Generation of a bimolecular complementation marker pair for the quantification of mitochondrial fusion
In order to develop a high-throughput mitochondrial fusion assay, we adapted a bimolecular complementation strategy that would allow multiple outputs [15]. cDNAs were generated encoding mitochondrial targeted split Venus and Renilla luciferase separated by a leucine zipper. Fusion between mitochondria expressing each of these chimeric proteins would lead to their dimerization through the leucine zipper motifs, promoting the functional assembly of both luciferase and Venus yellow fluorescent protein (YFP) [16, 17]. These constructs were designated N-MitoVZL (venus-zipper-luciferase) and C-MitoLZV (luciferase-zipper-venus; Figure 1a).
The constructs were used to generate two populations of stable suspension HeLa (sHeLa) cell lines, each expressing one of the split constructs. One litre of each stable line of sHeLa cells was grown to confluence and harvested by centrifugation and mitochondria were purified by differential centrifugation and flash frozen at -80°C.
As mitochondria are known to function quite well in vitro, most notably for metabolic and import reactions, we based the conditions of our assay on established buffering and energizing conditions [18, 19]. Mitochondria of each population (50 μg each) were incubated with 5 mg/mL cytosol, GTP (0.5 mM) and an adenosine triphosphatase/succinate mixture, in a final volume of 25 μL. After mixing, the mitochondria were first concentrated by centrifugation and incubated on ice for 30 min. As reported in the yeast cell-free mitochondrial fusion assay, this step helps to facilitate mitochondrial docking [5], thereby enhancing the reaction. Following this step, the mitochondria were resuspended within the reaction buffer and incubated at 37°C for 30 min. Following the reaction, the extent of mitochondrial matrix content mixing could be detected by confocal or electron microscopy (Figure 1), or using a luciferase assay (Figure 2). For confocal imaging, the fusion reaction included MitofluorRed633 to label all of the respiring mitochondria. In order to quench potential background complementation that might occur due to the presence of ruptured mitochondria, 25 μg trypsin was added following the fusion reaction and incubated on ice for 20 min. This treatment also resulted in the dispersal of mitochondria, which helped to resolve the fused mitochondria by microscopy (not shown). The trypsin was then inactivated by incubation with soy bean trypsin inhibitor for a further 20 min on ice. Finally, the mixture was mounted on slides and imaged by confocal microscopy (Figure 1b). The mitochondria were first identified using the potentiometric dye and fused mitochondria, which represented 12% (+/- 2.4%) of the entire population, were observed through the visualization of the Venus YFP fluorescence (Figure 1b). These data reveal the complementation of Venus YFP within the matrix of fused mitochondria. In order to test whether mitochondrial fusion is nucleotide dependent, we added apyrase directly into the fusion reaction to deplete NTPs. This control reaction in the absence of energy revealed significantly fewer fused mitochondria within a field (4.0 +/- 2.0%).
As a second output to quantify mitochondrial fusion, we decided to quantify any changes in the size of the mitochondria [5]. In order to examine this, we prepared fusion reactions performed at 4°, where docking should be facilitated, and reactions performed at 37°, for analysis by electron microscopy (EM). The trypsinization of the fusion reaction was omitted for this EM analysis to minimize the required steps. As can be seen in Figure 3a, the light membrane preparation of mitochondria used in our assays do contain other organelles. Western blot analysis showed that this preparation was highly enriched for mitochondria (anti-TOM20) and there was an absence of contaminating cytosol as indicated by the loss of tubulin signal (Additional File 1: Figure S1). Analysis of other organelle markers revealed the preparation was devoid of early endosomes and golgi, but retained ER and lysosomal markers (Additional File 1: Figure S1). The mitochondria within the preparation were largely intact, as indicated by the asterisks shown in Figure 3a. The majority of the mitochondria in these preparations are circular, with over 80% having a length:width ratio of 1:1 (not shown). When incubated at 4°, the mitochondrial diameter averages 500 nm. This diameter is increased upon incubation at 37°, with organelles regularly spanning 1 to 1.5 microns (Figure 3b). We quantified the size of 300 mitochondria from a fusion reaction completed at 37°C or 4°C and binned them into four categories. The data show that upon incubation at 37°C, there is an increase in the pool of mitochondria greater than 600 nm, from 21% to 44% of organelles counted (Figure 3b). This number is higher than the 12% observed by confocal imaging, however it should be noted that the complementation of Venus YFP can only be seen with the fusion of the N-MitoVZL with C-MitoLZV, where any homotypic fusion events will not be scored. However, the EM analysis scores all mitochondria, which may explain the increased numbers of larger organelles.
Mitochondrial fusion is dependent on nucleotide hydrolysis and requires a charge gradient across the inner membrane
The confocal imaging confirmed the complementation of the Venus fluorescent protein (Figure 1), and the EM analysis demonstrated an increase in overall mitochondrial length (Figure 3). While these two outputs confirm the specificity of the assay, our main objective was to utilize the Renilla luciferase activity to provide a robust and high throughput method to score for mitochondrial fusion, with a very high signal-to-noise ratio. Figure 2a shows a representative reaction, where the basal reaction, containing cytosol, GTP and adenosine triphosphate regenerating system results in a signal around 35,000 relative luciferase units (RLU). Since our assay conditions also support mitochondrial protein import [18, 20], it was possible that the signal we obtained may be due to the import of any free, cytosolic marker into the complementary mitochondria. This was not the case since the incubation of N-MitoVZL with cytosols taken from cells expressing C-MitoLZV, or visa versa, did not result in any signal (Additional File 1: Figure S2).
The absolute value of RLU obtained in the assay depended primarily on two factors. First, each mitochondrial preparation had a different load of the labels. Second, and perhaps more importantly, over the course of this study it became apparent that the length of time the mitochondria were left on ice prior to starting the reaction could reduce the efficiency of fusion. The dependencies on temperature and energy were unaffected, but the total values would decrease. Therefore it was important to use the mitochondria within the first 30-60 minutes of thawing them out (data not shown).
In order to test whether mitochondrial fusion is nucleotide and temperature dependent, we again added 0.1 U/μL apyrase, and performed a reaction at 4°C. As expected, mitochondrial fusion was not supported in the absence of energy, with ~400 RLU. Similarly, mitochondrial fusion required physiological temperature, with a reaction performed on ice resulting in only ~1200 RLU (Figure 2a). These results establish that the enzymatic amplification of the signal provides a very clear signal-to-noise ratio, which is essential for a high-throughput design. In order to estimate the efficiency of mitochondrial fusion in vitro, we disrupted all of the mitochondria within the reaction using 5 min pulses of sonification to allow maximum complementation at 37°, independently of a fusion event. The trypsin step, which is required to remove any background complementation coming from broken mitochondria after the reaction, was omitted in this reaction. With this, we detected ~227,500 RLU. Given that the basal reaction in this experiment gave 35,000 counts, it indicates that approximately ~15% of mitochondria were fusing under basal conditions (Figure 2a). In all of the reactions performed with this system, the efficiency has been seen to range between 5%-20%, depending primarily on the mitochondrial preparation (not shown). This is similar to other cell free fusion assay systems, including the yeast mitochondrial fusion system [5, 21, 22]. As mentioned above, this number is likely an under-representation of the total efficiency since it only scores for 'heterotypic' fusion events.
In a time course to examine the kinetics of mitochondrial fusion, we observe a steady increase in the fusion signal in the first 30 min. However, longer incubation times led to a decrease in the fusion signal, possibly due to degradation of the marker pairs (Figure 2b). The addition of non-hydrolysable analogues ATPγS and GTPγS inhibited mitochondrial fusion, consistent with a requirement for nucleotide hydrolysis. Omission of exogenous GTP led to a twofold reduction in fusion, suggesting that endogenous GTP within the mitochondria is limited and required for mitochondrial fusion (Figure 2c).
In order to determine the effect of the membrane potential across the inner membrane on mitochondrial fusion, we added 2 μM of the protonophore CCCP (carbonylcyanide-3-chlorophenylhydrazone). Fusion was abolished (Figure 2d) which is in agreement with previous studies in intact cells [8, 23] and the yeast cell-free mitochondrial fusion assay [5]. To further dissect the membrane potential requirements, we examined the effect of valinomycin (1 μM) and nigiricin (2 μM). Valinomycin is an electrogenic potassium ionophore which abolishes the charge gradient. Electroneutral potassium/proton ionophore nigiricin retains the charge, but ablates the proton gradient. Valinomycin diminished the fusion reaction by 97%, whereas nigiricin, even at higher concentrations, inhibited the reaction by only 27% (Figure 2d). We conclude that mammalian mitochondrial fusion in vitro requires the charge gradient across the mitochondrial inner membrane but is independent of the chemical gradient.
Mitochondrial fusion is modulated by signalling pathways
Similar to the yeast mitochondrial fusion reaction, the reaction was efficient in the absence of exogenous cytosol. Nonetheless, there was a profound consequence of cytosol addition, depending on the state of the cell at the moment of cytosol preparation. Figure 4a shows that the addition of 0.1, 0.3 and 3.0 mg/mL (final) cytosol prepared from non-confluent sHeLa cells undergoing normal proliferation stimulated the reaction in a concentration-dependent manner. However, the addition of 0.3 and 3 mg/mL purified placenta cytosol led to an inhibition of mitochondrial fusion in a dose-dependent manner. This cytosol inhibition was not due to osmotic effects, since 3 mg/mL BSA (bovine serum albumin) did not influence the reaction (Figure 4a). The results demonstrate that the amount of mitochondrial fusion can be modulated by cytosolic factors either in a negative or positive way depending on the source of cytosol. In this way, although we have efficient fusion in the absence of cytosol, the efficiency of the reaction is cytosol-dependent. One potential difference between placenta and sHeLa may be the status of signalling pathways that are activated upon the isolation of cytosol. In to test whether phosphoproteins may be required for mitochondrial fusion, we added Shrimp Alkaline Phosphatase (SAP) into the reaction. The data show that the dephosphorylation of proteins leads to a profound and concentration dependent inhibition of the reaction (Figure 4b). Heat-inactivated phosphatase was not inhibitory (Figure 4b).
In order to further probe the impact of signalling pathways on the fusion reaction, we isolated cytosols from drug-treated sHeLa cells to activate different pathways. We treated sHeLa cells with staurosporin (STS, 1.5 μM) for 6 h to induce apoptosis, collected the apoptotic cytosol by ultracentrifugation and added it back into the fusion reaction. As expected from previous studies [10, 24], and the microscopic analysis of treated cells (Figure 5a), apoptotic cytosol inhibited the fusion reaction (Figure 5b). This demonstrates how cytosolic factors can modulate mammalian mitochondrial fusion. In contrast to the inhibition of fusion during cell death, it has been shown that the activation of protein kinase A (PKA) through the cAMP pathway leads to the phosphorylation of DRP1 and an inhibition of mitochondrial fission, leading to a more connected mitochondrial reticulum [25–27]. Upon addition of 10 μM forskolin to HeLa cells, we also observed an increased interconnectivity of the mitochondrial reticulum within the first 30 minutes (Figure 5a). In order to test whether this phenotype may reflect an increase in fusion that may accompany DRP1 inhibition, we harvested cytosol from forskolin-treated sHeLa cells and added this to the cell-free fusion reaction. Relative to control cytosol, forskolin-treated cytosol led to a 1.5 - to twofold stimulation in mitochondrial fusion (Figure 5b and 5c). Addition of forskolin directly into the reaction did not have any effect on the efficiency of fusion (Figure 5c). Further incriminating the adenylate cyclase pathway in promoting fusion, stimulation by forskolin could be prevented by a prior incubation with the PKA inhibitor H89 (1 h, 10 μM; Figure 5c). This confirms that forskolin acts through the established plasma membrane-bound adenylate cyclase whose downstream signalling mechanisms converge on the mitochondria to activate mitochondrial fusion.