Lipids modulate the BH3-independent membrane targeting and activation of BAX and Bcl-xL

Significance

Here, we characterize a mechanism that modulates the membrane targeting, refolding, and activation of the apoptotic regulators BAX and Bcl-xL in the absence of their canonical activators (BH3-only effector proteins). The critical component of the suggested mechanism is the enrichment of the mitochondrial outer membrane with anionic lipids, primarily cardiolipin. This lipid-dependent activation is found to be coupled to the presence of physiologically relevant concentrations of divalent cations. Membrane insertion of Bcl-xL results in the release of the BH4 regulatory domain from the folded structure, leading to inhibition of pore formation by BAX. These results provide a mechanistic rationale for recent cellular studies that demonstrate the mitochondrial localization and activity of BAX and Bcl-xL in cells lacking BH3-only effectors.

Abstract

Regulation of apoptosis is tightly linked with the targeting of numerous Bcl-2 proteins to the mitochondrial outer membrane (MOM), where their activation or inhibition dictates cell death or survival. According to the traditional view of apoptotic regulation, BH3-effector proteins are indispensable for the cytosol-to-MOM targeting and activation of proapoptotic and antiapoptotic members of the Bcl-2 protein family. This view is challenged by recent studies showing that these processes can occur in cells lacking BH3 effectors by as yet to be determined mechanism(s). Here, we exploit a model membrane system that recapitulates key features of MOM to demonstrate that the proapoptotic Bcl-2 protein BAX and antiapoptotic Bcl-xL have an inherent ability to interact with membranes in the absence of BH3 effectors, but only in the presence of cellular concentrations of Mg²⁺/Ca²⁺. Under these conditions, BAX and Bcl-xL are selectively targeted to membranes, refolded, and activated in the presence of anionic lipids especially the mitochondrial-specific lipid cardiolipin. These results provide a mechanistic explanation for the mitochondrial targeting and activation of Bcl-2 proteins in cells lacking BH3 effectors. At cytosolic Mg²⁺ levels, the BH3-independent activation of BAX could provide localized amplification of apoptotic signaling at regions enriched in cardiolipin (e.g., contact sites between MOM and mitochondrial inner membrane). Increases in MOM cardiolipin, as well as cytosolic [Ca²⁺] during apoptosis could further contribute to its MOM targeting and activity. Meanwhile, the BH3-independent targeting and activation of Bcl-xL to the MOM is expected to counter the action of proapoptotic BAX, thereby preventing premature commitment to apoptosis.

Apoptosis or regulated cell death is a highly conserved process that is required for proper tissue homeostasis and development (1–3). As such, dysregulation of apoptosis is an etiologic factor in a broad range of diseases including cancer, immunodeficiency, neurodegeneration, and autoimmunity (4). Commitment of a cell to apoptosis features the permeabilization of the mitochondrial outer membrane (MOMP) (5, 6) by pore-forming members of the Bcl-2 protein family (e.g., BAX) (7, 8). The formation of oligomeric BAX pores, and subsequent apoptosis, is inhibited by BAX interaction with antiapoptotic Bcl-2 proteins, such as Bcl-xL, at the mitochondrial outer membrane (MOM). A third subset of the Bcl-2 family, the BH3-only proteins (e.g., Bid) activates and targets proapoptotic and antiapoptotic proteins to the MOM (7, 8) (Fig. 1A, Left).

Fig. 1.

BH3-independent membrane interactions of BAX and Bcl-xL. (A) Schematic representations of the BH3-dependent and BH3-independent membrane targeting and activation of BAX and Bcl-xL. The former requires activation by BH3-only effector proteins [e.g., Bid or Bim (8, 9)], while the latter is observed in the absence of these proteins. Cytosolic [Mg²⁺] and [Ca²⁺] ([Mg²⁺]_c, [Ca²⁺]_c) as well as mitochondrial [Ca²⁺] ([Ca²⁺]_M) are indicated as reference (61, 62). The latter represents the upper bound under stimulus induced [Ca²⁺] accumulation, while resting [Ca²⁺]_M is ∼0.0001 mM (61). (B–D) The BH3-independent membrane interactions of BAX and Bcl-xL are induced under physiological-like conditions (37 °C, pH 7.5) in a lipid-dependent manner (brown, 1TOCL:2POPC; green, 1TOCL:6POPC; cyan, 1POPS:2POPC; yellow, POPC). (B) BAX-mediated permeabilization of ANTS/DPX containing LUV, triggered by the addition of 1 mM Mg²⁺. (C) BAX-mediated leakage of 1TOCL:6POPC LUV in the presence of various Mg²⁺ (blue) or Ca²⁺ (red) concentrations. (D) Membrane insertion of full-length Bcl-xL (Bcl-xL FL) measured with the environmental-sensitive fluorophore NBD selectively attached to a single Cys at W169C mutant in α6.

A prevailing model, termed “embedded together,” proposes that functional interactions between members of the Bcl-2 family (i.e., Bcl-xL and BAX) require the cytosol-to-MOM redistribution of these proteins, leading to their refolding and activation (8–10). The mechanism and protein conformations required for these interactions, however, remain unclear. Several models of membranous BAX conformations have been suggested, each requiring a substantial refolding of the soluble structure (11). In the case of Bcl-xL, two different membranous conformations have been identified: “membrane-anchored” and “membrane-inserted.” In the former, the C-terminal helix anchors Bcl-xL to the bilayer while the remaining helices retain their soluble fold (12–14). In contrast, “membrane-inserted” Bcl-xL shows deep membrane penetration of several regions outside of the C-terminal helix and a substantially refolded conformation characterized by the release of its N-terminal BH4 helix (15, 16).

The redistribution of Bcl-xL and BAX to the MOM traditionally has been thought to be triggered by their interactions with BH3-only effector proteins (Fig. 1A, Left) (10, 17, 18). However, recent studies by Luo and coworkers (19, 20) have demonstrated that, in cells lacking any BH3-only effector proteins, Bcl-xL and BAX still redistribute to mitochondria, where they reside in an active state. These results have challenged the long-standing notion that BH3-only effectors such as Bid or Bim are essential for the targeting and activation of Bcl-2 proteins (10, 17, 18) and have led to the introduction of a new model, referred to as membrane-mediated spontaneous activation. This model proposes that MOM lipid composition is sufficient to target and activate Bcl-xL and BAX (21). However, results of previous studies with model membrane systems have not been consistent with this model, as they have shown a requirement for BH3-effectors for the membrane targeting of Bcl-xL and BAX (22, 23). The discrepancy between cellular and model membrane studies suggested that lipid composition alone may not be sufficient to explain the BH3-independent targeting and activation of Bcl-xL and BAX in cells and that a yet-unknown component could be required.

Here, we describe a mechanism for the selective targeting and activation of BAX and Bcl-xL in the absence of BH3-only effectors, henceforth referred to as the BH3-independent mechanism (Fig. 1A, Right). We demonstrate that the apparent discrepancy between the properties of BAX and Bcl-xL targeting and activation in cells and in model membrane systems can be explained by a requirement for divalent cations, which were not present in previous model systems (22, 23). Under physiological levels of Mg²⁺/Ca²⁺, BAX and Bcl-xL have a significant propensity to interact with anionic membranes, consistent with the membrane-mediated spontaneous activation model (21). These interactions are facilitated by anionic lipids, particularly by the mitochondrial specific lipid cardiolipin. Lipid composition in the presence of divalent cations not only modulates the targeting of Bcl-xL and BAX but also leads to their refolding into multiple active conformations.

Our studies do not imply physiological regulation by Mg²⁺, which maintains a constant cytosolic concentration at 1 mM. Instead, they point to variations in MOM cardiolipin concentration as the likely key regulatory mechanism of this pathway, echoing the “membrane (lipids) mediated spontaneous activation model” presented in the literature (21). This regulation can occur either by the changes in local MOM cardiolipin content [i.e., enrichment at contact sites with the inner membrane (24–26)] or via the global changes in MOM cardiolipin levels during apoptosis (27, 28). It is also possible that transient increases in cytosolic Ca²⁺ near the MOM will also contribute to the regulation via cardiolipin. Our results provide insights into how changes in lipid composition modulate the cytosol-to-MOM redistribution and activation of BAX and Bcl-xL independent of BH3-only proteins.

Results

Membrane Permeabilization by BAX in the Absence of BH3 Effector Proteins.

Membrane permeabilization by the pore-former BAX was studied using leakage assays with large unilamellar vesicles (LUVs) encapsulating the fluorophore/quencher pair ANTS/DPX, an accepted in vitro model of MOMP (22, 23). The addition of BAX to vesicles containing cardiolipin (TOCL), a mitochondrial specific lipid, at a 1:2 ratio with phosphatidylcholine (POPC) as the matrix lipid did not result in leakage (Fig. 1B, gray). This result is consistent with the previously reported low membrane affinity of BAX in the absence of BH3-only effectors (22, 23). The presence of cytosolic levels of Mg²⁺ (1 mM), however, greatly increases membrane permeation by BAX (Fig. 1B, brown). Control measurements demonstrating that Mg²⁺ by itself does not induce membrane leakage or LUV aggregation under these conditions are presented in SI Appendix, Fig. S1.

The efficiency of membrane permeation in the presence of Mg²⁺ was modulated by the lipid composition of the vesicles. For example, no leakage was observed in the absence of anionic lipids (Fig. 1B, yellow); while the presence of the anionic lipids cardiolipin (Fig. 1B, green and brown) or phosphatidylserine (POPS) (Fig. 1B, cyan) resulted in leakage. The larger levels of leakage observed in 1TOCL:6POPC cardiolipin containing membranes compared POPS (1POPS:2POPC), both of which have a similar molar charge of ∼30%, points to selectivity within different anionic lipids. This indicates that higher cardiolipin concentrations at the hotspots of apoptotic regulation in the MOM (24–26), as well as MOM cardiolipin increases during apoptosis (27, 28), can modulate the levels of membrane bound and active BAX in the MOM through a BH3-independent mechanism.

Leakage measurements performed at various Mg²⁺ or Ca²⁺ concentrations revealed that both cations induce equivalent levels of membrane leakage at equimolar concentrations (Fig. 1C). Effects of divalent cations on BAX leakage were not replicated by monovalent cations (SI Appendix). The dependence of BAX leakage on Mg²⁺ concentration should not be taken to mean that Mg²⁺ is an active regulator of BAX (Mg²⁺ concentration is constant in the cytosol). Instead, this dependence shows that divalent cations are a crucial component that links bilayer lipid properties to the membrane interactions of BAX. At constant Mg²⁺ levels, it is therefore the changes in lipid composition that are the modulators of the BH3-independent mechanism. Transient fluxes of Ca²⁺ can also be a contributing factor of this lipid-dependent regulation.

Membrane Interactions of Full-Length Bcl-xL in the Absence of BH3 Effector Proteins.

The membrane interaction of Bcl-xL, a BAX inhibitor, was characterized in the absence of BH3 effectors using the environmentally sensitive fluorophore nitrobenzoxadiazole (NBD) introduced at position W169C in a single-Cys variant of full-length Bcl-xL (Bcl-xL FL). This residue is located in the middle of the helix α6, which penetrates the bilayer in the membrane-inserted conformation of Bcl-xL (16) and remains outside the bilayer in its membrane-anchored conformation (12–14).

In the absence of divalent cations, the addition of cardiolipin containing LUV led to a blue shift in the NBD position of the maximum from 544.5 nm in the absence of LUV to 541 and 542.5 nm, respectively, when 1TOCL:2POPC (SI Appendix, Fig. S3A) or 1TOCL:6POPC (SI Appendix, Fig. S3B) were introduced. The spectral blue shifts were accompanied by an increase in fluorescence intensity. Together, these spectroscopic changes indicate the transition of the NBD probe to the bilayer and thus the membrane interaction of α6, suggesting the formation of the membrane-inserted state of Bcl-xL.

The presence of Mg²⁺ promoted Bcl-xL FL membrane interactions up to 1.5 mM Mg²⁺, at which point the NBD signal reached a plateau in 1TOCL:2POPC LUV (Fig. 1D, brown). As in the case of BAX, the BH3-independent membrane interactions of Bcl-xL FL were modulated by lipid composition and promoted by higher molar ratios of cardiolipin. At cytosolic 1 mM Mg²⁺, for example, a 95% membranous Bcl-xL FL fraction is observed in 1TOCL:2POPC LUV (Fig. 1D, brown). However, an equimolar [Mg²⁺] leads to a 70% membranous fraction in 1TOCL:6POPC (Fig. 1D, green) and 0% in POPC (Fig. 1C, yellow). This lipid dependence was confirmed by substituting cardiolipin for POPS (1TOCL:6POPC vs. 1POPS:2POPC), which only yielded an approximate 10% membranous Bcl-xL FL fraction at 1 mM Mg²⁺ (Fig. 1C, cyan). Control measurements in the absence of LUV showed no changes in the NBD spectra of Bcl-xL FL after the addition of Mg²⁺ (SI Appendix, Fig. S4).

Below, we explore in detail the coupled effects of membrane lipid composition and Mg²⁺/Ca²⁺ concentrations on the BH3-independent anchoring, insertion, refolding, and activity of Bcl-xL.

BH3-Independent Membrane Interactions of Bcl-xL Are Independent of Anchoring via C-Terminal Helix.

The characteristic feature of the membrane-anchored conformation of Bcl-xL FL is the transmembrane orientation of its C-terminal helix (12–14). [Note that the anchored conformation does not spontaneously anchor itself upon mixing with the vesicles and is usually prepared in vitro by the formation of nanodiscs around the anchoring helix (12–14).] To determine whether the BH3-independent membrane targeting of Bcl-xL FL produces an anchored conformation, the intensity of NBD-labeled Bcl-xL FL at its C-terminal tail (G222C) was measured in response to the irreversible soluble quencher dithionite, often used in topology measurements of NBD-labeled sites on membrane proteins (29, 30).

In order to determine the range of protection of the NBD probe attached to Bcl-xL FL from dithionite, we have compared two samples: 1) soluble protein in the absence of membranes, and 2) tail-anchored protein, in which lipid nanodiscs were formed around the C-terminal helix. In the absence of membranes, the addition of dithionite led to a rapid decrease in intensity to a zero level (Fig. 2, gray) due to complete exposure of the fluorophore in the soluble form of Bcl-xL FL. In contrast, the anchoring of the C-terminal helix in nanodiscs prevented signal quenching due to the protection of the NBD probe by the bilayer (Fig. 2, green). This difference establishes clear benchmarks for the bilayer protection of the protein anchoring into LUV. As expected, addition of 1TOCL:2POPC LUV in the absence of divalent cations resulted in no protection (Fig. 2, black). The addition of 2 mM Mg²⁺ or Ca²⁺ (Fig. 2, blue and red) produces an intermediate level of protection. The latter cannot be attributed to partial binding, as under these conditions the binding of Bcl-xL FL is complete (Fig. 1D). The relatively high quenching observed in the BH3-independent membranous form of Bcl-xL FL (i.e., divalent cations + LUV) can be explained by different possibilities that will be explored in the future, such as 1) poor Mg²⁺/Ca²⁺-induced insertion efficiency of the C-terminal helix at pH 7.5, 2) a deeply penetrating interfacial helix, or 3) a dynamic equilibrium for the C-terminal helix between an interfacial and a transmembrane confirmation. Nonetheless, these results suggest that the BH3-independent membrane targeting of Bcl-xL FL does not involve anchoring via C-terminal helix.

Fig. 2.

Membrane anchoring of full-length Bcl-xL via C-terminal α8 helix in 1TOCL:2POPC LUV (pH 7.5, 37 °C). The transmembrane insertion of the anchoring α8 helix was tested by measuring the protection of the NBD fluorophore attached in the middle of α8 at G222C from the soluble quencher dithionite. Addition of dithionite results in a rapid and complete decrease in fluorescence in the absence of membrane interactions (gray, black). The formation of nanodiscs in the presence of Bcl-xL provides strong protection from quenching (green), consistent with the transmembrane orientation of α8 under such conditions (12–14). The insertion of full-length Bcl-xL into vesicles in the presence of divalent cations results in intermediate quenching (Mg²⁺, blue; Ca²⁺, red).

Modulation of BH3-Independent Bcl-xL Membrane Insertion.

To confirm the lack of C-terminal helix involvement in the BH3-independent membrane targeting of Bcl-xL FL, we studied the Mg²⁺/Ca²⁺-induced membrane interactions of a construct with a C-terminal helix deletion. This construct will be referred to as Bcl-xL to differentiate it from Bcl-xL FL used in Figs. 1 and 2. Bcl-xL was labeled with NBD at position W169C in α6 (same position used for measurements in Fig. 1D).

In the absence of LUV at pH 8, Bcl-xL W169C-NBD presents a position of maximum of 529 nm (Fig. 3A, gray) that remains mostly unaffected after the addition of 1TOCL:2POPC LUV (Fig. 3A, black). The addition of 2 mM Mg²⁺, however, leads to a twofold intensity increase accompanied by a blue shift to 525 nm (Fig. 3A, red). These spectral changes indicate the transition of Bcl-xL α6 to the bilayer and confirm that the BH3-independent membrane targeting of Bcl-xL does not require its C-terminal helix. These effects were reversed by the addition of excess EDTA (SI Appendix, Fig. S5A) and enhanced by the protonation of Bcl-xL (Fig. 3A, brown), a known membrane insertion trigger (15, 16). This pH dependence indicates that although Mg²⁺ is a potent trigger of Bcl-xL membrane insertion it does not induce the redistribution of the entire Bcl-xL population at pH 8 and suggests a possible relationship between Mg²⁺ and protonation. The lack of significant spectroscopic changes in the absence of LUV (SI Appendix, Fig. S6) confirms that the Mg²⁺-dependent effects are due to the transition of Bcl-xL α6 to the bilayer and not caused by protein aggregation in solution, consistent with Bcl-xL FL measurements (SI Appendix, Fig. S4).

Fig. 3.

BH3-independent membrane insertion of Bcl-xL in 1TOCL:2POPC LUV. The membrane insertion of Bcl-xL α6 helix was tested using NBD-labeled Bcl-xL W169C. (A) Representative NBD spectrum before and after the addition of Mg²⁺ at 25 °C. The Mg²⁺-dependent insertion Bcl-xL was reversed by the addition of EDTA (SI Appendix, Fig. S6A). (B) Relative membrane insertion of Bcl-xL quantified using the increases in NBD fluorescence intensity at 510 nm. The membrane-inserted populations were proportional to [Mg²⁺]/[Ca²⁺].

The coupling between divalent cations and protonation was characterized by performing pH titrations in the presence of variable concentrations of Mg²⁺ and Ca²⁺ (Fig. 3B). All measurements were performed in 1TOCL:2POPC LUV at 25 °C. The relative levels of Bcl-xL insertion were determined using changes in NBD intensity at 510 nm. In the absence of divalent cations, Bcl-xL insertion is strongly pH dependent (Fig. 3B, black), which is consistent with published observations (15, 16, 31). The presence of divalent cations resulted in increased insertion at neutral and basic pH, flattening the pH dependence without changes in the apparent pK_a (Fig. 3B, color-coded data). Note, that both Mg²⁺ and Ca²⁺ have the same effect on insertion of Bcl-xL, which was also the case for activation of BAX (Fig. 1C). As in the case of BAX, the effects of Mg²⁺ and Ca²⁺ were not replicated by Na⁺ or K⁺ (SI Appendix).

BH3-Independent Refolding of Membrane-Inserted Bcl-xL: Release of N-Terminal BH4 Helix.

As demonstrated in our previous Förster resonance energy transfer (FRET) studies, the membrane-inserted form of Bcl-xL induced by acidic pH is characterized by the release of its N-terminal BH4 helix (15, 16). Here, we employed the same FRET methodology and construct to determine whether the BH3-independent membrane insertion of Bcl-xL leads to a similar structural rearrangement. These measurements rely on a Bcl-xL N-terminally conjugated to the fluorescent protein mCherry and the fluorophore AlexaFluor488 introduced at position D189C in Bcl-xL. The deletion of the C-terminal helix was previously shown not to affect the ability of Bcl-xL to refold and release the BH4 helix in its membrane-inserted state (16).

In the absence of LUV, the close proximity of the donor Alexa Fluor 488 to the acceptor mCherry in the soluble fold of Bcl-xL results in FRET between both fluorophores (Fig. 4A, gray). No significant intensity changes are observed after the addition of 1TOCL:2POPC LUV at pH 7.5 in the absence of divalent cations (Fig. 4A, black). The presence of either Mg²⁺ or Ca²⁺, however, leads to increases in donor Alexa Fluor 488 intensity that are accompanied by decreases of the acceptor mCherry signal (Fig. 4A, red and blue). This loss of FRET is caused by the increase in distance between both fluorophores and indicates that the BH3-independent interaction of Bcl-xL with membranes leads to its refolding and the release of its BH4 helix. Control measurements in the absence of LUV showed no changes in FRET as a function of divalent cation concentration (SI Appendix, Fig. S9), confirming that the Mg²⁺/Ca²⁺ effects require the interaction of Bcl-xL with membranes.

Fig. 4.

BH3-independent refolding of membrane-inserted Bcl-xL in 1TOCL:2POPC LUV at pH 7.5 at 25 °C. FRET measurements between mCherry conjugated at the N terminus of Bcl-xL and an Alexa 488 fluorophore at Bcl-xL D189C, as previously described (15, 16). (A) Representative emission spectra showing FRET between both fluorophores. The addition of either Mg²⁺ (blue) or Ca²⁺ (red) leads to loss of FRET in the presence of LUV, indicative of BH4 release. Bcl-xL refolding was reversible by the addition of EDTA at 37 °C (SI Appendix, Fig. S6B). (B) Representative smFRET distributions of Bcl-xL measured using fluorescence correlation spectroscopy (FCS) at different [Mg²⁺] concentrations: 0 mM Mg²⁺ (black) soluble Bcl-xL, 2 mM Mg²⁺ (purple) membrane-inserted Bcl-xL with BH4 released, and 0.5 mM Mg²⁺ (orange) membrane-inserted Bcl-xL with intermediate compactness. (C) FRET efficiencies calculated for ensemble (pink) and FCS (green). (D) Lipid-dependent modulation of Bcl-xL refolding in membranes with increasing cardiolipin content.

The BH3-independent release of the BH4 helix was further characterized by single-molecule fluorescence correlation spectroscopy (FCS) measurements using 1TOCL:2POPC LUV at pH 7.5. In the absence of divalent cations, Alexa Fluor 488-labeled mCherry-Bcl-xL D189C presents a single-molecule FRET (smFRET) efficiency centered at 0.4 (Fig. 4B, gray), consistent with the expected theoretical FRET efficiency of this construct (15). A lower smFRET efficiency of 0.05 (Fig. 4B, purple) was observed in the presence of 2 mM Mg²⁺, while a distribution in between those determined at 0 and 2 mM Mg²⁺ was determined at 0.5 mM Mg²⁺ (Fig. 4B, orange). Multiple intermediate smFRET efficiencies were observed at various [Mg²⁺] (Fig. 4C, green), indicating that the BH3-independent release of the BH4 helix involves multiple Bcl-xL conformations with different degrees of compactness. The presence of multiple Bcl-xL conformations in the bilayer is consistent with the protonation-dependent refolding of Bcl-xL (15). These results were confirmed by ensemble FRET measurements (Fig. 4C, pink), which yielded equivalent FRET efficiencies as those determined by FCS (Fig. 4C, green).

The lipid modulation of this refolding was determined using LUV with increasing molar contents of cardiolipin. Similar to the BH3-independent insertion of Bcl-xL (Fig. 1C and SI Appendix, Fig. S10), the release of the BH4 helix was significantly higher in membranes with larger contents of cardiolipin (Fig. 4D). For example, in 1TOCL:2POPC LUV, the presence of cytosolic 1 mM Mg²⁺ led to 60% of the Bcl-xL population to release its BH4 helix, while only 10% refolding was observed in 1TOCL:6POPC LUV and 0% in POPC LUV. This indicates that changes in mitochondrial lipid composition in the presence of divalent cations not only affects the propensity for membrane bound Bcl-xL but also serves as a switch between different conformational states.

Bcl-xL Membrane Insertion and Refolding Kinetics in the Absence of BH3 Effectors.

The kinetics of Bcl-xL insertion and refolding were determined in the presence of 1 mM Mg²⁺ or Ca²⁺ to characterize the time response of the BH3-independent mechanism. Measurements were performed in 1TOCL:2POPC LUV, which showed the most favorable refolding of Bcl-xL under steady-state conditions (Fig. 4D). Both divalent cations produced similar kinetic traces for membrane insertion (Fig. 5A, triangles) and refolding (Fig. 5A, circles), providing further evidence for the equivalency of Mg²⁺ and Ca²⁺ for the BH3-independent membrane interactions of Bcl-xL. These measurements also showed a stark difference in the rates for both events, where the refolding of Bcl-xL occurs after its membrane insertion. This suggests that Bcl-xL can be present in the bilayer (in an inserted state) without incurring extensive refolding (BH4 helix release), trapped by a relatively high energy barrier.

Fig. 5.

Kinetics of Bcl-xL membrane insertion and refolding in 1TOCL:2POPC LUV at pH 7.5. Time dependence of Bcl-xL membrane insertion (triangles) and BH4 release (circles) in the presence of 1 mM Mg²⁺ (blue) or 1 mM Ca²⁺ (red). (A) The presence of Mg²⁺ or Ca²⁺ did not affect the kinetics of either event, consistent with our steady-state measurements. Insertion and refolding do not occur simultaneously; instead, Bcl-xL refolding in the bilayer is delayed relative to its membrane insertion. (B) Arrhenius plot of Bcl-xL BH3-independent refolding kinetics measured between 25 and 42 °C in the presence of 1 mM Mg²⁺ or Ca²⁺. The shaded area represents a 95% confidence band for the slope estimate.

We have estimated the activation energy required for refolding by measuring the rates of refolding as a function of the temperature (SI Appendix, Fig. S11). It appears that the refolding kinetics can be represented by a single transition rate, k, which rapidly increased with increasing temperature. This temperature dependence is presented in Fig. 5B as an Arrhenius plot, lnk vs. 1/T, where T is an absolute temperature. The activation energy, estimated by multiplying the slope of the Arrhenius dependence by universal gas constant, equals E_a = 30 ± 5 kcal/mol and is of the same range reported for refolding of other membrane proteins (32). Temperature did not affect measurements under steady-state conditions (SI Appendix, Fig. S12).

Our refolding FRET studies suggest that the BH3-independent insertion of Bcl-xL leads to multiple membrane-inserted Bcl-xL conformations in the bilayer, with cardiolipin concentration serving as the switch between them. Two of these conformations can be readily identified upon membrane insertion: 1) high cardiolipin: membrane-inserted Bcl-xL with a released BH4 helix and 2) low cardiolipin: membrane-inserted Bcl-xL with a nonreleased BH4 helix.

Multiple Conformations of Membrane-Inserted Bcl-xL Inhibit BAX Pores.

The activity of the membrane-inserted conformations induced by the BH3-independent insertion of Bcl-xL at low and high cardiolipin content were tested by measuring their ability to block BAX-mediated leakage.

To determine whether the Bcl-xL conformation that releases its BH4 helix is active, we incubated 1TOCL:2POPC LUV containing the fluorophore/quencher pair ANTS/DPX with either Bcl-xL FL (Fig. 6A, blue) or Bcl-xL (Fig. 6A, red) in the presence 1 mM Mg²⁺ for 30 min at 37 °C. This incubation time ensured its maximal refolding in the bilayer (Fig. 5A, circles). Under these conditions, the addition of BAX (at 1BAX:5Bcl-xL) only produced 10% leakage regardless of the Bcl-xL construct used, while an 85% leakage was produced in the absence of Bcl-xL/Bcl-xL FL (Fig. 6A, gray). The large decrease in BAX-induced leakage in the presence of Bcl-xL demonstrates that the BH3-independent membrane-inserted form of Bcl-xL in high cardiolipin membranes (BH4 released) is active and able to block BAX. Furthermore, the equivalency between the inhibitory effects of both Bcl-xL variants demonstrates that the C-terminal helix is not required for the BH3-independent antiapoptotic activity of Bcl-xL, providing further validation for its deletion in this study. Measurements at multiple BAX:Bcl-xL ratios revealed that at least three Bcl-xL monomers per BAX are required for the complete inhibition of BAX-mediated leakage (Fig. 6B).

Fig. 6.

Pore-inhibiting action of Bcl-xL in the absence of BH3 effectors (pH 7.5, 37 °C). Inhibition of BH3-independent BAX mediated leakage by Bcl-xL FL and Bcl-xL in the absence of BH3-effectors in the presence of 1 mM Mg²⁺. Measurements were performed in LUV with high (1TOCL:2POPC) and low (1TOCL:6POPC) cardiolipin to capture the refolded and compact conformations, respectively, of membrane inserted Bcl-xL (Fig. 4D). (A) The refolded form of Bcl-xL (BH4 released) in 1TOCL:2POPC LUV is active (blue) and able to inhibit BAX. The presence or absence of the C-terminal membrane anchor helix did not affect inhibition of BAX leakage (red vs. blue). (B) Measurements at various BAX:Bcl-xL ratios in 1TOCL:2POPC LUV revealed that at least three Bcl-xL units are required per one BAX to achieve complete inhibition (red). (C) The compact conformation of membrane inserted Bcl-xL in low cardiolipin was also active and able to inhibit BAX-mediated leakage.

In the case of 1TOCL:6POPC LUV, the addition of BAX at 1 mM Mg²⁺ results in 50% of the maximal leakage possible (Fig. 6C, gray). In the presence of Bcl-xL, however, BAX-mediated leakage is severely inhibited and only 10% leakage is observed at a 1BAX:5Bcl-xL ratio (Fig. 6C, red). Similar results were observed at all [Mg²⁺] tested, with a maximal 10% BAX-induced leakage observed even at 2 mM Mg²⁺ when both proteins were present (Fig. 6C). These results demonstrate that the membrane-inserted conformation of Bcl-xL in low cardiolipin membranes (BH4 not released) is also active and able to inhibit BAX-mediated leakage. BAX-mediated leakage is therefore inhibited by multiple conformations of membrane inserted Bcl-xL, with lipid composition serving as a switch between its different active conformations in the absence of BH3 effectors (Fig. 7).

Fig. 7.

Schematic representation of the proposed mechanism of BH3-independent MOM targeting and activation of Bcl-2 proteins. The results presented here demonstrate that changes in membrane lipid composition, coupled with cellular levels of divalent cations, result in the membrane targeting of BAX and Bcl-xL. This targeting leads to the activation of both proapoptotic BAX and antiapoptotic Bcl-xL in the absence of BH3-only effector proteins, which explains the results reported for cell lines lacking BH3-only effectors (19, 20). At normal low concentrations of cardiolipin, Bcl-xL is targeted to the bilayer in a compact conformation, consistent with its canonic inhibition of apoptosis. Increases in cardiolipin content at the MOM [such as those observed during apoptosis (27, 28)] or local enrichment at mitochondrial contact sites (24–26) results in the insertion of Bcl-xL in a refolded conformation, which leads to the release of its BH4 helix, involved in noncanonic apoptotic inhibition (55). Both Bcl-xL conformations, unreleased BH4 at low cardiolipin and BH4 released at high cardiolipin, are active and able to block BAX-mediated leakage (Fig. 6).

Discussion

Every tissue, cell, and organelle have specific electrochemical properties resulting from different concentrations of protons, small ions, and charged macromolecules. Delivering proteins to these different environments can trigger functionally relevant structural rearrangements. A notable example of such conformational switching results in the transition of soluble proteins into lipid membranes, which features prominently in many physiological processes, including the regulation of apoptosis by the Bcl-2 protein family (8–10, 33–35).

Bcl-2 proteins regulate and execute MOMP, recognized as the point of no return in intrinsic apoptosis, via a series of protein–protein and protein–membrane interactions (10, 36, 37). For example, the membrane recruitment and activation of proapoptotic Bcl-2 proteins results in their oligomerization and the formation of multimeric pores responsible for MOMP; while the role of antiapoptotic Bcl-2 proteins is to inhibit this process by making nonproductive heterodimers with proapoptotic proteins (38, 39). Alterations to their function contribute to cancer pathogenesis and chemotherapy resistance, making them important targets for drug discovery (4, 40–43). The functions of the individual family members are largely determined by specific combinations of their four conserved motifs known as Bcl-2 homology (BH) regions (35). For example, the proapoptotic pore-forming proteins (e.g., BAX, Bak, Bok) contain BH regions 1–3; while the anti-apoptotic proteins (e.g., BCL-2, Bcl-w, Bcl-xL) contain all four BH regions, and the BH3-only effector proteins (Bid, Bim, Bad, Puma, etc.) only contain the BH3 region. The consensus from cellular studies is that the action of Bcl-2 proteins is intimately associated with their membrane interactions, which produce their active conformations (8, 10, 37). The nature of the modulation of regulatory interactions within the Bcl-2 family by membranes, however, remains elusive.

A crucial component of the mitochondrial recruitment of Bcl-2 proteins is the lipid composition of the mitochondrial membranes, known to undergo significant changes during apoptosis (27, 28). Particularly, the content of the mitochondrial-specific lipid cardiolipin has been linked to the regulation of mitochondrial permeabilization (27, 28). Under healthy cellular conditions, the bulk of the mitochondrial cardiolipin resides in the inner mitochondrial membrane and accounts only for ∼4% of the lipids in MOM. Cardiolipin is, however, redistributed to the MOM (from the inner mitochondrial membrane) during apoptosis (28). Local content of cardiolipin under healthy conditions is also variable and can reach up to 30% in specific contact sites between inner and outer membranes (44). These sites, with enriched cardiolipin levels, are believed to be the hot spots of apoptotic regulation, altering the recruitment and activation of Bcl-2 proteins (24–26).

The traditional model of intrinsic apoptosis states that BH3-only effectors are necessary to recruit proapoptotic and antiapoptotic Bcl-2 proteins to the MOM. Recent cellular studies where all BH3-only effector proteins were deleted has put in question this view by demonstrating that Bcl-xL and BAX are selectively targeted to mitochondria, where they are present in an active state, even in the absence of BH3-only effectors (19, 20). These results imply an alternative mechanism for the membrane targeting of these proteins. Understanding the involvement of BH3-only effectors on the targeting and activation of Bcl-xL and BAX is particularly important due to their role in cancer, which has led to the development of several peptides and molecules that mimic of BH3-only effectors that aim at controlling apoptosis (45–47).

The results presented here are in agreement with the BH3-independent targeting of Bcl-xL and BAX observed in cells (19, 20). Furthermore, they provide a rationalization for these results and their discrepancy in previous studies with model membranes. The key to this mechanism is the modulation of membrane lipid composition, which provides the trigger for the membrane partitioning of Bcl-xL and BAX and their conformational switch into their active conformations. Interestingly, these effects are only observed in the presence of cytosolic levels of divalent cations (previously omitted in model membrane studies) which couple lipid-specific effects to protein–membrane interactions. The specificity of divalent cations on this process is supported by the inability of monovalent cations like Na⁺ or K⁺ to reproduce or affect the effects of divalent cations on the targeting and activation of BAX or Bcl-xL (SI Appendix, Figs. S2, S7, and S8).

In terms of regulatory implications, our work points out to the role of cardiolipin as a signaling molecule, linking the BH3-independent regulation of apoptosis (21) to changes in MOM lipid composition (27, 28). While the molecular action of cardiolipin is mediated by divalent cations, this is different from the latter playing a direct regulatory role. Many cellular processes are regulated by Ca²⁺ signaling, including the apoptotic cascades downstream of MOM permeabilization (48, 49). Such regulation is carried out by systems that have high selectivity for Ca²⁺ and ensure that micromolar changes in Ca²⁺ levels are detectable against the background of invariable cytosolic concentrations of Mg²⁺ that are on the order of 1 mM. This is clearly not the case here, as we demonstrate that similar Ca²⁺ and Mg²⁺ have equivalent effects on lipid-dependent recruitment and activation of both BAX (Figs. 1 and 6) and Bcl-xL (Figs. 3 and 6). Under standard levels of Ca²⁺ and Mg²⁺ concentrations in the cytosol, however, the cardiolipin-dependent regulation is expected to be coupled primarily to the latter.

For example, the presence of cytosolic 1 mM Mg²⁺ led to ∼50% BAX-induced leakage in low cardiolipin 1TOCL:6POPC LUV (Fig. 1B, green) and a 70% inserted population of full-length Bcl-xL (Fig. 1D, green). The latter represents an active conformation of Bcl-xL capable of inhibiting BAX-mediated leakage (Fig. 6C). At this cardiolipin concentration, the active form of Bcl-xL is present in a compact state with an unreleased BH4 (α1) helix (Figs. 4D and 7). An alternative refolded conformation of Bcl-xL in which the N-terminal BH4 helix is released from the rest of the protein identified at an ∼30% molar content of cardiolipin (Figs. 4 and 7) was also determined to be active (Fig. 6A). Higher levels of cardiolipin also led to higher levels of membrane inserted Bcl-xL FL and BAX-mediated leakage (Fig. 1). The identification of these conformations indicates that 1) Bcl-xL is present at multiple conformations in the bilayer, 2) at least two different Bcl-xL conformations are able to inhibit BAX-mediated leakage, 3) these conformations can be formed in the absence of BH3 effectors and are modulated by lipid composition, and 4) this lipid-dependent modulation requires the presence of cytosolic levels of divalent cations.

The stable, high levels of cytosolic Mg²⁺ do not necessarily preclude Ca²⁺ from being involved in BH3-independent targeting and activation of Bcl-xL and BAX by cardiolipin. It is feasible that various Ca²⁺ bursts (50) could amplify the BH3-independent targeting and activation of Bcl-2 proteins. Similar effects could also arise from the accumulation of Ca²⁺ near the MOM due to disruption of Ca²⁺ storage organelles during apoptosis (51). For example, the initial release Ca²⁺ from mitochondria in the early stages of MOMP can serve as an amplifying factor to recruit BAX, leading to the release of larger apoptotic factors [e.g., cytochrome c, Smac/DIABLO (51, 52)]. This amplification would, however, still be dependent on the lipid composition of the MOM. Interestingly, Bcl-xL has been implicated in the regulation of Ca²⁺ fluxes through the interaction of its BH4 helix, released in the presence of divalent cations in a lipid-dependent fashion (Fig. 4), with Ca²⁺ channels such as VDAC (53, 54). The BH4 helix of Bcl-xL has also been directly linked to apoptotic regulation by providing an alternative mode to bind and inhibit BAX (noncanonical model) (15, 55). The regulation of Bcl-xL refolding by the BH3-independent mechanism described here could therefore modulate various aspects of apoptotic regulation.

The possible clues to understanding the molecular mechanism behind the coupled effects of lipids and divalent cations on Bcl-xL and BAX effects may come from our recent study of the cancer-targeting peptide pHLIP (pH-low insertion peptide) (56). It demonstrates that this short anionic peptide, with a well-characterized pH-dependent transition from a soluble to a transmembrane state (57), also inserts into membranes at neutral pH in the presence of Mg²⁺ or Ca²⁺. This pH-independent insertion can be attributed to a decrease of the thermodynamic penalty for partitioning of anionic residues into the membranes. In the case of pHLIP, the Mg²⁺/Ca²⁺-mediated coordination of Glu and Asp sidechains with lipids depends strongly on the nature of the lipid, with headgroups playing the central role (56). We suggest that a similar coupling between anionic residues and lipid headgroups mediated by divalent cations may constitute a generic mechanism for targeting of Bcl-2 proteins to the MOM. The ability of cells to regulate the number of Mg²⁺/Ca²⁺ coordination sites present in membranes by changing the fractional content of anionic lipids provides them with a convenient trigger to modulate protein–membrane interactions and activity. For example, MOM lipid composition is modulated by different events such as apoptosis, which lead to extensive lipid composition changes, including changes in MOM cardiolipin (a lipid with two anionic moieties in its headgroup) (27, 28). We hypothesize that apoptosis-related changes in MOM cardiolipin content disrupt the soluble vs. membranous balance, conformation and activation of BAX and Bcl-xL (Fig. 7). Additionally, regions of the MOM at its contact sites with the inner mitochondrial membrane (often called hot spots) are locally enriched in cardiolipin (24–26) and could provide further amplification of BAX/Bcl-xL interactions. The BH3-independent membrane targeting and activation of Bcl-xL and BAX, however, by no means undermines the physiological role of BH3-only effectors in transmitting an apoptotic signal.

The results on Bcl-xL and BAX, presented here, along with those we reported for cancer targeting peptide pHLIP (56), point to a crucial role of physiological levels of Mg²⁺ and Ca²⁺ on protein–membrane interactions. Thus, the inclusion of physiological levels of Mg²⁺/Ca²⁺ into in vitro experimental protocols is critical for realistic representations of protein–membrane interactions and conformational switching occurring in many physiologically important systems. Additionally, because all members of the Bcl-2 protein family share significant sequence and structural similarities (58), it would not be unreasonable to assume that divalent cations will be essential for the action of many other proteins in the family, including the BH3-only effectors. We expect that the reported effect of divalent cations (i.e., bridging of anionic moieties in lipids and amino acids) can promote the membrane binding and even insertion of proteins/peptides previously not considered to be capable of membrane binding.

Materials and Methods

Materials.

POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine), cardiolipin (TOCL: 1′,3′-bis[1,2-dioleoyl-sn-glycero-3-phospho]-glycerol), and POPS (1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-L-serine) were purchased from Avanti Polar Lipids. The fluorescent dyes ANTS (8-aminonaphthalene-1,3,6-trisulfonic acid), DPX (p-xylene-bis-pyridinium bromide), Alexa Fluor 488-maleimide, and IANBD amide [N,N′-dimethyl-N-(iodoacetyl)-N′-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)ethylenediamine] were obtained from Invitrogen.

Cloning, Expression, and Labeling.

All full-length Bcl-xL (Bcl-xL FL) and Bcl-xL (lacking C-terminal helix) constructs used were cloned and purified as previously described (15, 16). The traditionally used phosphate buffer was substituted for Hepes buffer in all protein purifications to prevent the chelation of Mg²⁺ and Ca²⁺ by phosphate. Fluorescent labeling with IANBD or Alexa Fluor 488 was performed using a standard labeling protocol for thiol-reactive dyes (59), and excess dye was removed by gel filtration in a 1 × 30-cm Superose 6 column.

LUVs.

The appropriate volume of chloroform lipid stocks was dried using a nitrogen stream and placed under high vacuum to dry overnight. The dried lipid films were resuspended in 20 mM Hepes buffer plus 100 mM NaCl, pH 8, to a final LUV concentration of 20 mM. The rehydrated lipid films were vortexed, and the LUVs were prepared by extrusion using a Mini-Extruder (Avanti Polar Lipids) with 0.1-µm nucleopore polycarbonate membranes (Whatman) and stored at 4 °C (60). LUVs containing entrapped ANTS/DPX were prepared by rehydrating the dried lipid film with Hepes buffer containing 1 mM ANTS and 10 mM DPX. The rehydrated samples were subjected to 10 freeze–thaw cycles followed by extrusion. ANTS/DPX-containing LUVs were prepared at a concentration of 100 mM to maximize the loading of the solutes. Nonincorporated ANTS and DPX molecules were removed by FPLC using a Superose 12 1 × 30-cm column.

Ensemble Fluorescence Measurements.

Fluorescence emission was measured on a SPEX Fluorolog FL3-22 steady-state fluorescence spectrometer (Jobin Yvon) equipped with double-grating excitation and emission monochromators. All experiments were made using 2 × 10-mm cuvettes oriented perpendicular to the excitation beam and collected after a 20-min incubation period. Experimental temperature was maintained constant using a Peltier device from Quantum Northwest.

NBD measurements were collected from 490 to 650 nm using an excitation wavelength of 470 nm, while FRET measurements between mCherry and Alexa Fluor 488 were collected between 500 and 650 nm using an excitation wavelength of 465 nm. All emission spectra were collected using 5-nm slits on both monochromators at 1-nm steps. The plotted spectra represent the average of two to three scans. All samples contained 0.3 μM protein and 1 mM LUV. Experiments between pH 8 and 10 were carried out using 20 mM borate buffer, while measurements between pH 6 and 8 used 20 mM Hepes buffer. These buffers were selected to prevent chelation or salt formation with the added Mg²⁺ or Ca²⁺. Sample acidification was achieved by the addition of small aliquots of 0.5 M acetate. The NBD intensity changes at 510 nm were used to calculate relative insertion, and the decreases in fluorescence intensity of the acceptor mCherry measured at 605 nm were used to estimate BH4 release percentages.

Ensemble FRET efficiencies presented in Fig. 5 were determined using the following formula:

$$E=1-\frac{F_{DA}}{F_D},$$ [1]

where F_D and F_DA represent to the fluorescence intensity of the donor Alexa Fluor 488 in the absence or presence of the acceptor mCherry, respectively. The F_D parameter was determined using Alexa Fluor 488-labeled Bcl-xL D189C lacking mCherry.

The pH-dependent data were fitted to the following equation:

$$I=\frac{I_N+I_L\hspace{1em}\left({10}^{m\left(\mathrm{p}{\mathrm{K}}_{\mathrm{a}}-\text{pH}\right)}\right)}{1+{10}^{m\left(\mathrm{p}{\mathrm{K}}_{\mathrm{a}}-\text{pH}\right)}},$$ [2]

where I is the intensity at 510 nm, I_N and I_L are the limiting intensities at high and low pH, respectively, and m is the transition slope.

All NBD kinetic measurements (dithionite quenching of C-terminal helix and insertion of α6) were measured by following NBD intensity changes at 510 nm. The kinetics of BH4 release was measured by following the increase in donor Alexa Fluor 488 intensity at 518 nm.

Single-Molecule FCS.

smFRET measurements were performed as previously described (15) using a MicroTime 200 confocal microscope (PicoQuant) with a 60×, N.A. 1.2 Olympus water-immersion objective and 50-μm confocal pinhole, resulting in a 1-fL confocal detection volume. All measurements employed 0.1 µM Alexa Fluor 488-labeled mCherry-Bcl-xL D189C and 1 mM LUV and a collection time of 20 min.

A pulsed picosecond diode laser LDH-P-C-470 at 40 MHz was used to excite the donor (Alexa Fluor 488). The fluorescent signal was then split into two using a 50/50 beam splitter cube onto two single-photon avalanche diodes (SPCM—AQR—14; Perkin-Elmer). The donor and acceptor signals were detected separately using an AHF/Chroma HQ 520/40 emission filter for the donor and a HQ 550LP long-pass band filter for the acceptor. A TimeHarp 200 board was used to detect the emitted signal using time-correlated single photon counting and store the data in the time-tagged time-resolved mode.

smFRET efficiency, E, was calculated using the following formula:

$$E=\frac{I_A}{I_A+\gamma \cdot I_D},$$ [3]

where I_D and I_A are the number of photons detected in the donor and acceptor channels, respectively, and γ is a correction factor to account for differences in detection efficiencies between the donor and acceptor photomultipliers.

LUV Light-Scattering Measurements.

The aggregation of LUV at pH 7.5 in the presence of Ca²⁺ or Mg²⁺ was measured by light-scattering changes at 400 nm as previously described (56). Samples were performed on a 4 × 10-mm quartz cuvette containing 0.1% LUV in 20 mM Hepes buffer plus 20 mM NaCl.

Data Availability

All study data are included in the article and/or SI Appendix.