Mitochondria respond to environmental cues and stress conditions. Additionally, the disruption of the mitochondrial network dynamics and its distribution is implicated in a variety of neurodegenerative diseases. Here, we reveal a new function for Myo19 in mitochondrial dynamics and localization during the cellular response to glucose starvation. Ectopically expressed Myo19 localized with mitochondria to the tips of starvation-induced filopodia. Corollary to this, RNA interference (RNAi)-mediated knockdown of Myo19 diminished filopodia formation without evident effects on the mitochondrial network. We analyzed the Myo19–mitochondria interaction, and demonstrated that Myo19 is uniquely anchored to the outer mitochondrial membrane (OMM) through a 30–45-residue motif, indicating that Myo19 is a stably attached OMM molecular motor. Our work reveals a new function for Myo19 in mitochondrial positioning under stress.
Mitochondria are found in almost all eukaryotic cells and play a role in processes such as ATP production, Ca2+ homeostasis, lipid synthesis and apoptosis signaling (Nunnari and Suomalainen, 2012). Mitochondria are organized as a network that undergoes continuous events of fission and fusion, processes that are crucial for their cellular function (Chen and Chan, 2009). The balance between the fission–fusion cycles is essential, as its disruption is implicated in human neurodegenerative diseases (Lin and Beal, 2006; Chen and Chan, 2009; Federico et al., 2012). The mitochondrial network responds to changes in physiological conditions, such as starvation and oxidative stress, as reflected in morphological rearrangement that results in hyperfusion and fragmentation, respectively (Fan et al., 2010; Rambold et al., 2011). Additionally, the mitochondrial network responds to various cues by changing its intracellular positioning, supplying local ATP demand and providing a localized signaling platform (Tait and Green, 2012; Desai et al., 2013; Sheng, 2014).
Mitochondrial motility is primarily based on microtubules (MTs), utilizing plus-end-directed kinesin motors and the minus-end-directed dynein (Pilling et al., 2006). However, the actin cytoskeleton has also been shown to be involved in mitochondrial motility, localization and cellular distribution (Bradley and Satir, 1979; Chada and Hollenbeck, 2004). Myo19, an actin-based molecular motor, was discovered as a new mitochondria-localized myosin in vertebrates. Myo19 localizes to mitochondria through its C-terminal tail domain and nearly doubles the mitochondria motility when ectopically expressed (Quintero et al., 2009). The finding of a myosin that localizes to the mitochondria suggests that it acts as the link between mitochondria and the actin cytoskeleton. Knockdown of Myo19 by RNA interference results in apparent defects of mitochondrial segregation during division and incorrect movement of the mitochondria to the spindle poles during anaphase (Rohn et al., 2014).
The interaction between the actin cytoskeleton, myosin motors and their function in mitochondria dynamics, morphology and cellular localization is now beginning to emerge (Korobova et al., 2014). A new function for actin-based motors could emerge as regulators of cellular adaptations to stress, linking actin cytoskeleton remodeling and mitochondria (Jayashankar and Rafelski, 2014). These recent studies demonstrate an essential role for both actin cytoskeleton and myosins for mitochondria function in homeostasis and pathological conditions.
To this end, we examined Myo19 involvement in the mitochondrial network response to stress conditions induced in human U2OS cells. We discovered that Myo19 localizes with mitochondria to tips of starvation-induced filopodia in a process that depends on its ATPase activity and the actin cytoskeleton. We show that RNA interference (RNAi)-mediated knockdown of Myo19 attenuates the formation of starvation-induced filopodia. Deletion of 45 amino acids in Myo19 tail domain (Myo19Δ851-895–eGFP) impairs its localization to the mitochondria. We further dissected Myo19 binding to the mitochondria, revealing that it is a monotopic membrane protein anchored to the OMM with both the N- and C-terminus facing the cytoplasm. To our knowledge, this is the first vertebrate myosin that is directly bound to a membrane.
Glucose starvation induced localization of mitochondria and Myo19 to foci at cell periphery protrusions
Mitochondria function as an intracellular biosensor that responds to environmental changes, stress cues and physiological stimuli (Jayashankar and Rafelski, 2014). Hence, it is very intriguing to test the Myo19 interaction with the mitochondrial network dynamics and morphology in cells under stresses, such as starvation, that induce mitochondrial responses. We therefore ectopically expressed Myo19 fused to eGFP (Myo19–eGFP) in U2OS cells, and followed Myo19 and mitochondria localization in response to glucose starvation. Under complete medium conditions, Myo19 localized to mitochondria and there was also some diffuse cytosolic appearance, which might be due to the ectopic expression, similar to in previous work (Fig. 1A). Notably, expression of Myo19–eGFP caused mitochondria to clump together into globular structures in a motor-dependent manner (Fig. S1A, white arrow; Fig. S1B). In contrast to complete medium conditions, glucose starvation of U2OS cells resulted in the localization of Myo19 together with mitochondria to foci in protrusions extending from the cells (Fig. 1A). We quantified that >80% of Myo19–eGFP foci were positive for mitochondria by calculating the ratio between the number of MitoTracker-Red-stained Myo19–eGFP foci and the number of Myo19–eGFP foci (Fig. 1A, n=160 from 14 cells). Notably, these protrusions formed randomly around the cell periphery, showing no preference of directionality, which is most likely due to missing cues such as mechanical forces or chemotactic molecules (Petrie et al., 2009). The localization of Myo19 to the starvation-induced foci depended on an active full-length Myo19 (Fig. 1A). Neither eGFP alone, Myo19 tail (Myo19824–970–eGFP) nor ATPase-dead full-length Myo19 (Myo19G135R–eGFP) (Adikes et al., 2013) were able to localize to the protrusions, linking Myo19 enzymatic function to protrusion localization. To test whether the localization required a mitochondria-binding motif, we repeated the experiment with a cytosolic mutant of Myo19 (Myo19Δ851–895–eGFP), which was able to localize to these protrusions; however, no mitochondria colocalized with this deletion mutant, demonstrating the requirement for mitochondria-binding motif for colocalization with the mitochondria (Fig. 1A). Verification of expression of the Myo19 constructs was performed by western blotting (Fig. S1C). Supplementation of the starvation medium with glucose completely prevented both protrusions and foci formation, indicating that it is a response to glucose starvation rather than other nutrients (Venter et al., 2014) (Fig. 1C).
We further examined the growth dynamics of these starvation-induced protrusions and the emergence of Myo19 foci within them by performing live imaging of starved cells ectopically expressing Myo19–eGFP (Fig. 2). Most of the foci formed within the cell boundary at time zero of imaging (+20 min of starvation), which is marked by yellow line (70%, n=100 from 14 cells), whereas fewer extended beyond the cell boundary (30%, n=100 from 14 cells). However, we cannot conclude from these experiments the order of events between protrusion formation and Myo19 foci dynamics. The relationship between the protrusions and Myo19 foci can also be seen in time-lapse DIC images, which clearly show the linkage between them (Fig. S2A). Similar protrusions were observed in wild-type (WT) cells under starvation (Fig. S2B; see also Fig. 5F).
To this end, we demonstrate quantitatively the link between Myo19, its mitochondria-binding motif and glucose starvation to localization to starvation-induced foci formation in the cell periphery protrusions. Moreover, these observations raised the notion that Myo19 might influence these protrusions and prompted us to test the involvement of the actin cytoskeleton in these structures.
The actin cytoskeleton is essential for starvation-induced Myo19 foci formation
To support our results that an active motor is required for Myo19 foci formation, we tested whether disruption of the actin cytoskeleton would prevent their formation. Treating Myo19–eGFP-expressing cells with 0.2 µM latrunculin B (LatB) 30 min prior to starvation prevented the foci formation (∼70% of the cells showed less than two foci), further supporting that Myo19 foci formation is through the actin cytoskeleton (Fig. 3A). Alternatively, we show that nocodazole treatment of Myo19–eGFP-expressing cells induced Myo19 foci formation under complete medium conditions (>80% of the cells). This is similar to what is seen in cells ectopically expressing mDia1 and treated with nocodazole, which show a strong shift towards actin-based mitochondrial motility (Fig. 3B) (Minin et al., 2006). However, the molecular basis for this is still unknown.
Starvation-induced protrusions possess filopodia markers
We had strong evidence that the foci are related to the actin cytoskeleton. Therefore, we tested the starvation-induced localization of Halo-tagged Myo19 (Myo19–Halo) with actin–eGFP in live cells or Alexa-Fluor-488–phalloidin-stained actin in fixed cells, revealing that Myo19 foci localized at the tips of actin protrusions (Fig. 4A, inset and line profile; Fig. S3A). To characterize these actin protrusions, we tested Myo19 localization with focal adhesion and filopodia markers, as these structures might provide further insight on the nature of these protrusions. Vinculin and paxillin were used as markers to test whether Myo19 foci are related to focal adhesions. Ectopically expressed Emerald–vinculin localized to the base of the protrusions; however, Myo19 was localized further towards the tip (Fig. S3B, insets and line profile). Notably, expressing paxillin–eGFP inhibited protrusion formation, in accordance with published literature; however, it was not localized with Myo19 in the few protrusions that formed (Fig. S3C, insets and line profile).
The protrusions resembled filopodia in their structure, therefore we tested Myo19 colocalization with the filopodia markers fascin and Drf3 (also known as mDia2), revealing that these actin protrusions are indeed growing filopodia and that Myo19 is present at their tip (Fig. 4B, inset, line profile). Glucose-starvation-mediated induction of filopodia formation is not limited to U2OS cells, as we also show that this also occurs in the HeLa cell line (Fig. S3D), suggesting that this is a universal phenotype. This finding strongly encouraged us to test the contribution of Myo19 to filopodia formation.
Myo19 is a de novo effector of starvation-induced filopodia
To examine the contribution of Myo19 to filopodia formation, we utilized RNA interference to knockdown Myo19 from U2OS cells and follow starvation-induced filopodia formation. We ectopically expressed fascin–eGFP in these cells, allowing us to follow filopodia formation (Fig. 5A; Movie 1). Interestingly, filopodia formed both at the dorsal side of the cell and at the cell periphery. However, Myo19 localized only to the peripheral filopodia, but not to the dorsal filopodia, therefore only these were measured (Fig. 6A). To correctly measure the filopodia length using wide-field fluorescent microscopy, we verified that most of the peripheral filopodia were parallel to the plane by measuring their angle at the z-axis to be <2° (Fig. 6B). Evidentially, knockdown of Myo19 (∼80%, Fig. 5B) resulted in significantly fewer (from 42±7.9 to 24±10.8 filopodia per cell, mean±s.e.m.) and shorter filopodia (from 2.9±1.22 µm to 1.9±0.74 µm in length) (Fig. 5C,D). In addition, we quantified the filopodia length distribution and observed a clear shift towards shorter filopodia in Myo19-knockdown versus mock-treated cells (Fig. 5E). The small interfering RNAs (siRNAs) utilized to knockdown Myo19 were targeted against the 3′ UTR, allowing us to confirm that the effects are specific to Myo19 by performing a rescue experiment, where we knocked down Myo19 and ectopically expressed fascin–eGFP and Myo19–Halo. The rescue reversed the knockdown of Myo19 as measured by restoration of starvation-induced filopodia number to 39±6.5 filopodia per cell and length to 2.8±0.93 µm, similar to in mock-treated cells (Fig. 5D). Moreover, the rescue restored the filopodia length distribution to a similar distribution to that of mock-treated cells (Fig. 5E).
Furthermore, we observed two differences between the mock- and Myo19-siRNA-treated cells. Mitochondria were present in a minority of the filopodia of mock-treated cells, which were completely absent from filopodia of Myo19 knockdown cells (Fig. 5F, arrow). Additionally, Myo19 knockdown cells had visible patches of fascin–eGFP at the cell periphery (Fig. 5G), which might represent a possible failed filopodia formation site. Comparing the rate of filopodia growth between mock-treated and Myo19-knockdown cells indicates that Myo19-knockdown cells feature a pronounced longer lag before reaching a similar steady-state rate of filopodia growth (0.072±0.032 µm/min and 0.079±0.032 µm/min, for Myo19-knockdown and mock-treated cells, respectively), suggesting that Myo19 role is important for the dynamic development of filopodia (Fig. 6C).
We tested the effect of Myo19 knockdown on the mitochondria network morphology; however, we did not notice any major changes compared to mock-treated cells (Fig. 5H). Our collective observations strongly implicate Myo19 in starvation-induced filopodia formation. We next pursued the characterization of the physical linkage between Myo19 and the mitochondria.
Endogenous Myo19 co-purifies with mitochondria
To examine the nature of the physical interaction between Myo19 and the mitochondria, we purified mitochondria from HEK293 cells using differential centrifugation in order to biochemically verify endogenous Myo19 subcellular localization. The differential centrifugation generates three crude fractions: a mitochondria-rich fraction, called the heavy mitochondrial fraction (HMF), a mitochondria-poor, called the light mitochondrial fraction (LMF), and a cytosolic fraction (CYT). Myo19 was present exclusively in the HMF (Fig. 7A). To verify the mitochondrial localization, the HMF was further fractionated in a self-forming Iodixanol density gradient. Mitochondria and ER co-fractionated to some extent (fraction 7–10), as seen by the existence of the mitochondrial and ER markers VDAC and GRP94, respectively. Myo19 only distributed in the fractions of mitochondria and ER, but was absent from the fraction of ER lacking mitochondria, indicating that it interacts with the mitochondria (Fig. 7B). Notably, we did not detect any dissociation of Myo19 throughout the rigorous purification and fractionation procedure, suggesting that this interaction is stable.
Myo19 is anchored to the outer mitochondrial membrane
We next sought to determine the molecular basis underlying the physical interaction of Myo19 with mitochondria. Several possibilities exist for protein–organelle interactions including sub-mitochondrial localization, interaction through a receptor or mediator protein or direct binding to the outer mitochondrial membrane (OMM). The sub-mitochondrial localization of Myo19 was determined by protease protection assay using proteinase K. Myo19 was completely digested in the absence of any detergent, indicating that it resides on the outer mitochondrial membrane (OMM) (Fig. 7C, left). To test the membrane topology of Myo19, we purified mitochondria from Myo19–eGFP-expressing cells, where the eGFP tag is fused to the C-terminus of Myo19, and repeated the protease protection assay. The proteinase-K-treated sample was not recognized by anti-GFP antibody, indicating that the C-terminus faces the cytoplasm (Fig. 7C, right). ENDO G or ATP synthase 5A (ATP5A), both mitochondrial intramembrane space (IMS) proteins, were used as a control to confirm that the IMS was protected from the protease. To address the possibility of protein–protein interactions between Myo19 with an adaptor protein or OMM receptor, we attempted to biochemically extract endogenous Myo19 from purified mitochondria. The Myo19–mitochondria interaction was resilient to high salt (2 M NaCl) and urea (2 M), indicating that the interaction is not limited to electrostatics or to protein–protein interaction but that there is probably an additional binding feature. As a control for the salt and urea extraction, we utilized ATP5A, which can be extracted by salt and urea when mitochondria are subjected to freeze–thaw cycles (Fig. 7D). Interestingly, Myo19 could only be extracted by carbonate extraction from the OMM, suggesting that it is a peripheral membrane protein (Fujiki et al., 1982) (Fig. 7D). Furthermore, Myo19 separated to the detergent phase in a Triton X-114 phase separation of purified mitochondria, whereas the soluble protein GRP94 was found in the aqueous phase (Fig. 7E). These findings suggest that Myo19 is a monotopic membrane protein that does not transverse the OMM, with both the N- and C-terminus facing the cytosol.
Residues 860–890 of Myo19 mediate the interaction with the OMM
To explore the mode of Myo19–OMM interaction, we searched for putative membrane-binding motifs in Myo19 tail domain, which has been shown to localize to mitochondria when ectopically expressed (Quintero et al., 2009). Using the DAS prediction webserver, we identified such a motif between amino acids 860 and 890 (Cserzo et al.) (Fig. S4A). We ectopically expressed Myo19860–890–eGFP as the predicted membrane-binding motif and verified that this motif was sufficient to target eGFP to mitochondria (Fig. 8A). To test whether this domain is essential for Myo19 OMM localization, we ectopically expressed a deletion mutant lacking this region [Myo19824–970(Δ860–890)–eGFP]. This deletion mutant was mostly distributed throughout the cytosol, although it was not fully excluded from mitochondria (compared to eGFP, Fig. S4B). Therefore, we generated a larger deletion, Myo19824–970(Δ851–895)–eGFP, which was further excluded from mitochondria but not completely. These results strongly support the hypothesis that the Myo19860–890 motif is both essential and sufficient for OMM localization; however, we cannot exclude the presence of a minor binding region or other protein–protein region within the tail domain of Myo19 (Fig. 8A).
To test whether the Myo19860–890 membrane motif only dictates OMM targeting or whether it also participates in Myo19 OMM anchorage, we performed biochemical extractions on purified mitochondria from cells overexpressing Myo19860–890–eGFP.
In accordance to our finding in cells (Fig. 8A), Myo19860–890–eGFP was co-purified with mitochondria; however, unlike the endogenous Myo19, it appeared to be more sensitive to extraction by salt. Carbonate completely extracts Myo19860–890–eGFP from purified mitochondria, similar to endogenous Myo19 (Fig. 8B). The difference in salt sensitivity might be due to an insufficient length of the motif meaning it is not fully anchored in the membrane compared to in the WT endogenous protein. Another possibility is that a certain part of Myo19860–890–eGFP is correctly anchored, whereas the rest was present there nonspecifically (e.g. through dimerization of eGFP). Nevertheless, we can conclude from these experiments that this motif is both essential and sufficient for OMM localization.
To further explore the OMM targeting, we examined the motif sequence. Although there is no consensus sequence for targeting proteins to the OMM, it has been shown that a moderately hydrophobic motif flanked by positive residues is required for OMM localization (Rapaport, 2003). Myo19 membrane motif contains a moderately hydrophobic region between residues 864 and 880, flanked by basic positively charged residues (Table S1, flanked by ‘*’). Therefore, we sought to assess whether Myo19 OMM localization follows these criteria. Point mutations of the basic residues, R882S and K883S, singly or in combination, resulted in a dramatic shift of the intracellular localization of Myo19 to the ER (Fig. 8C; Table S1, Fig. S4C). Increasing the hydrophobicity of the hydrophobic region by the point mutation P865V resulted in dual localization to both the ER and the mitochondria. Analysis of the P865V mutation shows that the hydrophobicity of the motif increased from 1.141 to 1.482 (Kyte and Doolittle, 1982) and, hence, this residue might participate in dictating the correct mitochondrial localization of Myo19 (Fig. 8C; Table S1). Both results are in accordance with similar studies on tail-anchored OMM proteins (Kanaji et al., 2000; Borgese et al., 2001, 2007; Kaufmann et al., 2003; Wattenberg et al., 2007). In addition, the membrane motif with surrounding residues is predicted to contain two α-helices at 856–878 and 882–892 according to the MemBrain webserver (Fig. S4D) (Yang et al., 2013). By plotting this region as a helical wheel using HeliQuest (Gautier et al., 2008), one could observe that the N-terminal portion of the helix has amphiphatic characteristics, which can be found in many membrane-associated proteins (Hristova et al., 1999) (Fig. S4E). Collectively, these results show that the Myo19–OMM interaction is highly specific and mutations in this motif dramatically disrupt this interaction.
Myo19-derived peptide binds to OMM-mimicking vesicles
To characterize the Myo19 membrane-binding motif, we studied the binding of a synthetic peptide representing Myo19 residues 851–895 (Table S2) to small unilamellar lipid vesicles (SUVs) with a phospholipid composition similar to the OMM [55% phosphatidylcholine, 30% phosphatidylethanolamine, 13% phosphatidylinositol and 2% dioleoyl phosphatidylserine (de Kroon et al., 1997)]. Binding was followed by fluorescence anisotropy, relying on two tryptophan residues that are present in the peptide. The binding of Myo19 peptide to vesicles resulted in both an increase in the anisotropy and enhanced fluorescence signal (Fig. 8D). Equilibrium binding of Myo19 peptide to vesicles was monitored by titrating peptide versus increasing vesicle concentration. This resulted in a hyperbolic binding curve of both the fluorescence anisotropy and the fluorescence total intensity (FTI) (Fig. 8D). Therefore, to account for the FTI change during the fluorescence anisotropy measurements, we applied global fitting to both signals arising from peptide binding to vesicles to compensate for the change in fluorescence yield during fluorescence anisotropy measurements. Globally fitting the data to a hyperbolic binding equation (see Eqn 3 in the Materials and Methods) yielded an apparent binding constant of KD=281±50 μM (mean±s.d., n=3). To verify our results that endogenous Myo19 cannot be released from the OMM by high salt in vitro, we examined the salt-dependence of the binding interaction between the Myo19 peptide and the vesicles. As expected, equilibrium binding of Myo19 peptide to vesicles as a function of [NaCl] up to 1 M exhibited no effect on the magnitude of binding (Fig. 8E, inset, slope≈0), supporting our above findings (Fig. 7D). We performed additional binding experiments using a shorter peptide derived from Myo19 membrane motif that contains the peak in predicted helix propensity and the corresponding ER mutant (Fig. S4D, Table S2, Myo19858-883 and Myo19858-RK882/3SS, respectively) to either OMM or ER mimicking vesicles [20% phosphatidylethanolamine, 66% phosphatidylcholine, 9% phosphatidylinositol, 3% dioleoyl phosphatidylserine (Watanabe et al., 1996)]. Both peptides showed similar affinities to OMM-mimicking vesicles (45.0±14.1 and 49.0±17.2 μM for WT or mutant peptide, respectively, Fig. 8F; mean±s.d., n=2). The smaller Myo19-derived peptide exhibited tighter affinity than the longer one. This observation might reflect the predicted higher specificity that resides in the shorter peptide than the longer peptide. The WT peptide had also a similar affinity to ER-mimicking vesicles (50.5±22.0 μM), not showing evident preference for OMM-mimicking vesicles. In vitro binding of a hydrophilic peptide (Zeta1, Table S2), was much weaker (∼2.5 fold) compared to the Myo19 peptide (124.4±18.7 μM), supporting the notion that Myo19–OMM binding is mediated by hydrophobic interactions (Fig. 8F). Although our newly identified membrane motif is essential and sufficient for OMM localization in cells, the cellular specificity might be achieved by unique, yet unidentified, cytosolic or mitochondrial components.
Mitochondria are a key regulator of many stress conditions that cells might encounter, such as deprivation of various nutrients. To survive, cells must possess mechanisms that enable them to regain homeostasis. We therefore focused our investigation on the function of Myo19 in the cellular response to stress by glucose starvation, a major metabolite that is key to many cellular signaling pathways and a stress condition that occurs in solid tumors (Acker and Plate, 2002; Schroeder et al., 2005; Tredan et al., 2007; Hirayama et al., 2009; Moruno et al., 2012). We demonstrate a new function for Myo19 in filopodia formation and mitochondria localization under the cellular glucose-starvation stress response.
Glucose starvation of U2OS cells ectopically expressing Myo19 resulted in striking localization of Myo19 and mitochondria to tips of newly formed starvation-induced filopodia in a motor-dependent manner. Interestingly, these experiments demonstrated a direct link between active Myo19 and the actin cytoskeleton for their localization to the tips of starvation-induced filopodia. Knockdown of Myo19 by siRNA revealed its contribution to the formation and elongation of these filopodia, strongly supporting the hypothesis that Myo19 plays a role in starvation-induced filopodia formation. Finally, we show the molecular details of Myo19–mitochondria interactions, which is achieved through a unique ∼30–45-residue motif that enables it to be directly anchored to the OMM. To the best of our knowledge, this is the first myosin that is anchored in the membrane, in contrast to other myosins, which interact with phospholipids through electrostatic interactions (Hokanson and Ostap, 2006; Spudich et al., 2007). This distinctive feature most likely facilitates strong force bearing ability, preventing detachment of Myo19 from the mitochondria and overcoming high drag forces and competing interactions.
Filopodia are actin-based protrusions that extend from the lamellipodia and have been shown to be important in cancer cell migration, metastasis and angiogenesis (Folkman, 2002; Machesky, 2008; Shibue et al., 2012). Filopodia formation depends primarily on ATP-actin polymerization at the growing tip, providing the energy to push the membrane forward (Mallavarapu and Mitchison, 1999). Molecular modeling performed by Papoian and coworkers (Lan and Papoian, 2008) has modeled the rate of filopodia growth as being dependent on three factors: the resistance of the cell membrane, the mechanical buckling forces that lead to breakage of filopodia filaments, and the transport of ATP-actin monomers to the growing tips. The model also predicts a length-dependence of ATP-actin diffusion from the lamellipodium to the filopodia tips, where eventually the polymerization of actin will be limited by ATP-actin concentration. Ectopically expressed class III myosins have been shown to localize to tips of stereocilia, transporting espin-1, which results in a boost of their growth (Salles et al., 2009; Merritt et al., 2012). Myosin X has been found to undergo intrafilopodial motility and function in filopodia formation (Berg and Cheney, 2002; Tokuo and Ikebe, 2004; Bohil et al., 2006).
We propose that Myo19 localization to the fast growing ends of the filopodia promote their growth probably by mediating the localization of mitochondria to supply the high local ATP demand for ATP-actin polymerization. Here, we noticed that not all filopodia (<20%) contain mitochondria, and we speculate that there are small mitochondria at the filopodia tips, which are below our detection limit. Another possibility is that not all filopodia are being actively elongated or fail to continue to mature. In support of this, it has been shown that mitochondria stalling at the axon branching sites are necessary for efficient branching, an actin-polymerization-dependent process (Spillane et al., 2013). However, we cannot exclude a role for mitochondria in filopodial Ca2+ regulation (Davenport et al., 1996) or other signaling events leading to filopodia formation.
Glucose starvation increases the cellular AMP:ATP ratio, triggering the activation of AMP-activated protein kinase (AMPK) to meet the energetic demand of the cell (Hardie et al., 2012). We found that the minimal glucose concentration that prevents U2OS starvation-induced filopodia is extremely low. Complete medium contains 25 mM glucose, whereas supplementation of the starvation medium with as little as 5 µM prevents formation of starvation-induced filopodia (Table S3). This concentration is much lower than necessary for glycolysis (Faulkner and Jones, 1978); therefore, we propose that the prevention of Myo19–mitochondria localization to starvation-induced filopodia is mediated by glucose signaling events rather than metabolic homeostasis. In support of this, glucose starvation results in activation of a unique signature of phosphotyrosine signaling associated with focal adhesions, actin remodeling and F-actin stabilization (Arber et al., 1998; Miki et al., 1998; Graham et al., 2012). Glucose starvation increases reactive oxygen species (ROS) production (Diaz et al., 2009; Weaver, 2009; Graham et al., 2012; Taulet et al., 2012; Olguin-Albuerne and Moran, 2015). ROS have been shown to be essential for the formation of actin protrusions and well-characterized invadopodia, which have been linked to increased cancer invasiveness (Diaz et al., 2009; Weaver, 2009; Graham et al., 2012; Taulet et al., 2012; Olguin-Albuerne and Moran, 2015). Therefore, it could well be that the effect of Myo19 on starvation-induced filopodia formation is regulated through ROS. It will be exciting to study whether glucose starvation and oxidative stress induces Myo19 post-transcriptional modifications (PTMs), such as phosphorylation, that might modulate its motor activity. Here, we found that, under complete medium growth conditions, ectopic expression of Myo19 causes mitochondria to form globular structures. Previous reports have indicated that ectopic expression of Myo19 affects mitochondrial morphology, causing single mitochondria to obtain tadpole morphology; however, such globular structures have not previously been reported (Quintero et al., 2009). Our analysis shows that these globular structures are in fact clumped mitochondria, rather than hyperfused mitochondria; otherwise loss of MitoTracker intensity would be throughout the entire globular structure (Fig. S1E). This indicates that Myo19 does not promote fusion of mitochondria; however its effect on mitochondrial fission and fusion has not yet been determined. Myo19 knockdown did not cause any obvious morphological changes in the tubular mitochondria network. This inclines that Myo19 is not essential for tubular network morphology; however, we cannot exclude that it might also play some role in intrinsic mitochondrial dynamics or that the reduced levels are still sufficient to maintain the mitochondrial morphology. Corollary to our findings, knockdown of Myo19 has been shown to be important for correct segregation of mitochondria in a subset of dividing cells, whereas the rest were correctly dividing (Rohn et al., 2014).
The nature of Myo19 interaction with the mitochondria is essential for understanding the linkage between mitochondria, Myo19 and the actin cytoskeleton. We demonstrate that Myo19 residues 860–890 are essential and sufficient to cause it to localize to the OMM in cells. This region is moderately hydrophobic, with flanking positive residues, which are essential for correct mitochondrial localization. This targeting system has been shown in several tail-anchored proteins. Tail-anchored proteins have a membrane-targeting domain in their last 30 residues, which is nested in the ribosome when the protein completes its translation, whereas the Myo19 membrane-targeting domain is followed by ∼80 residues. The precise targeting mechanism of Myo19 to the OMM still remains to be determined. Myo19 might compete between the ER and OMM membranes, between targeting proteins such as Asna1/TRC40 and a yet unidentified OMM targeting protein, or transported through the ER to the OMM. Myo19 membrane interaction is highly stable as deduced from our biochemical extractions of endogenous Myo19 and from our binding experiments of the OMM binding motif to vesicles. In addition to the increase in fluorescence anisotropy, we observed strong tryptophan fluorescence increase upon binding of peptide containing the OMM-binding motif to vesicles. Tryptophan fluorescence increase is associated with a shift of the residue from a hydrophilic to hydrophobic environment, suggesting some degree of membrane insertion of the OMM-binding motif (Tsai, 2006). We propose with high likelihood that Myo19 is permanently bound to the OMM through its entire lifetime. This implicates that the regulation of Myo19 activity is most likely performed by a mechanism that modulates its ATPase activity while it is anchored to the OMM. It will be fascinating in the future to study the mechanism by which Myo19 localizes to the OMM from the moment of its translation as it anchors to the membrane. Myo19 enzymatic adaptation has not yet been characterized. It will be crucial to study its mechanochemical properties and how they relate to its cellular function (De La Cruz and Ostap, 2004; Geeves et al., 2005; Henn and Sadot, 2014).
Myo19 has recently been implicated in gliomagenesis (Jalali et al., 2015). Moreover, mitochondria were found in filopodia of glioblastoma cells that actively penetrate extracellular matrix (Arismendi-Morillo et al., 2012). Taken together with our observation that knockdown of Myo19 attenuates filopodia formation without evident effects on the mitochondrial network organization, we propose that Myo19 might be a therapeutic candidate for cancer therapy.
MATERIALS AND METHODS
All chemicals, reagents and media were purchased from Sigma (St Louis, MO) unless otherwise indicated.
Myo19 cloning and plasmids
Myo19 was cloned from an EST library (NIH clone ID 3629870) cDNA. The amplified PCR product was missing the region encompassing nucleotides 1318–1916 (database sequence FLJ22865). Therefore, this region was artificially synthesized and cloned into pFC14K (Promega). Full-length Myo19 and its tail domain (residues 824–970) were subcloned to peGFP-N1b using restriction enzymes and used as a template for truncations and mutagenesis. Table S4 contains the primers used for the cloning. peGFP-N1 was a gift from Nabieh Ayoub (Technion Israel Institute of Technology, Haifa, Isreal), fascin–eGFP was a gift from Alexander Bershadsky (Weizmann Institute of Science, Rehovot, Isreal). mEmerald–Vinculin-N-21 was a gift from Michael Davidson (Addgene plasmid number 54304) (Burnette et al., 2011). pRK-GFP-Paxillin was a gift from Kenneth Yamada (Addgene plasmid number 50529). eGFP-tagged full-length mDia2/Drf3 was a kind gift from Klemens Rottner and Jan Faix (TU Braunschweig and Hanover Medical School, Germany, respectively).
Cell culture, cell lines, starvation conditions and cytoskeleton interfering drugs
U2OS cells (gift of Nabieh Ayoub) were grown at 37°C and under 5% CO2 in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum (FCS, Biological Industries, Beit Haemk, Israel), 4.5 g/l glucose, 2 mM L-Glutamine, 20 mM HEPES-KOH pH 7.4, 100 U/ml penicillin, 100 µg/ml streptomycin and 0.25 µg/ml amphotericin B. HEK293SF-3F6 (ATCC) for purification of mitochondria and cell organelles were grown in suspension in EX-CELL 293 (Sigma) at 37°C and 5% CO2. For starvation conditions, cells were rinsed twice with PBS and incubated in starvation medium (glucose-free DMEM supplemented with 20 mM HEPES-KOH pH 7.4, 5 mg/ml BSA, 100 U/ml penicillin, 100 µg/ml streptomycin and 0.25 µg/ml amphotericin B) for 2 h. Drugs that interfered with the cytoskeleton (final concentration: 0.2 µM Latrunculin B or 0.75 µg/ml nocodazole) were first diluted in medium and then added to the cells. Cells were tested for mycoplasma contamination and found clean (EZ-PCR mycoplasma detection kit, Biological Industries).
Purification of mitochondria
HEK293SF-3F6 cells were harvested by centrifugation at 200 g, washed twice with PBS and once in homogenization medium (0.25 M sucrose, 1 mM EDTA, 20 mM HEPES pH 7.4). The pellet was resuspended in homogenization medium containing protease inhibitors (0.1 mM benzamidine, 0.055 mM phenanthroline, 1 mM PMSF and, from AG Scientific, 0.01 mM bestatin, 0.02 mM leupeptin and 0.005 mM pepstatin A), homogenized using a teflon-glass pestle and centrifuged at 1000 g to pellet nuclei. The resulting post-nuclear supernatant was centrifuged for 20 min at 10,000 g to pellet mitochondria and obtain a heavy mitochondrial fraction (HMF). The supernatant from the 10,000 g centrifugation was taken and centrifuged for 20 min at 17,000 g to pellet light membranes and obtain a light mitochondrial fraction (LMF). The supernatant from the 17,000 g was taken to obtain a cytosolic fraction (CYT).
Additional purification of mitochondria
To obtain pure mitochondria we used isopycnic ultracentrifugation. The HMF was resuspended in homogenization medium containing protease inhibitors and adjusted to 17.5% OptiPrep. The HMF was centrifuged at 270,000 g for 3 h in a Sorvall T-890 or Beckman Type 70 rotor to form a self-forming density gradient. Following ultracentrifugation, three bands appeared: light membranes, mitochondria and a DNA-rich band (from top to bottom). The second band was taken using a pasture pipette, diluted 1:1 with homogenization medium containing protease inhibitors and centrifuged for 30 min at 17,000 g to pellet mitochondria. The mitochondria were resuspended in fresh homogenization medium containing protease inhibitors at ∼5–6 mg/ml, concentration was measured using a Bradford assay according to the manufacturer’s directions (BioRad). Purified mitochondria were flash frozen in liquid nitrogen and stored at −80°C. All steps were performed at 4°C.
Protease protection assay
For proteinase K digestion, mitochondria were washed three times with homogenization medium to remove protease inhibitors. Proteinase K was added at 50 µg/ml and incubated on ice for 30 min. Reactions were terminated by addition of PMSF to 2 mM and centrifugation at 10,000 g for 3 min to pellet mitochondria and to remove the protease. Sample buffer ×5 (final concentration, 50 mM Tris-HCl pH 6.8, 2% SDS, 10% glycerol, 1% β-mercaptoethanol, 0.02% Bromophenol Blue) was added to the samples and they were resolved by SDS-PAGE.
Purified mitochondria were resuspended in either 100 mM NaCO3 pH 11.5, 2 M NaCl in homogenization medium, or 2 M urea in 20 mM MES pH 6.5 for 30 min on ice, during which the NaCl and urea samples were subjected to ten freeze–thaw cycles in liquid nitrogen to mechanically break the mitochondria. The samples were then centrifuged for 1 h at 150,000 g in a Sorvall S100-AT3 rotor. The supernatant was collected and the pellet was resuspended in an equal volume of 100 mM HEPES pH 7.4. To precipitate the extracted proteins, we performed trichloroacetic acid (TCA) and acetone precipitation on the NaCO3 and NaCl supernatant, TCA was added to 12% from a 100% stock and allowed to incubate for 1 h at 4°C. The samples were centrifuged for 30 min at 17,000 g, resuspended with ice-cold acetone and left overnight at −20°C. The samples were then centrifuged for 30 min at 17,000 g, supernatant was removed and the acetone was allowed to evaporate by heating the samples for 10 min at 90°C. Urea-containing samples were desalted using Zeba spin desalting columns 7K MWCO according to the manufacturer’s protocol (Thermo Fisher Scientific). Sample buffer ×5 (as described in the protease protection assay section) was added to the samples and they were resolved by SDS-PAGE.
Triton X-114 extraction
Purified mitochondria were resuspended in cold phase separation buffer [10 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% Triton X-114 (MP Biomedicals); protease inhibitors as described in the purification of mitochondria section] and incubated on ice for 5 min, insoluble material was removed by centrifugation at 10,000 g for 10 min. The lysate was loaded on top of a cold sucrose cushion (10 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.06% Triton X-114, 6% sucrose, protease inhibitors), incubated at 37°C for 3 min and then centrifuged at 300 g for 3 min. The upper phase was placed in a new tube and Triton X-114 was added to 0.5%; the solution was then incubated on ice until it cleared, loaded on the sucrose cushion, incubated at 37°C for 3 min and then centrifuged at 300 g for 3 min. This process was repeated again with the exception that Triton X-114 was added to 2%. The upper and the lower phase were collected and corrected to the same volume. Sample buffer ×5 (described in the protease protection assay section) was added to the samples and they were resolved by SDS-PAGE.
Following SDS-PAGE, gels were washed in Milli-Q water. Nitrocellulose or PVDF membranes (PVDF membranes were hydrated in 100% methanol for 1 min) were incubated for 1 min in transfer buffer and transfer was performed for 10 min using a semi-dry blotter (GenScript). Transferred membranes were washed with Milli-Q water and blocked using EZblock (Biological Industries) for 1 h at room temperature. Following blocking, primary antibody was added for 1 h at room temperature or left overnight at 4°C [0.5–1 µg/ml anti-VDAC antibody abcam ab14734, anti-Myo19 antibody Sigma HPA021415 or Abcam ab174286, anti-GRP94 antibody (a kind gift from Christopher Nicchitta, Department of Cell Biology and Biochemistry, Duke University, North Carolina), anti-HSP60 antibody (Santa Cruz Biotechnology sc-59567), anti-ENDO G antibody (Santa Cruz Biotechnology sc26923), anti-ATP5A antibody (abcam ab14748) anti-GFP antibody (Roche 11814460001)]. The membrane was washed with TBS-T (TBS, 50 mM Tris pH 7.5, 150 mM NaCl with 0.02% Tween-20) and horseradish peroxidase (HRP)-conjugated secondary antibody was added for 1 h, washed with TBS-T and developed using 0.01% H2O2, 1.25 mM Luminol and 0.04 mM p-Cumaric acid in 0.1 M Tris-HCl pH 8.5. Images were acquired using a ImageQuant LAS4000 imager (GE).
Transfections and microscopy
Transfections were performed using polyethylenimine (PEI, Weis Scientific-PolySciences). Adherent U2OS cells were plated a day before transfection on plastic or glass bottom dishes and allowed to adhere overnight. Plasmid DNA and PEI were diluted separately in 150 mM NaCl and then combined, and complex formation was allowed to occur for 25 min at room temperature before addition to the cells and incubation overnight. Hoechst 33342 (0.75 µg/ml), MitoTracker (30 nM, Molecular Probes) and propidium iodide (1 µg/ml) were added 15 min prior to imaging. HaloTag constructs were stained by incubating the cells with HaloTag TMR Ligand ON (25 nM, Promega). Cells were imaged using Confocal Zeiss LSM 700 or Confocal Zeiss LSM 710 or GE InCell Analyzer 2000 at ×63 magnification in an environmental chamber.
Cells grown on coverslips were washed with PBS, and fixed with 4% PFA in PBS (EMS) at room temperature for 15 min. The PFA was removed by washing with PBS, and the cells were blocked for 1 h with EZblock (Biological Industries). To visualize actin and nuclei, the cells were incubated with Alexa-Fluor-488–phalloidin and Hoechst 33342 for 30 min. The cells were then washed with TBS and mounted on slides using Fluoromount-G.
Transfections were performed using Lipofectamine 3000 (Life Technologies). Adherent U2OS cells were plated at 70–90% confluence a day before transfection on plastic dishes and allowed to adhere overnight. Mock siRNA (Universal negative control siRNA, IDT) or mix of three Myo19 siRNAs (10 nM each, reference numbers 118568291, 118568285, 118568288, IDT), and Lipofectamine were separately resuspended in OptiMEM (Life Technologies). The RNAi mix was added to the Lipofectamine mix and incubated for 5 min at room temperature. The medium was replaced with DMEM without serum and the transfection mix was added. Complete medium was added after 4 h. On the next day, the cells were trypsinized and plated on glass bottom fluoridish (WPI) or optic bottom 24-well plates, for microscopy, or plastic dishes, for western blotting at 72 h post transfection.
Prediction webservers used in this study
Image analysis and calculation of mitochondria-positive Myo19–eGFP foci
Western blot band intensity, filopodia length measurements and line plot analysis were performed using ImageJ (NIH). Filopodia angle measurements were performed using IMARIS V8.0.0 (Bitplane AG, Zurich, Switzerland).
To calculate the fraction (f) of mitochondria positive Myo19-eGFP foci (Fd) we calculated the ratio between the number of MitoTracker-Red-positive Myo19–eGFP foci (NFd) to the total number of Myo19–eGFP foci (NFt) according to the equation:
Determination of filopodia growth rate
To determine the filopodia growth rate from the time-lapse imaging, we took into consideration only the region where the filopodia growth was constant over time (i.e. between 80 and 110 min). We used the following equation to calculate individual filopodia growth rate vf: where l110 is filopodia length at the time of 110 min, l80 is filopodia length at the time of 80 min and elapsed time (Δt) is 30 min. N=30 for each treatment.
Phosphatidylcholine, phosphatidylethanolamine, phosphatidylinositol and dioleoyl phosphatidylserine (Avanti) were mixed in chloroform and evaporated under nitrogen. The vesicles were then resuspended in binding buffer (40 mM HEPES pH 7.1, 150 mM NaCl), subjected to ten freeze–thaw cycles in liquid nitrogen and kept at −80°C. The vesicles were thawed and sonicated for 5 min at low intensity; insoluble material was pelleted by centrifugation at 100,000 g for 20 min. Vesicles were used within 2 days after sonication.
Microirradiation of mitochondria
U2OS cells were loaded with 25 nM of the MitoTracker dye (sensitive to mitochondrial membrane potential) for 15 min at 37°C. Unbound dye was removed by three washes with complete medium. A 0.5×0.5 µm square was irradiated using a 405 nm laser at 100% intensity, causing single mitochondria to depolarize and lose staining.
Fluorescence anisotropy measurements and KD determination
Equilibrium binding by fluorescence anisotropy measurements was performed with a PC1 spectrofluorimeter (ISS, Champaign, IL) designed as a T-format for simultaneous acquisition on two emission channel monochromators equipped with automatic polarizers. Samples were equilibrated (60 min, room temperature) and then measurements were performed by following intrinsic fluorescence of the peptide with λex=280 nm using vertical polarized light, and the emitted vertical and horizontal polarized light was monitored at 90° with double emission monochromators at λem=335 nm. The G-factor for correction of the different gain between the dual PMT detectors was calculated as described by the instrument manufacturer. The fluorescence total intensity (FTI) and total fluorescence anisotropy are given by: (1)and (2)where the I|| is the parallel fluorescence intensity and I⊥ is the perpendicular fluorescence intensity.
Our binding model for a simple bimolecular reaction is: Under the condition of Ptot≪KD then the general solution for this equilibrium binding scheme is in the form of the following quadratic equation: (3)where Ptot is monitoring species; Vtot is titrating species and [PV] is the bound species.
Statistical significance between mock and treated groups was performed using Student's t-test, the different groups were considered to have the same variance because the groups were taken from the same pool of cells. Statistical significance was scored as *P≤0.05, **P≤0.01, ***P≤0.001, ****P≤0.001. All presented error bars indicate standard deviation.
We acknowledge Dr Dan Cassel for his assistance in the vesicle preparation, and critically reading the manuscript. We thank Drs Yehuda G. Assaraf and Nabieh Ayoub for their invaluable comments on the manuscript. We are grateful for Dr Nitzan Dahan's assistance with the fluorescence microscopy imaging from Life Sciences and Engineering Infrastructure Unit at the Technion. We thank Drs Michael Davidson and Kenneth Yamada for providing the focal adhesion marker plasmids. We thank Drs Klemens Rottner and Jan Faix (TU Braunschweig and Hanover Medical School, Germany, respectively) for providing the eGFP-tagged full-length mDia2/Drf3 plasmid.
The authors declare no competing or financial interests.
B.I.S. designed and carried out experiments, analyzed the data and co-wrote the manuscript. M.U. designed and carried out experiments and analyzed the data. A.H. supervised the project, designed and analyzed the data and co-wrote the manuscript.
M.U. is supported by the Lady Davis Fellowship and Israeli Government Ministry of Foreign Affairs Scholarship. A.H. acknowledges the Marie Curie Career Integration [grant number 1403705/11].
Supplementary information available online at http://jcs.biologists.org/lookup/suppl/doi:10.1242/jcs.175349/-/DC1
- Received June 9, 2015.
- Accepted December 5, 2015.
- © 2016. Published by The Company of Biologists Ltd