Research ArticleCell Biology

ROMO1 Is an Essential Redox-Dependent Regulator of Mitochondrial Dynamics

See allHide authors and affiliations

Sci. Signal.  28 Jan 2014:
Vol. 7, Issue 310, pp. ra10
DOI: 10.1126/scisignal.2004374


The dynamics of mitochondria undergoing fusion and fragmentation govern many mitochondrial functions, including the regulation of cell survival. Although the machinery that catalyzes fusion and fragmentation has been well described, less is known about the signaling components that regulate these phenomena. We performed a genome-wide RNA interference (RNAi) screen and identified reactive oxygen species modulator 1 (ROMO1) as a redox-regulated protein required for mitochondrial fusion and normal cristae morphology. We showed that oxidative stress promoted the formation of high–molecular weight ROMO1 complexes and that knockdown of ROMO1 promoted mitochondrial fission. ROMO1 was essential for the oligomerization of the inner membrane guanosine triphosphatase (GTPase) OPA1, which is required to maintain the integrity of cristae junctions. As a consequence, cells lacking ROMO1 displayed fragmented mitochondria and loss of cristae, causing impaired mitochondrial respiration and increased sensitivity to cell death stimuli. Together, our data identify ROMO1 as a critical molecular switch that couples metabolic stress and mitochondrial morphology, linking mitochondrial fusion to cell survival.


Mitochondria form a dynamic reticulum that is sculpted by cycles of ongoing fusion and fission events of individual organelles, a phenomenon known as mitochondrial dynamics (13). Several mitochondrial shaping proteins that catalyze the opposing processes of membrane fusion and fission have been identified. In mammals, these include MFN1 and MFN2 (Fzo1 in yeast), as well as OPA1 (Mgm1 in yeast) (4), which are required during fusion for the integration of outer and inner mitochondrial membranes, respectively, and DRP1 (Dnm1 in yeast), which catalyzes fragmentation after recruitment to mitochondria by the mitochondrial fission factor (MFF) (5, 6). DRP1 is required for cell death during development (7), as well as for focal delivery of energy at neuromuscular junctions (8), and is related to neurodegeneration and ischemia/reperfusion in the heart (912). Mutations in MFN2 underlie Charcot-Marie-Tooth 2a (13), and OPA1 mutations are associated with autosomal dominant optic atrophy (14); down-regulation of MFN2 also occurs in obesity-related insulin resistance associated with type 2 diabetes (15), highlighting the pathological consequences of dysfunctional mitochondrial dynamics (16). During programmed cell death, the mitochondrial network fragments to facilitate opening of junctions of inner mitochondrial membrane cristae and the consequent metabolic catastrophe (3, 1721). In contrast, mitochondrial fusion after starvation or stress is thought to attenuate mitophagy and/or apoptosis (2225). Regulated interactions between MFN2 and BAX and BAK, two proapoptotic members of the Bcl2 family, appear to govern the earliest events of the apoptotic cascade, demonstrating an intimate relationship between mitochondrial morphology, mitochondrial integrity, and cell survival (2631). However, the signals and protein effectors that control the activities of the shaping proteins, and how mitochondrial dynamics is integrated into cell death signaling cascades, are still unclear (32, 33).


Genome-wide RNA interference screen identifies ROMO1 as a regulator of mitochondrial fusion

To identify regulators of mitochondrial morphology, we performed a genome-wide RNA interference (RNAi) imaging screen in HeLa cells (fig. S1A), using positive control small interfering RNAs (siRNAs) targeting DRP1 to induce elongation or a pool of siRNA targeting MFN1 and MFN2 to induce fragmentation (Fig. 1A). We applied an algorithm that calculates the length/width or the aspect ratio of mitochondria in screen images and reproducibly distinguished between control, fused, and fragmented mitochondria (Fig. 1B). Many regulators of mitochondrial dynamics, including DRP1, OPA1, MFF, and YMEL1, were identified by the screen (Fig. 1C). We considered siRNA pools that generated robust z scores within the range observed for the internal siRNA controls for further validation. Three of four siRNAs targeting reactive oxygen species (ROS) modulator 1 (ROMO1) produced robust fragmentation (Fig. 1, D and E). ROMO1 is a 79–amino acid mitochondrial transmembrane protein that promotes mitochondrial ROS production (3437), proliferation (38, 39), and senescence (35).

Fig. 1 Genome-wide RNAi imaging screen identifies ROMO1 as a novel regulator of mitochondrial fusion.

(A) Representative fields of control, fused (siDRP1), or fragmented phenotypes (siMFN1/2). Regions highlighted in orange were rejected by the algorithm to avoid perinuclear condensation artifacts. Images are representative of three independent experiments and 50 cells per condition. Scale bar, 20 μm. (B) Scatter plot of length/width (aspect) ratio shows robust separation of cells with control, fused (DRP1 siRNA), and fragmented (MFN1/2 siRNA) mitochondria. (C) Table showing the robust z score, the distance of the value from the median in SD units, of ROMO1 and known regulators of mitochondrial dynamics in the RNAi screen. (D) Photomicrographs of HeLa cells transfected with the indicated siRNAs. Scale bar, 20 μm (inset at the bottom, 5 μm). (E) Histogram showing quantitation of aspect ratio for the individual ROMO1 siRNAs used as a pool in (D) and for knockdown of MFN1, MFN2, or DRP1. Error bars represent SEM of three independent experiments and 50 cells per condition. *P < 0.05, ***P < 0.001, two-tailed unpaired t test. (F) Rescue of ROMO1 knockdown phenotype with siRNA-resistant ROMO1 wild-type (WT) cDNA, but not with the nonfunctional ROMO1-FFAA mutant (in which Phe67 and Phe70 were mutated to Ala) and vector control. (G) Histogram showing the aspect ratio for cells expressing empty vector or siRNA-resistant WT and FFAA constructs in control and ROMO1 knockdown cells shown in (F). *P < 0.05, **P < 0.01. n.s., not significant. n = 3 independent experiments and 50 cells per condition. (H) In vitro mitochondrial fusion assay showing luciferase values generated using mitochondria from cells transfected with control or ROMO1 siRNA or cells treated with CCCP. n = 3 independent biological replicates. ***P < 0.001.

To rule out off-target siRNA effects, we introduced an siRNA-resistant ROMO1 complementary DNA (cDNA) into U2OS cells using lentivirus, which restored the normal network aspect ratio in ROMO1 knockdown cells (Fig. 1, F and G, and fig. S1B), and validated knockdown at the mRNA level by quantitative polymerase chain reaction (fig. S1C). We next generated a mutant ROMO1 cDNA (ROMO1-FFAA) carrying two Ala substitutions at two highly conserved Phe residues, Phe67 and Phe70. The FFAA mutant could not rescue normal morphology in cells lacking ROMO1. Next, we analyzed in vitro fusion rates of mitochondria isolated from control and ROMO1 knockdown cells using a luciferase complementation assay (40). Mitochondria from cells lacking ROMO1 showed a 50% lower fusion rate than those from control cells (Fig. 1H), indicating that loss of ROMO1 results in fragmentation of the mitochondrial network due to a reduction in fusion. Although previous work has suggested that ROMO1 overexpression drives mitochondrial fragmentation (37), we found that cells expressing a ROMO1 construct tagged at the C terminus, but not at the N terminus, display fragmentation, suggestive of a dominant negative effect of C-terminally tagged ROMO1 (fig. S1, D and E). Together, our data demonstrate that ROMO1 is required for mitochondrial fusion.

ROMO1 is a redox sensor that promotes mitochondrial fusion

ROMO1 has been reported to promote the generation of ROS, and we found that superoxide concentrations increased twofold compared to control cells as measured by dihydroethidium (DHE) staining (Fig. 2A). Treatment of control cells with the complex III inhibitor antimycin A, which reduces efficiency of electron flux through the electron transport chain (ETC), led to a similar increase in ROS concentrations. Treatment of ROMO1 knockdown cells with antimycin A resulted in an additive 4.5-fold increase in superoxide over untreated control cells. We hypothesized that ROMO1 may couple ROS signaling to mitochondrial fusion and sought to identify potential redox-sensitive domains within ROMO1. Although ROMO1 was previously predicted to have a single membrane-spanning domain (36, 37), our computational analyses comparing sequences of ROMO1 from metazoan and nonmetazoan species revealed two putative transmembrane helices, as well as two highly conserved cysteine residues that could serve as redox exchangers (fig. S2). Consistent with the latter possibility, wild-type FLAG-ROMO1 isolated from cells under steady-state conditions migrated as a single species of ~10 kD on reducing SDS–polyacrylamide gel electrophoresis (SDS-PAGE) gels, but on nonreducing SDS-PAGE, a substantial proportion of FLAG-ROMO1 resolved as a 20-kD species, with several additional oxidized species including some that were >100 kD (Fig. 2B). Treatment with menadione, which induces superoxide formation through redox cycling (41), substantially increased the abundance of the slow-migrating oxidized ROMO1 species (Fig. 2B), which disappeared upon removal of menadione (fig. S3A). Mutation of the highly conserved amino acids Phe67 and Phe70 to Ala (FFAA) appeared to increase the abundance of these higher–molecular weight species; however, these species did not form when the two most highly conserved Cys residues, Cys15 and Cys79, were mutated to Ser (ROMO1-2CS). These data suggest that ROMO1 may function as a redox switch serving to couple redox state in the inner membrane to cristae integrity. Redox switches are sensitive either to the glutathione- or to the thioredoxin-reducing systems (42). To investigate this possibility, we knocked down glutathione reductase (GSR), which substantially increased the abundance of high–molecular weight complexes containing ROMO1 (Fig. 2C). In contrast, knockdown of mitochondrial thioredoxin (TXN2) did not increase the abundance of ROMO1-containing high–molecular weight complexes. Steady-state ROS concentrations (Fig. 2D) and morphology (Fig. 2E and fig. S3B) were rescued in ROMO1 knockdown cells by expression of the non-oxidizable, monomeric ROMO1-2CS construct, but not of the FFAA mutant, which was heavily oxidized. Furthermore, unlike the wild-type and FFAA constructs, ROMO1-2CS induced mild elongation of the mitochondrial network, although it was lower in abundance (Fig. 2F and fig. S3, B and C). Together, these data indicate that ROMO1 is a redox-sensitive factor that forms inactive high–molecular weight complexes in response to oxidative stress, and in its reduced, monomeric form drives mitochondrial fusion.

Fig. 2 ROMO1 is a redox sensor that promotes mitochondrial fusion.

(A) DHE fluorescence emitted from cells transfected with control or ROMO1 siRNA and vector or siRNA-resistant ROMO1. Effect of treatment with the complex III inhibitor antimycin A (Ant A) and the ROS scavenger N-acetyl cysteine (NAC) is shown. Error bars represent SEM of three independent experiments. *P < 0.05, two-tailed unpaired t test. (B) Western blot of FLAG-ROMO1 WT, FFAA, or 2CS mutants resolved on nonreducing (−DTT) and reducing gels (+DTT) after treatment of human embryonic kidney (HEK) 293T cells with vehicle control (H2O) or menadione in the presence of QVD, and voltage-dependent anion channel (VDAC) loading control. n = 3 independent biological replicates. (C) Western blot of FLAG-ROMO1 resolved on nonreducing (−DTT) or reducing (+DTT) from control, GSR, or TNX2 knockdown cells. n = 3 independent biological replicates. shRNA, short hairpin RNA. (D) Histogram of aspect ratio in control and ROMO1 knockdown cells expressing either vector control or siRNA-resistant ROMO1 WT, FFAA, or 2CS mutants. n = 3 independent biological replicates. *P < 0.05, two-tailed unpaired t test. (E) Histogram of aspect ratio in cells expressing vector control or siRNA-resistant ROMO1 WT, FFAA, or 2CS mutants. n = 3 independent biological replicates. **P < 0.01, two-tailed unpaired t test. (F) Histogram of aspect ratio in cells expressing vector control or siRNA-resistant ROMO1 WT, FFAA, or 2CS mutants. n = 3 independent biological replicates. *P < 0.05, two-tailed unpaired t test.

ROMO1 is required for the integrity of mitochondrial cristae junctions

Western blot analysis indicated that the abundance of mitochondrial subcompartment marker proteins was unchanged in ROMO1 knockdown cells, suggesting that ROMO1 abundance does not affect mitochondrial mass or the abundance of outer membrane remodeling machinery (Fig. 3, A and B). Because previous reports have localized ROMO1 to both outer and inner mitochondrial membranes (36), we performed a protease protection assay using purified mitochondria (Fig. 3C). In intact mitochondria, proteinase K treatment reduced the abundance of the outer membrane protein TOMM20, but had no effect on that of ROMO1 or OPA1, which resides in the inner membrane and inner membrane space. In contrast, when proteinase K was added after an osmotic shock step that ruptures the outer mitochondrial membrane, both ROMO1 and OPA1 were degraded. Although residence in the intermembrane space cannot formally be ruled out, when taken together with evidence for two transmembrane α helices in ROMO1, we propose that ROMO1 is an inner mitochondrial membrane protein with its N and C termini facing the intermembrane space.

Fig. 3 ROMO1 is required for the integrity of mitochondrial cristae junctions.

(A) Western blot for the mitochondrial markers apoptosis-inducing factor (AIF), ATP synthase F1 β subunit (F1β), COX1, CORE2, and actin as the loading control. n = 2 independent biological replicates. (B) Top: Western blot for components of the outer membrane dynamics machinery (MFN1, MFN2, and DRP1) with VDAC1 as the loading control in cells transfected with control siRNA or siRNAs against ROMO1, DRP1, or siMFN1 and siMFN2 together. Bottom: Histogram showing quantitation of three independent biological replicates. (C) Western blot of a proteinase K (PK) protection assay on purified U2OS mitochondria before or after osmotic shock (OS) showing submitochondrial localization of FLAG-ROMO1. TOMM20 was used as a marker of the outer membrane, OPA1 as a marker of the inner membrane and inner membrane space (which resolves less efficiently after osmotic shock), and SOD2 as a matrix marker. (D) Electron micrographs of control or ROMO1 knockdown U2OS cells showing altered cristae in cells lacking ROMO1. Magnification, ×1500 (left) and ×12,000 (right). Scale bars are indicated at the bottom left of each image (left = 2 μm; right = 250 nm). (E) Histograms showing number of cristae per mitochondrion pixels in U2OS cells transfected with control or ROMO1 siRNA (upper left panel, ***P < 0.001); the number of cristae per unit length of mitochondria in pixels (n = 100 mitochondria per condition) in U2OS cells transfected with control or ROMO1 siRNA (upper right panel, ***P < 0.001); and the percentage of U2OS cells with cristae with normal, detached (stacks), or abnormal appearance after transfection with control or ROMO1 siRNA (lower panel; *P < 0.05). Statistical significance was determined by two-tailed unpaired t test. n = 2 independent biological replicates and 10 cells. (F) Oxygen consumption in control or ROMO1 knockdown cells under basal conditions or after oligomycin (O), FCCP (F), and antimycin A (A) treatment. ROMO1 rescue is shown in dashed line, empty circles. *P < 0.05, **P < 0.01, two-tailed unpaired t test. n = 3 independent biological replicates. (G) Oxygen consumption in control or ROMO1 knockdown cells in the presence or absence of pyruvate and glutamine (pyr/gln), treated with oligomycin, FCCP, and antimycin A as indicated. *P < 0.05, **P < 0.01, two-tailed unpaired t test. n = 3 independent biological replicates.

Analysis of inner mitochondria ultrastructure showed that ROMO1 knockdown cells had mitochondria with fewer or no cristae (Fig. 3, D and E). A small subset of mitochondria also displayed cristae stacks, suggesting loss of connectivity to the inner membrane, reminiscent of yeast mutants lacking the yeast ortholog of mitofilin, fcj1 (43). In support of these data, two studies have identified the ROMO1 yeast homolog Mgr2 as part of a large protein complex known as mitochondrial inner membrane organizing system (MINOS) or mitochondrial organizing structure (MitOS), which establishes boundaries between cristae and juxtaposed boundary regions of the mitochondrial inner membrane (43, 44). Exogenously expressed ROMO1 and mitofilin, a central component of MINOS, also formed a complex in mammalian cells (fig. S4), suggesting that ROMO1 localizes to cristae junctions (45). However, silencing of ROMO1 did not affect the protein abundance of the core MINOS components mitofilin or CHCHD3, or altered the size of the complex as measured by gel fractionation, indicating loss of ROMO1 does not disrupt the integrity of the MINOS complex (fig. S4, B and C).

Despite these alterations in mitochondrial morphology, adenosine 5′-triphosphate (ATP) concentrations were unchanged under both glycolytic and oxidative conditions, and mitochondria in cells lacking ROMO1 remained positive for MitoTracker Red and thus maintained potential (fig. S5). Oxygen consumption rates under basal conditions or after treatment with the ATP synthase inhibitor oligomycin were also unaffected (Fig. 3F). In contrast, maximal (uncoupled) respiration elicited by FCCP (carbonyl cyanide p-trifluoromethoxyphenylhydrazone) treatment was abrogated in the absence of ROMO1 and was restored by reintroduction of siRNA-resistant ROMO1 (Fig. 3F) or by adding pyruvate and glutamine to the assay medium (Fig. 3G). Together, these data demonstrate that the cristae defects in ROMO1 knockdown cells do not interfere with normal function of the trichloroacetic acid (TCA) cycle or with ETC capacity but do reduce the maximal capacity of the respiratory chain.

ROMO1 is required for balance in OPA1 isoform abundance

Because ROMO1 localizes to the inner mitochondrial membrane and its silencing promotes fragmentation, we explored the effect of loss of ROMO1 on OPA1, the inner membrane fusion guanosine triphosphatase (GTPase) that promotes cristae junction integrity. OPA1 is found as multiple isoforms within the cell, denoted a to e (as in Fig. 4B). Long isoforms a and b are integral to the inner membrane, and proteolytic processing of these produces soluble short isoforms c to e that reside in the intermembrane space. Together, long and short isoforms of OPA1 are required for mitochondrial fusion (4648). Both long and short isoforms of OPA1 immunoprecipitated with wild-type FLAG-ROMO1 or the FFAA or 2CS mutants (Fig. 4A and fig. S6A). We noted that silencing ROMO1 led to an imbalance in OPA1 isoform abundance that favored the accumulation of isoform c (Fig. 4B), which has been previously identified as a cleavage product of a transmembrane form of OPA1 and denoted as isoform 1 (49). Although the balance in normal isoform abundance was restored by reintroduction of siRNA-resistant wild-type FLAG-ROMO1, expression of FLAG-ROMO1-FFAA caused the accumulation of long isoforms, suggesting that this mutant had a gain-of-function effect (fig. S6B). In addition, isoform c abundance was increased in wild-type mouse embryonic fibroblasts (MEFs) in which ROMO1 was silenced (Fig. 4C). Silencing ROMO1 in OPA1–/– MEFs expressing the OPA1 long isoform OPA1v1 led to the preferential accumulation of the processed form of OPA1, similar to what we observed with the endogenous protein in U2OS cells (Fig. 4D). We attribute these effects to enhanced processing of OPA1 long isoform in cells lacking ROMO1.

Fig. 4 Loss of ROMO1 prevents OPA1 oligomerization and increases sensitivity to cell death.

(A) Western blot showing coimmunoprecipitation of FLAG-ROMO1 with OPA1. Representative of three independent biological replicates. (B) Western blot showing relative abundance of OPA1 isoforms in control and ROMO1 knockdown cells with VDAC1 loading control. The vertical axis of the bottom OPA1 blot was stretched to enable visualization of the OPA1 isoforms (denoted a to e). Quantitation of isoform band intensity is shown at right. n = 3 independent biological replicates. *P < 0.05. (C) Western blot showing the effect of ROMO1 knockdown on OPA1 isoform balance in WT MEFs. (D) Top: Western blot showing the amount of OPA1 long isoform v1 and its proteolytic cleavage product isoform c in OPA1–/– MEFs stably reconstituted with OPA1v1 compared to vector control (Vec). Bottom: Quantitation of long and short isoform band intensity in control and ROMO1 knockdown cells. n = 3 biological replicates. *P < 0.05. (E) Left: Short and long exposures of Western blots showing formation of OPA1 oligomers and VDAC loading control after treatment of control or ROMO1 knockdown U2OS cells with the cross-linker BMH. Right: Histogram showing percentage of total OPA1 band intensity present in oligomers. n = 3 biological replicates. ***P < 0.001. (F) Histogram showing the effect of ROMO1 knockdown on cytochrome c release after infection with adeno-tBID. Knockdown of OPA1 shown as control. n = 3 biological replicates. **P < 0.01, ***P < 0.001. (G) Histogram showing the effect of increasing dose of etoposide on apoptosis in U2OS cells as quantified by condensed, apoptotic nuclear morphology in control, ROMO1, or DRP1 knockdown cells. Error bars represent SEM of three independent experiments. *P < 0.05, **P < 0.01.

Cells lacking ROMO1 are more sensitive to apoptotic stimuli

These results prompted us to determine whether the alteration in OPA1 isoform balance in cells lacking ROMO1 affected the oligomerization status of OPA1. Disruption of OPA1 oligomers results in opening of cristae junctions and cytochrome c release after translocation of proapoptotic truncated BID (tBID) to mitochondria during the initiation stage of apoptosis (50, 51). Treatment of U2OS cells with a cell-permeable cross-linker BMH (1,6-bismaleimidohexane) resulted in the appearance of high–molecular weight OPA1 oligomers (Fig. 4E). However, the total amount of OPA1 protein present in oligomers was reduced from 35% in control cells to 5% in ROMO1 knockdown cells (Fig. 4E), suggesting that ROMO1 regulates cristae junction dynamics by promoting OPA1 oligomerization. Knockdown of OPA1 enhanced tBID-induced cytochrome c release, which has been previously shown to be due to opening of cristae and fragmentation of the outer membrane (52). Similarly, knockdown of ROMO1 also enhanced tBID-induced cytochrome c release, consistent with the basal reduction in OPA1 oligomerization (Fig. 4F). This sensitization to cytochrome c release in ROMO1 knockdown cells correlated with an increase in cell death in response to treatment with the chemotherapeutic drug etoposide (Fig. 4G). The sensitization to cell death was accompanied by the loss of both monomers and oligomers of OPA1 protein, but not a change in isoform balance (fig. S6, C and D). Together, we conclude that ROMO1 controls the generation of OPA1 isoforms and that OPA1 is a downstream effector of ROMO1 function.


Here, we report a genome-wide screen in mammalian cells to identify regulators of mitochondrial dynamics using an unbiased quantitative analysis of mitochondrial morphology. We focused on the role of ROMO1, silencing of which leads to fragmentation of the mitochondrial network due to a defect in mitochondrial fusion. ROMO1 regulates the generation of ROS within cells, linking it to cellular processes such as proliferation, senescence, and cell death (35, 36, 38, 39). However, the molecular explanation for this has been unclear. We show that loss of ROMO1 increases ROS and that ROMO1 itself forms disulfide bridges and incorporates into higher–molecular weight species in response to oxidative stress, which are counteracted by the glutathione system.

How does reduced ROMO1 regulate mitochondrial morphology? We envision that formation of disulfide bridges within ROMO1 in response to oxidative stress is inhibitory, resulting in a fragmented mitochondrial network allowing for containment of oxidative damage and removal of damaged mitochondria. The reversibility of ROMO1 oxidation states appears critical for its function because the expression of the non-oxidizable ROMO1-2CS mutant drives mitochondrial elongation, suggesting that ROMO1 is active in its monomeric, reduced state. Consistent with this idea, ROMO1-FFAA, which is more oxidized under steady-state conditions, is unable to support mitochondrial fusion. Our computational analysis of ROMO1 orthologs in primordial species provides mechanistic insight as to how ROMO1 could be modulated by oxidative stress. We propose that with two transmembrane helices, ROMO1 would adopt an architecture in which the most highly conserved Cys residues in metazoans would be juxtaposed in the same mitochondrial subcompartment, thus permitting the formation of both homotypic and/or heterotypic Cys-Cys bridges. These Cys residues are absent in the yeast isoform, suggesting that ROMO1 may have obtained the capacity to sense oxidative stress in the intermembrane space or matrix as an additional mode of regulation of OPA1, cristae junctions, and inner membrane fusion in higher organisms.

We provide evidence that ROMO1 is essential for the integrity of mitochondrial cristae, supported both by ultrastructural analysis and functionally by the reduced threshold for cytochrome c release in cells lacking ROMO1. In the absence of ROMO1, mitochondria either lack cristae or have a minor population of detached cristae stacks, reminiscent of cristae in MINOS mutant yeast strains (43). Consistent with this observation, mitochondria from cells without ROMO1 display impaired respiration during uncoupling. We attribute these effects to defective compartmentalization of fuel and of ETC components in cristae invaginations in the absence of ROMO1, resulting in a less efficient ETC with a subsequent increase in ROS production.

Several lines of evidence point to a pathway in which ROMO1 couples mitochondrial ROS to OPA1 oligomerization and cristae junction integrity. First, in the absence of ROMO1, the balance in OPA1 isoform abundance is changed, likely due to altered processing. Perhaps, as a consequence of this disruption in OPA1 processing, the formation or maintenance of oligomers of OPA1 is lost upon ROMO1 silencing. In the absence of ROMO1, OPA1 oligomerization is reduced to that seen in apoptotic cells, and this reduced oligomerization contributes to enhanced sensitivity to apoptotic insults. Second, ROMO1 and OPA1 reside in the inner mitochondrial membrane and form a complex in cells. Third, silencing or ablation of ROMO1 or OPA1 phenocopies each other, with respect to cristae junction integrity and inner membrane fusion (53). Last, loss of both ROMO1 and OPA1 sensitizes cells to cytochrome c release and cell death (20). Together, our work has also provided new insights into how changes in mitochondrial redox state, often examined in the context of cell stress, may lead to the opening of cristae junctions and fragmentation of the mitochondrial network (54, 55). OPA1 may regulate cristae dynamics in response to changes in mitochondrial redox state upon inhibition of the ETC (56), which we propose involves ROMO1. Impaired mitochondrial fission contributes to chemotherapy resistance (57). Like OPA1, ROMO1 is increased in abundance in a subset of human tumors (5861) (, which could confer resistance to apoptosis. We propose that targeting the ROMO1-OPA1 pathway to open mitochondrial cristae in drug-resistant cancer cells may be of therapeutic benefit.


Antibodies and reagents

Antibodies directed against TOMM20 or AIF were from Santa Cruz Biotechnology. Antibodies directed against F1β, CORE2, or COX1 were from MitoSciences. Antibodies directed against FLAG M2 were from Sigma. Antibodies directed against mitofilin, CHCHD3, or Opa1 were from Abcam. Antibodies directed against VDAC1 were from Cell Signaling Technology. Antibodies directed against Opa1 were from BD Transduction Laboratories. Antibodies directed against cytochrome c were from BD Pharmingen. Lipofectamine RNAiMAX, Hoechst, and MitoTracker Red and DHE were all from Invitrogen. Polyethylenimine, CCCP, antimycin A, N-acetyl cysteine, Sepharose 4B, and etoposide were all from Sigma. Oligomycin and FCCP were from Seahorse Bioscience. The caspase inhibitor QVD was from Calbiochem. Halt Protease Inhibitor Single-Use Cocktail was from Thermo Scientific. BMH cross-linker was from Fisher Scientific.

Imaging screen and candidate validation

Pools of four siRNA SMARTpool duplexes targeting 18,255 human genes (Dharmacon) were spotted into 384-well plates at 10 nM and reverse-transfected into 450 HeLa cells (p34 to p36) using 0.025 μl of Lipofectamine RNAiMAX (Invitrogen). Seventy-two hours later, cells were fixed and stained with anti-TOMM20 antibody and Hoechst (1 μg/ml) to visualize mitochondria and cell nuclei. Eight fields per well were acquired with an automatic Cellomics VTI microscope equipped with a 40× objective. Robust z scores were calculated as [x − median (sample)]/[median absolute deviation × 1.4826]. The algorithm to detect mitochondrial network length uses TOMM20 staining signal and negates perinuclear signal to avoid condensation artifacts. The average length/width (aspect ratio) of measurements across multiple fields was used to quantify mitochondrial morphology on a population basis. We applied the MEAN_ObjectEqEllipseLWRCh1 metric within the Morphology BioApplication provided with the Cellomics vHCS Scan image analysis software to determine the aspect ratio of objects stained positive for TOMM20. After algorithm analysis, the efficacy of deconvoluted siRNA pools was then determined for candidates for which images from 10 wells were verified manually to display an elongated or fragmented state. Analysis of three data sets revealed a mean Pearson correlation coefficient of 0.82, demonstrating high plate-to-plate reproducibility.

In vitro fusion assay

HeLa cell lines stably overexpressing the N-MitoVZL or C-MitoLZV markers were silenced for Romo1 (D-015268-04, Dharmacon) or with control siRNA (D-001810-01, Dharmacon). For each cell line, five 15-cm tissue culture plates were seeded at 25,000 cells/cm2 and silenced with 20 nM siRNA and DharmaFECT I according to the manufacturer’s protocol. After 4 days of silencing, cells were harvested, and mitochondria were isolated by differential centrifugation and stored at −80°C as previously described (40). Mitochondrial fusion assays were performed with a total of 100 μg of mitochondria as previously described (25). Briefly, mitochondria were pelleted and incubated at 4°C for 30 min to promote docking. After resuspension, mitochondria were incubated at 37°C for 10 min to allow fusion. Next, mitochondria were treated with trypsin for 10 min at 4°C to remove protein from broken mitochondria, and the trypsin was inhibited with SBTI (soybean trypsin inhibitor). Finally, mitochondria were lysed and luciferase activity was measured with the Renilla Luciferase Assay System (Promega) and GloMAX Luminometer (Promega) according to the manufacturer’s protocol.

Plasmid constructs

The human ROMO1 open reading frame was cloned into pLenti6 destination vector (Invitrogen) and tagged at the 5′ end with a FLAG tag. siRNA-resistant, FFAA, and 4CS mutants of FLAG-ROMO1 were generated by the QuikChange Mutagenesis Kit (Stratagene).

Immunoprecipitation, SDS-PAGE, and Western blotting

HEK293T cells were transfected with polyethylenimine (Sigma). Forty hours later, cells were collected in phosphate-buffered saline (PBS), pelleted, and processed as described (62).

Protease protection assay

U2OS cells from five 95% confluent 15-cm dishes were trypsinized and pelleted. Cells were washed with PBS and then resuspended in mitochondrial isolation buffer (MIB) + protease inhibitor cocktail: 200 mM mannitol, 68 mM sucrose, 10 mM Hepes-KOH (pH 7.4), 10 mM KCl, 1 mM EDTA (pH 8.0) in KOH, 1 mM EGTA (pH 8.0) in KOH, 0.1% bovine serum albumin. Resuspended cells were incubated on ice for 30 min and then lysed by passing cells through needles as follows: 2 × 25 gauge, 3 × 27 gauge, and 35 × 30 gauge. Lysis was stopped once cells reached 70% trypan blue positivity. Lysate was centrifuged at 600g for 10 min to clear unlysed cells and debris. The supernatant was centrifuged at 600g for 10 min, and then at 5500g for 15 min to obtain the heavy membrane fraction. Heavy membrane pellet was resuspended in 400 μl of MIB and then divided equally into four tubes treated as follows: tube 1, untreated; tube 2, 20 μg of proteinase K; tube 3, osmotic shock (700 μl of 20 mM KCl added to 100 μl of mitochondria); tube 4, osmotic shock + 80 μg of proteinase K. Reactions were incubated for 75 min and then centrifuged at 15,000g for 15 min. Mitochondrial pellet was resuspended in 100 μl of MIB + 5 μl of 100% TCA to precipitate protein. Protein lysates were analyzed by SDS-PAGE and Western blotting.

Electron microscopy

In a six-well dish, 125,000 cells were reverse-transfected with siRNA, washed 72 hours later with PBS, pelleted, and resuspended in 2% glutaraldehyde/PBS for embedding as described (63).

Mitochondrial stress test

Cells were reverse-transfected on Seahorse XF24 cell culture microplates on day 0. Where applicable, lentivirus was added to wells 4 hours later. On day 3, cells were washed three times with 500 μl of Seahorse XF assay medium, and oxygen consumption was measured with oligomycin, FCCP, and antimycin A (all at 0.5 μM) as controls. Protein was quantified with BCA (bicinchoninic acid; Thermo Scientific) for normalization.


Cells seeded in 384-well imaging plates were stained as described (64) and imaged with a 60× water objective (numerical aperture = 1.2) on an Olympus FluoView FV1000 confocal laser scanning microscope.

ROS measurements

Seventy-two hours after siRNA transfection, cells were stained with 20 μM DHE for 30 min, and fluorescence was analyzed with the Olympus FluoView FV1000 confocal laser scanning microscope. Total fluorescence intensity was normalized to cell number to report relative DHE fluorescence. Where indicated, cells were pretreated with 10 mM N-acetyl cysteine for 4 hours or with 20 μM antimycin A for 1 hour.

ATP measurements

Cellular ATP content was measured with CellTiter-Glo Luminescent Cell Viability Assay (Promega) using a BioTek Synergy 2 plate reader. Values were normalized to cell number.

Gel chromatography

Cells were lysed with 500 μl of coimmunoprecipitation buffer containing 1% Triton X-100 and protease inhibitors without dithiothreitol (DTT), sonicated, and centrifuged at 13,000 rpm for 20 min. Supernatants were applied to a Sepharose 4B column, and 250 μl of fractions was collected every 8 min for SDS-PAGE and Western blotting analysis.

Quantitation of OPA1 oligomers

Seventy-two hours after knockdown, ROMO1 cells, where applicable, were treated with adeno-tBid for 8 hours, and then all cells were treated with the cell-permeable cross-linker BMH (50 μM) for half hour. Cells were then washed with PBS + 2 mM DTT and then lysed and run on an SDS-PAGE followed by Western blotting for OPA1 and VDAC1 as a loading control. Percentage of OPA1 oligomers was quantified by densitometry, whereby % oligomers = {(total intensity of bands greater than 100 kD)/[total intensity of all bands (75 kD and above)]} × 100.

Cytochrome c release

Cells were infected with adenovirus expressing a tetracycline (TET)–inducible tBid (gift from G. Shore, McGill University) with doxycycline (1 μg/ml) and 12.5 mM QVD, and stained with cytochrome c and TOMM20 antibodies before imaging on the Cellomics ArrayScan VTI. Cytochrome c release from TOMM20-positive structures was quantified with a colocalization algorithm.

Condensed nuclei death assay

Cells were treated with etoposide for 20 hours, fixed and stained with Hoechst (10 μg/ml), and then imaged with an OPERA confocal imager (PerkinElmer). Condensed nuclei were quantified with an algorithm based on nuclear size and intensity of Hoechst staining.

Statistical analyses

Adjustment for multiple testing in analysis of variance (ANOVA) models was performed with R version 3.0.1 with the multicomp and sandwich packages (6567). All other P values were determined with a Student’s unpaired, two-tailed t test.


Fig. S1. The high-throughput mitochondrial morphology screen identifies ROMO1 as a regulator of mitochondrial fusion.

Fig. S2. Full-length multiple-sequence alignment of ROMO1 with orthologs.

Fig. S3. Reversible oxidation of ROMO1 and rescue of mitochondrial morphology with siRNA-resistant ROMO1 constructs.

Fig. S4. ROMO1 associates with, but does not affect, integrity of MINOS.

Fig. S5. Normal cellular ATP content and mitochondrial membrane potential in ROMO1 knockdown cells.

Fig. S6. Effect of ROMO1 status on OPA1 isoform balance.

References (66, 67)


Acknowledgments: We thank P. Rippstein for assistance with electron microscopy and N. Barrowman for statistical advice. Funding: This work was supported by grants from the Ontario Institute for Cancer Research, Canadian Institutes of Health Research (CIHR; MOP #97772), and Canadian Foundation for Innovation to R.A.S.; the CIHR (MOP #43935) to H.M.M.; Helmholtz Alliance on Systems Biology to M.A.A.-N.; Ontario Graduate Scholarship studentship to M.N.; and Natural Sciences and Engineering Research Council of Canada and Frederick Banting and Charles Best Canada Graduate Scholarship from the CIHR to A.C.-H.N. R.A.S. holds the Canada Research Chair in Apoptotic Signaling. Author contributions: M.N. carried out all the experiments unless otherwise stated, interpreted the data, and wrote the manuscript. A.C.-H.N. developed the RNAi screen assay and interpreted the data. S.B. performed the screen imaging analysis. A.C.-H.N., A.D., and S.B. carried out the RNAi screen. S.B. carried out the imaging analysis. T.S. carried out the in vitro mitochondrial fusion assay. N.M. and M.A.A.-N. carried out the informatic analysis of ROMO1 and prepared the ROMO1 alignment. H.M.M. planned the project, analyzed the data, and wrote the manuscript. R.A.S. planned the project, designed the experiments, prepared the figures, and wrote the manuscript. All authors edited the manuscript. Competing interests: The authors declare that they have no competing interests.
View Abstract

Navigate This Article