- Top of page
- Materials and methods
- Electronic supplemental material
- Supporting Information
Regulation of programmed cell death by Bcl-xL is dependent on both its solution and integral membrane conformations. A conformational change from solution to membrane is also important in this regulation. This conformational change shows a pH-dependence similar to the translocation domain of diphtheria toxin, where an acid-induced molten globule conformation in the absence of lipid vesicles mediates the change from solution to membrane conformations. By contrast, Bcl-xLΔTM in the absence of lipid vesicles exhibits no gross conformational changes upon acidification as observed by near- and far-UV circular dichroism spectropolarimetry. Additionally, no significant local conformational changes upon acidification were observed by heteronuclear NMR spectroscopy of Bcl-xLΔTM. Under conditions that favor the solution conformation (pH 7.4), the free energy of folding for Bcl-xLΔTM (ΔG°) was determined to be 15.8 kcal·mol−1. Surprisingly, under conditions that favor a membrane conformation (pH 4.9), ΔG° was 14.6 kcal·mol−1. These results differ from those obtained with many other membrane-insertable proteins where acid-induced destabilization is important. Therefore, other contributions must be necessary to destabilize the solution conformation Bcl-xL and favor the membrane conformation at pH 4.9. Such contributions might include the presence of a negatively charged membrane or an electrostatic potential across the membrane. Thus, for proteins that adopt both solution and membrane conformations, an obligatory molten globule intermediate may not be necessary. The absence of a molten globule intermediate might have evolved to protect Bcl-xL from intracellular proteases as it undergoes this conformational change essential for its activity.
The Bcl-2 proteins regulate programmed cell death by acting in the cytosol and organellar membranes (Adams and Cory 1998; Chao and Korsmeyer 1998; Green and Reed 1998; Ng and Shore 1998; Harris and Thompson 2000; Hengartner 2000). Some Bcl-2 proteins act by adopting at least two different structural conformations: a solution conformation and an integral membrane conformation. For example, pro-apoptotic Bax is a monomeric, helical bundle protein localized in the cytosol until an apoptotic signal causes translocation to the mitochondrial outer membrane (Suzuki et al. 2000). At the mitochondrial outer membrane, Bax inserts and folds into a large, multimeric integral membrane protein that is thought to regulate the release of cytochrome c (Hsu et al. 1997; Wolter et al. 1997; Shimizu et al. 1999; Antonsson et al. 2000, 2001; Saito et al. 2000).
Similar to Bax, the anti-apoptotic protein Bcl-xL is a soluble, primarily monomeric, helical bundle protein local-izedinparttothecytosol (Muchmore et al. 1996). However, in contrast to Bax, Bcl-xL inserts into the mitochondrial outer membrane and folds into a small, integral membrane protein (Hsu et al. 1997; Minn et al. 1997). For both proteins, the solution-to-membrane conformational change has been reconstituted in vitro with only recombinant proteins and vesicles from synthetic lipids, suggesting that this conformational change might not be receptor-mediated (Minn et al. 1997; Basanez et al. 2001).
The dual structural nature of the Bcl-2 proteins allows for dual mechanisms for their biological activity. In the case of Bcl-xL, the solution conformation acts by sequestering pro-apoptotic factors in the cytosol as a water-soluble helical bundle (Sedlak et al. 1995; Sattler et al. 1997). By contrast, the membrane conformation acts as a small, moderately selective cationic channel in the mitochondrial outer membrane (Minn et al. 1997; Vander Heiden et al. 2001). The exact mechanism of Bcl-xL activity in the membrane is still under debate, but mutants of Bcl-xL that possess altered ion channel properties also have altered apoptotic activities, confirming a biological role for the membrane conformation (Minn et al. 1997, 1999; Xie et al. 1998; Losonczi et al. 2000; Basanez et al. 2001; Vander Heiden et al. 2001). These properties of Bcl-xL were actually demonstrated with Bcl-xLΔTM1 that lacks the C-terminal hydrophobic anchor, which is not required for biological activity or ion channel activity (Muchmore et al. 1996; Minn et al. 1997).
The dual mechanisms for the anti-apoptotic activity of Bcl-xLΔTM in the cytosol and in the membrane raise the question of how a water-soluble protein undergoes a conformational change to become an integral membrane protein. This conformational change appears to have two requirements: acidic pH and presence of lipid vesicles (Basanez et al. 2001). In fact, Bcl-xLΔTM is only weakly associated with lipid vesicles at pH 7.0 but is fully associated at pH 4.5 as measured by a sedimentation assay (Basanez et al. 2001). The ion channel activity of Bcl-xLΔTM also shows a similar pH dependence with conductance occurring readily under acidic conditions but not readily at pH 7.0 (Minn et al. 1997). Because these experiments were performed in vitro with recombinant protein and vesicles derived from synthetic lipids, they suggest that the amino acid sequence alone can specify both solution and membrane conformations.
Many bacterial toxins also undergo a pH-dependent conformational change from solution to membrane conformations including the translocation domain from diphtheria toxin and the colicin family of pore-forming toxins (Parker et al. 1990; Lakey et al. 1992; London 1992; Lacy and Stevens 1998; Zakharov and Cramer 2002a). The solution structures of many of these proteins are known and they share a common helical bundle topology (Parker et al. 1989, 1992; Choe et al. 1992; Elkins et al. 1997). In fact, the motivation to explore ion channel properties of Bcl-xLΔTM arose, in part, from the structural similarity it shares with the translocation domain of diphtheria toxin (Fig. 1), which also binds to lipid vesicles in a pH-dependent manner (Sandvig and Olsnes 1980; Muchmore et al. 1996). Some of these helical bundles retain a helical conformation in the membrane (Oh et al. 1996; Lindeberg et al. 2000; Chenal et al. 2002; Zakharov and Cramer 2002b), but no high-resolution structure of a membrane conformation of these proteins has been determined. The topology of this helical membrane conformation must be quite different from the solution conformation, because the polar or charged residues on the surface of the solution conformation would need to be sequestered from the hydrophobic milieu of the membrane bilayer. Based on these considerations, the solution to membrane conformational change has been referred to folding inside-out (Lesieur et al. 1997).
A change in the solution conformation in the absence of lipid vesicles is known for many proteins to lower the activation energy for the solution to membrane conformational change (Lesieur et al. 1997; Zakharov and Cramer 2002a). This change in the solution conformation can be large, such as a change in quaternary structure, or it can be small, such as a change in the tertiary structure commonly referred to as a molten globule conformation (Bychkova et al. 1996). A molten globule conformation is characterized by native-like secondary structure without the well-packed hydrophobic core found in native-like proteins (Ptitsyn et al. 1990; van der Goot et al. 1991). For example, an acid-induced conformational change in the pore-forming toxin Colicin A, even in the absence of lipid vesicles, results in formation of a molten globule conformation that more readily associates with lipid vesicles than the solution conformation that predominates at pH 7.4 (van der Goot et al. 1991). In fact, an acid-induced molten globule formation is the dominant mechanism for membrane insertion of many other proteins that undergo a solution to membrane conformational change including annexin 6, TRAIL, StAR, diphtheria toxin, and other toxins (Blewitt et al. 1985; van der Goot et al. 1991; Bychkova et al. 1996; Song et al. 2001; Chenal et al. 2002; Nam and Choi 2002).
For Bcl-xLΔTM, a mechanism for the solution to membrane conformational change is not known beyond the requirement for lipid vesicles and acidic conditions. Therefore, we first asked what changes are occurring to this protein under acidic conditions in the absence of lipid vesicles. Specifically we tested whether lowering the pH induces a change in the tertiary or quaternary structure by examining changes in the thermodynamic stability and structural properties of Bcl-xLΔTM.
- Top of page
- Materials and methods
- Electronic supplemental material
- Supporting Information
Here, we present data that suggest the driving force for the solution to membrane conformational change of Bcl-xL is different than the translocation domain from diphtheria toxin by examining the thermodynamic and structural properties of Bcl-xLΔTM as a function of pH in the absence of lipid vesicles. In contrast to diphtheria toxin translocation domain, we find little change in the thermodynamic stability or structure of Bcl-xLΔTM from pH 7.4 to 4.9, while in the presence of lipid vesicles this pH change results in complete association of Bcl-xLΔTM with lipid vesicles (data not shown). The results presented here suggest that this protein does not insert through an obligatory molten globule intermediate that has been observed in the absence of membrane vesicles for other proteins such as cytochrome c, TRAIL, StAR, diphtheria toxin, and other toxins including colicin A, exotoxin A, and equinatoxin (Blewitt et al. 1985; van der Goot et al. 1991; Bychkova et al. 1996; Song et al. 2001; Chenal et al. 2002; Nam and Choi 2002).
The difference in the free energy of folding (ΔΔG°) of Bcl-xLΔTM between pH 7.4 and 4.9 in the absence of lipid vesicles is only 1.2 kcal · mol−1, which is within the uncertainty in the measurements and is small relative to its thermodynamic stability (14.6 kcal · mol−1 at pH 4.9). To confirm these observations, we repeated these measurements several times on different preparations of the protein. Therefore, even at a pH that favors membrane insertion, the solution conformation of Bcl-xLΔTM is still quite stable. We conclude that acid-induced destabilization of the solution conformation does not contribute to the energetics of the solution to membrane conformational change. Therefore, membrane insertion of Bcl-xLΔTM must derive from an increase in the number of acidic side chains that partition into the membrane upon protonation at lower pH.
This result is notable because it differs from many proteins with solution conformations and acidic isoelectric points similar to Bcl-xLΔTM (calculated pI of 4.4), such as diphtheria toxin translocation domain, colicin A, and colicin B. In these cases, the primary driving force is the acid-induced destabilization of the solution conformation and not simply the acidification of side chains that favors the membrane conformation as we find for Bcl-xLΔTM (Ramsay et al. 1989; London 1992; Schendel and Cramer 1994; Lesieur et al. 1997; Sathish et al. 2002; Zakharov and Cramer 2002a)). For example, the rate-limiting step for the solution to membrane conformational change of colicin A is the acid-induced unfolding rate of the solution conformation in the absence of lipid vesicles, and not membrane binding (van der Goot et al. 1991). While the driving forces behind membrane insertion differ between Bcl-xLΔTM and many other toxins, our data do not exclude the possibility that the structural conformations involved for these proteins are similar. These toxins are thought to initially insert into membranes via a hydrophobic helical hairpin (Fig. 1) and then form an umbrella-like intermediate before becoming fully inserted. Our data do not address this issue and, given the structural similarity of the proteins, we expect that the Bcl-xLΔTM follows a similar pathway.
While the ΔΔG° is neglible, we did observe a pH-dependence to the denaturant dependence to the free energy of folding as reflected by the mG value. The mG value decreases from 4.3 ± 0.3 kcal · mol−1 · M−1 at pH 7.4 to 3.5 ± 0.2 kcal · mol−1 · M−1 at pH 4.9. Such a decrease in mG can be interpreted as the presence of an intermediate that is stabilized at acidic pH conditions, leading to a decrease in two-state character and a lower mG value (Whitten et al. 2001). This result might reflect an ensemble-based description of protein structure (Frauenfelder and Leeson 1998; Frauenfelder and McMahon 1998; Whitten et al. 2005). In this description, the macroscopic thermodynamic state of a protein is comprised of an ensemble of microstates that is quite heterogeneous. In this sense, protonation does not lead to changes in the macroscopic free energy of unfolding but can affect the distribution of microstates that could be reflected in a difference in the mG value that we observe.
Our results with Bcl-xLΔTM have implications for the full-length molecule. The requirement for acidic pH conditions in vitro for the solution to membrane conformational change of Bcl-xLΔTM in the presence of lipid vesicles is to potentially increase the likelihood that the protein lacking the C-terminal transmembrane anchor will associate with the membrane (Schendel et al. 1998). This acidic pH requirement might not be necessary for the full-length molecule. Or perhaps a lower decrease in pH is necessary for the full-length molecule to drive the equilibrium from solution to the membrane conformations in vivo. Interestingly, a slight decrease in the pH of the cytosol from ∼7.4 to 6.8 during the initial phases of apoptosis has been reported (Matsuyama et al. 2000). Also, the anionic lipids on the surface of a membrane, like the mitochondrial outer membrane, cause a negative surface potential, increasing proton concentration near the surface and effectively lowering the local pH near the membrane surface (McLaughlin 1989; Menestrina et al. 1989; van der Goot et al. 1991; Murray et al. 1999).
Perhaps the reason for a difference between diphtheria toxin and Bcl-xL relates to the biological process that triggers these conformational changes. While the exact trigger for Bcl-xL is unknown, diphtheria toxin enters the cell via a clathrin-coated endosome that becomes acidic during maturation of the endosome (Draper and Simon 1980; Sandvig and Olsnes 1980). By contrast, Bcl-xL is exposed to a cytosolic environment that is more susceptible to intracellular proteases than the endosome. Therefore, the avoidance of a molten globule intermediate might have evolved to prevent unregulated degradation of Bcl-xL by intracellular proteases that would cleave a molten globule intermediate state more efficiently than the native state of Bcl-xL. Such proteases would not be present in the endosome and might have allowed the evolution of a diphtheria toxin that capitalizes on the acidic nature of the endosome. Based upon the results presented here, Bcl-xL evolved a different driving force for membrane insertion.
The significance of these results extends beyond the Bcl-2 and toxin field because many proteins link their protonation state to functional transitions such as hemoglobin (Ackers 1998). Typically, such changes in protonation state are accompanied by conformational changes; however, as demonstrated here, such conformational changes may not derive from protonation alone. Indeed, pH dependent cooperativity can be quantitatively described without invoking conformational changes for cases when strong electrostatic interactions exist between ligand binding sites and protonation sites (Spassov and Bashford 1998). It appears the Bcl-xL may be an example of this phenomenon.
In summary, to gain a clearer understanding of the pH-dependent solution to membrane conformational change of Bcl-xLΔTM we first analyzed the thermodynamic stability and solution conformation of Bcl-xLΔTM upon acidification in the absence of membranes. Contrary to many other membrane-insertable proteins, we find no evidence for acid-induced unfolding of the Bcl-xL solution conformation. Therefore, the main driving force for membrane insertion must derive from the free energy of binding to the membrane that only occurs upon protonation of acidic residues.