Intrinsic Order and Disorder in the Bcl-2 Member Harakiri: Insights into Its Proapoptotic Activity

Harakiri is a BH3-only member of the Bcl-2 family that localizes in membranes and induces cell death by binding to prosurvival Bcl-xL and Bcl-2. The cytosolic domain of Harakiri is largely disorder with residual α-helical conformation according to previous structural studies. As these helical structures could play an important role in Harakiri's function, we have used NMR and circular dichroism to fully characterize them at the residue-atomic level. In addition, we report structural studies on a peptide fragment spanning Harakiri's C-terminal hydrophobic sequence, which potentially operates as a transmembrane domain. We initially checked by enzyme immunoassays and NMR that peptides encompassing different lengths of the cytosolic domain are functional as they bind Bcl-xL and Bcl-2. The structural data in water indicate that the α-helical conformation is restricted to a 25-residue segment comprising the BH3 domain. However, structure calculation was precluded because of insufficient NMR restraints. To bypass this problem we used alcohol-water mixture to increase structure population and confirmed by NMR that the conformation in both milieus is equivalent. The resulting three-dimensional structure closely resembles that of peptides encompassing the BH3 domain of BH3-only members in complex with their prosurvival partners, suggesting that preformed structural elements in the disordered protein are central to binding. In contrast, the transmembrane domain forms in micelles a monomeric α-helix with a population close to 100%. Its three-dimensional structure here reported reveals features that explain its function as membrane anchor. Altogether these results are used to propose a tentative structural model of how Harakiri works.


Introduction
The form of programmed cell death, known as apoptosis, is essential for the correct development of multicellular organisms by inducing unwanted or damaged cells to commit suicide [1]. The dysfunction of apoptotic mechanisms is implicated in several pathologies including cancer, autoimmune and neurodegenerative disorders [2]. This cell death machinery is guarded by proteins of the Bcl-2 family, which promote and inhibit mitochondrial apoptosis [3,4]. The prosurvival members (Bcl-2, Bcl-x L , Bcl-w, Mcl-1, A1) contain up to four Bcl-2 homology domains (BH1-BH4) together with a C-terminal hydrophobic sequence that typically directs them to the outer mitochondrial membrane. Prosurvival proteins can be antagonized by heterodimerizing with proapoptotic Bcl-2 members [5][6][7][8] that share homology solely in the BH3 region and are therefore grouped in the so-called BH3only subfamily (Fig. 1A).
BH3-only proteins show significant structural diversity, including a single member (Bid) known to have a well-defined fold [9,10], whereas others (i.e. Bim, Bad and Bmf) have been shown to be intrinsically unstructured [11]. In addition, several members are predicted to contain a TM domain (Fig. 1A). These characteristics turn structural studies on BH3-only proteins daunting and therefore scarce, despite their important role in cell death. For instance, only one study has been reported up to date addressing the role of the TM domain of a BH3-only protein (BNip3) at the atomic level. In this work, the three-dimensional (3D) structure of the TM domain of BNip3 was determined revealing the formation of a dimer of a-helices in the presence of lipid bicelles, thus suggesting a role in membrane permeabilization [12]. Kindred studies are necessary to help establish whether this is a common role in BH3-only TM domains.
Other structural studies have centered at complexes between prosurvival partners and peptides encompassing the BH3 region of BH3-only members. In all known 3D structures of these complexes the BH3 region forms an a-helix that is inserted in a hydrophobic groove of the antiapoptotic member [13][14][15]. These studies have recently raised significant interest because the reported structures serve as templates for the design of BH3 analogues to be used as drugs for cancer treatment [16][17][18]. However, little is known on the conformational preferences at atomic resolution of BH3-only proteins in isolation. Research in this direction can provide important insights into the mechanism of binding, thus strengthening the grounds for rational drug design. Moreover, such studies will help improve our understanding on the operating mode of functional intrinsically unstructured proteins, a field that is emerging as a new frontier in the concept of protein structurefunction relationship.
Within the BH3-only subfamily we have focused our studies on the protein Harakiri (Hrk), which localizes in membranes in the cellular context and exerts proapoptotic activity by interacting with survival Bcl-x L and Bcl-2 [19]. The BH3 region in Hrk, that forms part of its cytosolic domain, has been shown to be key for these interactions and thus for the killing activity [19]. In addition, Hrk possesses a stretch of approximately 30 residues at the Cterminus predicted as a TM domain [19]. Infrared studies on a peptide comprising this region have shown that it populates the ahelix conformation in the presence of lipids [20]. Most likely because of its toxicity, Hrk has proven to be difficult to produce by recombinant methods in other labs [11,21] and ours. Thus, in our previous studies on the characterization of the interaction between Hrk and the Bcl-2 member Diva, the latter was produced recombinantly whereas we used for Hrk synthetic fragments encompassing the cytosolic domain [22]. These studies revealed that the cytosolic region can bind to Diva and that, at the structural level, is largely disordered with residual a-helical conformation at a population of ,13% [22]. Hrk is therefore a potential model example combining both intrinsic disorder and membrane binding capabilities.
Up to date high-resolution structural studies on BH3-only proteins in isolation are lacking except for Bid [9,10], the only known member with a defined fold. In an effort to improve our knowledge on the operating mode of BH3-only proteins we have characterized Hrk by NMR and circular dichroism (CD) to obtain atomic-detailed information on the conformational preferences of the cytosolic and TM domains. We have used synthetic constructs that encompass: 1) the entire cytosolic domain, 2) a smaller fragment including the BH3 motif and short flanking sequences at its N-and C-termini, 3) the TM domain. Firstly, we show using enzyme immunoassays and NMR that the cytosolic constructs are capable of binding Bcl-2 and Bcl-x L , the two proteins known to be implicated in Hrk's apoptotic function [19]. In addition, by combining previous structural data on the entire cytosolic domain [22] and new data on the smaller construct in water and wateralcohol mixture we find that a-helical propensity is confined in a 25-residue long fragment that comprises the 15 amino acid long BH3 motif. We have determined the 3D structure of the shorter cytosolic construct, revealing that it is strikingly similar to the structure of BH3-peptides bound to prosurvival proteins. This result suggests that intrinsic residual structure in disordered BH3only proteins could play a fundamental role in the binding mechanism, as preformed conformations involving the BH3 domain are likely implicated in the interaction with survival partners.
Furthermore, we found that the TM domain of Hrk is poorly soluble in aqueous solution albeit it readily inserts into micelles. This result agrees with the reported localization of Hrk in membranes of intracellular organelles [19], suggesting that membrane binding is an important determinant of Hrk's function. We also show that the TM domain is monomeric in the presence of micelles, which points to a role as membrane anchor, in contrast to the TM domain of BNip3 found to function as a membrane permeabilization device [12]. Finally, we report the 3D structure of Hrk's TM domain in micelles revealing that it forms an a-helix followed by a well-ordered turn. The length of the helix and turn is ,30Å , which is similar to the thickness of cellular membranes. The electrostatic surface of Hrk-TM structure is mainly apolar except for a small positively charged patch located at the Cterminal ordered turn, suggesting possible electrostatic interactions with the lipid heads of the membrane.

Results
Design of Hrk constructs and binding assays with Bcl-2 and Bcl-x L Hrk was studied using synthetic fragments that together encompass the full-length protein (Fig. 1B). Its TM domain, as often times observed in other proteins, likely operates independently of the rest of the amino acid sequence, which could explain the reported localization in cellular membranes [19]. On these grounds we designed three synthetic constructs to perform structural and binding studies that could provide insight into the function of intact Hrk (Fig. 1C). Construct Hrk-DTM spans the entire cytosolic domain and construct Hrk-22_53 is a smaller fragment encompassing only the BH3 region and two flanking sequences predicted to form a-helices by the program PredictProtein [23]. Finally, construct Hrk-TM comprises the TM domain.
Previous enzyme-linked immunosorbent assays (ELISA) and NMR binding data from our group report on the interaction between the cytosolic constructs of Hrk and the Bcl-2 member Diva [22]. These results indicate that the synthetic protein fragments are able to bind. However, the biological role of the Harakiri/Diva interaction is still to be studied. Thus, we analyzed the binding of Hrk-DTM and Hrk-22_53 to Bcl-2 and Bcl-x L as these interactions have been reported to induce apoptosis [19]. Both cytosolic fragments show significant levels of interaction with the survival proteins relative to the control ( Fig. 2A). These levels typically increase at higher peptide concentration as expected for real binding in ELISA. However, the binding levels of Hrk-22_53 for both Bcl-2 and Bcl-x L do not increase beyond 12 mM ( Fig. 2A). This result indicates that the two constructs show differences in binding. In fact, the entire cytosolic domain displays higher interacting levels than the shorter fragment, which suggests that the additional sequence of the longer construct can play a role in the interaction. Higher ELISA binding levels to Diva, confirmed by NMR, were also observed for the entire cytosolic domain relative to the shorter construct [22]. The ELISA data indicate that the cytosolic constructs of Hrk are able to bind prosurvival Bcl-2 members in the absence of the TM. Thus, both domains likely function independently.
The binding between Bcl-x L and Hrk was also studied by NMR. The 15 N-HSQC spectrum of unbound Bcl-x L changes significantly by adding Hrk-22_53 at equimolar amount indicating direct interaction (Fig. 2C). An analogous result is obtained by the addition of Hrk-DTM (Fig. 2D). Moreover, the titration of Bcl-x L with either Hrk-22_53 or Hrk-DTM leads to almost identical changes in the NMR spectra suggesting that the interaction is similar at the structural level. This behavior was also observed for the Diva/Hrk complexes mentioned above [22]. By increasing the concentration of the Hrk fragments in excess relative to Bcl-x L , no additional changes in the spectra were observed. This result indicates binding saturation at equimolar amounts of Bcl-x L and the Hrk constructs, and the formation of 1:1 Bcl-x L /Hrk-22_53 and Bcl-x L /Hrk-DTM complexes. Furthermore, two sets of NMR signals are present in the spectra of Bcl-x L /Hrk mixtures at 1:0.3 molar ratio. One set corresponds to the spectrum of free Bcl-x L and the second set to the complex, as inferred from signal overlap of the spectra at 1:0.3 molar ratio with both the spectra of free Bclx L and the one resulting from the equimolar mixture (Fig. 2C, D). This result indicates that exchange is in the slow regime relative to the chemical shift time scale. Binding saturation and slow exchange suggest high affinity constants for these complexes, in agreement with previous fluorescence data reporting a dissociation constant value of 92 nM for a complex between Bcl-x L and a short 20 residue-long peptide comprising the BH3 domain of Hrk [24]. Nevertheless, further studies are necessary to understand the structural, dynamics and energetic factors implicated in the Bclx L /Hrk interaction.
a-Helical structure in Hrk-DTM is localized in the region 29-53 encompassing the BH3 motif Previous structural data from our group show that the cytosolic domain of Hrk is largely disordered [22]. In particular, small dispersion of amide 1 H chemical shifts (Fig. S1) and NMR signals associated to methyl groups indicate that the protein is unfolded and lacks a hydrophobic core [22]. However, previous CD data on Hrk-DTM indicate the formation of approximately 13% population of a-helical structure. This population increases to 35% upon TFE addition at 35% (v/v) [22], a solvent known to promote secondary structure formation in protein fragments and peptides with intrinsic conformational propensity [25,26].
To further characterize structurally Hrk-DTM by NMR is necessary to reduce signal overlap originating from both the small chemical shift dispersion resulting from the low structure population and the numerous signals of the rather long protein fragment. By augmenting the helical content is possible to increase chemical shift dispersion, as NMR signals are weight-averaged according to the population of the different conformations when they interconvert fast on the NMR experimental time. Unfortunately, NMR chemical shift dispersion in the presence of TFE is still small resulting in crowded spectra with severe signal overlap (Fig. S1), thus precluding further structural studies. Hrk-22_53 is a smaller fragment that will likely result in less complex spectra facilitating NMR analysis. However, it is important to check whether the helical propensity in Hrk-DTM is mainly located in the protein region flanking the BH3 motif (construct Hrk-22_53. Fig. 1B, C).
The CD spectrum of Hrk-22_53 indicates a significant degree of disorder albeit with approximately 17% population of a-helical structure (Fig. 3A). This value is similar to that shown by Hrk-DTM [22], thus, taking into account the difference in peptide length, a simple calculation indicates that the helix spans a common region in both peptides. In other words, most helical population of Hrk-DTM is located in residues 22-53. A similar conclusion is derived by comparing the helical population of Hrk-22_53 and Hrk-DTM in the presence of TFE; 60% and 35% population respectively. However, the CD data indicate that both in aqueous milieu and in TFE, the helical content of Hrk-DTM is slightly larger than for the shorter fragment considering the difference in sequence length. Hrk-22_53 results in less crowded NMR spectra ( Fig. S2A-D, Fig. 4), thus allowing chemical shift assignment. As protein 13 C a and 1 H a chemical shifts are particularly sensitive to secondary structure, their deviation from values tabulated for unstructured peptides (random coil values) is routinely used to identify extended and helical conformations [27]. Chemical shifts differences relative to random coil values indicate that Hrk-22_53 adopts in water ahelical structure from residues 29 to 53 ( Fig. 5A and Fig. S3 for 13 C a and 1 H a deviations, respectively). This result contrasts with the prediction program [23], which suggests that the helix extends to residue 22 (Fig. 1B). The a-helix seems to bend around G41 as in this region the chemical shifts are closer to the random coil values ( Fig. 5A and Fig. S3). Together with 13 C a and 1 H a chemical shifts, NOE data connecting local nuclei are also solid proof of the formation of protein secondary structure [28]. For ahelices the following NOEs are expected: [28]. To obtain more detailed information on the structure formed by Hrk-22_53 in water NOE NMR experiments were recorded. For instance, H a -H b (i, i+3) NOEs connecting Ala 39 to Asp 42, Asp 42 to His 45, and His 45 to Thr 48, indicate the formation of helical structure (Fig. 4A). Other NOE cross-peaks between protons H a -H N (i, i+2) and (i, i+3) correlating residues Asp 42-Leu 44, Gly 41-Glu 43, Leu 40-Glu 43, Gly 41-Leu 44 (Fig. 4B) indicate a-helix formation centered at G41-D42, despite being the chemical shifts closer to random coil values than in other regions. In addition, H N -H N (i, i+1) NOEs are observed from residues Thr 33 to Arg 52 (Fig. S2E). Altogether, the NOE data indicate that Hrk-22_53 adopts helical structure in water from residues Thr 33 to Arg 53 (Fig. 4A,B, Fig. S2E), in very good agreement with the information obtained from chemical shifts ( Fig. 5A and Fig. S3).
The chemical shift and NOE data unambiguously show that the cytosolic domain of Hrk forms an a-helix in water in the region 29-53. However, because of the low structure population and because some NOEs cannot be observed due to chemical shift overlap, the number of NOE connectivities is insufficient to reach an ensemble of conformers representing the helix when included in the structure calculation protocol (data not shown), resulting in ,3Å r.m.s.d for the 20-conformer ensemble. In contrast, by TFE addition the 1 H a and 13 C a chemical shifts of Hrk-22_53 show larger deviations from random coil values as expected for a larger structure population (Fig. 5A, Fig. S3). The profiles of 1 H a and 13 C a chemical shift deviations with and without TFE are equivalent, which indicates that TFE increases the population in the region with intrinsic propensity without inducing significant helix formation in other segments. In addition, NOE peaks observed in water and in TFE that are characteristic of the a-helix are analogous (Fig. 4). However, these peaks are more intense and abundant in water/alcohol mixture as a result of the larger population (Fig. 4C, D). The chemical shift and NOE comparison in water and TFE indicates that the structure is equivalent in both media albeit with different population. The larger population in TFE led to more NMR restraints resulting in a single ensemble of conformers. The structural details are explained below.
The 3D structure of unbound Hrk-22_53 in TFE closely resembles BH3-peptide structures complexed to prosurvival Bcl-2 members Structural information at atomic resolution on the unbound form of BH3-peptides or intrinsically disordered BH3-only proteins has not been reported to date. However, this information is important to better understand the determinants of BH3-only protein binding. The comparison of the structure of Hrk-22_53 to the known structures of BH3-peptides complexed with their prosurvival partners could provide insights into this subject. To this aim we have calculated the solution structure of Hrk-22_53. Structural statistics of the resulting ensemble of the 20 conformers with lowest energy are shown in Table S1. This ensemble forms an a-helix spanning residues 29-51 with significantly well defined backbone (r.m.s.d = 0.5Å ) and side chains (Fig. 6A). In contrast, the absence of a single conformation for residues 22-28 indicates that this region is disordered and residues 52-53 also appear to be more flexible (Fig. 6A). One side along the longitudinal axis of the a-helix is mainly hydrophobic, albeit surrounded by a significant number of positively (Arg, Lys) and negatively charged residues (Asp, Glu) forming an amphipathic helix (Fig. 6B). The apolar face of the helix most likely will interact with the hydrophobic groove of the antiapoptotic partner upon complex formation, leaving the polar side exposed to the solvent.
The structure of unbound Hrk-22_53 is compared to two reported BH3-peptide/prosurvival complex structures (Mcl-1/ Bid-BH3 [15] and Bcl-x L /Bim-BH3 [29]) in Fig. 6C-F. Although the length of the different helices formed by the BH3-peptides varies, the structure of Hrk-22_53 can be superimposed to these helices using as reference two highly conserved residues in the BH3 domain of the three proteins (Gly 41 and Asp 42 in Hrk) (Fig. 6C, D). According to the reported 3D structures of the survival/BH3-peptide complexes, 4 conserved hydrophobic residues in the BH3 domain of the BH3-peptides form part of the interface and help to anchor the peptide to the survival protein  hydrophobic groove [13]. The superposition of the structures in Fig. 6E, F shows that the side chain position of these conserved residues (Thr 33, Leu 37, Leu 40 and Leu 44 in Hrk) forming part of the interface is equivalent in Hrk-22_53 and the bound forms of Bid-BH3 and Bim-BH3.
The structures of unbound Hrk and the complexed peptides are significantly similar. This result suggests that the preformed helical conformation can play an important role in the binding process. In fact, previous studies on peptides derived from the BH3 domain of the BH3-only protein BAD have shown that affinity towards Bclx L increases by augmenting the a-helical content through mutations [14]. It is currently unknown whether the a-helical population in isolation might increase by folding upon binding once the partner is present, or only the preformed helices are able to bind shifting in turn the folding equilibrium towards the structured form. At this point it is worth to recall the higher binding levels of the long Hrk construct to Diva [22], Bcl-2 and Bcl-x L . However, the NMR and ELISA studies on the Diva/ Harakiri interaction suggest that the interacting surface in Diva is structurally similar when complexed to both Hrk-DTM and Hrk-22_53 [22]. These results indicate that the additional sequence of the longer construct is influencing binding albeit without modifying the interacting region. The unstructured sequences flanking the helix could enhance binding by forming transient contacts with the prosurvival protein or by augmenting the helical propensity of the binding region in Hrk. In fact, the entire cytosolic domain shows larger a-helical population according to the CD data (vide supra).
The factors involved in the binding of intrinsically disordered proteins to their partners are still poorly understood. Recent findings indicate that this mechanism is highly complex and can vary greatly depending on the intrinsically disordered protein studied [30]. For instance, several or all of the following; preformed elements of secondary structure, tertiary contacts between the interacting partners and the dynamics of the bound and unbound forms, can be key players in the binding process [30][31][32][33]. In this context the behavior of Hrk is different from other intrinsically disordered proteins that do not adopt preformed structural elements [34], or when these are present they do not match the final structure adopted upon complex formation [30].

3D structure of Hrk-TM in micelles
In an effort to characterize the solution structure of Hrk-TM at atomic resolution we have tested different conditions that could render both high-quality NMR spectra and emulate the membrane environment. Hrk-TM is barely soluble in aqueous solution as expected from the hydrophobic nature of its amino acid sequence (Fig. 1C), thus precluding NMR studies in water. However, it readily dissolves in the presence of micelles formed by SDS and DPC detergents. In both micellar systems NMR signals are significantly broad complicating NMR analysis. However, spectra quality is slightly better in SDS. It has been reported that the addition of alcohol to detergent micelles increases micelle flexibility, as alcohol molecules apparently interact with the detergent hydrophobic chains affecting their mobility [35,36]. Higher micelle flexibility can result in narrower NMR signal linewidth and therefore Hrk-TM was dissolved in SDS micelles mixed with the simplest alcohol, methanol. Under these conditions the resulting NMR spectra were suitable for analysis. However, SDS and DPC are detergents with significantly different chemical properties, thus CD experiments were performed first to test whether the conformational behavior of Hrk-TM varies depending on the micellar system used. The CD data indicate that Hrk-TM adopts similar structure in both milieus, corresponding to ,60% population of a-helix in SDS/methanol and ,55% in DPC/methanol micelles (Fig. 3B).
Once checked that Hrk-TM is able to form a large population of a-helical structure in micelles, the next step was to investigate the oligomerization state in this medium. It is well known that TM a-helices migrate in SDS-polyacrilamide gel electrophoresis (SDS-PAGE) according to their native-like oligomeric sate [37,38]. Therefore, the migration of Hrk-TM in SDS-PAGE was compared to other proteins of known molecular weight (Fig. 7).
Our results indicate that Hrk-TM migrates in SDS-PAGE according to its molecular weight (,3 kDa), thus faster than the two proteins (Lysozyme and gpW) used in the experiment with  NMR can be used to get information on whether TM domains get inserted in the micellar environment. For instance, proteins in aqueous solution typically show in [ 1 H-1 H]-NOESY spectra crosspeaks at amide proton and water chemical shifts, which originate from chemical exchange between them (Fig. S4). The absence of these cross-peaks indicates that amide protons are protected from the solvent, for example by the surrounding detergent molecules [37]. Only Asn 90 at the C-terminus of Hrk-TM shows a NOE cross-peak between its amide proton and water in the NOESY spectra of Hrk-TM in micelles (Fig. S4). This result indicates that Hrk's TM domain penetrates the micelle.
The 1 H a and 13 C a chemical shifts of Hrk-TM in the micellar environment consistently indicate the formation of helical structure spanning residues Trp 69 to Leu 91 (Fig. 5B). However, no regular secondary structure is observed from residues Ala 61 to Tyr 68. The population of a-helix in solution can be estimated from the magnitude of C a chemical shifts relative to tabulated random coil values [39]. Excluding the 8 N-terminal residues, which do not adopt regular secondary structure, the a-helical population estimated from NMR data is 95%. However, when all 31 residues are included the population drops to 61%, which is in very good agreement with the population calculated from the CD data that, in contrast to NMR, can only report information averaged for all residues.
The NMR structure of Hrk-TM (structural statistics in Table  S2) shows that the a-helix actually spans residues Trp 69 to Leu 85 and is followed by a well-ordered turn from residues Leu 86 to Leu 91 (Fig. 8A). The backbone and in particular the side chains of aromatic residues are very well defined ( Fig. 8A and Table S2). The helix is slightly curved showing in its concave part bulky hydrophobic residues such as Leu and Val, and in the convex part less hydrophobic or bulky amino acids (Ala, Gln or Cys) (Fig. 8A). The length of the helix including the well-ordered turn is 31Å , which is very close to the thickness of the hydrophobic inner part of a lipid bilayer. This result suggests that Hrk-TM could fully span the bilayer leaving outside the N-terminal sequence that does not form part of the helix. Aromatic residues, specifically Trp, are particularly abundant in protein TM domains and are not evenly distributed throughout the domain, rather are concentrated at the edges [40]. The role of these hydrophobic and bulky residues is believed to prevent the TM helix from slipping across the membrane. Hrk-TM contains 3 Trp and 1 Tyr. To illustrate the position of these residues in the membrane context the structure of Hrk-TM is placed on a model of a lipid bilayer on Fig. 8B. As these studies were done in micelles, the precise angle that the long axis of the TM helix forms with the normal of the lipid bilayer is not known. However, NMR data on amide 1 H protection from the solvent suggest that the TM domain would be fully inserted in the membrane. Additionally, the two Arg residues at the Cterminus (Fig. 8B) are unlikely buried in the hydrophobic part of the bilayer, as this would require a high energetic cost. These residues form a positively charged patch in the otherwise apolar electrostatic surface of Hrk-TM that could potentially form electrostatic interactions with the polar heads of the bilayer. These results suggest that Hrk-TM likely adopts a position close to parallel to the bilayer normal as shown in Fig. 8B, C. Summarizing, the monomeric state of Hrk-TM and its structural characteristics suggest that it functions as a membrane-anchoring device.

Discussion
This study reveals high-resolution structural characteristics of both the cytosolic and the TM domain of the Bcl-2 protein Hrk. The cytosolic domain is largely unstructured in solution [22], which indicates that Hrk shares features commonly found in intrinsically disordered proteins. Three other members of the BH3-only subfamily have been shown to be unstructured [11]. Thus, intrinsic disorder most likely plays an important role in the function of these proteins. However, no structural characterization at atomic resolution of a natively unfolded Bcl-2 member has been reported up to date. In the case of Hrk, our study shows that out of the 59 residues in the cytosolic domain only a stretch of 25 (residues 29-53), including the BH3 region, significantly populates the a-helical conformation. This helical structure very closely resembles the a-helix formed by BH3-peptides in complex with survival partners, which suggests that preformed elements of secondary structure are major players in the binding mechanism. This result agrees with recent structural studies on Protein Phophatase 1 regulators indicating that the transient a-helix conformation likely plays a key role in binding [41]. However, other intrinsically disordered proteins that fold upon binding do not adopt residual secondary structures in isolation [34].
In addition, the low solubility of Hrk-TM in aqueous milieu and its capability of forming ,100% populated helical structure in micelles agree with previous reports indicating the localization of Hrk in intracellular organelle membranes [19]. It is worth noting that Hrk is predicted to have an N-myristoylation site at Gly 63 (at the N-terminus of Hrk-TM) according to prediction programs [23]. Myristoylation is known to play a vital role in membrane targeting. Furthermore, the monomeric state of Hrk-TM in micelles and its 3D structure suggest that it likely functions as a membrane anchor. Up to date high-resolution structural information on the TM domain of BH3-only proteins besides Hrk-TM is known only for BNip3, which has been suggested to function as a membrane permeabilization device [12]. The disturbance of the mitochondrial membrane is a recognized mechanism to trigger cell death, therefore the different behavior of Hrk and BNip3, together with the structural heterogeneity found in the BH3-only subfamily point to a wide variety of mechanisms to prompt apoptosis in this subfamily.
Finally, ELISA and NMR data indicate that the cytosolic domain of Hrk can bind Bcl-2 and Bcl-x L independently of the TM domain, which according to our structural results functions as an anchoring device. Based on these considerations and combining the results on both domains, we propose a tentative structural model to help explain how Hrk works. According to this model, Hrk is anchored to the mitochondrial membrane through its TM helix exposing the disordered N-terminus to the cytosol (Fig. 9). This domain folds into an a-helix from residues 29-53 upon binding to the prosurvival partner that could be already anchored to the membrane or in the cytosol. The intrinsic flexibility of the disordered domain likely facilitates binding by increasing the capture radius for prosurvival Bcl-2 members. The advantageous effect of flexibility in the binding of intrinsically disordered proteins has been theoretically predicted before as a ''fly-casting'' mechanism [31], and suggested for multi-domain proteins with long semi-flexible linkers and protein-binding function based on structural and dynamics studies [42]. This tentative mode of operation for Harakiri agrees with the recently proposed ''membrane-embedded-together'' model for apoptotic mechanisms [43], which emphasizes that many interactions within the Bcl-2 family occur only in the membrane.

Bcl-x L cloning, expression and purification
A human Bcl-x L construct comprising residues 1-209 (including the flexible ,50 residue-long loop) was cloned by TopGen Technologies (Canada) into a pET21a vector (Novagen) between the NdeI and NotI restrictions sites, thus without the histidine tag. The protein was expressed in C41(DE3) E. Coli strain during 3 hours at 37uC by the addition of isopropyl b-D-thiogalactopyranoside (1 mM concentration) at an OD 600 of 0.6-0.7. Uniformly 15 N-labeled Bcl-x L was produced using 15 NH 4 Cl (Cambridge Isotopes) as sole nitrogen source. The cells were harvested by Figure 9. Tentative structural model of Harakiri's operating mode. The survival protein (Bcl-2 or Bcl-x L ) [13,14] and Hrk are represented as red and green ribbons, respectively. The arrow connects the states before (Hrk's cytosolic domain is shown largely disordered with preformed secondary structure) and after (Hrk's cytosolic domain is forming a helix) the interaction between Hrk and the survival partner. The hydrophobic and polar parts of the bilayer are colored in light and dark yellow, respectively. doi:10.1371/journal.pone.0021413.g009 centrifugation and resuspended in a buffer containing 10 mM Tris_HCl at pH 7.0, 0.1 mM protease inhibitor cocktail (Sigma) and 1 mM TCEP (Tris(2-carboxyethyl)phosphine). Cells were lysed by sonication at 4uC and centrifuged at 25,000 rpm for 1 hour. The soluble protein was purified by anion exchange chromatography using a HiTrap Q sepharose Fast Flow column (GE Healthcare), followed by reverse phase chromatography (C8 column) in water-acetonitrile mixtures and lyophilization of the protein solution. Mass spectrometry was used to confirm the 15 Nlabeled Bcl-x L sample (residues 1-209).

Enzyme-linked immunosorbent assays (ELISA)
Microplates (Costar Ltd., US) were coated with 50 ml of Bcl-2 or Bcl-x L both lacking the TM domain (purchased to R&D Systems, Inc., US) at a concentration of 7.5 mg/ml in PBS and incubated overnight at 4uC. Plates were then washed three times with distilled water and blocked with 2.5% BSA (Sigma) in PBS for 2 hours at 37uC. Stock solutions of Hrk constructs (Hrk-DTM and Hrk-22_53) were prepared at fixed initial peptide concentrations derived from absorbance measurements at 280 nm. The final peptide solutions used in the binding studies contained PBS, 2.5% BSA and 0.1% Tween to avoid aggregation and non-specific binding. Hrk samples were loaded and incubated overnight at 4uC. BSA at 2.5% in PBS was used as negative control. After thorough wash with distilled water and 0.1% Tween, a polyclonal rabbit anti-Harakiri BH3 domain antiserum was added (1:1000) (Abcam, UK). Following a streptavidin peroxidase conjugate antirabbit Ig (1:1000) (Dako) as secondary antibody, the peroxidase activity was detected by the addition of 3,39,5,59-tetramethylbenzidine dihydrochloride peroxidase substrate (Sigma). Colour development was then allowed to proceed for 15 minutes at room temperature and stopped adding 0.1 M H 2 SO 4 . The optical density was read at 450 nm on an ELISA plate reader (Labsystems Multiskan BICHROMATIC). All assays were done in duplicate. The absorbance of the control at 450 nm (,0.1) is within typical background values in ELISA experiments.

NMR spectroscopy
Hrk-DTM and Hrk-22_53 NMR samples were prepared at ,1.0-1.5 mM peptide, 20 mM phosphate buffer, 5 mM d 16 13 C natural abundance in a cold probe to alleviate signal overlap. Experiments were processed with NMRPipe [47] and analyzed with PIPP [48] and Sparky [49].

Structure calculation
Peak intensities from NOESY experiments on Hrk-22_53 with and without TFE were translated into a continuous distribution of interproton distances. A summation averaging ((gr 26 ) 21/6 ) was used to represent distances involving methyl groups, aromatic ring protons and non-stereospecifically assigned methylene protons [50]. Errors of 35% of the distances were applied to obtain lower and upper limits. Distance restraints for a total of 17 hydrogen bonds (r NH-O = 1.9-2.5Å , r N-O = 2.8-3.4Å ) were defined according to the experimentally determined secondary structure of the protein only for Hrk-22_53 in TFE. The TALOS program [51] was used to obtain 38 and 18 dihedral restraints for Hrk-22_53 with and without TFE, respectively, for those residues with statistically significant predictions. Peak-intensities from NOESY experiment on Hrk-TM in micelles were classified in 4 categories: strong (2.2Å , 20.3, +0.7), medium (3.0Å , 20.9, +0.9), weak (3.5Å , 21.1, +1.1) and very weak (4Å , 21.0, +1.0). Distance restrains corresponding to 12 hydrogen bonds derived from the secondary structure were used. w and y dihedral angle restraints for 16 residues were derived from statistically significant TALOS [51] predictions. Structures were calculated with the program X-PLOR-NIH 2.16.0 [52]. The starting structure was heated to 3,000 K and cooled in 20,000 steps of 0.002 ps during simulated annealing. The final ensemble of 20 NMR structures was selected based on lowest energy and no restraint-violation criteria. The 20 lowest-energy conformers have no distance restraint violations and no dihedral angle violations greater than 0.3Å and 3u, respectively for Hrk-22_53 in TFE, 0.4Å and 5.5u, respectively for Hrk-22_53 without TFE, 0.5Å and 5u, respectively, for and Hrk-TM in micelles. Structures were analyzed with MOLMOL [53] and their quality was assessed with PROCHECK-NMR [54]. Coordinates for Hrk-22_53 in TFE and Hrk-TM in SDS/methanol micelles were deposited in the Protein Data Bank with accession codes 2l58 and 2l5b, respectively.

Circular dichroism
Hrk-22_53 was dissolved in 20 mM sodium phosphate buffer at pH 5.8 in the presence and absence of 35% (v/v) trifluoroethanol (TFE). The concentration of Hrk-22_53 was 36 mM in the buffer without TFE and 37 mM in TFE, respectively. Hrk-TM was dissolved in 200 mM SDS, 35% (v/v) methanol, pH 3.2 and 200 mM DPC, 35% (v/v) methanol, pH 3.2 at 50 mM and 54 mM concentration, respectively. Concentration values were calculated by measuring absorbance at 280 nm. Circular dichroism measurements were acquired at room temperature using a JASCO model J-815 spectropolarimeter with a 1 mm cuvette. The CD signal at 222 nm was converted to mean residue ellipticity ([h] obs ) after subtracting the blank using the equation: where c is the peptide concentration (in mM), N is the number of peptide residues and l is the path-length (in cm).
The percentage of a-helical population was determined using the following equation: where [h] helix is the mean residue ellipticity of a complete helix, i.e., -42,500(1-(3/N)), and [h] coil is the ellipticity of a random coil, i.e. +640 [55,56].