We set out to annotate the minimal complement of genes encoding putative human E3s, based on the presence of signature “catalytic” domains, as well as of domains characteristic of substrate-recognition subunits of multi-subunit RNF-dependent E3s. In addition, we annotated the genes encoding A20 finger proteins and MALT1/paracaspase, which have recently been found to act as E3s [13], [14]. In contrast to two Ub E1 and the estimated fewer than 40 Ub E2 genes (our unpublished observations and [1]) and consistently with the role of E3s in conferring substrate specificity to ubiquitylation, we uncovered ∼617 genes encoding putative Ub E3s after searching a human proteome superset and annotating the respective encoding loci. The number of putative E3 genes is thus greater than the number of human genes for protein kinases (518 genes; Tables 1 and S1). We also found that ∼95% of all E3s are RNF-dependent, with RNF domain-encoding genes and multisubunit RNF-dependent E3s represented in similar numbers (309 and 270 genes, respectively). In S. cerevisiae, ∼80 genes were found to encode putative E3s –of these, 13 are essential for viability in rich medium, which mirrors the proportion of all yeast genes estimated to be essential under these conditions (17%; Tables 1– 3). Thus, a similar fraction of all yeast and human genes are dedicated to encode E3s (1–2%).

All but 75 human RNF proteins exhibit at least one other domain that could be identified using publicly available protein domain search databases. The 56 or more types of domains that we found associated with the remaining RNF E3s (Table 4) could play a variety of roles, including mediating interactions with substrate degradation signals (degrons). This variety of domains is consistent both with the diversity of proteins targeted for ubiquitylation and with the diversity of mechanisms involved in their recognition. Domain composition and protein homology have been previously utilized to define subfamilies of RNF proteins. The largest subfamily, TRIM/RBCC, has ∼76 representatives in humans (but none in S. cerevisiae) and is characterized by the presence of a B box and a coiled coil ([15] and Table 5). The second largest human RNF subfamily is the RBR/TRIAD [16], with at least 14 members exhibiting two RNFs and an in-between RING (IBR) domain (e.g., Parkin). Interestingly, many RNF E3s are associated with other zinc-binding domains –particularly zinc fingers of the C2H2, C3H1, UBR1 and RBZ types (Table 4). Alignment of the RNF domain alone did not reveal any new subfamilies of RNF proteins (unpublished observations).

The discovery of new mechanisms of cellular regulation involving the Ub-proteasome system (UPS) could benefit from the development of functional genomics tools that allow its specific manipulation, such as by interfering with E3 levels. With this in mind, we hit-picked cDNA and plasmid-encoded RNAi (shDNA) expression constructs representing nearly the entire human and mouse E3 families from publicly available genome-wide collections, and individually arrayed these constructs in 384-well plates for functional screens [17].

In one such screen, we set out to identify novel E3s that regulate mitochondrial dynamics. Mitochondria divide, fuse and move around cells [18]–[20], processes whose relevance is underscored by the observation that mutations in the cognate proteins can lead to disease, in particular neuronal dysfunction [19]. Mitochondria utilize the cytoskeleton as tracks for directional movement –in mammalian cells, they are distributed mainly along microtubules, with kinesins (KIF) 1B and 5B having been identified as the anterograde (plus end) trafficking motors [21]. A phenotype that is characteristic of defective mitochondrial anterograde trafficking is the perinuclear clustering of the organelle, which we observed following transfection of HeLa cells with siRNA against KIF5B (a gift of Dr. Jiang Wei, Burnham Institute; [22]); in contrast, mitochondrial trafficking was not impaired in cells transfected with siRNA to KIF11, which has not been implicated in mitochondrial trafficking (Fig. 1A).

Having validated a morphology-based imaging assay for mitochondrial dynamics, we next screened our E3 collections by transfecting HeLa cells together with a plasmid encoding a mitochondrial-targeted red-fluorescent protein (MT-RFP, consisting of RFP fused to the presequence of human cytochrome c oxidase subunit VIII). This allowed visualization of mitochondria specifically in E3-transfected cells. Following hit-picking and reconfirmation of the results from the primary screens, we uncovered MULAN, a putative E3 whose ectopic expression led to defects in both mitochondrial morphology and subcellular distribution (Fig. 1A). The phenotype of HeLa cells ectopically expressing MULAN is characterized by the fragmentation of otherwise long, tubular-shaped mitochondria (Figs. 1A and 1B). In addition, as in the KIF5B knockdown, mitochondria in MULAN-transfected cells clustered around the perinuclear region. These phenotypes were MULAN dose-dependent (Fig. S1) and were observed with all cell lines examined: HeLa, 293, NIH3T3, COS7, H9c2 and HL1 (unpublished observations). Notably, the phenotypes resulting from MULAN ectopic expression were manifested rapidly, within 24 h of transfection.

The effects of MULAN on mitochondrial dynamics could be indirect, e.g., as a result of the disruption of the microtubule network. To examine this possibility, HeLa cells were co-transfected with MT-RFP and either MULAN or the vector control, and microtubules were visualized by staining with α-tubulin antibody. The results showed no obvious differences in the microtubule network between vector- and MULAN cDNA-transfected cells (Fig. 1B). Thus, the altered mitochondrial distribution in cells ectopically expressing MULAN was apparently not due to microtubule disruption. However, these results do not rule out the possibility that the E3 exerts more subtle effects on microtubules that could not be detected with these analyses.

MULAN is a 40-kDa protein with orthologs identified from flies to vertebrates, and in plants. As MarchV, it does not have an obvious ortholog in yeast. Reciprocally, Mfb1p and Mdm30p, budding yeast F-box E3 subunits that are essential for mitochondrial dynamics, are apparently not conserved in higher eukaryotes [6]. One possibility to explain the lack of conservation of such critical regulators of what is thought to be an essential process for both yeast and human cells is that there are important distinctions between the molecular mechanisms underlying mitochondrial dynamics in these organisms. Of relevance is the observation that mitochondrial trafficking in yeast largely relies on the actin cytoskeleton network, while in mammalian cells mitochondria are transported predominantly along microtubules [18], [21]. Moreover, E3 regulators of mitochondrial dynamics could function as mediators of the response to organism-specific signaling, as is the case of organellar trafficking along long motor neuron axons, a process regulated both by synaptic activity and by secreted molecules such as NGF and TNF (e.g., [21]).

MULAN exhibits two predicted transmembrane domains (TMDs) as well as an evolutionarily conserved C-terminal RNF domain (Fig. 1C and Fig. S1A, B). To determine whether MULAN is a bona fide E3, a GST fusion with its RNF was tested for the ability to mediate ubiquitylation in reactions utilizing the promiscuous E2, Ubc4. The RNF from MULAN, under these conditions, was nearly as potent for ubiquitylation as that of c-Cbl (Fig. 1D). As is often the case in this type of assay, GST blots revealed that reaction products consisted of, at least in part, the auto-ubiquitylated GST-RING finger protein itself, and their formation required E1, E2, E3 and Ub (unpublished observations). We also examined whether the integrity of the MULAN RNF domain was required for E3 activity. C339, one of the Cys residues that are essential for maintaining the RNF structure through zinc coordination, was mutated to Ala, resulting in abolished E3 activity (Fig. 1D). This result led us to ask whether the mitochondrial morphology phenotypes associated with expression of MULAN were also dependent on an intact RNF. NIH3T3 cells transfected with MULAN wild-type or with RNF mutant constructs were analyzed for mitochondrial distribution (Figs. 1E and S1D). While mitochondria in ca. 52% of wild type MULAN-transfected cells presented a trafficking defect, this number was markedly reduced in cells expressing either the RNF-deleted E3 (1-301; 1%), or the C339A mutant (3%) (Fig. 1E). These mutants were expressed at similar or higher levels than the wild-type (Figs. S1D and S2A) and their lack of effect on mitochondrial dynamics was even more evident when normalized to the wild-type expression level (Fig. S1D). Thus, the simplest interpretation of these results is that regulation of mitochondrial dynamics by MULAN is dependent on its E3 activity. The results also suggest that the mitochondrial phenotype that we observe due to MULAN expression is not just a non-specific effect due to increased expression of any protein on the mitochondrial surface.

To confirm that endogenous MULAN plays a role in mitochondrial dynamics, we generated two different siRNA, siMULAN1 and siMULAN2, which led to greater than 70% reduction of the endogenous mRNA level two days after transfection of HeLa cells (Fig. 2A; knockdown was also observed at the protein level, Fig. S2C). The phenotype of MULAN knocked-down cells exhibited similarities to that of cells ectopically expressing the E3, which is not unprecedented (e.g., [23], [24]) and could be due to a dominant negative effect of the ectopically expressed protein (Fig. 2 and Fig. S3). Mitochondria in 80–90% of MULAN siRNA-transfected cells were clustered around the nucleus, although clustering was less compact compared to that seen in response to ectopic MULAN expression; in contrast, only 5–12% of cells transfected with the control non-silencing siRNA exhibited this phenotype (siScrambled; Figs. 2A and 2C). Immunostaining with anti-α-tubulin antibody in cells knocked down for MULAN did not reveal an obvious disruption of the microtubule network (Fig. 2B). The fact that the knockdown phenotype was specific to siRNA against MULAN and that it was observed with two different siRNA sequences argues against an off-target effect and confirms a role for MULAN in the regulation of mitochondrial dynamics.

With regard to mitochondrial morphology, the phenotype of MULAN-knockdown cells was not as evident as for cells ectopically expressing the E3 (Fig. 2C and Fig. S3). In the latter, mitochondrial clustering is accompanied by a high degree of fragmentation and by the collapse of the mitochondrial network, leaving little discernible tubular structure remaining; in MULAN-knockdown cells, tubular-shaped mitochondria remain visible and the increase in fragmentation relative to control cells is not as marked. These results suggest that the increased mitochondrial fragmentation in response to MULAN ectopic expression could represent either a gain-of-function phenotype or a secondary consequence of severely disrupted mitochondrial trafficking.

To begin elucidating the mechanism by which MULAN affects mitochondrial dynamics, we determined its subcellular localization. Earlier results had already revealed that the protein's distribution was reminiscent of mitochondrial staining patterns. Several lines of evidence indeed indicate that MULAN is a mitochondrial protein: (i) the signal of C-terminal Flag-tagged MULAN completely overlapped with that of co-transfected MT-RFP (Fig. 3A, upper panels). The untagged MULAN 1-301 protein, whose levels were high enough to be detected by immunocytochemistry using anti-MULAN antibody, also colocalized with MT-GFP (Fig. 3A, lower panels). Thus, both an untagged and the C-terminal tagged E3 co-localized with mitochondria. The specificity of MULAN localization is also revealed by the observation that neither protein colocalized with the Golgi marker lectin-Alexa 488 (Fig. S1E); (ii) lysates of 293 cells overexpressing MULAN-Flag were subjected to centrifugation in sucrose gradient. Gradient fractions were run in SDS-PAGE and analyzed by western-blot with different antibodies, revealing that MULAN's sedimentation was similar to that of the mitochondrial protein Tom20 but not of the Golgi marker, Golgin 97 (Fig. 3B, left panel); (iii) detergent-free lysates of untransfected 293F cells were first fractionated by differential centrifugation. Mitochondria in the heavy membrane (HM) fraction were further purified by sucrose gradient centrifugation. Samples of the whole cell extract, “mitochondrial” fraction, and cytosolic and light membrane (LM) fraction were normalized for protein amount and analyzed by western-blot using antibodies against cytochrome c, Golgin 97, calreticulin/Sec61, tubulin or EEA1 (markers for mitochondria, Golgi, ER, cytosol and endosomes, respectively; Fig. 3B, right panel). Consistently with the previous observations, the results show that endogenous MULAN was abundant in the fraction enriched for mitochondria, but was not detected in the fraction enriched for Golgi, ER, cytosol and endosomes. Although some Golgi material contaminated the “mitochondrial” fraction, immunocytochemistry analyses also revealed that MULAN-Flag did not detectably colocalize with lectin in intact cells (Fig. S1E). In the case of the ER, there was also contamination of the mitochondrial fraction with soluble ER luminal material (calreticulin) and the ER signal in immunocytochemistry was too disperse to allow complete resolution from mitochondria (Fig. S1F). However, we do not expect a significant fraction of MULAN to be localized to the ER, since this organelle's membrane-resident marker Sec61 was not detected in the purified mitochondrial fraction enriched for MULAN, and MULAN was not detected in the LM fraction enriched for the ER (Fig. 1B); (vi) lastly, immuno-electron microscopy analyses demonstrated that the MULAN signal concentrated specifically around the circumference of mitochondria (Figs. 3C and S1). The finding that MULAN localizes to mitochondria is interesting because only one other transmembrane E3 has been shown to reside in this organelle [9], [11]. These results suggest that MULAN's effects on mitochondrial dynamics are direct.

We then set out to determine the localization and topology of MULAN within mitochondria. To confirm that MULAN's predicted TM domains are functional, we generated MULAN-Flag mutants with deletions of sequences encompassing the TMD1 (amino acids 9-29), TMD2 (amino acids 242-259), or both. Fig. 3D shows that deletion of either TMD partially affected the co-localization of MULAN with MT-RFP, and that deletion of both TMDs completely mistargeted the protein. These results are consistent with the prediction that MULAN is a transmembrane protein, but do not distinguish whether it sits at the outer or the inner mitochondrial membranes. Access to other components of the Ub system would require MULAN to be localized to the MOM with the RNF domain facing the cytosolic side. To test this possibility, we performed a limited protease susceptibility experiment with purified mitochondria. 293 cells expressing MULAN-Flag were homogenized in isotonic sucrose buffer and intact mitochondria were isolated by sucrose gradient purification. Mitochondria were incubated with or without trypsin for various times and used for western blots to detect MULAN-Flag, the intermembrane space protein Smac, and the MOM protein Tom20. The results (Fig. 3E) revealed that MULAN's C-terminal Flag tag was readily susceptible to trypsin digestion, its signal disappearing within 5 min of treatment. Similar results were obtained with Tom20, while Smac remained trypsin-resistant unless mitochondria were permeabilized with 0.5% Triton X-100 prior to addition of the protease (Fig. 3E, lower panel). Together, the results presented so far suggest that MULAN is MOM-anchored through two transmembrane domains, exposing a short N-terminus and the RNF-containing C-terminus to the cytosol. This model predicts that its RNF should have access to the remaining components of the Ub system and thus function as a bona fide E3 in vivo (Fig. 3F).

We next investigated the mechanisms that target MULAN to mitochondria and found that, as for certain other MOM proteins, transmembrane sequences with moderate hydrophobicity and net positive charge in flanking regions, known as signal-anchor or TMD+ domains are involved ([25] and Figs. S4 and S5, Text S1). For example, mutation of two positively charged residues C-terminal to TMD2 (R260A/K261A) led to MULAN's targeting to the endoplasmic reticulum (Fig. S4C); MULAN's N-terminus also seemed to influence proper targeting, since an N-terminal Flag tag led to the protein's cytosolic localization (Fig. S4D). To determine whether modulation of mitochondrial dynamics by MULAN requires its proper localization to the organelle, the TMD2-deleted mutant, the R260A/K261A mutant, as well as N-terminal Flag tagged protein were evaluated for the ability to promote perinuclear clustering of mitochondria. The results in Fig. 4 show that mistargeted MULAN mutants were defective in eliciting perinuclear clustering, despite that all were expressed at levels comparable or higher to the wild-type (Fig. S2B). Likewise, MULAN's isolated cytoplasmic domain (amino acids 260-352) had no effect on mitochondrial distribution or morphology (unpublished observations). Interestingly, although the TMD2-deleted mutant (Δ242-259) still remained partially localized to mitochondria, it was completely defective in promoting perinuclear clustering of the organelle, perhaps suggesting that an intact MULAN structure is required for interaction with other components of its pathway. The results above are consistent with a model in which MULAN's regulation of mitochondrial dynamics involves signaling effectors that are either integral to, or can peripherally associate with, mitochondria.

To determine MULAN's expression pattern, we used MULAN cDNA to probe a Northern blot of RNA from several human tissues. The result revealed a single, 2.4-Kb mRNA in most human tissues tested, with highest levels in the heart (Fig. 5A). Consistently, semi-quantitative RT-PCR analyses revealed that MULAN is also expressed at high levels in isolated mouse and rat primary cardiomyoctyes (Fig. 5B, left panels) as well as in the adult mouse cardiomyocyte cell line HL-1 when compared to NIH3T3 cells or to the rat skeletal muscle-like cell line H9c2 (Fig. 5B, right panels).

To identify putative E3-encoding genes, the human proteome deduced from a superset combining 5 databases [RefSeq peptides, predicted transcripts of the human genome assembly #36, Ensembl release 39, transcripts and corresponding proteins predicted by Celera (release R27), and a 2006 version of UniProt] was searched with Hidden Markov Models designed for the individual domains, using HMMER 2.3.2 (based on the Interpro v.13.1, including but not limited to Pfam database v.20). The related PHD, PIAS and LIM domains were excluded from the analyses. Iterative BLAST searches (v.2.2.12) of the UniGene, RefSeq and Ensembl were used to compile a complete list of “synonyms” (alternative accession numbers), and using database cross-references subsequently unified by non-redundant identifiers –EntrezGene IDs (or a Celera hCG in the absence of an Entrez gene). In some cases, domain assignment or verification involved manual comparison to the predicted domains of the proteins' orthologs listed in Homologene. In a limited set, some manual editing was also used to analyze and combine de novo gene model predictions from genomic sequences (using the Genscan program, Celera and Ensembl gene models) with split predictions into a full-length gene model. Prediction of additional domains in RNF E3s was performed with SMART (http://smart.embl-heidelberg.de/).

Human full length MULAN cDNA (GenBank accession no. NM_024544) constructs in the pME vector were previously described [12]. Mouse full length MULAN cDNA (GenBank accession no. NM_026689) was obtained from the Mammalian Genome Collection. Vectors used to generate tagged MULAN constructs were p3xFlag-CMV-14 (Sigma, Saint Louis, MO), pFlag, EGFP-N3 and EGFP-C2 (Clontech, Mountain View, CA). Point mutations were created using the QuickChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). Internal deletions were generated using the inverse PCR method. All mutations were verified by sequencing.

MULAN polyclonal antibodies were generated by immunizing rabbits with a synthetic peptide corresponding to amino acids 57-76 (conserved between human and mouse), EAPGKCVPYAVIEGAVRSVK. The peptide antibody was further affinity-purified using an antigen column. Other primary antibodies used in this study were: Rabbit polyclonal antibodies against Ub (Z0458, DAKO, Carpinteria, CA), Tom20 (sc-11445, Santa Cruz Biotechnology, Santa Cruz, CA), Smac (Imgenex Corp., San Diego, CA), Sec61 (Dr. Nicholas Gekakis, TSRI) and Calreticulin (405-417, Calbiochem, San Diego, CA); Monoclonal antibodies against Golgin-97 (CDF4, Molecular Probes, Carlsbad, CA), the Flag tag (M2, Sigma), EEA1 (Transduction Laboratories, Lexington, KY), cytochrome-c (7H8.2C12, BD Pharmingen, La Jolla, CA) and alpha-tubulin (DM1A, Sigma). Secondary antibodies used were: Alexa Fluor 488 goat anti-rabbit and goat anti-mouse IgGs (Molecular Probes), Cy3 donkey anti-mouse IgG (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA).

HEK293, NIH3T3 and HeLa cells were maintained in DMEM supplemented with 10% FBS at 37°C in 5% CO2. DNA transfection was performed with lipofectamine 2000 (Invitrogen, Carsbad, CA) or Fugene 6 (Promega, Madison, WI). siRNAs were transfected using lipofectamine 2000.

The E3 cDNA library was arrayed in a 384-well plate format, ready for reverse transfection [17]. For the screen reported, each cDNA was co-transfected with MT-RFP. 24 h after transfection, cells were fixed in 4% formaldehyde and screened for cDNAs that led to obvious changes in mitochondrial morphology under the fluorescence microscope at 20× magnification.

siRNA oligos utilized were: siMULAN1, GAGCGCCUGUGUAGUGUGUUU (Dharmacon, Lafayette, CO); siMULAN2, CCGCGCCUUGCCAGAACCCAA (Qiagen, Valencia, CA); siScrambled (Qiagen # 1022076); and siCyclophilinB (Dharmacon). HeLa cells were transfected with siRNA on day 1 (at a final concentration of 50–100 nM in the culture media) followed by MT-RFP on day 2. Cells were harvested on day 3 either to collect RNA, or for fixation with 4% formaldehyde. Fixed cells were used to score for defective trafficking based on the perinuclear clustering of mitochondria, or were subjected to immunocytochemistry as described below.

For immunocytochemistry, cells were plated out on 35-mm glass bottom dishes (MatTek Corp., Ashland, MA). Cells were fixed with 4% paraformaldehyde in phosphate-buffered saline (PBS) for 15 min, washed with PBS, permeabilized with 0.2% Triton X-100 for 5 min, washed four times with PBS and blocked with 3% bovine serum albumin, all at room temperature. Cells were then incubated with primary antibodies for 2 h at room temperature, washed three times with 0.2% Triton X-100, incubated with secondary antibodies for 30 min and washed again. Samples were mounted using Prolong Antifade (Invitrogen) and analyzed by confocal microscopy using an Olympus Fluoview 500 laser scanning confocal on an Olympus IX61 upright microscope. GFP was imaged with the 488-nm line of the Argon laser, and the emission filter was a 505–525 bandpass filter. RFP was imaged with 543 nm laser line from a HeNe green laser, and the emission filter was a 560–600 bandpass filter.

Immuno-gold electron microscopy was performed at the University of California, San Diego, EM core. COS7 cells expressing MULAN-Flag were fixed in PBS containing 2% paraformaldehyde and 0.2% glutaraldehyde. Fixed cells were washed with 0.15 M glycine/PBS, embedded in 10% gelatin/PBS and infused overnight with 2.3 M sucrose/PBS at 4°C. 1-mm³ cell blocks were mounted onto specimen holders and snap frozen in liquid nitrogen. Ultracryomicrotomy was carried out at −100°C on a Leica Ultracut UCT with EM FCS cryoattachment (Leica, Bannockburn, IL) using a Diatome diamond knife (Diatome US, Fort Washington, PA). 60–70 nm frozen sections were picked up with a 1∶1 mixture of 2.3 M sucrose and 2% methyl cellulose and transferred onto Formvar and carbon-coated copper grids. Immunolabeling was performed by a slight modification of the “Tokuyasu technique.” Briefly, grids were placed on 2% gelatin at 37°C for 20 min, rinsed with 0.15% glycine/PBS and the sections were blocked using 1% cold water fish-skin gelatin. Primary antibody against Flag was diluted 1/100. Incubation with primary antibody for 1 h at room temperature was followed by 5-nm gold-conjugated goat anti-mouse IgG and IgM (RPN 430, Amersham Pharmacia Biotech, Piscataway, NJ) and 10-nm gold-conjugated goat anti-rabbit IgG (RPN 421), both diluted 1/25 in 1% BSA/PBS at room temperature for 30 min. Grids were viewed and photographed using a JEOL 1200EX II transmission electron microscope (JEOL, Peabody, MA).

For gradient centrifugation, HEK293 cells expressing MULAN-Flag were harvested from a 10-cm dish, washed in PBS and collected by centrifugation at 600×g for 5 min. Cells were then washed with HE buffer (10 mM Hepes-KOH, pH 7.5, and 1 mM EDTA) containing 10% (wt/vol) sucrose. Cells were suspended in 1 ml of the HE buffer with 20 µg/ml α2-macroglobulin and protease inhibitor cocktail. Cells were passed through a 27 gauge needle five times for homogenization and centrifuged at 600×g for 10 min to obtain a post-nuclear supernatant. The supernatant was layered over a discontinuous gradient of 40% and 60% sucrose in HE buffer (6.6 and 2.2 ml, respectively). The gradient was centrifuged at 100,000×g for 3 h. 1ml aliquots were collected, 100 µl of each fraction was concentrated using microcon columns (Millipore, Billerica, MA) and subjected to SDS-PAGE. Immunoblotting was performed using anti-Golgin 97, anti-Tom20 and anti-Flag antibodies. Highly purified mitochondria were isolated using methods described by Rezaul and colleagues [30]. Cells were harvested by centrifugation at 1000 rpm for 5 min at room temperature, then resuspended in solution A (0.25 M sucrose, 20 mM HEPES-KOH, pH 7.5, 10 mM KCl, 1.5 mM MgCl₂, 1 mM EDTA, 1 mM EGTA, 1 mM DTT) with protease inhibitor cocktail (Sigma). The cellular suspension was homogenized with a glass Dounce homogenizer with 40 times up and down passes of the pestle. The homogenate was centrifuged at 750×g for 15 min. The supernatant was centrifuged at 12,500×g for 25 min to collect crude mitochondria. The pellet was resuspended in solution B (0.25 M sucrose, 1 mM EDTA, 10 mM Tris-HCl pH 7.4) then layered on a 1 M/1.5 M discontinuous sucrose gradient and centrifuged at 60,000×g for 40 min. Mitochondria were collected from the interface and washed with at least 5 volumes of solution A for downstream applications. For trypsin digest, 50–100 µg of highly purified mitochondria were incubated with or without 25 µg/ml of trypsin protease in 50 µl total volume at 37°C for 5 sec, 5 min and 15 min, with or without 0.5% Triton-X 100. Reactions were stopped with trypsin inhibitor and run on SDS-PAGE to probe with anti-Flag, anti-Tom20 and anti-SMAC antibodies.

In vitro ubiquitylation was performed as described [31], with His-Ubc4 and MULAN GST fusion proteins. GST fusion proteins were expressed and isolated as described [31].

RNA was isolated using the RNeasy kit (Qiagen). Semi quantitative RT-PCR was performed using the Superscript II reverse transcriptase (Invitrogen). TaqMan qRT-PCR was performed using the one-step Superscript III platinum reagent (Invitrogen). Samples were run in triplicate as multiplexed reactions with a normalizing internal control (36B4; probe and primer were gifts of Dr. E. Saez, TSRI). MULAN probe and primer were ordered from Applied Biosystems (Foster City, CA). Northern blot was performed with a commercial 12-lane multiple human tissue blot (Clontech) using full length human MULAN cDNA to generate the probe.