|Home | About | Journals | Submit | Contact Us | Français|
Mammalian Bcl-xL protein localizes to the outer mitochondrial membrane, where it inhibits apoptosis by binding Bax and inhibiting Bax-induced outer membrane permeabilization. Contrary to expectation, we found by electron microscopy and biochemical approaches that endogenous Bcl-xL also localized to inner mitochondrial cristae. Two-photon microscopy of cultured neurons revealed large fluctuations in inner mitochondrial membrane potential when Bcl-xL was genetically deleted or pharmacologically inhibited, indicating increased total ion flux into and out of mitochondria. Computational, biochemical, and genetic evidence indicated that Bcl-xL reduces futile ion flux across the inner mitochondrial membrane to prevent a wasteful drain on cellular resources, thereby preventing an energetic crisis during stress. Given that F1FO–ATP synthase directly affects mitochondrial membrane potential and having identified the mitochondrial ATP synthase β subunit in a screen for Bcl-xL–binding partners, we tested and found that Bcl-xL failed to protect β subunit–deficient yeast. Thus, by bolstering mitochondrial energetic capacity, Bcl-xL may contribute importantly to cell survival independently of other Bcl-2 family proteins.
Bcl-xL is an antiapoptotic Bcl-2 family member that is required for embryonic development and can contribute to cancer cell survival (Letai, 2008; Hardwick and Youle, 2009). The traditional viewpoint is that anti- and proapoptotic Bcl-2 family proteins actively engage each other to determine cell fate after a death stimulus (Galonek and Hardwick, 2006; Youle and Strasser, 2008). The best-characterized cell survival activity of Bcl-xL is its ability to inhibit Bax-induced pores in the outer mitochondrial membrane (Billen et al., 2008). In this manner, Bcl-xL prevents release of mitochondrial cytochrome c into the cytoplasm, where cytochrome c induces apoptosome formation to trigger caspase-dependent death of mammalian cells. Attention has been focused on the functional interactions and the binding specificities between anti- and proapoptotic Bcl-2–related proteins, leading to new therapeutic strategies (Oltersdorf et al., 2005).
The evolutionary conservation of Bcl-2–like proteins cannot be uniformly linked to apoptosis regulation (for example, the Bcl-2 homologues of Drosophila melanogaster and viruses; Bellows et al., 2002; Graham et al., 2008; Galindo et al., 2009). Many other binding partners have been reported for human Bcl-xL, linking Bcl-xL to other cellular processes including mitochondrial dynamics, energetics, and autophagy (Vander Heiden et al., 2001; Levine et al., 2008; Li et al., 2008). Thus, Bcl-2 proteins may have alternative biochemical functions independent of their proapoptotic Bcl-2 family binding partners, or they may participate in other machineries before engaging classical apoptosis.
One nonapoptosis role of Bcl-2 family proteins in mammals and worms is regulation of mitochondrial fission and fusion (Karbowski et al., 2006; Berman et al., 2009; Montessuit et al., 2010; Hoppins et al., 2011). This role appears to contribute importantly to Bcl-xL–induced mitochondrial localization at neuronal synapses, neuronal activity, and seizure behaviors (Fannjiang et al., 2003; Li et al., 2008). However, regulation of fission and fusion rates is not sufficient to explain the ability of endogenous and overexpressed Bcl-xL to increase mitochondrial biomass (Berman et al., 2009). Therefore, we pursued alternative functions of Bcl-xL in mitochondria. Consistent with an evolutionarily conserved function, Bcl-2 family proteins have been linked to control of mitochondrial energetics by regulating the voltage-dependent anion channel in the outer membrane or the adenine nucleotide transporter (ANT)/adenine nucleotide carrier in the inner membrane, which are the primary conduits through which ATP and ADP are exchanged between the cytosol and the mitochondrial matrix (Vander Heiden et al., 2001; Belzacq et al., 2003; Cheng et al., 2003). The relative contributions of antiapoptotic activity versus alternative functions of Bcl-xL for overall cell survival are unclear.
The mitochondrial F1FO ATP synthase synthesizes ATP in the mitochondrial matrix using cytosolic ADP and phosphate as substrates (Hong and Pedersen, 2004). This process requires a potential across the inner mitochondrial membrane that is generated by pumping out protons via the electron transport chain (ETC; or respiratory chain) fueled by NADH. Reentry of protons into the mitochondrial matrix via the FO ring (oligomycin-sensitive fraction) embedded in the inner membrane drives rotation of the central stalk against the catalytic F1, a ring of three α and three β subunits (Walker and Dickson, 2006). In this manner, proton flux through FO is coupled to ATP synthesis. Because mitochondrial membrane potential is required for essential functions other than ATP synthesis, there are alternative strategies for building a potential. Reversal of the F1FO ATP synthase hydrolyzes cytoplasmic ATP produced by glycolysis, reversing the flow of protons through FO to stabilize a potential (Nicholls and Ferguson, 2002; Abramov et al., 2007). A membrane potential is also required for mitochondrial fusion, and depolarization of the potential leads to Parkin-dependent mitophagy (Narendra et al., 2008; Twig et al., 2008). Although mitochondrial energetics are linked to mitochondrial morphology changes, the details are complex (Benard and Rossignol, 2008).
By analyzing bcl-x–deficient neurons, we uncovered a new function of Bcl-xL. We found that Bcl-xL can localize to the inner mitochondrial membrane/matrix, which is contrary to current opinion. Importantly, Bcl-xL is required to stabilize the membrane potential across the inner mitochondrial membrane. By decreasing excess ion flux across the inner mitochondrial membrane, Bcl-xL increases overall energetic efficiency, which is consistent with the limited reserve capacity of bcl-x–deficient neurons and their susceptibility to cell death. This function of Bcl-xL involves the mitochondrial F1FO ATP synthase.
To explore the function of Bcl-xL in healthy neurons, several mitochondrial parameters were analyzed by two-photon laser-scanning fluorescence microscopy, comparing control and bcl-x conditional knockout (cKO) cortical neuron cultures (Berman et al., 2009). Both unfloxed and bcl-x–floxed littermates express neuron-specific knockin NEX-Cre recombinase starting around embryonic day 12 (E12) to delete bcl-x. Staining for Cre recombinase serves as a positive marker for the survival of bcl-x–deficient (and control unfloxed) cortical neurons (Fig. 1 A; Berman et al., 2009). Mitochondrial membrane potential (ΔΨm) was assessed in immature cortical cultures with the potentiometric dye tetramethylrhodamine methyl ester (TMRM; nonquench mode), revealing higher fluorescence intensity in the mitochondria-enriched regions of bcl-x knockout cortical neurons (Fig. 1, B and C [left]). This is not a result of increased mitochondrial biomass because bcl-x–deficient neurons have lower, not higher, mitochondrial biomass in these and other cell types based on several criteria (Kowaltowski et al., 2002; Berman et al., 2009). Thus, it appears that bcl-x deficiency may result in a higher mitochondrial membrane potential.
In respiring cells, three direct mechanisms (Fig. 1 D, dashed boxes) of proton flux across the inner membrane (Fig. 1 D, blue arrows) are main determinants of ΔΨm: (1) the ETC, (2) the F1FO ATP synthase, and (3) uncoupling proteins or other molecularly undefined leak mechanisms (protons not used for ATP synthesis; Nicholls and Ferguson, 2002). To further explore these parameters in bcl-x knockout neurons, reactive oxygen species (ROS) production by the ETC was assessed in the same mitochondrial areas where TMRM was evaluated. Unexpectedly, bcl-x knockout neurons have lower mitochondrial ROS. This suggests either a high rate of electron flow through the respiratory chain or that bcl-x knockout neurons are more dependent on glycolysis than on mitochondrial respiration for energy production (Fig. 1, A and B). The same mitochondrial areas of bcl-x knockout neurons have modestly higher mitochondrial NAD(P)H levels, which is consistent with an earlier study (Schwartz et al., 2007). Higher levels of the complex I substrate NADH indicate more than adequate supplies of NADH either because the respiratory chain is inactive or other metabolic processes are altered, such as decreased anaplerosis (Abramov et al., 2007; Cheng et al., 2011). However, no inherent defects in respiratory chain activity were detected when complexes I–IV, II–IV, and IV were assessed by measuring oxygen consumption in isolated brain mitochondria (Fig. S1, A and B). The relative contributions of glycolysis versus the mitochondrial F1FO ATP synthase to the levels of total cellular ATP were also similar between bcl-x knockout and control cultures. ATP levels (relative to total protein) were slightly reduced in the bcl-x knockout cortical cultures, though this was a result in part of 15% lower cell viability compared with controls (Fig. S1, C and D). In sum, no defects were detected to explain the altered mitochondrial parameters of bcl-x knockout neurons.
To pursue the role of Bcl-xL in regulating mitochondrial parameters, we determined the subcellular localization of endogenous Bcl-xL protein in neurons of the brain. Endogenous Bcl-xL in HeLa cells resides predominantly in the cytosol as a homodimer and translocates to mitochondria via its C-terminal transmembrane domain after a death stimulus (Jeong et al., 2004). However, crude fractionation of mouse cortex suggests that a significant proportion of endogenous Bcl-xL localizes to mitochondria in the brain (Fig. 2 A), which is consistent with an earlier finding (Soane et al., 2008). Deletion of bcl-x (except in interneurons and glial cells where NEX-Cre is not expressed; Berman et al., 2009) did not significantly alter other mitochondrial markers (Fig. 2 A).
Immunogold EM was used to more precisely determine the subcellular localization of endogenous Bcl-xL in tissue slices of mouse brain hippocampus. Approximately 90% of gold-labeled anti–Bcl-xL (BioCarta) is associated with membranes, and at least half of these membranes (54%) can be clearly identified as mitochondria (Fig. 2, B and C). Surprisingly, most of the mitochondrial staining was inside mitochondria, where the colabeled ATP synthase β subunit was also found (Figs. 2 [B and C] and S2). The frequency of labeled mitochondria with inner membrane/matrix Bcl-xL gold label (58%) argues strongly against the possibility of contamination from the outer membrane as a result of edge skimming, folding of the slice preparation, or random background. Gold particles on mitochondria were also detected in polar clusters (Fig. 2 C, arrowheads) and on membranes adjacent to mitochondria, possibly marking mitochondrial fission/fusion sites or where the outer mitochondrial membrane may be tethered to the ER (Fig. 2 C, line arrows), though patchy staining can reflect the uneven epitope accessibility in ultrathin cryosections. Importantly, this staining is specific for Bcl-xL based on multiple parallel preparations of the same brain regions from bcl-x cKO mice, in which total anti–Bcl-xL immunogold label was reduced by ~90%, and mitochondrial Bcl-xL gold was reduced ~99% (Fig. 2, B and D).
To support these findings, protease susceptibility of Bcl-xL was analyzed in purified rat brain mitochondria. Unlike the outer membrane protein Tom20, a portion of Bcl-xL (Abcam antibody) is resistant to proteases even after treatment with digitonin to permeabilize the cholesterol-containing outer membrane (Fig. 2 E). As expected, matrix-localized β subunit was protected from digestion until addition of Triton X-100 to disrupt the inner membrane, when both Bcl-xL and β subunit were completely digested. The proportion of Bcl-xL protected from proteases by the inner membrane can be higher depending on the antibody used (Fig. S3). We conclude that a portion of endogenous Bcl-xL localizes to the inner mitochondrial membrane and/or matrix. These findings are in sharp contrast to the widely held view that Bcl-2 family proteins are localized only to the cytoplasmic side of the outer, not inner, mitochondrial membranes, though inner membrane localization of Bcl-xL and Bcl-2 has been previously reported (Hockenbery et al., 1990; Kharbanda et al., 1997; Motoyama et al., 1998; Gotow et al., 2000; Belzacq et al., 2003).
The possibility that Bcl-xL regulates mitochondrial membrane potential by acting at the inner mitochondrial membrane led us to revisit our earlier yeast two-hybrid screen (Chau et al., 2000). Seeking to identify prosurvival functions distinct from antiapoptotic functions of Bcl-xL in an unbiased screen, the BH1 domain mutant of Bcl-xL (mt1; F131V/D133A), which inhibits cell death without binding prodeath family members Bax or Bak (Cheng et al., 1996), was used to screen a human B cell library (Chau et al., 2000). Among the six hits, we identified an unexpected Bcl-xL–binding partner, the β subunit of the mitochondrial F1FO ATP synthase. This interaction was confirmed in a secondary yeast two-hybrid screen in which the β subunit interacted with wild-type Bcl-xL and Bcl-2 but did not interact with mutants lacking antideath activity (Bcl-xL mt7 and mt8) and did not interact with Bax or Bak (Fig. 3 A). Because Bcl-xL mt1 could potentially inhibit mammalian cell death by binding BH3-only proteins (Billen et al., 2008), we assayed the function of mt1 and mt8 in yeast, which lack Bcl-2 and BH3-only proteins. Bcl-xL mt1 but not mt8 protected yeast from dose-dependent cell death (Fig. 3 B).
An independent biochemical purification scheme also identified the β subunit as the prominent binding partner of endogenous Bcl-xL. WEHI 7.1 membrane preparations were solubilized with CHAPS to avoid detergent-induced dimerization with Bax during extract preparation (Hsu and Youle, 1997). Bcl-xL–containing complexes were purified by sequential ion exchange and immunoaffinity chromatography followed by preparative SDS-PAGE (Fig. 3 D). The only major Coomassie-stained species copurifying with Bcl-xL was ~54 kD. Direct sequencing of two tryptic peptides derived from this extracted band yielded exact matches with the human/mouse F1FO ATP synthase β-subunit residues 244–253 (NDLYHEMIES) and 389–404 (IAELGIYPAVDPLDST).
A screen of 80 detergents yielded a strategy for purifying enzymatically active F1FO ATP synthase from mitoplasts isolated from rat liver mitochondria for the purpose of 3D structure determination (Ko et al., 2003). Immunoblot analyses of these preparations revealed monomeric Bcl-xL, which decreased in abundance with purification as a band approximately the size of Bcl-xL dimers was enriched with purification (Fig. 3 C). Detergents likely induced SDS-stable dimers of Bcl-xL, which are frequently encountered with purified Bcl-xL (O’Neill et al., 2006). On parallel blots, both Bcl-xL bands were eliminated when the antibody was preadsorbed with recombinant Bcl-xL protein, indicating that Bcl-xL is enriched in highly purified preparations of the ATP synthase from liver. To determine whether Bcl-xL is monomeric or present in larger complexes inside cells, CHAPS-solubilized lysates were separated by column chromatography, revealing that all of the Bcl-xL was in complexes >70 kD that overlap fractions containing the β subunit (Fig. 3 E). An association of Bcl-2 with the F1FO ATP synthase has also been observed by the laboratories of J. Downward (London Research Institute, London, England, UK), Y. Tsujimoto (Osaka University, Osaka, Japan), and S. Korsmeyer and G. Linette (Washington University in St. Louis, St. Louis, MO; personal communication).
Because the mitochondrial F1FO ATP synthase is an important control point for proton flux across the inner mitochondrial membrane, mitochondrial membrane potential was further evaluated by time-lapse imaging (3.5-s intervals). TMRM intensity in mitochondria-enriched regions fluctuates modestly in control neurons, which is consistent with an earlier study (Vergun et al., 2003). However, bcl-x–deficient neurons exhibited a striking fluctuation in TMRM fluorescence intensity over irregular intervals in time (Fig. 4 A), across a single bcl-x knockout cell (Fig. 4 B), and in individual mitochondria (Fig. 4 C). Thus, the increase in mean mitochondrial potential in bcl-x–deficient neurons (Fig. 1 B) represents the mean of a time-varying potential that fluctuates predominantly to higher (more negative) potentials than controls. Therefore, the presence of Bcl-xL stabilizes the inner mitochondrial membrane potential.
Because Bcl-xL can bind to the inositol triphosphate receptor in the ER to regulate calcium gating by the inositol triphosphate receptor (White et al., 2005), we investigated a role for calcium in mitochondrial membrane potential fluctuation. We found that basal cytosolic calcium levels were uniformly steady in cultured bcl-x knockout and control cortical neurons (Fig. 4 D). Although compiled data indicate a small but significant calcium elevation in bcl-x knockout neurons, fluctuations in potential appear not to be controlled by paired fluctuations in cytosolic calcium.
To investigate the possibility that Bcl-xL has a direct role in stabilizing the mitochondrial membrane potential, tetramethylrhodamine ethyl ester (TMRE) fluorescence intensity was monitored in cultured hippocampal neurons treated only briefly with ABT-737, a specific inhibitor of Bcl-xL designed to fit the binding pocket on Bcl-xL and block its antiapoptotic function (Oltersdorf et al., 2005). Individual mitochondria exhibited greater fluctuations in TMRE fluorescence intensity after only 10 min of ABT-737 (in 0.1% DMSO) compared with DMSO alone (Fig. 4, E and F). To confirm the specificity of ABT-737, hippocampal neurons expressing scrambled or bcl-x–specific short hairpin RNA (shRNA) were monitored for TMRE fluorescence intensity in time (Fig. S4), and the SD of fluorescence intensities was significantly greater for the bcl-x knockdown than the control (Fig. 4 G). These data again suggest that Bcl-xL stabilizes the mitochondrial membrane potential by limiting total ion flux across the mitochondrial membrane.
It is known that when any chemical system is not at thermodynamic equilibrium, as is the case for respiring mitochondria, the occurrence of persistent fluctuations or oscillations can only be maintained by expending energy (Nicolis and Prigogine, 1977). Moreover, the additional time-dependent flux of ions across the inner membrane that drives these fluctuations in potential can result in an overall ion flux (both inward and outward directions) that is greater than what is required simply to maintain a nonfluctuating membrane potential at a steady negative value. Thus, the fluctuations in mitochondrial membrane potential in bcl-x knockouts imply that more energy is required to maintain ion gradients across the inner membrane. To illustrate this concept, we constructed a simple numerical model to investigate the effect of fluctuations on the dissipation of ion gradients across the mitochondrial membrane. A vesicle (1 µm in diameter) was used to represent a mitochondrion (Fig. 5 A). This vesicle was equipped with active ion pumps (Fig. 5 A, b) capable of pumping out protons/ions (approximating the respiratory chain) to build a negative potential (−180 mV) and with ion channels (Fig. 5 A, a) that can partially dissipate this potential by allowing ions to reenter the vesicle (approximating the F1FO ATP synthase and nonproductive leaks). We first modeled steady-state conditions in which the inward flux and outward flux of ions are exactly matched in time, and the membrane potential does not fluctuate in amplitude. These conditions approximate the steady-state conditions of mitochondria in wild-type cells. Next, we modeled fluctuations in membrane potential by introducing small currents across the vesicle membrane (Fig. 5 A, c). These small currents (set arbitrarily at 5 ms with a fixed amplitude between 0 and 10 pA) were allowed to occur randomly (averaging 1/s) to drive fluctuations in the potential across the vesicle membrane. To assess the effects of these external current amplitudes (Fig. 5 A, c), we measured the magnitude of total ion flux through the pumps (Fig. 5 A, a) and the channels (Fig. 5 A, b). In all cases, the total amount of ion flux (measured in picocoulombs) was increased when current fluctuations were introduced and was further increased with increasing external current amplitude (Fig. 5 B). The additional amount of ion movement (Fig. 5 A, a and b) produced by the small transient current fluctuations (Fig. 5 A, c) represents a futile dissipation of the ion gradient that has to be balanced by pump activity to restore the mean membrane potential. Thus, Bcl-xL could improve mitochondrial energetics simply by preventing futile ion flux.
Collectively, our results suggest the possibility that Bcl-xL regulates an inner mitochondrial membrane ion-conducting channel and that this channel has an increased probability of opening in the absence of Bcl-xL. This increased channel opening (analogous to point c in Fig. 5 A) could result in the increased membrane potential fluctuations observed in the bcl-x knockout. To investigate the effects of such a Bcl-xL–regulated channel on mitochondrial membrane potential fluctuations, we made a second computational model that more closely represents known properties of the inner mitochondrial membrane. This enabled us to test explicitly the effect of very brief transient openings of a nonselective inner membrane channel on the mitochondrial membrane potential measured over time. A vesicle (1.5 µm in diameter) was equipped with a proton pump and a proton leak pathway as in Fig. 5 A (a and b). The steady-state proton concentration of the mitochondrion was further regulated by a fixed proton buffer and a proton–cation exchange pathway (Garlid and Paucek, 2003). Finally, we introduced a nonselective cation channel representing the one regulated by Bcl-xL, which is permeable to both protons and to other cations and has a reversal potential of 0 mV (Lam et al., 1998; Vander Heiden et al., 2001; Alavian et al., 2011). Opening of the nonselective cation channel (mean open time of 0.33 ms) was allowed to occur stochastically with different opening probabilities of 0–0.1. We found that opening of the nonselective channel produced fluctuations in the membrane potential that increased with increased probability of channel opening (Fig. 5, C and D). The simulation further reveals that the very brief increases in internal proton concentration produced by influx through the channel resulted in proton pump activation, resulting in an overall hyperpolarization of the membrane as the frequency of channel openings increased (Fig. 5, C and D). This is consistent with transient hyperpolarization of mitochondria in bcl-x knockout cells as a result of overshooting by the respiratory chain after the channel opens.
Our vesicle models predict that the increased membrane leakiness (productive and nonproductive ion flux) across the inner mitochondrial membrane in bcl-x–deficient neurons will result in decreased energetic performance. To test this prediction, cultured bcl-x cKO and control cortical neurons were energetically stressed by the addition of mitochondrial ATP synthase inhibitors and analyzed for ATP levels and for mitochondrial parameters by two-photon microscopy. Extensive genetic and biochemical evidence indicates that oligomycin inhibits mitochondrial ATP synthesis by acting on FO to disrupt the proton path (Walker and Dickson, 2006), and a crystal structure reveals that aurovertin B inhibits the enzymatic F1 subunit by binding near the ATP-binding site on β subunit (van Raaij et al., 1996). Treatment with oligomycin or with aurovertin B caused cellular ATP levels to decline similarly in control and knockout neurons (Fig. 6, A and B). Therefore, the F1FO ATP synthase was an important contributor to ATP production and concomitant dissipation of membrane potential in both genotypes before treatment.
In contrast to controls, bcl-x–deficient cortical neurons consistently underwent delayed mitochondrial depolarization 30–45 min after the addition of oligomycin (Fig. 6, C and E). Consistent with an energy-wasting crisis unique to bcl-x–deficient neurons, oligomycin also causes mitochondrial NAD(P)H levels to decline to ~50% of pretreatment levels in <1 h, whereas NAD(P)H levels rebound and stabilize after oligomycin treatment in controls (Fig. 6 F). These results suggest that bcl-x–deficient mitochondria continue to deplete the substrate of complex I, as would be expected for a leaky mitochondrial membrane that allows the respiratory chain to continue running. Consistent with this conclusion, rates of oxygen uptake by cells decrease with overexpression of Bcl-xL and increase with shRNA knockdown of Bcl-xL (Alavian et al., 2011). NAD(P)H depletion and membrane depolarization were not simply a result of inhibition of mitochondrial ATP synthesis because NAD(P)H levels and membrane potential were sustained for at least 1 h after aurovertin B treatment, although at lower steady-state levels relative to controls (see Discussion; Fig. 6 G). To verify that depletion of NAD(P)H and mitochondrial depolarization is not simply a marker of cell death, oligomycin was washed away from depolarized bcl-x–deficient neurons in a flow chamber. Upon washout, we observed simultaneous increases in NAD(P)H levels and TMRM intensity, indicating cell recovery (Fig. 7 A). The evidence presented suggests that Bcl-xL increases the efficiency of mitochondrial energetics by decreasing inner membrane leakiness, thereby preventing membrane potential fluctuations and the resulting energy deficits (Fig. 7 B).
To test whether Bcl-xL increases cell survival through a functional interaction with F1FO ATP synthase and independently of any other Bcl-2 family members, we tested the ability of human Bcl-xL to inhibit cell death of yeast lacking β subunit (Δatp2) of the F1FO ATP synthase. Using a novel heat ramp cell death assay (Teng et al., 2011), we found that Bcl-xL failed to protect yeast in which the ATP2 gene was deleted. In contrast, Bcl-xL protected yeast with mutations in both the mitochondrial fission protein FIS1 and WHI2, which lack mitochondrial fission and have respiratory function defects (Δfis1*; Figs. 7 C and S5; Fannjiang et al., 2004; Cheng et al., 2008). Yeast have no recognizable Bcl-2 family members or BH3-only proteins yet have a highly conserved F1FO ATP synthase. Our results indicate that Bcl-xL promotes cell survival through an interaction with the ATP synthase.
Our evidence indicates that endogenous Bcl-xL prevents a futile ion flux across the mitochondrial inner membrane, thereby preventing pronounced irregular fluctuations in mitochondrial membrane potential observed in bcl-x knockout cells. The additional energy required for fueling excessive ion flux across the mitochondrial membrane would place bcl-x–deficient cells at a distinct disadvantage during cell stress. Unable to sustain a potential across a more leaky inner membrane, bcl-x–deficient mitochondria depolarize and subsequently die. Prominent localization of endogenous Bcl-xL with the inner mitochondrial membrane is consistent with a close link between Bcl-xL and the membrane leak channels. Copurification of Bcl-xL with the F1FO ATP synthase raises the possibility that a novel leak channel could be within the ATP synthase itself or a functionally interacting component. This function of Bcl-xL can be expected to alter many other aspects of mitochondrial and cellular physiology, though, like many mitochondrial proteins, the mechanism by which Bcl-xL enters mitochondria is not known.
These findings are consistent with a conceptually simple mechanism in which Bcl-xL acts at the inner mitochondrial membrane to close a molecularly undefined leak, thereby preventing large swings in membrane potential. This could be achieved if Bcl-xL directly closes the leak channel. Our computational models predict that opening of this Bcl-xL–inhibited leak channel results in membrane potential fluctuations, which is consistent with fluctuations observed in bcl-x–deficient cells. The models further predict that the large fluctuations are a result of transient overcompensations by the respiratory chain, which is consistent with the observed transient hyperpolarizations in bcl-x–deficient cells. Our numerical simulations also suggest that the greater total flux of ions across the mitochondrial inner membrane in bcl-x–deficient mitochondria requires more energy to maintain ionic homeostasis, analogous to other fluctuating or oscillating biochemical systems (Kaczmarek, 1976). Even if the additional ion flux in bcl-x–deficient mitochondria was coupled to ATP synthesis by F1FO, additional energy would be required to move ions out of the matrix when the potential is fluctuating compared with a steady state with little or no fluctuations. Our simulations are consistent with the notion that the stabilizing effect of Bcl-xL on inner mitochondrial membrane potential contributes importantly to the efficiency of energy production. An expected negative consequence of excess ion flux is that a sudden deficit in nutrients or a sudden increase in energy demand cannot be readily satisfied by an inefficient system.
Seemingly contrary to these findings, recombinant Bcl-xL can induce ion channel activity in outer mitochondrial membranes and synthetic bilayers, although these channels are smaller than those formed by the related proapoptotic Bax protein (Lam et al., 1998; Basañez et al., 2002). Thus, the ion-conducting activity of Bcl-xL may not be related to the Bcl-xL functions under study here. However, more complex scenarios remain possible where Bcl-xL channels open to correct other ion-conducting activities in the inner membrane, thereby preventing large swings in potential. Though the detailed molecular events remain unclear, our conclusions are supported by patch clamp recordings of mitochondrial inner membrane vesicles with Bcl-xL (Alavian et al., 2011). It is conceivable that the Bcl-xL–inhibited mitochondrial leak channel is related to the Bax pores that trigger apoptosis, except Bax pores are not known to occur in inner membranes (Billen et al., 2008). The capacity of a non–Bax/Bak-binding mutant of Bcl-xL (mt1) to interact with the F1 β subunit and to inhibit cell death in mammalian cells argues against this possibility. Furthermore, Bcl-xL can inhibit cell death in wild-type yeast, which lack Bcl-2 family and BH3-only proteins, but Bcl-xL cannot protect yeast lacking F1 β, which shares 89% amino acid sequence homology with humans. The inner membrane function and the antideath function of Bcl-xL appear to be separable biochemical events (e.g., Fig. 7 A). However, yet unknown nonapoptotic activities of Bax/Bak could be involved, potentially those that regulate neuronal activity or inhibit neuronal cell death in vivo (Lewis et al., 1999; Fannjiang et al., 2003). Furthermore, Bcl-xL was recently reported to regulate acetyl-CoA levels in a Bax/Bak-independent manner (Yi et al., 2011).
Given the unexpected finding that mitochondrial ROS levels in bcl-x–deficient cells are lower than controls, our data are not consistent with Bcl-xL–mediated leak closure serving to reduce ROS levels (Jastroch et al., 2010). To the contrary, the leakier/fluctuating membrane potential could be expected to increase oxygen consumption in bcl-x–deficient cells, which is consistent with reduced oxygen consumption in Bcl-xL–overexpressing cells (Alavian et al., 2011).
Because Bcl-xL is not found in the 3D structures of F1Fo ATP synthase, we considered other potential functions for partnering. Given a structural resemblance between Bcl-xL and Diphtheria toxin, a polypeptide translocator, we considered that Bcl-xL could facilitate entry of the β subunit into mitochondria. However, we found that bcl-x–deficient mitochondria appear to have normal levels of β subunit. Bcl-xL could interact with fully assembled ATP synthase or participate in the assembly process, which requires many factors not present in active complexes (Rak et al., 2011). Alternatively, the effects of Bcl-xL on membrane curvature could influence the ATP synthase (Basañez et al., 2002; Paumard et al., 2002).
Our biochemical and pharmacological analyses are consistent with a Bcl-xL–regulated leak mechanism involving the F1Fo ATP synthase. However, the differential effects of two F1Fo inhibitors are somewhat puzzling. Both aurovertin B (acting on F1) and oligomycin (acting on FO) block ATPase/synthase activity and proton movement through Fo because of the coupling between F1 and Fo. Yet, only aurovertin B inhibited mitochondrial depolarization in bcl-x–deficient cells, possibly by triggering closure of the leak channel by binding the β subunit (van Raaij et al., 1996). In contrast, the Bcl-xL–regulated leak channel appears to be oligomycin resistant, though it is not clear whether this is the long sought-after leak channel that explains continued mitochondrial respiration with oligomycin treatment (Nicholls and Ferguson, 2002). We speculate that the F1FO ATP synthase is involved in leaking ions and that the regulation of this function is defective in bcl-x–deficient neurons. These studies further extend the long-standing link between Bcl-2 family proteins and the ATP synthase (Matsuyama et al., 1998; Vander Heiden et al., 2001; Belzacq et al., 2003).
Conditional bcl-x knockout cortical neuron cultures were prepared separately from individual E16.5 mouse embryos as previously described (Berman et al., 2009). Wild-type and floxed bcl-x mice/embryos were distinguished using PCR primers 5′-GCCACCTCATCAGTCGGG-3′ and 5′-TCAGAAGCCGCAATATCCC-C-3′. The NEX-CRE locus was identified with primers 5′-TCTTTTTCATGTGCTCTTGG-3′ and 5′-CCGCATAACCAGTGAAACAG-3′, and the wild-type allele was identified with 5′-CAAGTTGTCCTTCGAGGAAAGAGC-3′ and 5′-GATACAGACAAGAGGGAAGGG-3′. All experiments were performed on density-matched cultures. All animal procedures were approved by the Animal Care and Use Committee. For immunofluorescence microscopy, cortical neuron cultures were quickly washed with Locke’s buffer (154 mM NaCl, 5.6 mM KCl, 2.3 mM CaCl2, 1.0 mM MgCl2, 5 mM Hepes, and 10 mM glucose, pH 7.4), fixed for 15 min in 4% PFA, permeabilized for 5 min with 0.2% Triton X-100, blocked for 30 min at RT with 5% normal goat serum, and incubated with primary antibodies at 4°C overnight followed by 1 h at RT with secondary antibodies.
Potentiometric dye TMRM, which accumulates in the matrix according to its Nernst potential, was used at 100 nM (the lowest workable concentration; nonquench mode verified with carbonyl cyanide-p-trifluoromethoxyphenylhydrazone) to monitor mitochondrial membrane potential ΔΨm. Accumulation of ROS was monitored simultaneously with 2 µM CM-H2DCFDA (5-(-6)-chloromethyl-2’,7’-dichlorohydrofluorescein diacetate). Fluorescence probes were loaded into cortical neuron cultures (3–5 days in vitro [DIV3–5]) for at least 20 min, and images were recorded using a two-photon laser-scanning microscope (MRC-1024MP; Bio-Rad Laboratories) with an excitation at 740 nm (Tsunami Ti:Sa laser; Spectra-Physics) to measure fluorescence intensity of TMRM (605 ± 25 nm), CM-DCF (525 ± 25 nm), and intrinsic autofluorescence of endogenous NADH/NADPH (<490 nm; Aon et al., 2003). For single recordings, three to five fields per culture dish were imaged in immediate succession. For time-lapse recordings, images of the same field were captured every 3.5 s for up to 5 min using 50% laser intensity to limit photo damage to live samples. Region of interests (ROIs) were drawn and analyzed using ImageJ (National Institutes of Health) for all neurons per image. Mean fluorescence intensity per pixel in each ROI at the three emission wavelengths was calculated as arbitrary fluorescence units. Background from cell-free areas was subtracted for each wavelength. Photodamage-induced fluctuations specifically in knockout cells is unlikely, as fluctuations are evident at the earliest time points and with single-photon microscopes (Fig. 4), and treatment with antioxidant N-acetyl cysteine does not inhibit depolarization.
Mouse forebrain mitochondria were isolated from littermates of control and cKO mice (postnatal day 2–7 [P2–P7]) by modification of a standard protocol (Rosenthal et al., 1987). Mitochondria (primarily nonsynaptosomal) were prepared as previously described for subcellular fractionation and were further washed with mannitol sucrose (MS) buffer (without EGTA). Rates of oxygen consumption by purified mitochondria (0.5 mg/ml) were measured with a Clark-type oxygen electrode (Hansatech Instruments Limited) in KCl buffer (125 mM KCl, 20 mM Trizma base, 2 mM potassium phosphate, and 1 mM MgCl2, pH 7.2) plus substrates (Fig. S1 A), 1 mM MgCl2, and 0.25 mM EGTA and were calculated in nanomoles of O2 per mg protein per minute based on a KCl medium oxygen content of 195 nmol/ml O2 at 30°C.
Cerebral cortexes from P3 mice were rapidly dissected, minced on ice in 2 ml MS-EGTA buffer (225 mM mannitol, 75 mM sucrose, 250 µM EGTA, 1 mg/ml fatty acid–free BSA, and 5 mM Hepes, pH 7.4), and homogenized with 15 strokes in a 2 ml Dounce. The cortical suspension was clarified (at 2,800 g for 3 min at 4°C), and mitochondria were collected by centrifugation (15,000 g for 8 min), gently resuspended in MS-EGTA, and recentrifuged (for 10 min). The pellet was lysed in 100 µl MS buffer + 1% NP-40, and the supernatant was centrifuged at 100,000 g for 30 min to clarify the cytosolic fraction.
Mouse cerebral cortex was dissected on ice and passed 15 times through a 25-guage needle in 3 vol of radioimmunoprecipitation assay (RIPA) lysis buffer (50 mM Tris, 1% NP-40, 150 mM NaCl, and 1 mM EDTA, pH 7.4) plus 1 mM NaF, 1 mM Na3VO4 (sodium orthovanadate and phosphatase inhibitor), and protease inhibitors (PMSF, leupeptin, pepstatin, and apoprotinin). Cultured cortical neurons were washed quickly with Locke’s buffer and lysed in 150 µl of RIPA buffer. 50-µg lysates were separated by SDS-PAGE and blotted with specific antibodies (see figure legends).
Immunogold labeling of ultrathin cryosections was performed as previously described (McCaffery and Farquhar, 1995). Tissues from both genotypes were fixed in 4% PFA diluted in PBS, pH 7.4, for 15 min followed by 8% PFA for 1 h at RT. Samples were cryoprotected in 2.3 M sucrose plus 20% polyvinylpyrrolidone (for 1 h), mounted on aluminum cryopins, and frozen in liquid N2. 80-nm ultrathin cryosections cut on a microtome (Ultracut T; Reichert) equipped with an FCS cryostage were collected onto 300 mesh formvar/carbon-coated nickel grids. Grids were passed through several drops of PBS plus 2.5% FCS and 0.01 M glycine, pH 7.4, blocked in 10% FCS, and incubated overnight with mixed primary antibodies against ATP synthase β subunit (BD) and/or anti–Bcl-xL/S antibody, each at ~10 µg/ml. Washed grids were incubated for 2 h with one or both secondary antibody gold conjugates (1:50; Jackson ImmunoResearch Laboratories, Inc.). Grids were washed several times, first with PBS and then with double-distilled H2O. Grids were embedded (3.2% polyvinyl alcohol [molecular mass of 10 kD], 0.2% methyl cellulose [400 cps], and 0.2% uranyl acetate) and observed on a transmission electron microscope (EM 410; Philips Research Eindhoven), and images were collected with a digital camera (SIS Megaview III; Olympus). Figures were assembled in Photoshop (Adobe) with only linear adjustments in brightness and contrast.
Nonsynaptic brain mitochondria were isolated from adult male Sprague Dawley rats (weights of 300–350 g) by using a Percoll gradient centrifugation method as previously described (Kristian et al., 2007). In brief, rat forebrains were removed and homogenized in isolation medium (225 mM mannitol, 75 mM sucrose, 5 mM Hepes, 1 mM EGTA, 225 mM mannitol, 75 mM sucrose, 5 mM Hepes, and 1 mg/ml fatty acid–free BSA, pH 7.4, at 4°C). The resulting homogenate was centrifuged at 1,300 g for 3 min. The supernatant containing mitochondria was collected, and the pellet was resuspended and recentrifuged at 1,300 g. The pooled supernatants were then centrifuged for 10 min at 16,000 g. The crude mitochondrial pellet was resuspended in 15% Percoll and layered on top of a 40/20% Percoll gradient and then centrifuged for 12 min at 21,000 g. Nonsynaptic mitochondria were collected from the interface of the two bottom layers, diluted with isolation medium, and centrifuged at 16,000 g for 8 min. The purified mitochondrial pellet was resuspended in isolation medium and kept on ice until use. Protein concentration of mitochondrial samples was determined by using the BCA assay (Thermo Fisher Scientific) and BSA as standards. Mitochondrial protease digestion assays were performed by incubating freshly isolated mitochondria (1 mg/ml in isolation medium containing 1 mM MgCl2) for 30 min at 37°C with either 0.2–25 µg/ml Proteinase K or 25–200 µg/ml Trypsin. 0.01% digitonin was used to permeabilize the outer mitochondrial membrane. The reactions were stopped by addition of a protease inhibitor cocktail (Thermo Fisher Scientific), and then mitochondria were collected by centrifugation (16,000 g for 10 min at 4°C), and the mitochondrial pellets were lysed in RIPA lysis buffer (30 mM Tris-HCl, pH 7.4, 0.15 M NaCl, 1% NP-40, 0.1% SDS, 0.5% sodium deoxycholate, 1 mM EDTA, 1 mM DTT, and 2 mM MgCl2) containing protease inhibitors. The sensitivity of various mitochondrial proteins to protease digestion was then examined by immunoblot analysis using the following antibodies: anti–Bcl-xL (ab2568, 1:1,000; Abcam), anti–cytochrome c (clone 7H8.2C12, 1:1,000; Thermo Fisher Scientific), anti-Tom20 (sc-11415, 1:2,000; Santa Cruz Biotechnology, Inc.), and anti–ATP synthase β subunit (1:1,000). The ATP synthasome fractions used are as previously described (Ko et al., 2003) and were solubilized in lithium dodecyl sulfate for separation by SDS-PAGE.
WEHI 7.1 cells (~20 ml packed pellet) were lysed in 200 ml of hypotonic buffer (37.5 mM NaCl and 10 mM Hepes, pH 7.4) plus 25 µg/ml PMSF using a Dounce homogenizer. The membrane pellet (at 23,000 g for 30 min) was solubilized in 300 ml of isotonic buffer (150 mM NaCl and 10 mM Hepes, pH 7.4) plus 1% CHAPS and was clarified by centrifugation (for 15 min at 15,000 g). The supernatant was loaded onto a 10-ml trimethylaminoethyl anion exchange column and washed with 10 column volumes of isotonic buffer with 0.5% CHAPS, and bound proteins were eluted with a salt gradient (Bcl-xL eluted at 0.35 M NaCl). Bcl-xL–containing fractions were immunoaffinity purified with 1 ml anti–murine Bcl-xL antibody 7D9 (Hsu et al., 2003) bound to beads (2 mg antibody/ml Sepharose beads) for 3 h at 4°C. Beads were washed with isotonic buffer with 0.5% CHAPS, and Bcl-xL–containing complexes were eluted with 3 ml of 0.1 M acetic acid + 0.3% CHAPS. The sample was neutralized with 0.4 ml of 1 M Tris (pH 8.0), concentrated (Centricon-30), and separated by preparative SDS-PAGE.
108 HeLa cells stably expressing Bcl-xL (Hou et al., 2003) were lysed by Dounce homogenization in hypotonic buffer. The membrane pellet (at 31,000 g for 30 min) was solubilized in 3 ml of isotonic buffer plus 1% CHAPS and clarified by centrifugation (at 31,000 g for 15 min). The supernatant (0.4 ml) was loaded onto a 38-ml Superdex 200 gel filtration column (GE Healthcare) precalibrated with 67 kD BSA, 45 kD ovalbumin, and 24 kD chymotrypsinogen. 0.4-ml column fractions containing Bcl-xL were identified by immunoblot analysis with monoclonal 2H12 (Hsu and Youle, 1997).
Dissociated rat hippocampal neurons were prepared from E18 embryos and plated on poly-l-lysine–coated dishes in Neurobasal medium with B27 (Invitrogen; Li et al., 2008). Mature (DIV14–16) cultures were incubated at 37°C in recording buffer (5 mM KCl, 110 mM NaCl, 2 mM MgCl2, 10 mM glucose, 10 mM Hepes, 2 mM CaCl2, pH 7.4, and 310 mOsm) containing TMRE (5 nM final). Individual puncta containing mitochondria at the base of a dendrite near the soma were outlined and measured by averaging 4 × 4 pixels as previously described (Li et al., 2008). Fluorescent images were collected (1/s for 30 s) with fixed exposure times (300 ms) using an inverted microscope (Axiovert 200; Carl Zeiss) with a 63× oil objective. Background fluorescence was subtracted for each image, and data were analyzed using AxioVision software (version 4.3; Carl Zeiss). For analysis of SDs, a straight baseline was subtracted from each graphed line using OriginLab 8.0 software to eliminate any artifacts due to slight organelle movement during imaging.
Mouse cortical neurons (DIV3–5) grown on 15-mm coverslips were loaded with 2 µM cell-permeable Fura-2 acetoxymethyl ester at 37°C for 30 min, washed with culture medium, and incubated at 37°C for 20–30 min to allow complete hydrolysis of acetoxymethyl ester. Coverslips were mounted on a AttoFluor system (Carl Zeiss) and continuously infused with Locke’s buffer. Cells were sequentially excited at 340 nm/380 nm, and fluorescence intensities (510 nm) were determined for individual neurons from images captured at ~4-s intervals. The 340:380 ratios were converted to nanomoles Ca2+ using a video imaging system (Intracellular Imaging Inc.) and commercial reference standards (Invitrogen) by the formula [Ca2+]i = Kd ([R − Rmin]/[Rmax − R])(Fmax/Fmin), where R equals the ratio of 510-nm emission intensities excited at 340 nm relative to 380 nm, Rmin equals the ratio at zero free Ca2+, Rmax equals the ratio at saturating Ca2+ (39 µM), Fmin equals the fluorescence intensity excited at 380 nm for zero free Ca2+, and Fmax equals the fluorescence intensity excited at 380 nm in saturating Ca2+.
To estimate the flux of ions in the model vesicle (Fig. 5 A), we integrated the equation CdV/dt = Ich + If(t), where V is the membrane potential across the vesicle, C represents the capacitance of the vesicle, Ich is the ionic current flowing through the channel in the membrane, and If(t) is an additional fluctuating current that is applied across the membrane. Ich was defined by the equation Ich = gL(V0 − V), where gL is the conductance of the membrane and, under steady-state conditions, provides a membrane potential of −180 mV (V0 = −180 mV). If(t) was either fixed at 0 or was allowed to fluctuate from 0 to a value IFmax for 5-ms periods. These fluctuations occurred randomly with a mean period of 1 s. In different simulations, the value of IFmax was increased from 0.01 to 10 pA. Equations were integrated for a time span of 20 s, and the ion flux for the entire period was calculated in picocoulombs. Activity of pumps was not simulated explicitly in these models but was incorporated implicitly because the reversal potential for ion flux (V0 = −180 mV) was held fixed during the simulations. Parameters for the simulations were C = 0.314 picofarads and gL = 3.14 picosiemens.
For the second model, we tested the effects of a nonselective cation channel on the membrane potential across the inner mitochondrial membrane. For simplicity, we included a fixed proton buffer and an electroneutral proton–cation (K+) exchange pathway (Garlid and Paucek, 2003) such that, in the absence of any other channel activity, this model has a steady-state internal proton concentration of 20 nM (pH 7.7) and a membrane potential of −180 mV, as is typical for many mitochondria.
We integrated the following pair of coupled stochastic differential equations:
Ht is the total bound and unbound concentration of protons in the vesicle, and V is the voltage across the membrane. Hout is the H+ concentration outside the vesicle and was fixed at 40 nM, whereas Hin, the concentration of free protons in the vesicle, was related to the value of Ht and to B, the concentration of H+ buffer in the vesicle, by the following quadratic equation:
Here, B = 5 × 10−3 mM, and Keq = 2.3 × 10−5 mM.
The constant Av was equal to (1.036 × 104)/W, where W is the volume of the vesicle (1.767 mm2). The capacitance of the vesicle C was 5 × 10−9 nanofarads. The values for gH (the basal proton leak), gL (the basal leak of cations), and gCHAN (the conductance of the nonselective cation channel) were 5, 5.55 × 10−6, and 5 × 10−4 picosiemens, respectively. The values for kR (the rate constant for H+ pumping out of the vesicle) and kX (the rate of electroneutral cation/H+ exchange) were 2.5 and 5 × 10−5 ms−1, respectively.
l(t) is a stochastic function that takes on the value of 1 or 0 depending on whether the nonselective cation channel is open or closed. The open probability, P0, of the channel was determined by the rate constants for channel opening (kf) and closing (kb) and was given by the following relation:
The value of kb was fixed at 0.3 ms−1, which provided a mean open time of 3.33 ms. In the simulations of Fig. 5 (C and D), the values of kf were set at 0, 0.0005, 0.0101, and 0335 ms−1, providing mean open probabilities of 0, 0.0017, 0.323, and 0.1004.
Cortical cultures were harvested as for immunoblot analysis plus a phosphatase inhibitor; mouse cortex lysates were supplemented with 50 mM atractyloside. Samples were analyzed immediately, or time points were frozen instantly and analyzed together. Protein concentration/sample (BCA assay) and fresh ATP standards (0, 25 µM, 5 µM, 500 nM, 50 nM, and 5 nM) were used to calibrate every experiment.
Overnight cultures of yeast strains (MATa, his3Δ1, leu2Δ0, met15Δ0, ura3Δ0, and yfg::KanMX4; Invitrogen) transformed with modified pRS-PGK vector without/with human Bcl-xL were diluted and grown to midlog phase (synthetic complete–uracil medium) and plated before and after a heat ramp treatment to trigger cell death (30–40°C in 2 min, 40–51°C in 10 min, and held at 51°C for 5 min; Teng et al., 2011). Both the ATP2 and FIS1* knockout strains are more sensitive to cell death than wild type, in which Bcl-xL also protects (Fannjiang et al., 2004). For immunoblot analyses, lysates were prepared from overnight cultures in lysis buffer (0.05 M Tris-HCl, pH 7.5, 0.15 M NaCl, 1% NP-40, and 0.1 M PMSF) with glass beads and blotted with anti–Bcl-xL (1:5,000 rabbit monoclonal) and anti–rabbit IgG (1:20,000; GE Healthcare).
Fig. S1 shows that no respiratory defects were detected in bcl-x–deficient mitochondria. Fig. S2 shows coimmunogold EM for Bcl-xL and F1 β subunit. Fig. S3 shows protease digestion of mitochondria detected with Bcl-xL antibody. Fig. S4 shows an example of TMRE traces and Bcl-xL blots for shRNA knockdowns in Fig. 4 G. Fig. S5 shows expression levels of Bcl-xL protein in yeast. Online supplemental material is available at http://www.jcb.org/cgi/content/full/jcb.201108059/DC1.
We thank Dr. L. Boise (University of Miami, Miami, FL) for the Bcl-xL antibody.
This work was supported by research grants GM077875 and NS37402 (to J. Marie Hardwick), R37-HL54598 (to B. O’Rourke), NS40932 (to Y.-T. Hsu), DC01919 and NS018492 (to L.K. Kaczmarek), NS045876 (to E.A. Jonas), and CA010951 (to P.L. Pedersen) from the National Institutes of Health.