the MSP1 promoter. We expressed the wild‐type or E193Q mutant Msp1 in both the WT and get3Δ strains and purified Msp1 by two‐step purification. In examining potential substrates, we aimed to look for proteins that fulfill two criteria. First, they are purified more efficiently by the ‘trap mutant’ Msp1E193Q than by the wild‐type protein. Second, they exhibit stronger association with Msp1 in the get3Δ mutant strain than in the WT strain. As shown in Supplementary Table S1, Gos1 meets both criteria and is one of the most abundant proteins detected by mass spectrometry. Importantly, it is also a TA protein and functions as a v‐SNARE in Golgi vesicular trafficking (McNew et al, 1998). We first verified this result by conducting a directed co‐immunoprecipitation experiment. We expressed GFP‐Gos1 and Msp1‐HA fusion proteins under their endogenous promoters and performed an anti‐HA immunoprecipitation from crude mitochondrial lysate. As seen in Fig 5A, GFP‐Gos1 showed a stable association with Msp1E193Q. We also observed a stronger GFP‐Gos1 purification with Msp1E193Q in the get3Δ and get4Δ mutants compared to a strain with a functional GET system (WT). Note that Msp1E193Q consistently accumulated at a higher level compared to the wild‐type protein (Fig 5A – input panel; also see Fig 4C – input panel). Similar to Pex15, the steady‐state level of Gos1 was higher in the get3Δ and get4Δ mutant strains expressing the dominant‐negative Msp1E193Q mutant (Input panel, Fig 5A). To test Gos1 localization, we expressed GFP‐Gos1 under the endogenous GOS1 promoter and observed a partial overlap of GFP‐Gos1 signal with Mito‐RFP in both the get4Δ msp1Δ mutant strain and in the get3Δ mutant expressing the dominant‐negative Msp1E193Q (Fig 5B and Supplementary Fig S5A).
To test whether Msp1 facilitates the degradation of mislocalized Gos1, we used the GAL1 promoter‐based transcriptional shut‐off system to test its half‐life. Overexpression of Gos1 from the GAL1 promoter leads to its mislocalization to the mitochondria of the get3Δ mutant (T = 0 min, Fig 5D). The time‐course of Gos1 degradation in the get3Δ mutant expressing dominant‐negative Msp1E193Q is significantly slower than the get3Δ mutant, which expresses endogenous wild‐type Msp1 (Fig 5C). We also visualized the localization of GFP‐Gos1 throughout the time course. We observed that mitochondrial Gos1 was degraded faster in the get4Δ mutant compared to the strain expressing Msp1E193Q, which still retained significant mitochondrial Gos1 after 180 min (Fig 5D).
To further address whether misaccumulation of Gos1 contributes to the mitochondrial damage in the get3Δ msp1Δ mutant, we attempted to delete GOS1 and test for rescue of the growth phenotype. Unfortunately, loss of Gos1 in our W303‐1a strain background causes lethality (Supplementary Fig S5B). Instead, we reasoned that overexpression of Gos1 might overload mitochondria and exacerbate the growth phenotype caused by loss of Msp1 and/or the GET system. As shown in Fig 5E, overexpression of Gos1 either caused or exacerbated growth phenotypes in get1Δ, get2Δ, and get3Δ single mutants. The double mutants containing those deletions combined with an msp1Δ mutation already exhibited such severe growth defects that Gos1 had no additional consequence. However, the get4Δ msp1Δ, get5Δ msp1Δ, and msp1Δ sgt2Δ double mutants, which maintained much more mitochondrial function (See Fig 2), were markedly impaired by Gos1 overexpression, particularly in glycerol growth (Fig 5E). The effect of Gos1 overexpression is specific as overexpression of Sbh1, an endoplasmic reticulum TA protein that does not mislocalize to mitochondria, did not exacerbate the growth phenotypes (Supplementary Fig S5C) (Schuldiner et al, 2008).
In addition to Gos1, we also observed that Tom5, a native mitochondrial TA protein, was strongly purified by Msp1E193Q in the wild‐type and the get3Δ mutant strain (Supplementary Table S1). This observation raised the possibility that Msp1 might not only bind to non‐native, mislocalized mitochondrial TA proteins but also to native proteins. To test this hypothesis, we tagged Tom5 and all other mitochondrial TA proteins, Tom6, Tom7, Tom22, Fis1, Gem, and Fmp32 with a GFP at their N‐termini and performed co‐immunoprecipitation experiments (Supplementary Fig S6A). We observed no interaction between Tom5, Tom6, Fis1, Gem1, or Fmp32 and either wild‐type Msp1 or the Msp1E193Q mutant. Tom7 and Tom22 showed an interaction, but it was with both the wild‐type and mutant Msp1 protein. More importantly, we observed no significant delay in the rate of degradation of any of these native mitochondrial TA proteins in the msp1Δ mutant (Supplementary Fig S6B). Furthermore, overexpression of TOM5, TOM6, or TOM7 did not exacerbate the growth phenotype of the msp1Δ mutant strain (Supplementary Fig S6C). Therefore, it appears that none of the native mitochondrial TA proteins are efficient substrates of Msp1, at least in the conditions in which we conducted these experiments.
Based on our observations with Gos1, we performed two experiments to test whether GOS28, the mammalian ortholog of yeast Gos1, is subjected to mitochondrial misaccumulation, which is prevented by ATAD1. First, to test whether GOS28 misaccumulates on mitochondria upon ATAD1 depletion, we established a human dermal fibroblast (HDF) cell line stably expressing a GFP‐human GOS28 fusion protein and subjected it to knockdown with control and ATAD1 siRNAs. In control siRNA (scr)‐treated cells or cells treated with si#4 that fails to knockdown ATAD1 (Fig 6B), the majority of the GFP signal was localized to the peri‐nuclear Golgi apparatus and had no overlap with Mitotracker Red (Fig 6A). However, when ATAD1 was knocked down by si#1, 2, or 3 (Fig 6B), we clearly observed a partial redistribution of GFP‐GOS28 to overlap with Mitotracker Red (Fig 6A). Second, to directly test whether GOS28 physically interacts with ATAD1, we assessed co‐immunoprecipitation using crude mitochondrial extract from HepG2 cells stably expressing GFP‐GOS28 and empty vector, wild‐type ATAD1 or a substrate trap mutant ATAD1 (ATAD1E193Q). Only the ATAD1E193Q mutant purified detectable amounts of GOS28 (Fig 6C). Note that the steady‐state GOS28 level was reproducibly elevated in cells expressing the ATAD1E193Q (Fig 6C, lane 3), which is likely caused by its dominant‐negative activity. This is consistent with the observation that expression of ATAD1E193Q causes ectopic accumulation of GFP‐GOS28 on the mitochondria (Supplementary Fig S7A). Most importantly, we further observed increased steady‐state abundance of endogenous GOS28 in multiple ATAD1−/− mouse tissues, including brain, liver, and heart (Fig 6D and Supplementary Fig S7B). We therefore conclude that Gos1 and GOS28 are bona fide substrates of yeast Msp1 and mammalian ATAD1, respectively.
Based on the data presented herein, we conclude that yeast Msp1 and human ATAD1 belong to an evolutionarily and functionally conserved protein family. We demonstrate that they play a role in quality control of mitochondrial outer membrane proteins. Specifically, Msp1 or ATAD1 prevents a subset of TA proteins (yeast–Pex15 and Gos1; mammals–PEX26 and GOS28) that escape from the GET or TRC system from inappropriately accumulating on the mitochondria, likely by extracting them and facilitating their cytoplasmic degradation. The physiological consequences of combined impairment of Msp1 and the GET system include loss of mtDNA, loss of mitochondrial resident proteins, and severely altered mitochondrial morphology. The principal evidence supporting this model is enumerated below.
First, we have found that the targeting and insertion systems for TA proteins are quite imprecise. This necessitates quality control systems to manage the burden of TA proteins that are mislocalized to other cellular compartments. In mammals, the cytoplasmic Bag6 protein not only participates in chaperoning TA proteins to the ER for insertion, but also mediates ubiquitination of proteins that are inappropriately released from ribosomes (Mariappan et al, 2010; Hessa et al, 2011). We suggest that the Msp1 protein family represents a similar protein quality control system for the mitochondria. This system is necessitated by the fact that, even when the GET or TRC system is fully functional, TA proteins can escape and mislocalize to mitochondria. This is demonstrated by our observation that deletion of Msp1 alone caused ectopic accumulation of Pex15 protein on yeast mitochondria (Fig 4A). This is also true for both PEX26 and GOS28 when ATAD1 protein was depleted in multiple human cell lines (Figs 4G and 6A and Supplementary Fig S7A). For reasons that remain unclear, the mitochondrial membrane appears to be a receptive host for TA proteins that fail to target their native membrane. It is possible that this is because the mitochondrial outer membrane appears to be one of the only two organelles where TA proteins are directly targeted and inserted. Previous reports and our data (Fig 4A) clearly show that genetic impairment of the GET system causes mislocalization of a subset of TA proteins on the mitochondria (Schuldiner et al, 2008; Jonikas et al, 2009). Therefore, we suggest that cells have evolved a protein quality control system residing on mitochondria to handle this burden of mislocalized TA proteins.
Second, Msp1 physically interacts with mislocalized TA proteins. Our two‐step purification identified the TA protein Gos1 as a possible genuine substrate of Msp1, which interacts with the substrate trap mutant Msp1 (Supplementary Table S1 and Fig 5A). This Msp1 mutant can also stably bind to overexpressed Pex15, but for unknown reasons, we did not observe Pex15 in the Msp1 co‐purification experiment. Presumably, the substrate trap mutant Msp1 stabilizes the physical interaction with substrates, while wild‐type Msp1 may bind to them transiently. It is also possible, however, that the enhanced observed binding of mutant Msp1 is simply a function of its elevated accumulation. Beyond Gos1 and Pex15, we expect that additional substrates remain to be discovered. Interestingly, we also identified Tom5 as another potential substrate by Msp1 co‐purification; however, we failed to recapitulate this result using directed co‐immunoprecipitation experiments (Supplementary Fig S6A). It is possible that the observation of co‐purification is an artifact, and Tom5 and Msp1 have no relationship. It is also possible, however, that Tom5 is a bona fide substrate of Msp1, but we failed to obtain evidence for this either due to the specific conditions of our experiments or the use of a GFP‐Tom5 fusion protein. As a result, we do not firmly conclude that Msp1 cannot recognize native mitochondrial TA proteins as substrates. In addition, it is possible that Msp1 extracts a broader range of protein substrates, including non‐TA proteins, which is consistent with our observation of non‐TA mitochondrial proteins bound specifically to the Msp1‐E193Q mutant in the two‐step purification experiment (Supplementary Table S1). In fact, mammalian ATAD1 (possibly the cytoplasmic pool) was shown to extract or disassemble the AMPAR protein complex, which has no TA protein constituents, from the cell surface (Zhang et al, 2011). Therefore, while Msp1/ATAD1 clearly functions in the degradation of mitochondrially mislocalized TA proteins, it appears to function in a broader context as well.
Third, Msp1 is required for the normal rate of degradation of mislocalized TA proteins. We repeatedly observed that the steady‐state level of Msp1 substrate proteins is higher in Msp1‐ or ATAD1‐depleted cells. Furthermore, we conducted a chase experiment on Pex15 and Gos1 and demonstrated that their half‐life is extended in the absence of Msp1 (Fig 4E, Supplementary Figs S4B–E and S5C and D). We speculate that Msp1 unfolds and extracts Pex15 and Gos1 via ATP hydrolysis by the AAA+ domain. However, unlike m‐ or i‐AAA ATPase, which has an accessory proteolytic domain to digest the unfolded peptide, Msp1 possesses only the AAA+ domain. Therefore, the mechanisms whereby mislocalized TA proteins are degraded in the cytosol remain to be explored. It is known that the cytoplasmic proteasome is engaged to degrade membrane proteins from various organelles, including the ER and mitochondria (Neutzner et al, 2007; Karbowski & Youle, 2011; Taylor & Rutter, 2011). Therefore, it seems likely that proteasomes provide the major degradative activity for TA proteins once they are extracted from the mitochondria.
Fourth, the Msp1 and ATAD1 proteins are essential for the maintenance of mitochondria. We observed that loss of both Msp1 and the GET system causes loss of mtDNA and protein, and severe morphological defects (Fig 2). Mammalian cells, however, seem to be more susceptible to the depletion of ATAD1 as mutation of ATAD1 is sufficient to cause significant mitochondrial impairment (Fig 3A–D). Prior to this study, the major connection between the GET system and mitochondria related to TA protein mislocalization to mitochondria upon GET system disruption. Perhaps, the most profound mitochondrial defect observed is the loss of mitochondrial protein content in the get1Δ msp1Δ, get2Δ msp1Δ, and get3Δ msp1Δ mutant strains, which cannot be explained by the loss of mtDNA. Given that the TA proteins Tom5, Tom6, Tom7, and Tom22 are critical components of the TOM complex, perhaps ectopic TA protein accumulation causes defects in TOM complex formation. The ectopic TA proteins could either displace one or more of the tail‐anchored components of the TOM complex, rendering it inactive. Alternatively, hyper‐accumulation of TA proteins could impair the unknown mitochondrial TA protein import system, leading to decreased import efficiency of Tom5, Tom6, Tom7 and/or Tom22 and therefore impaired mitochondrial protein import. We also observed altered mitochondrial morphology in both msp1Δ yeast and ATAD1 knockdown cells. Intriguingly, before the function of the GET system was determined, Get1 (also known as Mdm39) was discovered in a genetic screen for genes important for mitochondrial distribution and morphology (Dimmer et al, 2002). Perhaps, this observation relates to the effect of mislocalized TA proteins, caused by mutation of GET1, to impair mitochondrial morphology. In fact, many mitochondrial shaping factors such as Fis1 (yeast and mammals), Mff (mammals), and Gem1 (yeast) are TA proteins. Therefore, as speculated for the TOM complex above, mitochondrial morphology and dynamics might be susceptible to aberrant accumulation of non‐mitochondrial TA proteins. For example, mislocalization of the Gos1 v‐SNARE protein might enable promiscuous fusion events with organelles displaying cognate t‐SNARE proteins.
Finally, our data suggest that the mitochondrial function of Msp1 protein family is conserved between yeast, flies, and mammals. Both Drosophila CG5395 and human ATAD1 suppressed the respiratory growth phenotype of the get3Δ msp1Δ mutant strains (Fig 3E). In addition, depletion of ATAD1 caused accumulation and mitochondrial localization of PEX26 and GOS28 as well as mitochondrial defects (Figs 3, 4G and 6, and Supplementary Fig S3D). In Drosophila, mitochondria undergo dramatic morphological changes during spermatogenesis. At the onion stage, mitochondria form a specialized structure called the nebenkern (Tokuyasu, 1975). Mutation of the fly MSP1 ortholog CG5395 (nmd mutant, no mitochondrial derivative) leads to a much smaller nebenkern and a slight loss of mitochondrial membrane potential (Noguchi et al, 2011). While these data do not implicate impaired protein quality control, it demonstrates that this gene is important in mitochondrial morphology and function. Furthermore, ATAD1 knockout mice exhibit severe neuronal defects, including deficiencies in learning and memory and a seizure‐like syndrome (postnatal day 19–25) (Zhang et al, 2011). Interestingly, we observed ectopic accumulation of GOS28 in multiple tissues from ATAD1−/− mice, particularly the brain (Fig 6D). Some of these neuronal deficits are likely due to the failure of disassembly of the AMPAR receptor, but similar neuronal phenotypes are observed with impaired mitochondrial quality control in mice and humans (Rugarli & Langer, 2012). We speculate that the mitochondrial defects caused by loss of ATAD1 might partially contribute to the phenotype of the ATAD1−/− mice.
In conclusion, we suggest that the Msp1/ATAD1 protein family performs an evolutionarily conserved role in quality control of the mitochondrial outer membrane. This function, while perhaps including additional activities, is likely to extract and promote the degradation of TA proteins that have been mislocalized to mitochondria. The consequences for failure of this activity, particularly under a stress to the GET system, are devastating for mitochondrial morphology and function. Given the critical importance of mitochondrial quality control in human physiology and pathophysiology, we anxiously await a broader and deeper understanding of this protein family in health and disease.
Materials and Methods
Yeast strains and growth conditions
Saccharomyces cerevisiae W303‐1a (MATa, his3 leu2 met15 trp1 ura3) was used in this study. The standard PCR‐based homologous recombination method was used to generate all mutant strains. Briefly, drug selection cassette (KanMX4, hphMX4, or natMX4) flanked with 45‐bp fragments upstream and downstream of the gene of interest was PCR amplified and transformed into the wild‐type diploid (Goldstein & McCusker, 1999) (Wach et al, 1994). The haploid strain was generated by sporulation and tetrad analysis. The strains with chromosomally integrated PGAL1‐GFP‐PEX15 were generated as described in (Longtine et al, 1998). The genotype of the strain was verified by standard genotyping PCR. The genotypes of all yeast strains used in this study are listed in Supplementary Table S2.
For yeast transformation, the standard lithium acetate procedure was used (Gietz et al, 1992). Transformed yeast cells were grown in synthetic complete dextrose (SD) medium lacking the appropriate amino acid(s) for auxotrophic selection purposes at 30°C. Media used in this study include standard YP and synthetic minimal medium supplemented with 2% glucose, 2% raffinose, or 3% glycerol. Solid plates contain 2.2% (w/v) agar.
To produce plasmids expressing non‐tagged, C‐terminal His6/HA3 or GFP‐tagged Msp1, the MSP1 ORF flanked with its upstream 500‐bp promoter region was PCR amplified from yeast genomic DNA and ligated into the pRS416, pRS426, or pRS416 vector containing either a C‐terminal His6/HA3 or GFP tag. Msp1E193Q construct was made by site‐directed mutagenesis of the Msp1‐His6/HA3 construct. To generate human ATAD1 and fly CG5393 constructs, ATAD1 and CG5393 ORF were amplified from HepG2 and fly cDNA, respectively, and ligated into pRS416‐based vector containing the ADH1 promoter (Mumberg et al, 1995). N‐terminal GFP‐tagged Gos1 construct was made by ligating PGOS1, GFP, and GOS1 ORF fragments together using the standard sewing PCR and ligating into the pRS414 vector. The PGAL1::GFP‐Gos1 construct was generated by ligating GOS1 ORF into pRS414‐based vector containing the GAL1 promoter and a N‐terminal GFP tag. The Retro‐X™Q Vectors (Clontech) were used to generate mammalian constructs. Human ATAD1 ORF fused with HA or GFP tag at the 3′‐end or PEX26 or GOS28 ORF fused with GFP at the 5′‐end was ligated into pQCXIP or pQCXIZ vector. Site‐directed mutagenesis was used to convert ATAD1‐HA into ATAD1E193Q‐HA. Peroxisomal fluorescent marker construct was made by fusing RFP protein with the peroxisomal targeting sequence (SKL) at the C‐terminus.
Isolation of yeast mitochondria
Yeast cells were harvested at mid‐log phase (OD600 = 2–3) unless indicated otherwise. Preparation of crude and purified mitochondria was as described previously (Boldogh & Pon, 2007). Yeast pellet was washed once with ddH2O, resuspended, and incubated in TD buffer (100 mM Tris‐SO4, pH 9.4 and 100 mM DTT) for 15 min at 30°C. Spheroplasts were obtained by incubating cells in SP buffer (1.2 M sorbitol and 20 mM potassium phosphate, pH 7.4) supplemented with lyticase (2 mg/g of cell pellet) (Sigma‐Aldrich) for 1 h at 30°C to digest the cell wall. Spheroplasts were gently washed once in ice‐cold SHE buffer (1.2 M sorbitol, 20 mM HEPES‐KOH, pH 7.4, 2 mM MgCl2, 1 mM EGTA, and 1 mM PMSF) and homogenized in ice‐cold SHE buffer that contains 0.6 M sorbitol with a Dounce homogenizer applied with 10‐20 strokes. The crude mitochondrial fraction was obtained by differential centrifugation. Protein concentration was determined using Bradford protein assay (Bio‐Rad).
Continuous Nycodenz gradients were used to purify crude mitochondria. To make a gradient, 2.1 ml of 5, 10, 15, 20, and 25% Nycodenz in SHE buffer was layered in 14 × 89 mm Ultra‐Clear centrifuge tubes (Beckman) and the tubes were sat at room temperature for 3–4 h to allow the Nycodenz to diffuse. Crude mitochondria were loaded on top of the chilled gradient and separated at 100,000 × g for 1 h at 4°C (SW41 rotor; Beckman). Intact purified mitochondria were recovered from a brown band at around 16% Nycodenz concentration.
Assessment of sub‐mitochondrial localization
The experiment was conducted as described previously (Chen et al, 2012). Briefly, proteinase‐free mitochondria were incubated in the isotonic SH buffer (0.6 M Sorbitol, 20 mM HEPES‐KOH) or hypotonic H buffer (20 mM HEPES‐KOH) with and without 1% Triton X‐100. Proteinase K (10 μg/μl) was then added and incubated on ice for 20–30 min. The digestion was stopped by adding phenylmethylsulfonyl fluoride (PMSF) to 2 mM. The reaction mixtures were denatured in 6× Laemmli buffer and resolved by 12% SDS‐PAGE, followed by immunoblot.
The high‐salt and alkaline extraction were adapted from a previously described protocol (Boldogh & Pon, 2007). Intact mitochondria were treated with salt (100 mM KCl in SHE buffer without sorbitol) or carbonate (100 mM Na2CO3 in SHE buffer) for 30 min on ice and subjected to ultracentrifugation (100,000 × g for 20 min) to separate soluble and insoluble proteins. Soluble fractions were precipitated by 15% trichloroacetic acid (TCA) and dissolved in 1× Laemmli buffer. An equal amount of lysate from both the soluble and insoluble fraction was analyzed by immunoblot.
Two‐step Msp1‐His6/HA3 purification
Crude mitochondria isolated from strains grown in synthetic raffinose medium to mid‐log phase (OD600 = 2.5) were solubilized in lysis buffer (20 mM HEPES, pH 7.4, 10 mM KCl, 1.5 mM MgCl2, 1 mM EDTA, 150 mM NaCl, 10 mM imidazole, 2.1 mg/ml NaF, 10.8 mg/ml 2‐glycerolphosphate, 0.5% digitonin and protease inhibitors) for 1 h at 4°C. For the first step of the purification, cleared mitochondria lysates were incubated with equilibrated Ni‐NTA beads for 1 h at 4°C. Ni‐NTA beads were washed 5 times with wash buffer (compositions are same as lysis buffer except for 20 mM imidazole and 0.05% digitonin). Proteins were eluted three times by 250 mM imidazole under non‐denaturing conditions. For the second step of the purification, final eluates of the first purification were combined and mixed with anti‐HA antibody‐conjugated agarose (A2095; Sigma) for 1 h at 4°C. The agarose was then washed five times, and proteins were eluted five consecutive times by 1 mg/ml HA peptides. Final eluates were precipitated in 15% TCA overnight at 4°C and analyzed by mass spectrometry.
Yeast samples were prepared by growing yeast cells to early log phase (OD600 = 0.8–1) in synthetic dropout medium with appropriate carbon sources at 30°C. To prepare mammalian cell samples, we cultured cells on Poly‐L‐Lysine‐coated Lab‐Tek® II chamber slides for a day to allow them to fully attach. They were either visualized directly or stained with 25 nM Mitotracker® Red CMXRos (Life Technologies) in FBS‐free medium for 5 min at 37°C before imaged. To examine morphological changes in mitochondria from ATAD1−/− MEFs, they were transiently transfected with Mito‐RFP to label the mitochondria. Transfected cells were fixed for 15 min with 4% PFA + 4% sucrose in PBS without permeabilization. Cells were washed three times with PBS buffer, stained with DAPI, and washed three more times with PBS buffer. Images of the cells were acquired by using a Zeiss LSM 710 laser‐scanning confocal microscope using a 40× oil‐immersion objective.
Cells were imaged on the Axio Observer. Z1 imaging system (Carl Zeiss) equipped with 40× and 100× objectives (oil‐immersion). Digital fluorescence and differential interference contrast (DIC) images were acquired using a monochrome digital camera (AxioCam MRm, Carl Zeiss). Z‐stacks of 0.35‐μm slides were obtained and deconvolved using the AxioVision software (Version 4.8, Carl Zeiss). 2D projection of the z‐stacks was performed using the ImageJ software. The final images were adjusted and assembled using Adobe Photoshop CS5.1. Brightness and contrast were adjusted only using linear operation on the entire image.
The yeast fluorescent images shown in this study are representative pictures from at least two independent experiments (n ≥ 2). The images of the mammalian cells are representative of three independent experiments (n = 3).
TCA pellets were resuspended in 8 M urea, 50 mM HEPES, and pH 8.8. Proteins were reduced by the addition of dithiothreitol to 5 mM and incubation at room temperature for 30 min. Cysteines were alkylated by adding iodoacetamide to 15 mM and incubating in the dark at room temperature for 60 min. Iodoacetamide was quenched with an additional 10 mM DTT. Samples were digested with lysyl endopeptidase (lysC) at 20 ng/μl overnight at room temperature after diluting the urea to 2 M by adding 3 volumes of 50 mM HEPES, pH 8.8. LysC‐digested peptides were subsequently digested with trypsin at 15 ng/μl for 60 min at 37°C. Doubly digested peptides were acidified by adding trifluoroacetic acid to 0.2% and desalted on hand‐packed C18 STAGE tips (Rappsilber et al, 2007). Peptides were eluted with 70% acetonitrile and 1% formic acid and dried under vacuum. Desalted peptides were resuspended in 5% formic acid, and each sample was analyzed in technical duplicate by reverse‐phase liquid chromatography electrospray mass spectrometry on a LTQ Orbitrap Velos Pro (Thermo Fisher Scientific). Peptides were analyzed on a LTQ Orbitrap Velos Pro mass spectrometer (Thermo Fisher Scientific). Nanospray tips were hand‐pulled with 100‐μm inside diameter fused‐silica tubing and packed with 20 cm of Maccel C18AQ resin (3 mm, 200 Å; Nest Group). Peptides were separated with a gradient of 6–30% CH3CN in 0.125% formic acid over 60 min at a flow rate of 300 nl/min. Peptides were detected using a data‐dependent Top20 MS2 method. For each cycle, one full MS scan of mass/charge ratio (m/z) = 400–1,200 was acquired in the Orbitrap at a resolution of 60,000 at m/z = 400 with automatic gain control (AGC) target of 1 × 106. Each full scan was followed by the selection of the most intense ions, up to 20, for collision‐induced dissociation (CID) and analysis in the linear ion trap. Selected ions were excluded from subsequent selection for 60 s. Ions with a charge of 1 or unassigned were also excluded from MS2 analysis. Maximum ion accumulation times were 100 ms for both full MS and MS2 scans. Lockmass, with atmospheric polydimethylsiloxane (m/z = 371.1012) as an internal standard, was used for internal mass calibration.
Peptide identification and filtering
MS2 spectra were searched using SEQUEST v.28 (rev. 13) against a composite database containing the translated sequences of all predicted open reading frames of S. cerevisiae (http://downloads.yeastgenome.org, downloaded 30 October 2009) and its reversed complement with the following parameters: a precursor mass tolerance of ± 20 parts per million (ppm); 1.0‐dalton product ion mass tolerance; combined lysC–trypsin digestion; up to two missed cleavages; a static modification of carbamidomethylation on cysteine (+57.0214); and a dynamic modifications of methionine oxidation (+15.9949). Peptide spectral matches were filtered to 1% FDR using the target‐decoy strategy combined with linear discriminant analysis (LDA) using several different parameters including the SEQUEST Xcorr and DCn′ scores, and precursor mass error (Elias & Gygi, 2007; Huttlin et al, 2010). The data were further filtered to control protein‐level FDRs. Peptides from all fractions in each experiment were combined and assembled into proteins. Protein scores were derived from the product of all LDA peptide probabilities, sorted by rank, and filtered to 1% FDR. The FDR of the remaining peptides fell markedly after protein filtering.
Steady‐state protein analysis
Yeast whole‐cell lysates were prepared from 1 OD600 of yeast cells harvested from raffinose medium at mid‐log phase. Yeast pellets were washed with ddH2O once, resuspended in 200 μl of 20 mM NaOH, incubated at room temperature for 5 min, and pelleted. Laemmli buffer was added to resuspend pellets, and they were denatured at 95°C for 5–10 min (Kushnirov, 2000). Anti‐Cox2, Cox3 (MitoSciences), Por1, and Pgk1 (3‐phosphoglycerate kinase) (Abcam) were used on immunoblots.
Animal tissue lysates were obtained from brain homogenates of 4‐ to 5‐week‐old wild‐type and ATAD1−/− mice (C57BL/6) (Zhang et al, 2011). Freshly isolated whole brains were powderized on dry ice and homogenized in lysis buffer (50 mM HEPES, pH 7.5, 150 mM NaCl, 1 mM DTT, 5% glycerol, 1% Triton X‐100) containing protease inhibitors (Sigma‐Aldrich). The brain extracts were kept on ice for 1 h and then centrifuged at 15,000 × g for 30 min. Protein concentration was determined using the BCA protein assay (Thermo Scientific). Twenty μg of lysates was resolved on 4–20% gradient NuPAGE (Invitrogen) and transferred to PVDF membranes. Immunoblot analyses were performed using ATAD1 antibodies (NeuroMab or abcam) and mitochondrial antibodies, TOMM20, COX1, COX4, hexokinase 1, hexokinase 2, VDAC1, pyruvate dehydrogenase, and S6 ribosomal protein (Cell Signaling Tech.). β‐actin, β‐tubulin, and MAP2 (SIGMA) antibodies were used for controls. All mouse experiments were performed under approved protocols of the Institutional Animal Care and Use Committee at Johns Hopkins University School of Medicine.
Measurement of oxygen consumption rate
Mitochondrial oxygen consumption rate (OCR) was assessed using WT and ATAD1 KO MEFs in an XF24 Extracellular Flux Analyzer (Seahorse Bioscience), as described previously (Cooper et al, 2012). Culture media of MEF cells plated at a density of ~0.5 × 106 per well in an XF24 cell culture microplates were replaced with XF24 Dulbecco's modified Eagle medium (DMEM) containing 10 mM glucose, 2 mM l‐glutamine (Life Technologies), and 2 mM sodium pyruvate (Life Technologies). OCR was measured at 37°C with 1‐min mix, 1‐min wait, and 5‐min measurement protocol. OCR was analyzed after 30 min incubation in a CO2‐free incubator. Oligomycin, carbonilcyanide m‐chlorophenylhydrazone (CCCP), and rotenone were sequentially injected into each well to assess basal respiration, coupling of respiratory chain, and mitochondrial respiratory capacity. OCRs were normalized relative to protein concentration in each well, and the data are presented as % change of control.
One milligram of crude mitochondria extracted from each strain was solubilized in 0.5% digitonin for 1 h at 4°C. Cleared mitochondria lysates were mixed with anti‐HA antibody‐conjugated agarose for 1 h at 4°C. Agarose was washed two times with buffer containing 250 mM NaCl and 0.05% digitonin and three times with buffer containing 600 mM NaCl and 0.05% digitonin. Laemmli buffer without β‐ME was added to the agarose to elute proteins. Final eluate and 4–5% of the mitochondria lysate were resolved by 12% SDS‐PAGE followed by immunoblot. Results shown in this study are representatives of two independent experiments (n = 2).
Mammalian cell culture
HepG2 (human hepatocellular carcinoma) cells were maintained in DMEM/F12 (Thermo Scientific) with 10% FBS. HEK293T, HDF (human dermal fibroblasts), and HeLa cells were maintained in DMEM (Thermo Scientific) with 10% FBS. Cells were cultured at 37°C with 5% CO2. To establish stable cell lines, retrovirus was produced in HEK293T cells that were co‐transfected with the vector containing the gene of interest, Gag‐pol, and VSVG (amount ratio is 3:2:1) using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Retrovirus was harvested from the medium at 48 h post‐transfection and applied to the target cells. After 24 h post‐infection, target cells were selected in 4 μg/ml puromycin and/or 150 μg/ml zeocin for 5–7 days. Stable cell lines were maintained in the appropriate medium with 0.5 μg/ml puromycin and/or 20 μg/ml zeocin. All knockdowns were performed by treating cells with 10 nM siRNA, using the Lipofectamine RNAiMax reagent, according to the manufacturer's instructions (Invitrogen). The All‐Stars non‐targeting siRNA (Qiagen) was used as the control (scr). siRNAs targeting human ATAD1 (NM_032810.2) were designed with the Dharmacon siDesign Center tool (http://www.thermoscientificbio.com/design-center/). Sequences of the sense strands of targeting siRNAs, which include a 3′ tt DNA overhang, are as follows: (#1) GAAGCAAAUUGGAGUGAAAtt, (#2) GAAUGAAGUUGUUGGUUUAtt, (3) CAUGUUACUUGGAGUGAUAtt and (4) GAUAAGUGGUAUGGAGAAUdtt. Cells were subjected to knockdown on day 0, again on day 3, and analyzed on day 6. Efficacy of ATAD1 knockdown was verified by immunoblot with mouse monoclonal anti‐ATAD1 antibody (1:1,000) (Abcam).
Mouse embryonic fibroblasts (MEFs) were prepared from embryonic day 13 mouse pups as described previously with some modification (Jozefczuk et al, 2012). MEFs were isolated from tissues by 0.05% trypsin/EDTA (Gibco, Invitrogen) dissociation. Cells were strained using 40‐μm cell strainer (BD Falcon) and resuspended in Dulbecco's modified Eagle's medium (Gibco, Invitrogen), supplemented with 10% fetal bovine serum, 100 μg/ml penicillin, and 100 g/ml streptomycin. Cells were plated on poly‐D‐lysine‐coated cell culture dishes. All mouse experiments were performed under approved protocols of the Institutional Animal Care and Use Committee at Johns Hopkins University School of Medicine.
Note added in proof
In agreement with our conclusions, Okreglak and Walter (2014) independently demonstrated that yeast Msp1 promote the degradation of the mistargeted Pex15 tail‐anchored protein on the mitochondrial outer membrane.
Okreglak V, Walter P (2014) The conserved AAA‐ATPase Msp1 confers organelle specificity to tail‐anchored proteins. Proc Natl Acad Sci U S A, in press
Y.‐C.C. and J.R. proposed project hypothesis, designed experiments and prepared manuscript. Y.‐C.C. performed and analyzed experiments of Figures 1, 2, 4, 5, 6 and [Link], [Link], [Link], [Link], [Link], [Link], [Link]. T.M.D., V.L.D., G.K.E.U. and S.A.A. designed, performed and analyzed experiments of Figure 3. N.D. and S.G. performed mass spectrometry and analyzed the result.
Conflict of interest
The authors declare that they have no conflict of interest.
Supplementary Figure S1
Supplementary Figure S2
Supplementary Figure S3
Supplementary Figure S4
Supplementary Figure S5
Supplementary Figure S6
Supplementary Figure S7
Supplementary Table S1
Supplementary Table S2
We thank members of the Rutter laboratory for helpful discussions; Tim Formosa, David Stillman, and Janet Shaw for yeast strains, reagents, antibodies, technical supports, and helpful discussions; Tammy Nguyen for assistance with mammalian live cell imaging; Adam Frost for discussing the MITO‐MAP data and critical comments on the manuscript; Adam L. Hughes for critical comments on the manuscript; Dennis Winge for Sdh1 and Sdh2 antibodies. This work was supported by the NIH grant R01GM094232 (to J.R.) and the NIH/NIDA DA000266 (to T.M.D., V.L.D.) and NIH/NINDS 5R01AG029368 (to V.L.D.) American Heart Association 12SDG9310031 (S.A.A.). T.M.D. is the Leonard and Madlyn Abramson Professor in Neurodegenerative Diseases.
- © 2014 The Authors