Structural Insights of tBid, the Caspase-8-activated Bid, and Its BH3 Domain*

Background: The membrane-associated tBid promotes Bax membrane insertion and activation. Results: tBid adopts an extended structure in 1-palmitoyl-2-hydroxy-sn-glycero-3-[phospho-RAC-(1-glycerol)] (LPPG) micelles with its six helices including the BH3 domain interacting with the micelles. Conclusion: An “on the membrane” binding mode was suggested for tBid interaction with Bax. Significance: Revealing tBid structure on the membrane is key to the understanding of tBid-mediated Bax activation. The Bcl-2 family proteins regulate mitochondria-mediated apoptosis through intricate molecular mechanisms. One of the pro-apoptotic proteins, tBid, can induce apoptosis by promoting Bax activation, Bax homo-oligomerization, and mitochondrial outer membrane permeabilization. Association of tBid on the mitochondrial outer membrane is key to its biological function. Therefore knowing the conformation of tBid on the membrane will be the first step toward understanding its crucial role in triggering apoptosis. Here, we present NMR characterization of the structure and dynamics of human tBid in 1-palmitoyl-2-hydroxy-sn-glycero-3-[phospho-RAC-(1-glycerol)] micelles. Our data showed that tBid is monomeric with six well defined α-helices in the micelles. Compared with the full-length Bid structure, a longer flexible loop between tBid helix α4 and α5 was observed. Helices in tBid do not pack into a compact-fold but form an extended structure with a C-shape configuration in the micelles. All six tBid helices were shown to interact with LPPG micelles, with helix α6 and α7 being more embedded. Of note, the BH3-containing helix α3, which was previously believed to be exposed above the membrane surface, is also membrane associated, suggesting an “on the membrane” binding mode for tBid interaction with Bax. Our data provided structural details on the membrane-associated state of tBid and the functional implications of its membrane-associated BH3 domain.

Apoptosis (programmed cell death) plays an important role in tissue development and maintenance (1). A group of apoptosis regulator proteins, the Bcl-2 (B-cell lymphoma 2) family proteins, were discovered to work coordinately to modulate mitochondrial outer membrane permeabilization (MOMP) 2 (2)(3)(4). Bcl-2 family members share at least one of four BH (Bcl-2 homology) domains and many of them possess a transmembrane domain at the C terminus. Based on their apoptotic function, Bcl-2 family proteins are classified into two groups: pro-apoptotic and anti-apoptotic proteins. A subgroup of the pro-apoptotic proteins, containing only the BH 3 domain (e.g. Bid, Bim, and PUMA), can directly interact with pro-apoptotic proteins Bax/Bak and promote their homo-oligomerization on the membrane and lead to MOMP (5)(6)(7). This represents the major process of the direct activation model for apoptosis initiation. Another competing model, the indirect activation (or displacement) model, suggests anti-apoptotic proteins (e.g. Bcl-2, Bcl-xL, Bcl-w, and Mcl-1) can heterodimerize with Bax/ Bak to suppress their pro-apoptotic activity (8). Within this model, a sensitizing role was implied for the BH 3 -only proteins (e.g. Bad), which can bind anti-apoptotic proteins, displace Bax/ Bak, and indirectly promote Bax/Bak oligomerization and MOMP (9 -16). Cumulative evidences imply mitochondrial membrane insertion of many Bcl-2 family proteins is critical for their apoptotic functions and the translocation from cytosol to mitochondrial membrane could account for totally different inter-protein interactions among Bcl-2 family members (13,17,18). To accommodate such effects in the context of membrane association, the "embedded together" model was proposed with the recognition that final steps of Bax/Bak activation and MOMP take place on the membrane (13). Currently, little structural information is available on the membrane-associated state of Bcl-2 family proteins compared with their well studied cytosolic forms (19).
The BH 3 -only protein, Bid, was shown to bridge the crosstalk between extrinsic and intrinsic pathways of apoptosis through its cleavage by caspase-8 (20 -22), a downstream mediator of Fas or the tumor necrosis factor (TNF) death receptor signaling pathway (23,24). The proteolytic cleavage of cytosolic Bid by activated caspase-8 results in two fragments: p 7 (7 kDa) and p 15 (15 kDa) (21,25). After cleavage, the C-terminal fragment (p 15 ), also known as tBid (truncated Bid), translocates to the mitochondrial outer membrane (21). The membrane asso-ciation of tBid was shown to be important for the efficient recruitment of cytosolic Bax to the mitochondrial outer membrane (18,26). Through interaction with the tBid BH 3 domain, Bax was suggested to undergo conformational changes, membrane insertion, and homo-oligomerization at the mitochondrial outer membrane, mediating MOMP and cytochrome c release (17, 18, 26 -28). However, studies also suggested tBid could bypass Bax/Bak and directly promote the mitochondrial permeabilization by tBid alone with the induction of the negative membrane curvature (29 -31).
In the cytosol, both human and mouse Bid contain eight ␣-helices with two hydrophobic helices (helix ␣ 6 and ␣ 7 ) forming the core of the protein (32,33). After caspase-8 cleavage, the Bid N-terminal p 7 fragment retains tight association with tBid through primarily hydrophobic interactions (32). Earlier studies suggested that tBid dissociated from the p 7 fragment upon contact with the hydrophobic membrane and exposed its hydrophobic p 7 binding pocket, promoting tBid to undergo conformational changes on the membrane (34 -36). In vitro, isolated tBid was shown to bind to membranes quickly and is competent to activate Bax-mediated membrane permeabilization (18). A previous solid-state NMR study suggested that tBid associated with the membrane with an orientation parallel to the membrane surface and without membrane-traversing helix insertion (37). Additionally, membrane-associated tBid showed helical propensity at Bid helices ␣ 6 -8 regions with helix ␣ 6 tilted into the membrane based on the EPR studies (34). However, these studies did not provide specific information on tBid helical boundaries, the tertiary-fold of tBid, and its overall interaction with the membrane. Currently, a proposed model indicated that tBid helices ␣ 6 -8 are membrane associated, leaving the tBid BH 3 domain fully exposed above the membrane surface for potential interactions with other proteins (35). This model is useful to explain some of the observed tBid functional data in solution but provided limited information regarding tBid function on the membrane (35). The exposed conformation of tBid BH 3 domain above the membrane surface is energetically unfavorable unless stabilized by other proteins. It is inconsistent with the finding that the tBid BH 3 domain peptide was membrane associated and fully competent to activate Bax (38). Thus, it is still not clear whether the tBid BH 3 domain forms a helix on the membrane and whether its hydrophobic residues that were believed to be involved in the protein interactions in solution behave differently on the membrane by interacting with the membrane instead (20).
To address these questions with more structural details, we chose to determine the full-length tBid structure on the membrane along with the characterization of tBid regions that are interacting with the membrane. Here, we used solution NMR methods to characterize human tBid in membrane mimic 1-palmitoyl-2-hydroxy-sn-glycero-3-[phospho-RAC-(1-glycerol)] (LPPG) micelles. Our data showed tBid preserved all six Bid helices (␣ 3 to ␣ 8 ), including the BH 3 -containing helix ␣ 3 . Compared to full-length Bid, the biggest difference was observed for tBid helix ␣ 5 , which is a shorter helix with a longer flexible loop connecting it to helix ␣ 4 . tBid possessed a C-shape conformation with a close but no direct contact between N and C termini of tBid in LPPG micelles as evidenced by paramag-netic relaxation enhancement (PRE) measurements. Furthermore, extensive interactions with LPPG micelles were observed for all six tBid helices. Specifically, the BH 3 -containing helix ␣ 3 , which was previously believed to be exposed outside the membrane, is also membrane associated. The protein-micelle interaction and protein backbone dynamics data suggested helices ␣ 6 and ␣ 7 are likely to be more embedded into the micelles than the rest of the helices. Our findings provide detailed structural information of membrane-associated tBid and suggest the membrane-associated feature of tBid BH 3 domain favors an "on the membrane" binding mode with Bax.

EXPERIMENTAL PROCEDURES
Cloning and Sample Preparation-The pET15b-Bid FL vector carrying human Bid cDNA was used to produce full-length Bid protein. To produce tBid in a more efficient manner without caspase-8 cleavage, we replaced the caspase-8 cleavage site with a TEV protease cleavage site by mutating the pET15b-Bid FL vector ( 53 YDELQTDG to 53 DENLYFQG). Thus the modified full-length Bid can be cleaved by TEV protease and the resulting cleavage product tBid will be the same as the native tBid. The modified plasmid (pET15b-Bid TEV ) was transformed into Escherichia coli BL21(DE3) cells to express recombinant protein. Isotopically ( 15 N or 15 N, 13 C) labeled proteins were produced in isotope-enriched M9 minimal media. The perdeuterated tBid sample was grown in 99% D 2 O minimal media. The E. coli cells were grown at 37°C, induced with 1 mM isopropyl 1-thio-␤-Dgalactopyranoside at A 600 ϳ0.8, and cell growth was continued overnight at 18°C. The cells were harvested by centrifugation, resuspended in lysis buffer (20 mM Tris-HCl, pH 8.0, 500 mM NaCl, 5 mM imidazole, 1 mM PMSF), and sonicated. The lysate was centrifuged at 34,000 ϫ g for 30 min at 4°C and the supernatant was loaded to Ni 2ϩ affinity chromatography. The His 6tagged full-length Bid TEV was eluted using elution buffer (20 mM Tris-HCl, pH 8.0, 500 mM NaCl, 250 mM imidazole) and dialyzed into buffer A (20 mM Tris-HCl, pH 8.0) before it was further purified by an ion-exchange chromatography on a HiTrap 1-ml Q column (GE Healthcare) with a linear salt gradient. The purified full-length Bid TEV was pooled and mixed with 200 units of AcTEV protease (Invitrogen) for cleavage by dialyzing against 1 liter of TEV cleavage buffer (50 mM Tris-HCl, pH 8.0, 0.5 mM EDTA, 1.0 mM DTT) overnight at room temperature. The cleaved protein mixture was loaded onto the Ni 2ϩ affinity chromatography and tBid was separated by washing with lysis buffer with 1.2% (w/v) n-octyl-␤-D-glucoside. The fraction containing purified tBid was collected and dialyzed against buffer (20 mM Tris-HCl, 100 mM NaCl, pH 8.0) to remove detergent n-octyl-␤-D-glucoside. After dialysis, tBid precipitated out of solution and was collected by centrifugation. The NMR sample of tBid at 0.5-1.0 mM was prepared by dissolving the protein pellet into NMR buffer (50 mM potassium phosphate, pH 6.6) supplemented with 8% (160 mM) LPPG, 10% D 2 O, and 0.02% sodium azide. The tBid samples in LPPG micelles were kept at 45°C at all times for better sample stability.
Characterization of Protein-Micelle Interaction-A 15 N-labeled perdeuterated tBid sample was expressed in M 9 minimal media supplemented with 99.9% D 2 O and 15 N ammonium chloride. The three-dimensional 15 N-edited NOESY-HSQC (t mix ϭ 70 ms) experiment was acquired to obtain the NOE distance information between the protein backbone amide protons (H N ) and the LPPG protons.
Site-directed Mutagenesis-tBid single cysteine mutants (S78C, Q136C, and Q180C) were prepared using a QuikChange site-directed mutagenesis kit (Stratagene). These point mutations were introduced using pET15b-Bid TEV expressing plasmid as template. All mutants were confirmed by DNA sequencing.
Paramagnetic Relaxation Enhancement-Because there is no cysteine in tBid, non-native cysteines were introduced by sitedirected mutagenesis at the respective sites mentioned above. The full-length Bid bearing the cysteine mutant was first expressed, purified, and cleaved as described above. After TEV cleavage, the protein sample was soaked with DTT at a final concentration of 20 mM to reduce any possible cysteine disulfide bonds. The protein was then exchanged into a reaction buffer (100 mM Tris-HCl, 1 mM diethylenetriaminepentaacetic acid, pH 8.5) using a PD-10 desalting column. Following that, parallel spin probe labeling reactions were carried out by mixing the protein sample with 20ϫ molar excess of the paramagnetic probe MTSL (1-oxyl-2,2,5,5-tetramethyl-3-pyrroline-3methyl) methanethiosulfonate (Toronto Research Chemicals Inc.) or diamagnetic analog dMTSL (1-acetyl-2,2,5,5-tetramethyl-3-pyrroline-3-methyl)methanethiosulfonate (Toronto Research Chemicals Inc.) at room temperature overnight. The stock of both MTSL and dMTSL reagents had been previously dissolved at 20 mg/ml in acetonitrile. After the reactions, excess spin label was removed by dialysis against a buffer (20 mM Tris-HCl, 100 mM NaCl, pH 8.0). Both the paramagnetic and diamagnetic protein samples were further purified in 1.2% n-octyl-␤-D-glucoside as described above. Experimental errors for the PRE were estimated as previously described (51). Labeling of the tBid samples was confirmed by liquid chromatography/mass spectrometry. A two-dimensional 1 H, 15 N-HSQC spectrum of each paramagnetic tBid sample was recorded and no obvious chemical shift changes observed, indicating spin labeling did not disrupt tBid structure. Solvent PRE experiments were performed by stepwise addition of the water-soluble spin label reagent gadodiamide (aqua [5,8- 13 ]gadolinium hydrate; trade name Omniscan, GE Healthcare) into the NMR sample to arrive at final concentrations of 0.25, 0.5, 1.0, 2.0, and 5.0 mM. A reference spectrum was recorded prior to the addition of spin label reagent.
Structure Calculation-Cross-peak intensities from NOESY experiments were translated into a continuous distribution of proton-proton distances. Based on the assigned tBid backbone chemical shifts, the TALOS program (52) was used to predict and dihedral angles. The statistically significant dihedral angles in regular ␣-helical secondary structure were used as structural restraints. Generic hydrogen bond distance restraints were imposed for residues located at well defined ␣-helical regions. Additionally, PRE relaxation rates were calibrated over residue pairs within one single helix and converted into distances for structure calculation. The backbone T 1 /T 2 relaxation ratios were also implemented into the structure calculation as previously described (53). Planar energy restraints for tBid residues with strong micelle interactions were also applied by mimicking a close distance with the membrane interface (54). tBid structures were calculated using simulated annealing protocol by the program Xplor-NIH (55). The final structure calculations employed 749 inter-residual and 180 intra-residual interproton distance restraints, 80 hydrogen bond distance restraints, 137 PRE distance restraints, 72 and 72 angle restraints, 59 T 1 /T 2 ratio restraints, and 28 planar energy restraints (Table 1). Structure presentation was prepared using PyMOL (56).

RESULTS
Secondary Structure of tBid in LPPG Micelles-tBid was produced by incubating purified modified full-length Bid with TEV protease instead of caspase-8 (25) (see "Experimental Procedures"). The choice of LPPG micelles to solubilize tBid for NMR characterization was made based on previous studies (37, 57) and a screen through various detergent micelles. The acquired two-dimensional 1 H, 15 N-HSQC spectrum of tBid in NMR buffer with 8% LPPG at 45°C showed varying peak intensities suggesting tBid residues might experience different effective rotational correlation times. This is likely due to their different states of association with the micelles. Analysis of the NMR datasets yielded near complete backbone and side chain resonance assignments of tBid. As many as 125 of 131 nonproline tBid backbone amides were assigned and only residues at the N terminus (Asn 62 to His 66 ) were not assigned due to the exchange with the solvent. The 1 H N NMR resonances of tBid showed a small dispersion, indicating the presence of mostly helical and random coil conformation. To be noted, the tBid C-terminal region (Ser 188 -Asp 195 ) branched with another set of distinct resonance peaks with slightly weaker intensities, suggesting the flexible C-terminal tail experienced at least two distinct conformations.
Backbone Dynamics of tBid in LPPG Micelles-The backbone 15 N T 1 and T 2 relaxations were measured for tBid at both 800 and 600 MHz proton resonance frequencies. The backbone 15 N T 2 relaxation data indicated that all tBid helical regions are rigid compared with the non-helical regions (Fig. 2B), and show similar diffusive behavior. This implies that these rigid helices tumble together with the micelle. In contrast, the loop between helices ␣ 4 and ␣ 5 (Asn 115 -Asp 126 ) are highly flexible with high T 2 and low T 1 values (Fig. 2), similar to flexible N and C termini. The average T 1 and T 2 relaxation times for the rigid helical regions at 800 MHz frequency are 1.55 Ϯ 0.21 s and 38.2 Ϯ 4.1 ms, respectively. The calculated T 1 /T 2 ratios (data not shown) gave an estimate of rotational correlation time ( c ) of 16.0 ns, which is roughly equivalent to a globular protein with a molecular mass of 43 kDa in solution. The variations in T 1 /T 2 ratios (data not shown) for the helices in tBid indicate that the molecule is quite anisotropic in LPPG micelles. It is also worth noting that tBid helices ␣ 6 and ␣ 7 have shorter 15 N T 2 values (33.3 Ϯ 1.1 ms) and higher T 1 /T 2 ratios (55.0 Ϯ 2.9) at 800 MHz frequency than the rest of the helices observed (Fig. 2). The similar T 1 /T 2 ratios also suggested helices ␣ 6 and ␣ 7 have comparable effective correlation times and orient similarly to each other with respect to the rotational diffusion axis. To identify residues that experience large amplitude fast motion, residue specific order parameters (S 2 ) were calculated by fitting the relaxation data obtained at the two magnetic fields (Fig. 2C). As expected the exposed loops between helices ␣ 3 -␣ 4 , ␣ 4 -␣ 5 , and ␣ 5 -␣ 6 have low S 2 , in addition to the termini. All helices show no evidence of large amplitude fast motion. The initial probe for possible conformational exchange was done by evaluating the ratio of T 2 values obtained at 600 versus 800 MHz (data not shown). Residues whose T 2 ratio lies outside the 1.  The consecutive large positive bars in 13 C␣, 13 CO and negative bars in 13 C␤, 1 H␣ suggest the presence of the ␣-helical conformation. The defined ␣-helices were also indicated by the characteristic medium range inter-proton NOE connectivities of H␣ i to H␤ iϩ3 and strong NOE connectivities of H N i to H N iϩ1 . The thickness of bars used to show NOE connectivities corresponds to the relative NOE intensity. The missing NOEs within the ␣-helical regions are due to overlapped peaks. The secondary structures of human Bid reported previously were also shown for comparison (32). Residue numbering is kept the same as the full-length Bid.
Although residues Leu 149 and Ala 156 are in ␣ 6 , and Gly 193 , Met 194 , and Asp 195 are at the C-terminal all show significant deviation. To identify additional residues that might be undergoing conformational exchange as well as to confirm those identified by the T 2 ratio at the two magnetic fields, relaxation dispersion experiments were carried out. The relaxation dispersion data (Fig. 2D) suggested a few residues (Asn 124 , Glu 145 , Ala 156 , Val 159 , Thr 173 , and Ile 178 ) undergo microsecond-millisecond time scale exchange. They are identified by choosing a cutoff of ⌬R 2 beyond the 1.5 times S.D. Residues Glu 145 , Ala 156 , and Val 159 suggest tBid helix ␣ 6 is likely to experience some conformational exchange, which is consistent with the T 2 ratio data. Of note, residue Asn 124 at the loop between helices ␣ 4 -␣ 5 and Thr 173 and Ile 178 at helix ␣ 7 are also shown to undergo conformational exchange.
tBid Adopts a C-shape Conformation in LPPG Micelles-The near complete assignments of the inter-proton cross-peak NOE dataset yielded 931 NOEs. Analysis of these inter-proton NOEs did not produce any long-range contacts (greater than 5 residues apart). The absence of long-range NOEs renders the determination of tertiary fold of the six ␣-helices tBid structure impossible. Therefore, longer distance oriented PRE experiments were performed. Paramagnetic spin labels were incorpo-rated at three tBid sites with cysteine mutations (S78C, Q136C, and Q180C). The backbone 1 H N -⌫ 2 rate measurements were carried out and the results showed residues close to the paramagnetic center experienced obvious enhancement (Fig. 3). For the S78C sample, weak enhancements were shown for residues at the tBid C terminus, suggesting a close conformation but with no NOE contact (Ͻ5 Å) between the tBid N and C terminus (Fig. 3). Consistently, weaker PRE effects were shown for tBid N-terminal residues to the C-terminal spin label at residue 180 (Q180C sample). For Q136C, weak PRE effects were observed for residues in helices ␣ 4 and ␣ 5 as well as some residues in the N and C termini (Fig. 3). Another possible contribution to the observed PRE effects is due to intermolecular contacts. To address whether tBid homooligomerizes in LPPG micelles, PRE measurements by mixing 15 N-labeled wild type tBid with spin-labeled 14 N-tBid at equal molar ratios were carried out and no obvious enhancement was observed for both S78C and Q136C samples (data not shown). This implied no inter-molecular contact between tBid monomers and therefore tBid does not homo-oligomerize and is monomeric in LPPG micelles.
All tBid Helices Are Involved in the Interaction with LPPG Micelles-To understand the molecular interaction between tBid and LPPG micelles, a perdeuterated 15 N-labeled tBid was Error bar for the 15 N T 1 and T 2 relaxation data were estimated as previously described (50). D, the residues undergoing microsecondmillisecond conformational exchange identified using the backbone 15 N R 2 -CPMG rate differences, ⌬R 2 , the difference between measurements at two different cp values (⌬R 2 ϭ R 2 (6 ms) Ϫ R 2 (0.3 ms)). The cutoff line corresponds to the ⌬R 2 threshold of 1.5 times the S.D. The missing values are due to resonance peak overlap. Secondary structure elements of tBid are indicated at the top for reference.
prepared in LPPG micelles. The three-dimensional 15 N-edited NOESY-HSQC experiment was carried out to detect direct NOE interactions between the protein backbone amide protons and protons from LPPG micelles. The LPPG molecule possesses a long acyl chain with abundant methylene protons, which gives a dominant 1 H NMR signal at a chemical shift of 1.26 ppm (Fig. 4A). Extensive analysis of the NOE cross-peaks at this chemical shift yielded a residue-specific protein-micelle interaction profile. Due to overlaps with the signals from some tBid Thr H␥ 2 or Ile H␥ 1 protons because of incomplete deuteration, we simply omitted those residues for further consideration of possible protein-micelle interactions. In the end, 28 tBid residues with obvious protein-micelle interactions were identified (Fig. 4B). These tBid residues are mostly hydrophobic residues and are scattered over all six tBid helices:  4B). This is consistent with previous EPR studies that showed tBid helices ␣ 6 -8 are membrane-associated (34). Of note, three positively charged residues (Lys 157 , Arg 168 , and His 172 ) in these two helices also showed strong micelle interactions. Surprisingly, the BH 3 domain containing helix ␣ 3 , which was previously proposed to be exposed and outside the membrane surface, also show extensive micelle interactions (Fig. 4A). Taken together, tBid helices ␣ 3-5 and ␣ 8 showed a consistent micelle interaction pattern for hydrophobic residues located at one side of the helix, suggesting tBid helices are embedded parallel to the membrane surface. In contrast, residues (Ser 76 , Glu 77 , Ser 78 , Gln 79 , Arg 118 , Ser 119 , Glu 120 , Glu 121 , Asn 124 , and Gly 193 ) in tBid flexible regions including the long loop (Asn 115 -Asp 126 ) between helix ␣ 4 and ␣ 5 clearly did not show any obvious interaction with the micelles. This observation was further corroborated with the strong protein-water interaction for this flexible region (data not shown). In addition, the solvent PRE experiments were carried out and showed consistent limited water accessibility, which is a weak solvent PRE effect, for residues that are lipid embedded for all tBid helices and a strong PRE effect for water-exposed residues (Fig. 4C). Structure of tBid in LPPG Micelles-Structure calculation of tBid were carried out using the following NMR restraints: NOE-derived distances, TALOS-derived dihedral angles, generic hydrogen bonds, PRE-derived distances, and T 1 /T 2 derived relaxation restraints (Table 1). With the assumption that tBid helices are embedded parallel to the membrane surface, tBid residues that showed a strong micelle interaction were also implemented into the structure calculation with residue-specific planar energy restraints. The calculated 20 lowest Residues experiencing significant PRE effects are mainly localized to residues that are close to the paramagnetic center. For the S78C sample (A), weak PRE effects for residues in the C terminus included helix ␣ 8 . C, albeit even weaker, similar weak PRE effects were also observed between the tBid N terminus and C terminus. For the Q136C sample (B), weak PRE effects were observed for residues in helices ␣ 4 and ␣ 4 as well as some residues in flexible N and C termini. Error bars were estimated as previously described (51). Secondary structure elements of tBid are indicated at the top for reference.  (Fig. 5A). The calculated tBid structures adopted an extended conformation with the N-terminal end being close but without direct atomic contact to the C terminus, thus forming a C-shape structure (Fig. 5, A and B). Relative to the membrane interface, tBid helices formed two faces that showed dramatic differences in elec-trostatic surface potentials (Fig. 5C). One side of the tBid structure that faces the membrane shows dominant hydrophobic patches, which correspond to hydrophobic tBid residues that showed strong interactions to the LPPG micelles (Fig. 5D). Distinct electrostatic surfaces were shown on the other side of the tBid structure with many charged surfaces including BH 3 domain helix ␣ 3 (Fig. 5E). With micelle-interacting residues (Ile 86 , Ala 87 , Leu 90 , Val 93 , and Gly 94 ) in helix ␣ 3 facing toward the membrane surface, charged or polar residues (Asp 81 , Arg 84 , Arg 88 , Gln 92 , and Asp 95 ) are left facing away from the membrane surface for potential electrostatic interaction with other proteins (Fig. 5, B and C). Of note, strong positively charged surface areas were formed along tBid helices ␣ 6 -␣ 8 (Fig. 5E), which could potentially form electrostatic interactions with the negatively charged LPPG micelle head groups. This electrostatic interaction would prevent these helices to be completely buried inside the micelles. This is consistent with a previously suggested tBid conformation on the membrane by solid-state NMR studies (37). Overall, these structural features displayed by tBid in LPPG micelles are consistent with the fact that most tBid helices are amphipathic.

DISCUSSION
The cleavage of full-length Bid by caspase-8 forms the activated tBid, which can be translocated to the mitochondrial outer membrane (21). The association of tBid to membrane is critical for the promotion of apoptosis (18). Upon interaction with the hydrophobic environment in vitro, tBid undergoes conformational changes for favorable lipid interactions (34). Similar to full-length Bid, tBid was shown to be highly helical in various membrane mimics (37,58,59). Previously, membraneassociated tBid was shown to be parallel to the membrane surface with no transmembrane helix and the region bound to the membrane was suggested to be located at helices ␣ 6 -␣ 8 (34,37). Those studies did not provide details on tBid helical boundaries Conversely, these residues did not show an obvious cross-peaks signal to water (4.61 ppm). B, relative peak intensities are calculated by taking the ratios of measured NOE intensities between protein backbone amides ( 1 H N ) and the LPPG acyl chain methylene (CH 2 ) groups at 1.26 ppm against the diagonal peak intensities. These relative peak intensity ratios are used to quantify the strength of interaction between various sites in tBid and micelles. Residues showing obvious protein-lipid interactions are from all six tBid helices (␣ 3-8 ). C, the experimental solvent PRE measurement for tBid backbone amide protons. The measured 1 H N -⌫ 2 PRE values in the presence of a water-soluble paramagnetic probe showed obvious relaxation enhancement for residues that are water exposed and weak enhancement for residues that are embedded in the lipid. Secondary structure elements of tBid are indicated at the top for reference. and the membrane-associated full-length tBid structure at the atomic level. Especially, no structural information on the functionally important BH 3 domain is yet available. This renders understanding of the precise apoptotic function of tBid and its molecular interactions with Bax/Bak or Bcl-xL on the membrane impossible. To fill this gap, we determined the solution structure of full-length human tBid in a lipid membrane mimic, LPPG micelles by NMR.
Secondary structure characterization by NMR demonstrated that tBid has six well defined helices, which are mostly preserved from the full-length Bid (Fig. 1). However, some obvious differences were observed between full-length Bid and tBid in LPPG micelles. For instance, tBid residues (Ser 119 -Asp 126 ) close to the N-terminal end of helix ␣ 5 did not form a rigid helix (Fig. 1). Compared to the full-length Bid structure, the loop between helices ␣ 4 and ␣ 5 is much shorter (32) than the loop in tBid. This extended tBid loop between helices ␣ 4 and ␣ 5 was shown to be highly flexible and did not show any interaction with the micelles, contrasting to the rigid helical conformation in Bid (Figs. 1 and 4). The highly flexible behavior of this loop is likely due to the presence of many charged residues (Arg 118 , Glu 120 , Glu 121 , Asp 122 , Arg 123 , Arg 125 , and Asp 126 ), which could potentiate electrostatic interactions of tBid with other protein(s). Recent peptide screening proposed an important role for this flexible tBid region (60), through which tBid can directly interact with the protein MTCH2 (mitochondrial carrier homologue 2) (36,61). The exposure of this flexible tBid loop enables a scenario where the interaction between tBid and MTCH2 is likely to occur at the mitochondria outer membrane. Overall (B) tBid adopts a C-shape structure with both N and C termini in close conformation. C, tBid helices (␣ [3][4][5][6][7][8] ) are parallel to the putative membrane surface with helices ␣ 6 and ␣ 7 more embedded. For the membrane-associated tBid helix ␣ 3 , charged or polar residues (Asp 81 , Arg 84 , Arg 88 , Gln 92 , and Asp 95 ) are facing away from the membrane surface and are exposed for potential interactions with other proteins. C, on the opposite side of helix ␣ 3 residues (Ile 86 , Ala 87 , Leu 90 , Val 93 , and Gly 94 ) with strong micelle interactions are facing toward the membrane surface. The flexible N terminus (Gly 61 -Ser 78 ) was omitted for clarity. D, electrostatic potential surface view of the side of the tBid structure facing the membrane shows dominant hydrophobic patches along the ordered tBid helices. E, the opposite side of the tBid surface shows strong charged areas along tBid helices with a continuous positive charged surface along the helices ␣ 6 -␣ 8 . D and E, the flexible loop with a highly charged surface between helix ␣ 4 and helix ␣ 5 contains a number of charged residues. DECEMBER 13, 2013 • VOLUME 288 • NUMBER 50

JOURNAL OF BIOLOGICAL CHEMISTRY 35847
The analysis of tBid NOE data did not produce any longrange NOEs, indicating that six well defined tBid helices do not form a compact structure (Fig. 5). Long distance-oriented PRE measurements, however, showed some distances between the tBid N and C termini within 20 -25 Å (Fig. 3). These distances impose a C-shape conformation for tBid on LPPG micelles surface. A rather similar result was also seen previously for the micelle-bound ␣-synuclein (62) with a contrast that tBid is more complex with 6 helices. The size of the LPPG micelle was estimated to have a diameter of ϳ68 Å (63). At the longest circumference, it is considerably too large for tBid to wrap around it and form a C-shape configuration as described (Fig.  5). Based on our experimental data, however, we believe the C-shape conformation of tBid is likely due to its characteristic interaction with the curved LPPG micelle membrane within one constrained area on the micelle hemisphere. This is consistent with previous findings that tBid could colocalize with cardiolipin, a mitochondrial lipid that was previously suggested to induce mitochondrial membrane curvatures (5), at the mitochondrial outer membrane contact sites with "curvature stress" (64). Therefore, tBid could be attracted to the contact sites with membrane curvatures on the mitochondria outer membrane, hence sensitizing the mitochondrial membrane for Bax-induced membrane permeabilization.
The tBid ␣ 6 helix contains a stretch of hydrophobic residues ( 148 MLVLALLLA 156 ) and were predicted to be membrane-associated. Our data indeed showed strong micelle interactions for hydrophobic residues in tBid helix ␣ 6 and ␣ 7 as well (Fig. 4). The actual tBid membrane interaction site was suggested to be at the mitochondrial contact site with clustered negatively charged lipid cardiolipin (64 -72). The negatively charged LPPG was chosen not only to provide the best working sample, but as a relevant mimic to the cardiolipin-rich mitochondrial membrane as well. Positively charged lysine residues (Lys 157 and Lys 158 ) in tBid helix ␣ 6 were suggested to be critical for a specific electrostatic interaction with the mitochondrial outer membrane (71,72). Our data confirmed this hypothesis by showing a strong micelle interaction for tBid residue lysine 157 as well as charged residues Arg 168 and His 172 in helix ␣ 7 (Fig. 4). Consistently, the electrostatic surface map of the tBid structure further showed continuous positively charged surfaces along helices ␣ 6 -8 (Fig. 5), which might form electrostatic interactions with negatively charged lipid cardiolipin headgroups on the membrane. Furthermore, this additional electrostatic interaction may account for our findings that both helix ␣ 6 and ␣ 7 are more membrane embedded than the rest of tBid helices (Figs. 2 and 4) and implicate a membrane recognition and docking role for these two helices. The recently reported fluorescence data showed tBid regions corresponding to Bid helix ␣ 4 -7 are membrane embedded and also suggested a conformational equilibrium between two major states of tBid (36). Our structure, which show all tBid helices are membrane associated, is consistent with their findings. Furthermore, to be in equilibrium of multiple conformations, the tBid structure must have a considerable degree of flexibility. Our structure and dynamic data support this possibility. We showed that tBid adopted an extended membrane-associated structure that contains flexible loops connecting helices ␣ 3 -␣ 4 , ␣ 4 -␣ 5 , and ␣ 5 -␣ 6 . In addition, we also showed that some residues in helix ␣ 6 (Leu 149 , Ala 156 , and Val 159 ) and at its N-terminal (Glu 145 ), the site of membrane targeting, undergo conformational exchange. This result seems to be supporting the conformational equilibrium observation by the fluorescence measurement.
A previous model proposed that the tBid BH 3 domain was fully exposed and located above the membrane surface, poised for protein interactions with Bax/Bak or Bcl-xL (35). However, our studies revealed that the important BH 3 domain containing helix ␣ 3 was membrane associated with hydrophobic residues facing toward the membrane surface (Figs. 4 and 5E). In contrast to the latest suggested solvent-exposed BH 3 domain (36), our data presented an energetically more favorable conformation for the tBid BH 3 domain that is membrane-associated due to its residue composition. Therefore, our structure supports a scenario that the interaction between the membrane-associated helical BH 3 domain and Bax happens on the membrane surface. The on the membrane interaction is consistent with previous observations that membrane-targeted tBid BH 3 peptides can still effectively interact with and activate Bax to permeabilize mitochondrial membranes (38,73). Previous studies also suggested that Bax helix ␣ 1 interacts with the tBid BH 3 domain as the first step in its tBid-induced activation (28,74). Upon contact with the membrane surface, Bax was suggested to experience a reversible conformational change and its N terminus can get exposed and recognized by the conformation-specific antibody 6A7 (75). The subsequent membrane insertion and activation of Bax was shown to involve the interaction with membrane-associated tBid (18). It is possible that the tBid BH 3 domain interacts with the transiently exposed Bax helix ␣ 1 on the membrane surface and this specific interaction can shift the reversible conformational change of membrane-encountered Bax into a membrane-associated state. In fact, a specific electrostatic interaction between tBid BH 3 domain residue Arg 84 and Bax helix ␣ 1 residue Asp 33 was reported to be important for tBid-induced Bax activation in a previous study (74). Therefore, our results of the tBid structure in LPPG micelles in combination with the above reported findings lead to the possibility that charged residues in the tBid BH 3 domain (Asp 81 , Arg 84 , Arg 88 , and Asp 95 ) that were shown to face away from the membrane could potentially form electrostatic interactions with charged residues (Asp 33 and Arg 34 ) in Bax helix ␣ 1 on the membrane surface. If the conventional BH 3 binding mode to Bax is required (19, 76 -78), then the interaction has to occur in such a way to allow the hydrophobic tBid BH 3 residues to be available to bind Bax. One possible scenario would be that the electrostatic interactions between the tBid BH 3 domain and Bax helix ␣ 1 could promote displacement of the tBid BH 3 residues from the lipid membrane, exposing them for binding to the Bax BH 3 binding pocket.
For Bcl-xL, it is difficult to discuss its interaction with tBid on the membrane with very limited information on membraneassociated Bcl-xL. However, it is likely that a similar interaction between tBid and Bcl-xL on the mitochondrial membrane was adopted as in the tBid-Bax interaction given the structural similarity between Bax and Bcl-xL.
Overall, Our data provided novel molecular information regarding the structure and dynamics of tBid in LPPG micelles, which leads to a first step in understanding the tBid apoptotic function on the mitochondrial membrane. Although our findings did not directly address the favorability of a direct or indirect activation model, the membrane-embedded tBid BH 3 domain suggested an on the membrane binding mode of tBid with Bax, thus leaning toward the previously proposed embedded together model (13). The recently solved x-ray structure of the Bax dimer promoted by the tBid BH 3 peptide and detergent provided valuable information on the tBid-induced Bax activation (79). The reported binding mode with the tBid BH 3 peptide sitting inside the hydrophobic BH 3 -binding cleft of Bax is essentially similar to inter-protein interactions among Bcl-2 family members in solution (19,76,77). Nevertheless, how the tBid BH 3 domain, shown in our study to be associated with the micelles, can go into the BH 3 binding pocket of Bax still remains unclear. Further elucidation of the tBid-Bax complex structure on the membrane will provide more molecular details on the mechanism of tBid-mediated Bax activation.