ABSTRACT
Mitochondria–ER contact sites (MERCs) enable communication between the ER and mitochondria and serve as platforms for many cellular events, including autophagy. Nonetheless, the molecular organization of MERCs is not known, and there is no bona fide marker of these contact sites in mammalian cells. In this study, we designed a genetically encoded reporter using split GFP protein for labeling MERCs. We subsequently analyzed its distribution and dynamics during the cell cycle and under stressful cellular conditions such as starvation, apoptosis and ER stress. We found that MERCs are dynamic structures that undergo remodeling within minutes. Mitochondrial morphology, but not ER morphology, affected the distribution of MERCs. We also found that carbonyl cyanidem-chlorophenyl hydrazone (CCCP) and oligomycin A treatment enhanced MERC formation. The stimulations that led to apoptosis or autophagy increased the MERC signal. By contrast, increasing cellular lipid droplet load did not change the pattern of MERCs.
INTRODUCTION
The evolution of eukaryotic cells led to the development of complex intracellular membrane organelles to enable compartmentalization of the cytoplasm for specific cellular activities. Homeostasis of the cell requires coordination among these membrane organelles, and this coordination is achieved through long-distance communications mediated by signaling factors (Butow and Avadhani, 2004; Raffaello et al., 2016) and direct physical contacts mediated by tethering molecules (Eisenberg-Bord et al., 2016; Phillips and Voeltz, 2016).
Close localization of ER and mitochondria has been observed repeatedly in ultrastructural electron microscopy studies (Franke and Kartenbeck, 1971; Morré et al., 1971). Subcellular fractionation studies have also revealed that a part of the mitochondrial membrane is tightly associated with the ER and can be isolated together with ER elements (Meier et al., 1981; Pickett et al., 1980). Mitochondria–ER contact sites (MERCs) play important roles in many different cellular processes (Lang et al., 2015; Phillips and Voeltz, 2016; Vance, 2014). The ER is the major site of lipid synthesis in the cell. Mitochondria need to import lipids from the ER, and non-vesicular mechanisms are known to be used in these lipid transfer processes. The physical contacts between the ER and mitochondria provide a way to control the exchange of phospholipids according to the metabolic needs of the cell. MERCs also mediate Ca2+ flux from the ER to mitochondria (Krols et al., 2016). The uptake of Ca2+ can regulate mitochondrial activity and buffer the possible cytotoxic effects of excessive Ca2+ in the cytosol. The mitochondrial Ca2+ transporter has a low affinity for Ca2+ and the close contacts of ER and mitochondria enable the efficient uptake of this high local Ca2+ concentration into mitochondria. In addition to these functions, MERCs are hotspots of autophagosome formation (Hamasaki et al., 2013), mitophagy (Böckler and Westermann, 2014), mitochondrial division (Friedman et al., 2011) and mitochondrial DNA synthesis coupling (Lewis et al., 2016). Changes in MERC activity have been observed in Alzheimer's disease (Hedskog et al., 2013; Schreiner et al., 2015) and metabolic diseases such as obesity (Arruda et al., 2014).
In yeast, a protein complex called the ER–mitochondria encounter structure (ERMES) was found to tether the ER and mitochondria (Kornmann et al., 2009). ERMES has four components: the outer-membrane mitochondrial proteins Mdm10 and Mdm34, cytosolic protein Mdm12 and integral ER protein Mmm1 (Lang et al., 2015). The ER protein complex EMC has also been proposed to form contacts between the ER and mitochondria (Lahiri et al., 2014; Voss et al., 2012). In addition, Ltc1/Lam6 regulates contacts between organelles, including the contacts between ER and mitochondria (Elbaz-Alon et al., 2015; Murley et al., 2015). MERCs have also been observed in mammalian cells. In contrast, orthologs of ERMES components have not been identified in mammals by means of homology searches (Murley and Nunnari, 2016; Vance, 2014). The interaction between mitochondrial porin voltage-dependent anion-selective channel protein 1 (VDAC1) and the ER Ca2+ channel IP3R has been reported to mediate MERCs in mammals (Szabadkai et al., 2006). It has been shown that homotypic interactions of MFN2 mediate the physical link between the ER and the mitochondrion (de Brito and Scorrano, 2008; Naon et al., 2016, Naon et al., 2017). Recently, it was reported that the binding of the ER protein VAPB to the mitochondrial protein PTPIP51 (also known as RMDN3) tethers the two organelles (Stoica et al., 2014). In addition, the cargo-sorting protein PACS2 (Simmen et al., 2005), mitochondrial fission protein Drp1 (also known as DMN1L; Friedman et al., 2011) and autophagy protein Atg14 (Hamasaki et al., 2013) have been reported to localize to the MERCs during certain cellular events.
Although more and more molecules have been found to localize to the MERCs, there is no specific marker of these contacts. A drug-inducible reporter was developed to visualize MERCs. Nonetheless, artifacts are produced by prolonged drug treatment (Csordás et al., 2010). Using a dimerization-dependent fluorescent protein (ddFP) tool, an indicator of MERCs was developed (Alford et al., 2012). However, these reporters are not widely used in the field. To date, most studies still rely on transmission electron microscopy (TEM) and on overlays between an ER marker and mitochondrial marker in confocal microscopy to evaluate the changes of the contacts. TEM is labor-consuming and can reflect only the static status of the cells. The overlays between ER and mitochondrial markers do not always reflect the proximity between these two organelles and the lack of reliable reporters of MERCs may explain why there are conflicting results in this field (Filadi et al., 2017; Leal et al., 2016; Naon et al., 2017). A proximity-ligation ‘Duo-Link’ assay was generated to assess MERCs (Paillusson et al., 2017; Tubbs and Rieusset, 2016). However, this method only worked in fixed cells. Recently, some new techniques have been developed to detect MERCs. Elgass et al. (2015) combined correlative cryogenic fluorescence microscopy and soft X-ray tomography to observe mitochondrial fission at the MERCs. Focused ion beam-scanning electron microscopy was used to generate 3D reconstructions of the intracellular organelles and the membrane contacts in neurons (Wu et al., 2017) and systems-level spectral imaging was applied to analyze the dynamics of organelle contacts in live cells (Valm et al., 2017). Although these techniques can be used to study MERCs, all require special equipment, and they are labor-consuming.
In this study, we developed a genetically encoded reporter using split GFP protein for labeling MERCs. We determined its properties and analyzed its distribution and dynamics during the cell cycle and under normal or stressful cellular conditions. Using this reporter, we demonstrated that MERCs are dynamic structures that undergo active remodeling under different cellular needs.
RESULTS
Designing a reporter of MERCs
High-resolution electron microscopy revealed that the distance between the ER and mitochondria is ∼10–30 nm (Csordás et al., 2006) at the contact points. We decided to label the ER membrane and mitochondrial outer membrane with two parts of the split GFP protein, spGFP1-10 and spGFP11, respectively (Magliery et al., 2005). If the ER and mitochondria are close enough, spGFP1-10 and spGFP11 will form functional GFP and glow (Fig. 1A). A fusion reporter using a similar principle has been used to detect MERCs in yeast cells (Eisenberg-Bord et al., 2016). In yeast, protein Ubc6 is distributed ubiquitously on the ER, and its N-terminus faces the cytosol and C-terminus harbors an ER-anchoring sequence (Kornmann et al., 2009). Tom70 is a mitochondrial outer-membrane protein whose N-terminus has mitochondrial localization sequences (a presequence and a transmembrane sequence) and its C-terminus faces the cytosol (Kornmann et al., 2009). Both Ubc6 and Tom70 are conserved from yeast to human. The human orthologs of Ubc6 and Tom70 are called UBE2J2 and TOMM70, respectively. ER and mitochondrial targeting sequences in UBE2J2 and TOMM70 were identified through protein sequence alignment and literature mining (Tiwari and Weissman, 2001; Young et al., 2003). We then fused six copies of Myc to the N-terminus of ER tail-anchor sequence derived from UBE2J2 (residues 228–259), and GFP was fused to the C-terminus of the mitochondrial localization sequence from TOMM70 (residues 1–59). The resulting proteins were named Myc–ERt and Mitot–GFP, respectively, and expressed in human osteosarcoma U2OS cells. As expected, Myc–ERt showed ubiquitous ER distribution, and it perfectly co-localized with the ER marker Sec61β–RFP (Fig. S1A,A′). Mitot–GFP was distributed evenly on the mitochondrial surface and co-localized with the mitochondrial marker HSP60 (Fig. S1B,B′). These data suggest that the targeting sequences were able to function efficiently in mammalian cells. We then fused spGFP1-10 to the N-terminus of UBE2J2 fragment (spGFP1-10–ERt) and spGFP11 to the C-terminus of the first 59 amino acids of TOMM70 (Mitot–spGFP11). To monitor the expression of spGFP1-10–ERt, a V5 epitope tag was inserted between spGFP1-10 and ERt fragment. Linkers of different lengths (∼3 nm or ∼10 nm) were inserted between the spGFP11 and TOMM70 fragment to ensure that spGFP11 and spGFP1-10 could reach each other. The resulting constructs were called Mitot-l1–spGFP11 and Mitot-l2–spGFP11.
Design of a reporter of mitochondria–ER contacts (MERCs). (A) spGFP1-10 was fused to the N-terminus of UBE2J2 fragment (spGFP1-10–ERt) and spGFP11 was fused to the C-terminus of the first 59 amino acids of TOMM70 (Mitot–spGFP11). Linkers with different lengths (∼3 nm or ∼10 nm) were inserted between spGFP11 and TOMM70 fragments to ensure that spGFP11 and spGFP1-10 could reach each other. When the distance between mitochondria and the ER is about 10–30 nm, spGFP1-10 and spGFP11 will refold and emit green fluorescence. (B) U2OS cells stably transfected with spGFP1-10–ERt and Mitot–2×spGFP11constructs (MERC reporter) were transfected with ER marker Myc–UBE2J2, fixed and stained with anti-Myc and anti-HSP60 antibodies. ER–mitochondria contact reporter signals (green, arrows) localized only to the junctions of the ER (indicated by Myc–UBE2J2, red) and mitochondria (indicated by HSP60 staining, blue). (C) Green channel of B. (D–H) U2OS cells with MERC reporter expressed were fixed and stained with anti-TOM20 antibody. 3D SIM images show that the contact reporter (green) forms thin lines or rings on the mitochondria, indicated by TOM20 staining (red). (E) Zoom of the green channel in the boxed region in D. (F) Red channel of D. (G) Green channel of D. (H) U2OS cells with MERC reporter (green) stably expressed were transfected with Mito–RFP (red). 3D SIM images were collected using GE DeltaVision OMX. The 3D rendering was made using Imaris NIS 4.3. MERCs are indicated by the green signals and mitochondria by the red signals. (I,J) TEM analysis of control U2OS cells (Ctrl) and U2OS cells with MERC reporter stably expressed (stable line). Yellow arrows indicate the contact regions (distance between ER and mitochondrion is less than 30 nm). (K) The contact sizes were normalized by dividing contact length with the circumference of the mitochondrion. The statistical analysis of the size of the contact shows that there is no dramatic difference between control and stable line; n=61 mitochondria were analyzed. The experiments shown were replicated twice. NS, no significance (two-tailed t-test). Data are presented as means±s.d. (see also Fig. S1).
As expected, expressing each of these proteins in U2OS cells did not yield any fluorescent signal. When Mitot–spGFP11 or Mitot-l1–spGFP11 was expressed with spGFP1-10–ERt, no green signal was observed in healthy cells (Fig. S1C). This is probably because the sizes of these reporters do not match the scales of the MERCs (10–30 nm). When Mitot-l2–spGFP11 was expressed together with spGFP1-10–ERt, faint but distinct green puncta were observed inside cells (Fig. S1C). To enhance the GFP signal intensity, we fused the first 59 amino acid of TOMM70 with two copies, four copies or seven copies of spGFP11 (Kamiyama et al., 2016), adjusted the linker length accordingly, and expressed them simultaneously with spGFP1-10–ERt (Fig. S1C). The GFP signals became bright enough when two copies of spGFP11 (2×spGFP11) were used. To avoid possible complications of multiple copies of spGFP11, we decided to use Mitot–2×spGFP11 for further experiments. We realized that transient expression of spGFP1-10–ERt and Mitot–2×spGFP11 sometimes led to changes in mitochondrial morphology. To overcome this problem, U2OS cells were stably transfected with these constructs and single-cell clones with normal mitochondrial morphology were picked for further analysis.
To test whether the reporter labels the MERCs specifically, we transfected the stable cell clones with the ER marker Myc–UBE2J2 and co-stained them with anti-Myc and anti-HSP60 antibodies. The reporter GFP signals were localized only to the junctions of the ER and mitochondria, indicated by positive signals for both Myc and HSP60, suggesting that the reporter was very specific (Fig. 1B,C). We also found that not every ER and mitochondrial junction was positive for the GFP reporter, which suggests that some ER and mitochondria are not close enough to form real contacts (Fig. 1B,C). Alternatively, the space between ER and mitochondria might be too tight to allow the refolding of spGFP at some of the contact sites.
In addition, we examined the MERC reporter using Delta Vision OMX 3D structured-illumination microscopy (SIM). Most reporter signals were found at the surface of mitochondria and formed thin lines. Some of the contacts wrapped the mitochondria and formed ring-like structures (Fig. 1D–H), suggesting that some ER cisternae indeed wrap mitochondria and form contacts as predicted in other studies (Friedman et al., 2011).
The interaction of spGFP1-10 and spGFP11 was reported to be irreversible under in vitro experimental conditions (Magliery et al., 2005; Pédelacq et al., 2006). It is not clear whether the interaction could be made reversible when spGFP1-10 and spGFP11 are fused to two proteins that could be pulled apart under certain conditions. If the interaction is irreversible, the new reporter may function as a tether and increase the contacts between ER and mitochondria. We therefore used TEM to quantify and compare the MERCs in U2OS cells, with and without reporter expression. We found that expression of the reporter components did not affect the frequency and the size of the MERCs in the stable cell lines (Fig. 1I–K), suggesting that the interaction of spGFP1-10 and spGFP11 is reversible in our experimental settings.
Characterization of the reporter of MERCs
To test whether the reporter could show the assembly and disassembly of MERCs, we examined its patterns under various conditions that were reported to either enhance or reduce association between the ER and mitochondria. We first made an artificial tether that could attach the ER to mitochondria. This tether was a fusion protein with a mitochondrial localization sequence from protein AKAP1 at the N-terminus (Csordás et al., 2010), a red fluorescent protein (RFP) in the middle and an ER anchor from protein SACM1L at the C-terminus (Csordás et al., 2010; Varnai and Balla, 2007). When the tether was expressed, TEM analysis indicated that mitochondria were wrapped with ER (Fig. S2A). In addition, signal quantifications suggested that the tether was functional (Fig. S2B). When the tether was expressed in the U2OS cells, which were stably expressing the contact reporter, we observed that the intensity of reporter GFP signals was greatly enhanced as compared with that in the cells expressing the reporter alone. In addition, when the tether was expressed, the GFP signals of the reporter were evenly distributed on the mitochondria and colocalized with the signals of the tether. These data suggest that the reporter is also able to monitor the artificial-tether-induced MERCs (Fig. 2A–C).
Characterization of the MERC reporter. U2OS cells with MERC reporter stably expressed were transfected with an artificial ER–mitochondria tether (B–B″,C), or PTPIP51–Myc and VAPB–RFP expression vectors (E–E″,F), or PTPIP51 and VAPB siRNAs (H–H″,I), respectively. The cells then were fixed and stained with anti-TOM20 antibodies to label mitochondria. Expression of the tether and VAPB are indicated by RFP fluorescence and expression of PTPIP51 by anti-Myc antibody staining. The expression of ER–mitochondria tether (tether, red in B) (A–C) or the overexpression of PTPIP51 (blue in E) and VAPB (red in E) (D–F) enhanced the size and intensity of the reporter. The RNAi of PTPIP51 and VAPB decreased the size and intensity of the MERC reporter (G–I). The middle panels of the fluorescence images (A′,B′,D′,E′,G′,H′) indicate the green channel. The right panels of the fluorescence images (A″,B″,D″,E″,G″,H″) show the heat map of the green channel. (C,F,I) Quantification of the fluorescence images. The sizes of the contacts were calculated by using the ratio between the areas of contacts to the areas of the mitochondria. n=50 cells were analyzed. The experiments shown were replicated three times. ***P<0.001 (two tailed t-test). Data are presented as means±s.d. (see also Figs S2 and S3).
It has been reported that VAPB and PTPIP51 interact with each other, and they are required for MERC formation in mammalian cells. When VAPB and PTPIP51 were overexpressed, the number of MERCs was increased (Stoica et al., 2014). In line with this finding, when VAPB and PTPIP51 were overexpressed in the cells expressing the reporter, the GFP signals of the reporter were intensified (Fig. 2D–F). When we blocked VAPB and PTPIP51 expression in cells that expressed the reporter using RNA interference (RNAi), the GFP puncta decreased dramatically (Fig. 2G–I). To exclude the possibility that the change in GFP signals is due to the change in spGFP protein expression, we quantified spGFP1-10 levels by western blotting. spGFP1-10–ERt was designed with a V5 tag inserted between spGFP1-10 and the ER targeting sequence (Fig. S3A). Using anti-V5 antibody to assess spGFP1-10–ERt expression, we found that spGFP1-10 levels did not change significantly when VAPB and PTPIP51 expression was reduced. This suggests that the decrease in GFP signals is not due to the decrease in spGFP levels (Fig. S3B,B′) and that the MERC reporter can accurately reflect an increase or decrease in such contacts in cells.
MERCs are dynamic structures
Both the ER and mitochondria are highly dynamic organelles (Pendin et al., 2011; Youle and van der Bliek, 2012) that keep changing their structures through fusion and fission. It would be interesting to find out whether MERCs are also dynamic structures. We transfected the cells expressing the reporter with Mito–RFP, did a time-lapse recording of the live cells, and then used particle tracing to analyze their dynamic patterns.
In comparison with the highly dynamic mitochondria, most of the contacts are not very active. They usually stay at the spot where they were first formed and occasionally show small movement. Approximately 27% of the contacts were short-lived, lasting less than 60 s from their appearance to disappearance and ∼22% of the contacts lasted between 60 and 250 s. The remaining 51% of the contacts were quite stable, lasting more than 250 s (Fig. 3A–C; Movie 1). These data suggest that although they are relatively stable, MERCs are dynamic structures that could move around and form or dissolve themselves frequently. The dynamic patterns of the contact reporter further support the notion that the interactions between spGFP1-10 and spGFP11 were reversible under this experimental condition.
MERCs are dynamic structures. (A) U2OS cells with MERC reporter stably expressed were transfected with Mito–RFP. Time-lapse movies recording the cells were collected with a Yokogawa spinning-disk confocal microscope and analyzed using the ImageJ plugin particle tracker. A typical frame is shown. The trajectories of each contact are indicated by lines in different colors. (B) The statistics of proportions of the MERCs with different durations (short lived <60 s, medium 60–250 s, long lived >250 s). n=6 cells were analyzed. Data are presented as means±s.d. (C) A typical time-lapse series to show MERCs with different durations (white arrow: a contact lasted less than 60 s; pink arrow: a contact lasted more than 60 s but less than 250 s; green arrow: a contact lasted longer than 250 s). (D) Mitochondrial fission sites are associated with MERCs (white arrow). Mitochondria are marked by Mito–RFP (red), MERCs by GFP (green) (see also Movie 1).
It has been reported that MERCs mark the mitochondrial fission spots (Friedman et al., 2011). Some MERCs indeed were associated with mitochondrial fission sites in our time-lapse images. After fission, the contact signal localized to one side of the mitochondria (Fig. 3D). We also observed that some contact sites persisted without mitochondrial fission (Fig. 3D).
Changes in MERCs during a mitotic cell cycle
During the mitotic cell cycle, mitochondria and the ER undergo dramatic morphological changes (Jongsma et al., 2015). However, the dynamics of MERCs are not known. The cells stably expressing the MERC reporter were synchronized. Next, the cells were harvested at different time points and stained with the mitochondrial marker TOM20. To distinguish G1–S and G2 phases, we stained the cells with anti-PCNA and anti-Aurora-B antibodies (Mitra et al., 2009) (Fig. 4). In the G1–S phase, the majority of mitochondria are known to become elongated and form networks (Mitra et al., 2009). In this phase, the MERC reporter signals were observed as small puncta that decorated the mitochondria (Fig. 4). In the G2 phase, mitochondria began to get fragmented, and GFP puncta of the contact reporter covered more of the mitochondrial surface than they did in the G1–S phase (Fig. 4). During prophase, metaphase, anaphase and telophase, the majority of the mitochondria were fragmented and clustered. The GFP signals of the contact reporter did not change much compared with the GFP signal level in the G2 phase (Fig. 4). Statistical analysis also revealed that during the G1–S phase, MERCs covered less mitochondrial surface than during the other the stages of cell cycle (Fig. 4B).
Changes of MERCs during the cell cycle. U2OS cells stably expressing MERC reporter were synchronized and different cell cycle phases were judged by DAPI, anti-PCNA and anti-Aurora B antibody staining. MERCs were indicated by GFP signals (green) and mitochondria were marked by anti-TOM20 staining (red). (A) At G1–S phase, MERCs covered less surface area of mitochondria than they did at other cell cycle stages. PCNA and Aurora B staining were used to distinguish G1–S and G2 phases. Other cell cycle stages were distinguished by DAPI staining. (B) Quantification of the size of contact in cells at different cell cycle phases. n=50 cells were analyzed. The experiments shown were replicated three times. **P<0.01, ***P<0.001 (two tailed t-test). Data are presented as means±s.d.
Mitochondrial defects trigger remodeling of MERCs
Because MERCs mediate the cross-talk between the ER and mitochondria, we wanted to check if there were any changes in its reporter signals when there were mitochondrial defects. Hence, we incubated the cells that stably expressed the MERC reporter with carbonyl cyanidem-chlorophenyl hydrazone (CCCP), a chemical that causes depolarization of the mitochondrial membrane and inhibition of oxidative phosphorylation (Minamikawa et al., 1999). After CCCP treatment, the mitochondria became fragmented, and the GFP signals of the MERC reporter occupied a greater proportion of the mitochondrial surface than they did in the cells subjected to mock treatments (Fig. 5A,B), suggesting that the association between ER and mitochondria was enhanced. Similar phenomena were observed when we treated the cells with oligomycin A (Fig. 5A,B), a chemical that reduces oxidative phosphorylation of ADP to ATP through inhibition of the proton channel subunit of ATP synthase (Jastroch et al., 2010). Both CCCP and oligomycin A treatment led to fragmentation of mitochondria, suggesting that the mitochondrial morphology change was crucial for the association between ER and mitochondria. To test this idea, we treated the reporter-expressing cells with mitochondrial division inhibitor 1 (mdivi-1), an inhibitor of Drp1, to block mitochondrial fission (Cassidy-Stone et al., 2008). The mitochondria became elongated after treatment with mdivi-1, and the contact signals were reduced significantly (Fig. 5A,B). To exclude the side effects of drug treatments, we also examined the contacts in cells expressing the dominant-negative form of Drp1 (Drp1-K38A). The expression of Drp1-K38A indeed reduced contact signals (Fig. S4A,C). However, the reduction of MERCs in Drp1-K38A-expressing cells was not as dramatic as that in the cells after mdivi-1 treatment. This is probably because only a small portion of mitochondria become elongated, but the rest remain clustered in these cells. GFP signals were reduced on the highly elongated mitochondria but did not change dramatically on the clustered mitochondria. We then used our reporter to examine MERCs in cells by OPA1 RNAi. OPA1 encodes a GTPase that is essential for mitochondrial inner membrane fusion. The loss of OPA1 leads to mitochondrial fragmentation (MacVicar and Langer, 2016). The contact signals were significantly increased in cells expressing OPA1 siRNA (Fig. S4B,D). This finding indicates that mitochondrial fragmentation promotes the association between ER and mitochondria. We assessed the levels of spGFP1-10 in the cells after CCCP and mdivi-1 treatments. CCCP treatment led to a slight, but not statistically significant, reduction in spGFP1-10 levels. Furthermore, Mdivi-1 treatment did not cause any statistically significant changes in spGFP1-10 levels (Fig. S3C). These data suggest that the reduction in GFP signals in the treated cells was not due to the changes in protein expression.
Mitochondrial defects affect MERCs. U2OS cells with MERC reporter stably expressed were treated with DMSO (Ctrl), CCCP (10 μM, 4 h), mdivi-1 (50 μM, 4 h) and oligomycin A1 (10 μg/ml, 4 h), respectively. The cells were fixed and stained with anti-TOM20 antibodies (red). MERCs are indicated by the green signals. (A) CCCP and oligomycin A treatment increased MERCs and mdivi-1 treatment reduced MERCs. Mitochondria were labeled with TOM20 staining. (B) Quantification of the size of contacts in cells treated with different drugs. n=50 cells were analyzed. The experiments were replicated three times. *P<0.05,**P<0.01 (two tailed t-test). Data are presented as means±s.d. (see also Figs S3 and S4).
A change in ER morphology does not affect MERCs
We then wanted to determine whether changes in ER morphology would remodel MERCs. Atlastin (ATL) proteins are a family of GTPases that are required for shaping ER morphology (Hu and Rapoport, 2016). To test the effects of ER morphology on MERCs, we overexpressed wild-type ATL1 or a GTP-binding mutant, ATL1-K80A. When ATL1 was highly expressed, the ER formed aberrant sheet-like structures in cells (Hu et al., 2009). In contrast, overexpression of ATL1-K80A resulted in long and unbranched ER tubules (Hu et al., 2009). We examined the MERC reporter in these cells and did not observe any obvious changes in the intensity or size of the GFP puncta (Fig. 6A,C). Therefore, we concluded that the MERCs are not sensitive to changes in ER morphology.
ER morphology changes and ER stress do not change MERCs significantly. U2OS cells with MERC reporter stably expressed were transfected with sec61β–RFP, ALT1–HA or ALT1-K80A–HA expression vectors. The cells were fixed and stained with anti-HA (red) and anti-TOM20 (blue) antibodies. MERCs are indicated by green signal. ALT1 or ALT1-K80A overexpression changed the ER morphology but did not affect the size and intensity of MERCs. Sec61β–RFP overexpression served as a control. (B) U2OS cells with MERC reporter stably expressed were treated with DMSO or tunicamycin (0.5 μg/ml, 4 h). The cells were fixed and stained with anti-TOM20 (red) antibodies and DAPI (blue). MERCs are indicated by green signal. Tunicamycin treatment does not change MERCs significantly. (C) Quantification of the size of MERCs in the cells with indicated treatments. n=50 cells were analyzed. The experiments shown were replicated three times. NS, not significant (two tailed t-test). Data are presented as means±s.d. (see also Fig. S5).
We also decided to test whether ER stress causes remodeling of MERCs. Tunicamycin is a drug that blocks N-linked glycosylation (N-glycans) during glycoprotein synthesis and therefore leads to the unfolded protein response (UPR), causing ER stress (Kaufman, 1999). Cells stably expressing the contact reporter were treated with 0.5 μg/ml tunicamycin for 4 h, which was sufficient to induce ER stress. We found that there were no dramatic differences in the GFP signals between the tunicamycin-treated or the mock-treated cells (Fig. 6B,C). When the tunicamycin concentration and the treatment duration were increased to 4 μg/ml and 10 h, respectively, only a slight increase in MERC signals was observed (Fig. S5). In agreement with this, a recent study that examined MERCs in tunicamycin-treated cells through TEM analysis did not observe any dramatic changes in the MERCs (Zhao et al., 2017). This suggests that ER stress does not induce dramatic remodeling of MERCs.
Starvation and apoptosis induction, but not lipid feeding and mtUPR, lead to the remodeling of MERCs
To test the responses of MERCs to different cellular stresses, we treated the U2OS cells that stably expressed the contact reporter with various stimuli and examined the changes in GFP signals from the reporter. We treated the cells with HBSS for 12 h or 20 h to starve the cells and found that the intensity and area of GFP signals of the contact reporter were greatly increased, suggesting that the association between ER and mitochondria was increased (Fig. 7A,B; Fig. S6). We examined the levels of spGFP1-10 protein in HBSS-treated cells by western blotting and did not observe any obvious changes in protein expression, suggesting that the increase in GFP signals was not due to the increase in spGFP protein expression (Fig. S3D,D′). We then treated the cells with HBSS for 12 h and then re-fed the cells with normal culture medium for another 12 h and detected the contact reporter. The GFP signals were significantly increased upon starvation and returned to normal levels after cell re-feeding (Fig. S6). These data further suggest that the interaction between spGFP1-10 and spGFP11 is reversible. When we treated the cells with staurosporine (STS) to induce apoptosis (Belmokhtar et al., 2001), we found that the GFP signals of the contact reporter covered a larger area of mitochondrial surface than they did in controls (Fig. 7A,B), suggesting that apoptosis induction enhances the association between the ER and mitochondria. The levels of spGFP1-10 in STS-treated cells were examined by western blotting. Although STS treatment reduced the levels of both spGFP1-10 and tubulin, the relative level of spGFP1-10 did not change significantly (Fig. S3C,C′). These data suggested that the change in the reporter signals was not due to the change in spGFP1-10 levels.
Starvation and induction of apoptosis, but not lipid feeding and mtUPR lead to MERC remodeling. U2OS cells with MERC reporter stably expressed were used for all the treatments. (A) For OA treatment, cells were incubated 200 mM oleic acid for 24 h or control medium (Ctrl). Cells were then fixed and stained with anti-TOM20 antibody (blue) to label mitochondria and 10 ng/ml Nile Red (red) to indicate the lipid droplets. MERCs are indicated by the green signal. OA feeding increased cytosolic lipid droplets, but did not change the size of MERCs. For starvation treatment, cells were incubated in HBSS for 20 h. For STS treatment, cells were incubated with 10 μM STS for 1 h. Cells were then fixed and stained with anti-TOM20 antibody (blue) to label mitochondria. MERCs are indicated by the green signal. Starvation and STS treatment could increase MERCs. To induce mtUPR, Myc-tagged OTC-D was transiently expressed in the cells. As a control, Myc-tagged OTC was also transiently expressed. Cells were then fixed and stained with anti-TOM20 (blue) and anti-Myc (red) antibodies. MERCs are indicated by the green signals. The mtUPR induced by OTC-D overexpression did not change MERCs. (B) Quantifications of the size of contacts in the cells with indicated treatments. n=50 cells were analyzed. The experiments shown were replicated three times. ***P<0.001; NS, no significance (two tailed t-test). Data are presented as means±s.d. (see also Figs S3 and S6).
It has been reported previously that obesity leads to an increased association of ER and mitochondria in the mouse liver (Arruda et al., 2014). We decided to test whether feeding the cells with oleic acid (OA) to induce acute lipid droplet accumulation could lead to MERC remodeling. OA treatment induced the accumulation of a large amount of lipid droplets as indicated by Nile Red staining (Singh et al., 2009). In contrast, there was no dramatic change in the GFP signal of the contact reporter and there was also no close association between the lipid droplets and the GFP puncta (Fig. 7A,B).
We then tested whether mtUPR could change the contacts between the ER and mitochondria. We overexpressed a mutant form of OTC, OTC-D, which is misfolded and has been reported to induce mtUPR (Zhao et al., 2002). Cells expressing wild-type OTC were used as control. No dramatic change in reporter expression patterns was observed when these proteins were expressed (Fig. 7A,B), suggesting that the mtUPR does not affect association between the ER and mitochondria.
DISCUSSION
MERCs play crucial roles in the exchange of Ca2+, lipids and other metabolic materials between the ER and mitochondria and mediate cross-talk between them. Many studies suggest that the contact regions serve as platforms for various cellular processes (Murley and Nunnari, 2016). Nonetheless, many more functions of this structure are yet to be identified and the molecular basis of this structure is not well understood in mammalian cells. One of the major obstacles to understanding the biology of MERCs is the lack of good markers.
In this study, we designed a reporter for labeling the MERCs. We showed that the reporter can reliably reflect the increase or decrease of the ER–mitochondria association without inducing obvious artifacts when it is transiently expressed at a moderate level. However, we did observe a change in mitochondrial morphology when the reporter was expressed at a high level. It is difficult to judge whether the change in morphology is due to some specific features of the reporter or occurs simply because of the effects of protein overexpression. Therefore, it is advisable to use stable cell lines expressing the contact reporter to assess MERCs.
It has been demonstrated that the interaction between the two components of split GFP is irreversible in vitro (Cabantous et al., 2005; Magliery et al., 2005). However, it has not been tested whether the interaction between spGFP1-10 and spGFP11 could be reversed if there were forces to pull them apart. In this study, we fused spGFP1-10 and spGFP11 with ER and mitochondrial targeting sequences, respectively, in order to examine MERCs. When two organelles are close enough, spGFP1-10 and spGFP11 bind to each other and emit fluorescence. We found that expression of the reporter did not increase the tether between ER and mitochondria. The GFP signals from the contacts were highly dynamic. They could appear and disappear within minutes. In addition, the contact GFP signals increased during cell starvation and deceased with cell re-feeding. All these data suggested that the interaction between spGFP1-10 and spGFP11 was reversible in our experimental settings, and thus the reporter could faithfully reflect the dynamic changes of MERCs. With this reporter, one can probably design a screening experiment to identify molecules required for the organization of the contacts, examine the dynamic changes in the contacts under different conditions or introduce the reporter into an animal model, and examine the contacts under different physiological conditions. The conduct of such research studies was very difficult or even impossible before our invention of the contact reporter.
We analyzed the contacts in live cells and found that they were dynamic structures. Nonetheless, the contacts seemed less dynamic than the highly dynamic mitochondria. Most contacts made movements only within its local region without long-distance traveling. They could appear or disappear within minutes. These data suggest that the contacts usually do not move along the mitochondria but rather assemble or disassemble locally. We found that many contact sites persisted without mitochondrial fission, which is in line with a previous study that showed that MERCs were not rate-limiting for mitochondrial division (Lewis et al., 2016).
In this study, we demonstrated that the morphology of mitochondria, but not that of the ER, affected the association between the ER and mitochondria. We found that MERCs increased when mitochondria were fragmented. Consistent with what we observed, it has been reported previously that when the hepatic mammalian target of rapamycin complex 1 (mTORC1) signaling pathway is inhibited, the density of mitochondrial cristae drops and MERCs double in length (Sood et al., 2014). It was surprising that ER morphology did not have a major influence on the contacts between the ER and mitochondria. As ER is more abundant than mitochondria inside the cell, mitochondria may be determining the position of the contacts. Further research is needed to elucidate whether there are changes in the fine structure or function of the contacts when the ER morphology changes.
Furthermore, our data suggest that fragmented mitochondria have more contacts with the ER. It would be interesting to test whether this phenomenon is due to a change in the distribution of key surface proteins and lipids or simply due to an alteration in the curvature of the membranes.
We found that HBSS-induced starvation led to an increase in MERCs in the culture cells. This increase is not due to mitochondrial fragmentation, as mitochondria are known to become elongated after this treatment (Gomes et al., 2011). In the mouse model, mTORC1 was inhibited without an activated autophagy program and the mitochondria network became fragmented (Sood et al., 2014). Even though MERCs were increased in both cases, different mechanisms might be involved.
As in any newly developed tool, the reporter of MERCs also has its potential caveats. The contacts with very close proximity between ER and mitochondria might not have enough space to allow GFP to re-fold properly. Therefore, these contacts might not be detected through fluorescence. In addition, although the interaction of spGFP1-10 and spGFP11 could be reversed in our experimental setting, the stable nature of the refolded GFP might make the reporter less responsive to the subtle changes in the contacts. Despite these potential problems, the reporter we designed here is very easy to use. It shows reliable responses to the gain and loss of MERCs and offers great spatiotemporal options for MERC research.
MATERIAL AND METHODS
Plasmids and siRNAs
Myc-UBE2J2 plasmid was constructed by inserting full-length UBE2J2 cDNA into pXF3HM-HIS-6MYC vector. RFP-Sec61β plasmid was constructed by inserting Sec61β cDNA into a modified pEGFPC1 vector in which the EGFP coding sequence was replaced with an RFP coding sequence. To construct Myc-ERt plasmid, cDNA encoding the residues 228–259 of UBE2J2 was inserted into pXF3HM-HIS-6MYC vector. Mitot-GFP plasmid was made by inserting cDNA encoding the fist 59 amino acids of TOMM70 into the pEGFPN1 vector. To construct plx304-spGFP1-10-ERt, cDNAs encoding spGFP1-10, a V5 tag, and the residues 228–259 of protein UBE2J2 were amplified and cloned into plx304 plasmids using Gateway cloning system. To construct pLVX-Mitot-spGFP11, cDNAs encoding the first 59 amino acid of TOMM70 and spGFP11 were inserted into pLVX-IRES-puro vector. To construct pLVX-Mitot-l1-spGFP11 and pLVX-Mitot-l2-spGFP11, cDNAs encoding the first 59 amino acid of TOMM70, linker and spGFP11 were inserted into pLVX-IRES-puro vector. The protein sequences of the linkers were GSGSNGSSGGGSGGGSG (pLVX-Mitot-l1-spGFP11) and GSGSNGSSGSGGSGGGGGGSRGGSGGGGSG (pLVX-Mitot-l2-spGFP11). To construct pLVX-Mitot-(1×, 2×, 4×)spGFP11, cDNA encoding the first 59 amino acid of TOMM70, one, two or four copies of spGFP11, and linker sequences GSGSNGSSGSGGSGGGGSGGSRGGSGGGGSGG were inserted into pLVX-IRES-puro vector. To construct pLVX-Mitot-7×spGFP11, cDNA encoding the first 59 amino acid of TOMM70 and 7 copies of spGFP11 were cloned into pLVX-IRES-puro vector. To construct ER-mitochondria tether, cDNA encoding the first 30 amino acids of AKAP1, a red fluorescent protein (RFP) protein, and an ER anchor from protein SACM1L (521–587 residues) were inserted into the modified pEGFPN1 vector in which the EGFP coding sequences were removed. Sequences encoding a linker AEAAAKEAAA KEAAAKA were inserted between AKAP1 fragment and RFP. Sequences encoding another linker GSGGSGSGSSGGSGS were inserted between RFP and SACM1L fragment. To construct PTPIP51-Myc, PTPIP51 cDNA was inserted into pcDNATM 3.1Myc-his(-)A. To construct VAPB-RFP, VAPB cDNA was cloned into a modified pEGFPN1 vector in which the EGFP coding sequences were replaced by the RFP coding sequences. To construct OTC-Myc and OTC-D-Myc plasmids, cDNA encoding OTC and OTC-D fragments were subcloned from plasmids OTC [Addgene, 71877; deposited by Nicholas Hoogenraad (Zhao et al., 2002)] and OTC-delta [Addgene, 71878; deposited by Nicholas Hoogenraad (Zhao et al., 2002)] into pcDNA-TM 3.1myc-his(-)A. To construct HA-ATL1 plasmids, ATL1 cDNA was inserted into pcDNA3.1-HA (N). HA-ATL1-K80A and mCherry-Drp1 K38A plasmids were obtained through site directed mutagenesis. The Mito–RFP construct was described previously (Stoica et al., 2014). The sequences of siRNA targeting PTPIPT1 and VAPB were used as described previously (Stoica et al., 2014). The sequences of siRNA targeting OPA1 were 5′-GCCUGACAUUGUGUGGGAAUU-3′.
Cell culture, transfection and generation of stable cell lines
U2OS cells were originally from ATCC and were recently authenticated and tested for contamination. Cells were cultured in DMEM supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin, at 37°C with 5% CO2. Cells were transfected with appropriate constructs using Lipofectamine 2000 according to the manufacturer's instructions. Similarly, cells were transfected with appropriate siRNAs by using Lipofectamine RNAiMAX according to the manufacturer's instructions.
The stable cell lines with ER–mitochondrial contact reporter expression were generated by lentiviral infection as described previously (Naviaux et al., 1996). U2OS cells were infected with pLVX -Mitot-spGFP11×2 and plx304-spGFP1-10-ERt and selected with the correspondent antibiotic puromycin and blasticidin. Then, single clones with moderate expression levels were selected.
Cell synchronization and drug treatment
The double thymidine block method was used for cell synchronization. Briefly, 25–30% confluenct U2OS cell cultures were incubated with 2 mM thymidine for 16 h (first block), followed by washing with 1× PBS and adding fresh DMEM for 8 h to release the cells. Then, the cells were incubated with 2 mM thymidine for another 16 h (second block) and thymidine was removed by washing the cells with 1× PBS. Then, the cells were released by adding fresh DMEM. Thymidine nocodazole block was also used. Briefly, 40% confluency of cell cultures were incubated with 2 mM thymidine in DMEM for 16 h. After removal of thymidine, fresh DMEM was added for 3 h to release the cells. Then 100 ng/ml nocodazole was added to the medium for 12 h and nocodazole was removed by washing with 1× PBS and fresh DMEM was added to release the cells. The different cell cycle phases were judged by DAPI, anti-PCNA and anti-Aurora B antibody staining.
The following drug concentrations and treatment durations were used: CCCP (10 μM, 4 h), mdivi-1 (50 μM, 4 h), STS (10 μM, 1 h), oligomycin A1 (10 μg/ml, 4 h), tunicamycin (0.5 μg/ml, 4 h or 4 μg/ml, 10 h) and oleic acid (200 mM, 24 h). For starvation, cells were incubated in HBSS for 12 h or 20 h. For re-feeding, HBSS was substitute with complete culture medium after 12 h of starvation. Cells were fixed immediately after each treatment for further analysis.
Antibodies and reagents
CCCP, mdivi-1, puromycin, blasticidin, thymidine, nocodazole, MG132, oligomycin A1, tunicamycin and Nile Red were obtained from Sigma. Anti-Tom20 antibody (rabbit) (1:100) was from Sigma (HPA011562). Anti-Tom20 (mouse) antibody (1:200) was from BD Biosciences (612278). Anti-Aurora B antibody (1:250) was from BD Biosciences (611082). Anti-HSP60 antibody (1:400) was from Cell Signaling (12165), anti-HA antibody (1:1000) was from Cell Signaling (3724) and anti-PCNA antibody (1:1000) was from Cell Signaling (2586S). Anti-Myc antibody (1:250) was from Santa Cruz (sc-40). Anti-MFN2 antibody (1:1000 for western blot) was from Abnova (H0009927-M03). Anti-V5 antibody (1:1000 for western blot) was from Life Technologies (46-0705). Anti-α-tubulin antibody (1:1000 for western blot) was from Boster (BM1452). Lipofectamine 2000 and Lipofectamine RNAiMAX were from Life Technologies.
Immunofluorescence
Cells grown on glass coverslips were fixed in 4% formaldehyde for 20 min. Then, they were permeabilized with 0.1% Triton X-100 for 10 min and blocked with 3% bovine serum albumin for 30 min. Cells for PCNA staining were fixed with 100% methanol for 5 min at −20°C. The cells were then incubated with appropriate primary antibodies at 4°C overnight followed by washing with PBS thoroughly. Then the cells were stained with fluorescent secondary antibodies for 1 h at room temperature. After washing, the samples were mounted for confocal microscopy. For lipid droplet staining, Nile Red (10 ng/ml) was added directly to the fixed cells and incubated for 5 min.
Fluorescence microscopy and live-cell imaging
Fluorescence images were acquired on Zeiss 710 confocal microscope and GE DeltaVision OMX. The 3D rendering was made by Imaris NIS 4.3 (BitPlane). Live-cell imaging was performed using a Yokogawa spinning-disk confocal microscope. Linear adjustments were made with ImageJ if necessary. Particle tracking was done by using the ImageJ plugin particle tracker (Sbalzarini and Koumoutsakos, 2005).
Transmission electron microscopy
TEM was performed as described previously (Zhang et al., 2016) using a Hitachi HT7700 electron microscope.
Statistical analysis
The areas of contact sites and mitochondria in each cell were measured and calculated by ImageJ. The size of contacts was quantified as the area of contact divided by the area of the mitochondria. Data were expressed as means±s.d. or means±s.e.m. Statistical analysis among groups was performed using Student's t-test. *P<0.05, **P<0.001, ***P<0.0001. The statistical analyses were performed using GraphPad Prism 5.
Acknowledgements
We are grateful to Dr Bin Zhao and Dr Zhefeng Gong for plasmids. Dr Tong is supported by National Key Research & Developmental Program of China (SQ2017YFSF080009, SQ2017YFSF080003), National Natural Science Foundation of China (31622034, 31571383), and Zhejiang Provincial Natural Science Foundation of China (LR16C070001).
Footnotes
Competing interests
The authors declare no competing or financial interests.
Author contributions
Conceptualization: C.T.; Methodology: Z.Y., X.Z., J.X., W.S., C.T.; Validation: X.Z., J.X., C.T.; Investigation: Z.Y., X.Z., J.X., W.S.; Data curation: Z.Y., X.Z., C.T.; Writing - original draft: Z.Y., X.Z., C.T.; Writing - review & editing: C.T.; Supervision: C.T.; Project administration: C.T.; Funding acquisition: C.T.
Funding
Dr Tong is supported by the National Key Research & Developmental Program of China (2017YFC1001500, 2017YFC1001100), the National Natural Science Foundation of China (31622034, 31571383), the Zhejiang Provincial Natural Science Foundation of China (LR16C070001) and the Fundamental Research Funds for the Central Universities.
Supplementary information
Supplementary information available online at http://jcs.biologists.org/lookup/doi/10.1242/jcs.208686.supplemental
- Received July 18, 2017.
- Accepted November 13, 2017.
- © 2018. Published by The Company of Biologists Ltd