Phylogenomics of Life-Or-Death Switches in Multicellular Animals: Bcl-2, BH3-Only, and BNip Families of Apoptotic Regulators

http://www.100md.com 分子生物学进展 2005年第12期

     * Laboratoire de Biométrie et Biologie Evolutive, Université Claude Bernard Lyon 1, 69622 Villeurbanne Cedex, France; and Laboratoire de Biologie Moléculaire de la Cellule, IFR 128 BioSciences Lyon-Gerland, Lyon Cedex, France

    E-mail: aouacher@biomserv.univ-lyon1.fr.

    Abstract

    In this report, we conducted a comprehensive survey of Bcl-2 family members, a divergent group of proteins that regulate programmed cell death by an evolutionarily conserved mechanism. Using comparative sequence analysis, we found novel sequences in mammals, nonmammalian vertebrates, and in a number of invertebrates. We then asked what conclusions could be drawn from phyletic distribution, intron/exon structures, sequence/structure relationships, and phylogenetic analyses within the updated Bcl-2 family. First, multidomain members having a sequence pattern consistent with the conservation of the Bcl-XL/Bax/Bid topology appear to be restricted to multicellular animals and may share a common ancestry. Next, BNip proteins, which were originally identified based on their ability to bind to E1B 19K/Bcl-2 proteins, form three independent monophyletic branches with different evolutionary history. Lastly, a set of Bcl-2 homology 3–only proteins with unrelated secondary structures seems to have evolved after the origin of Metazoa and exhibits diverse expansion after speciation during vertebrate evolution.

    Key Words: comparative genomics ? apoptosis ? Bcl-2 family ? BNips ? phylogeny

    Introduction

    Apoptosis is crucial for the proper development and functioning of multicellular organisms (Cory and Adams 2002; Opferman and Korsmeyer 2003). In vertebrates, this program of cell suicide functions in sculpturing organs, deleting unwanted structures, adjusting cell number, and eliminating harmful, abnormal, or misplaced cells. Abnormalities in cell death control can lead to a variety of disorders from degenerative diseases to cancer (Green and Evan 2002; Kirkin, Joos, and Zornig 2004).

    Pioneering genetic and biochemical studies in invertebrate models and mammals have led to the identification of many apoptotic players from diverse organisms and have shown that the cell death program has been conserved throughout evolution. For instance, the core apoptotic machinery of the nematode Caenorhabditis elegans consists of four proteins: three proapoptotic proteins, namely, CED-3, CED-4, and EGL-1, and a unique antiapoptotic molecule, CED-9 (Igaki and Miura 2004). Mammalian homologues for all these C. elegans molecules have been discovered. CED-3 was found to be related to the caspase family of death proteases, whereas the Apaf-1 adaptor molecule turned to be a mammalian CED-4 homologue. Structural and functional relatives of CED-9 and EGL-1 belong to the Bcl-2 family of apoptotic modulators, whose activity has been implicated since these initial reports in many physiological and pathological conditions.

    Bcl-2 family members are central regulators of programmed cell death because they integrate diverse survival and death signals generated outside and inside the cell and act like checkpoints through which these signals must pass before they determine the cell fate (Adams and Cory 2001). These proteins constitute an expanding and heterogeneous family that has been initially divided by function into either pro- or antiapoptotic members. Thus, antiapoptotic proteins such as Bcl-2 or Bcl-XL tend to prevent the release of apoptogenic molecules from mitochondria (e.g., cytochrome c) and subsequent caspase activation, whereas proapoptotic proteins such as Bax or Bak promote these deleterious events. Mutual interactions between these death inhibitors and promoters and their relative ratio are thought to arbitrate the life-or-death decision. On the other hand, the Bcl-2 family has been divided into three classes based on the conservation of typical motifs known as Bcl-2 homology (BH) domains (designated BH1, BH2, BH3, and BH4). All members possess at least one of the four BH domains, which roughly correspond to -helical segments on the proteins. Many Bcl-2 family proteins also contain a putative transmembrane (TM) domain at their C-termini, and at least to some members such as Bcl-2, this hydrophobic tail is critical for both subcellular localization and activity. Outside these regions, these proteins display considerable sequence diversity. In general, antiapoptotic multidomain members (the Bcl-2–like survival factors) share sequence similarity in all four conserved BH domains (BH1–4), whereas proapoptotic multidomain members (the Bax-like death factors) have similar BH1–3 domains. The third subset is formed by death proteins collectively known as the "BH3-only" subfamily (e.g., Bid, Bim, and Egl-1) that display sequence similarity only in the BH3 domain. The BNip proteins represent a rather marginal subgroup, first discovered as adenoviral E1B 19K/Bcl-2 interacting proteins and then classified into this subfamily due to some sequence similarity with the BH3 and TM motifs.

    Mechanistically, it is believed that BH3-only proteins act as sensors for distinct apoptotic pathways, whereas multidomain Bax subfamily members may constitute executioners of death signals downstream of the BH3-only proteins. Antiapoptotic Bcl-2–like family members appear to function, at least in part, by interacting with proapoptotic family members. Although the precise molecular mechanism of apoptosis regulation by Bcl-2 family proteins is still a matter of intense debate (Borner 2003; Cory, Huang, and Adams 2003; Coultas and Strasser 2003; Kuwana and Newmeyer 2003; Reed 2003), it is believed that the amphipathic -helix formed by the BH3 region of BH3-only proteins or proapoptotic multidomain members binds through its hydrophobic surface to an elongated hydrophobic groove formed on the surface of prosurvival Bcl-2 proteins by the combination of their BH1, BH2, and BH3 regions (Sattler et al. 1997). Noticeably, among the BH3-only proteins, Bid has a unique role in apoptosis signaling because its cleavage by caspase-8 links the death receptor pathway (the so-called "extrinsic" pathway of cell death) to the mitochondrial signaling pathway (or "intrinsic" pathway) regulated by the other members of the Bcl-2 family (Chou et al. 1999; McDonnell et al. 1999).

    With the recent completion of whole-genome sequencing efforts and progress in bioinformatics, more than 25 members of the bcl-2 gene family have been identified in humans (Koonin and Aravind 2002; Reed, Doctor, and Godzik 2004). Although understanding of activity and function has grown rapidly for several Bcl-2 family proteins (e.g., Bcl-2, Bcl-XL, Bid, and Bad), most members of this family have received only initial characterization. A straightforward way to enhance the understanding of one particular Bcl-2–related protein would be to use the information gained from the study of related members. However, optimal use of this information requires a detailed knowledge of the evolutionary relationships within the Bcl-2 family and its subgroups. We and others have previously included phylogenetic analyses in other studies concerning the Bcl-2 family (Aouacheria et al. 2001; Wiens, Diehl-Seifert, and Muller 2001; Koonin and Aravind 2002; Lanave, Santamaria, and Saccone 2004), but limited sequence diversity and recurrent omission of the BH3/BNip subgroups have restricted the usefulness of these reports. In this study, we undertook a search of all recognizable Bcl-2 family relatives in a large set of publicly available databases, including genome and expressed sequence tag (EST) databases. We present and compare by phylogenetic analysis the Bcl-2 family members found in humans, rodents, birds, fishes, arthropods, and worms, together with other counterparts discovered in simple multicellular animals. In order to determine the extents of family expansion in different evolutionary lineages, we defined and constructed a phylogeny for five subsets of Bcl-2 family members, namely, those sharing a same "pore-forming" helical bundle fold, BH3-only members, and BNip1–3 proteins. We also examined the domain organization of the multidomain Bcl-2 family members in an attempt to discuss the timing of their establishment.

    Materials and Methods

    Database Searching

    Ensembl and GenBank nucleotide and protein sequence databases and EST databases located at the PBIL (P?le Bioinformatique Lyonnais) were scanned for proteins related to known Bcl-2, BH3-only, and BNip proteins using either Psi-Blast (protein databases) or TBlastN (nucleotide databases) (Altschul et al. 1997). Fugu genome searches were performed using TBlastN and three Fugu genome databases. The first database is located at the Sanger Institute (http://www.ensembl.org/Fugu_rubripes/index.html), the second is located at the Institute for System Biology (http://genome.jgi-psf.org/fugu6/fugu6.home.html), and the third corresponds to the Fugu Genomics Project of the MRC-HGMP Resource Center (http://fugu.biology.qmul.ac.uk). Tetraodon genome searches were performed at the Genoscope by using Exofish (exon finding by sequence homology) (http://www.genoscope.cns.fr/cgi-bin/exofish.cgi) and with the MRC-HGMP blast server. Zebrafish genome searches were performed using the Ensembl Blast server (http://www.ensembl.org/Danio_rerio/index.html) and Exofish. After initial sequence collection, the full set of databases was probed with individual sequences to enhance the chances of finding sequences related to particular divergent proteins and better define the range of organisms containing these proteins. Although BH3 domains were reported in BRCC2, BRAG-1, BmP109, p193, APE1820, ITM2B(s), MAP-1, Rad9, Spike, SphK2, and SP1610 proteins, further scrutiny revealed these sequences to have been incorrectly classified (see Naohiro Inohara's list at http://www-personal.umich.edu/ino/List/2342.htm).

    Alignment and Tree Building

    Sequences were first aligned using ClustalW (at the PBIL) with BLOSUM alignment matrices, except multidomain members which were aligned using PRALINE with a PAM250 matrix (http://zeus.cs.vu.nl/programs/pralinewww/). The resulting initial alignments were optimized by adjusting the alignment parameters (gap penalties). Multiple alignments were subsequently edited and manually improved using the Seaview alignment editor (Galtier, Gouy, and Gautier 1996) when ClustalW failed. Phylogenetic analyses were performed on the conserved core of these final alignments. Amino acid sites where gaps exist in the alignments were excluded from the calculation. Ungapped alignments were computed to derive trees according to neighbor-joining and maximum parsimony methods using the Phylo_win program (Galtier, Gouy, and Gautier 1996). Bootstrap analysis (1,000 replicates) provided a measure of confidence for the detected relationships. The multiple alignments were also used for maximum-likelihood (ML) analyses using Phyml v2.4.4 (Guindon and Gascuel 2003) using a Jones-Taylor-Thornton model of sequence evolution and with the alpha parameter of the gamma distribution estimated from the data. Bootstrap support was based on 100 replicates using the programs SEQBOOT and CONSENSE (with majority-rule method) of the PHYLIP package (Felsenstein 1996) to generate data replicates and consensus tree, respectively. The log-likelihood values for the ML trees are: –17,117 (multidomain members), –1,906 (BH3-only), –3,952 (Bnip1), –8,875 (Bnip2), and –5,213 (Bnip3). Trees were displayed using the TreeView program (Page 1996), converted to image files, and then annotated using Adobe Photoshop. Alignments are available upon request.

    Species Abbreviations

    Aca, Ancylostoma caninum; Ace, Ancylostoma ceylanicum; Ag, Anopheles gambiae; Ap, Apis mellifera; At, Ambystoma tigrinum; Bm, Bombyx mori; Cb, Caenorhabditis briggsae; Ce, Caenorhabditis elegans; Ci, Ciona intestinalis; Cv, Crassostrea virginica; Di, Dirofilaria immitis; Dm, Drosophila melanogaster; Dr, Danio rerio; Gc, Geodia cydonium; Gg, Gallus gallus; Gr, Globodera rostochiensis; Hc, Haemonchus contortus; Hg, Heterodera glycines; Hm, Hydra magnipapillata; Hr, Halocynthia roretzi; Lr, Lumbricus rubellus; Mh, Meloidogyne hapla; Mi, Meloidogyne incognita; Mj, Meloidogyne javanica; Mp, Meloidogyne paranaensis; Ppa, Pristionchus pacificus; Ppe, Pratylenchus penetrans; Pt, Parastrongyloides trichosuri; Sd, Suberites domuncula; Sj, Schistosoma japonicum; Sp, Strongylocentrotus purpuratus; Ss, Strongyloides stercoralis; Tn, Tetraodon nigroviridis; Tr, Takifugu rubripes; Ts, Trichinella spiralis; Tv, Trichuris vulpis; Xl, Xenopus laevis; and Xt, Xenopus tropicalis.

    Results and Discussion

    Isolation of Bcl-2 Family Sequences from Expression and Genome Databases

    To isolate Bcl-2 family genes, we performed Blast searches on a number of genome and expression databases as described in Materials and Methods, using previously known human, mouse, nematode, Drosophila, and sponge sequences as an initial seed. All retrieved sequences were then used to make other Blast searches, and we eventually got 450 different sequences, which are listed in tables 1–3. A comparative analysis of human and rodent sequences is provided in table 1. Databases from genome projects of zebrafish D. rerio and of the pufferfishes T. rubripes (Aparicio et al. 2002) and T. nigroviridis (Jaillon et al. 2004) were screened using nonteleost sequences as queries, and the results were compiled in table 2. When possible, we confirmed the orthology of a Bcl-2 family member by strictly conserved synteny and contiguity, i.e., presence on the same chromosome in two organisms and conserved order and orientation with respect to an adjacent gene or genes. The existence of putative orthologues was further confirmed by complementary EST database analysis and study of chicken and chimp genomes (table 2). Table 3 lists the ESTs identified for Bcl-2 family members present in nematodes/arthropods and "simple" multicellular animals.

    [in this window]

    [in a new window]

    Table 1 List of Bcl-2 Family Members From Humans, Mouse, and Rat

    [in this window]

    [in a new window]

    Table 2 List of Bcl-2 Family Members in Complete Genomes

    [in this window]

    [in a new window]

    Table 3 List of Bcl-2 Family Members Present in Simple Metazoa

    Subclass Definition

    Three-dimensional (3D) structures of eight members of the Bcl-2 family have been determined, revealing striking resemblance to the pore-forming domains of bacterial toxins (Muchmore et al. 1996; Chou et al. 1999; Suzuki, Youle, and Tjandra 2000; Petros et al. 2001; Huang et al. 2002; Denisov et al. 2003; Huang et al. 2003; Woo et al. 2003). Despite the opposing biological functions and wide differences in amino acid sequences, we found that the experimentally determined structures of CED-9, Bcl-XL, Bax, and, surprisingly, BH3-only protein Bid were strikingly similar (Chou et al. 1999; McDonnell et al. 1999; Suzuki, Youle, and Tjandra 2000). Moreover, we easily modeled a number of other members of the Bcl-2 family (e.g., Bok, Buffy, and Bfl-1) on the same nuclear magnetic resonance or X-ray crystallographic coordinates, while other members were predicted to have similar two-dimensional structures, suggesting that all these members actually share a same -helical conformation and probably common ancestry (fig. 1). These observations prompted their analysis as a distinct subclass of probable "homologues," i.e., multidomain members expected to share a same pore-forming helical bundle structural fold. On the other hand, predicted structures for the other members of the family (BH3-only and BNip proteins) cannot be modeled on Bcl-2–like multidomain templates, and secondary structure analysis indicates that they probably do not share structural similarity with the pore-forming helical bundle class (fig. 1). Yet, sequence similarity strongly suggests that BNip1, BNip2, and BNip3 proteins form three homogenous (putatively monophyletic) groups within the Bcl-2 family. On the contrary, BH3-only proteins could not be defined as a clade because of their high sequence divergence outside the BH3 domain. Hence, we decided to treat them as Bcl-2 "analogs" having their own phylogenetic history. In this study, these relationships were used to distinguish each subgroup from the others, leading to the definition of five subclasses: multidomain members expected to share a same pore-forming structural fold (e.g., CED-9, Bcl-2, Bax, and Bid), Bnip1–3 proteins, and BH3-only members with unrelated secondary structures.

    [in this window]

    [in a new window]

    FIG. 1.— Schematic representation of Bcl-2, BH3-only, and BNip subfamily members, drawn to scale. For each member, structural features (black shapes) and domain organization (lines, on top) are depicted. Rectangles and arrows denote -helices and ?-strands, respectively. For a number of Bcl-2 members sharing the same pore-forming Bcl-2/Bax/Bid topology, predicted structures were obtained by computer simulation. In these cases, the PDB code of the template used for homology modeling is indicated in parentheses. For experimentally determined 3D structures, the Protein Data Bank code appears in italics directly after the acronym without any brackets. Otherwise, secondary structures were determined using the Psipred consensus program (Prof and PHD programs leading to similar results). Characteristic BH signatures reflect domain composition (legend at the top). "?" indicates that the BH domain considered is uncertain or controversial. Class assignments used in phylogenetic analyses are reported (right). All diagrams correspond to human sequences except CED-9, Egl-1, and ceBNip3 from Caenorhabditis elegans, Buffy and dBok from Drosophila melanogaster, and BHP1-2 from Geodia cydonium. Of note, mouse and rat Noxa proteins consist of two tandem repeats of the BH3 domain. Moreover, Mcl-1 and Bcl-G exhibit unusually long N-terminal extensions before the pore-forming structural module (white rectangles), whereas Bcl-rambo is presumably a mosaic protein in which the Bcl-2 module is separated from the TM by a unique amino acid insertion.

    Establishment and Expansion of the Various Bcl-2 Subfamilies

    Multidomain Members Expected to Share a Pore-Forming Helical Bundle Structural Fold

    Bcl-2–like genes were found in sponges and in all major subdivisions of the Bilateria: annelids, molluscs, platyhelminthes, nematodes, arthropods, tunicates, and vertebrates (table 3). In contrast, no Bcl-2–like multidomain sequence can be found in prokaryotes, fungi, and plants. It seems therefore that this Bcl-2 family clade was established in early metazoan evolution. Two Bcl-2 homologues, namely, BHP1 and BHP2 have been depicted in demospongiae (Wiens et al. 2000; Wiens, Diehl-Seifert, and Muller 2001; Wiens et al. 2004). These two paralogues likely result from the duplication of a unique Bcl-2 multidomain ancestor. In Cnidaria and Ascidiacea, descendants of this ancestral multidomain protein formed a multigene family of paralogues early in evolution, with up to five Bcl-2 relatives in H. magnipapillata and four members in C. intestinalis, a basal chordate species. Therefore, Bcl-2 relatives had been established in the very early evolution of animals before the parazoan-eumetazoan split, the earliest branching among extant animal phyla. Contrary to a previous report that failed to identify Bcl-2 family members in platyhelminthes (Verjovski-Almeida et al. 2003), we discovered two homologues in Schistosoma mansoni and its close relative S. japonicum.

    The complete Bcl-2 multidomain sets have been previously defined for D. melanogaster (dBok/Debcl/DRob-1/dBorg-1 and Buffy/dBorg-2) and C. elegans (CED-9). In Drosophila, despite sharing 62% similarity in amino acid sequence, the two homologues of Bcl-2 have opposite effect on cell survival (Brachmann et al. 2000; Colussi et al. 2000; Igaki et al. 2000; Quinn et al. 2003). It has been suggested that Buffy and dBok may constitute the functional counterparts of Bax-like and Bcl-2–like multidomain members, respectively. Notably, Buffy and dBok are most similar to each other than to any other Bcl-2 family member, raising the possibility that they originated from the duplication of a same gene, followed by divergence both in sequence and function (i.e., opposite effects on survival). In contrast to the fly, CED-9 is the sole multidomain Bcl-2 family member in C. elegans, and it has also been found in two other nematodes, namely, C. briggsae and D. immitis. Curiously, the nematode T. spiralis and the annelid L. rubellus contain several Bcl-2 homologues (that segregate with Bcl-2/XL and Mcl-1, respectively, see phylogenetic analysis in fig. 2). This may represent the emergence of new Bcl-2 members in these particular lineages or alternatively a loss of ancestral types in the C. elegans lineage. This may also be due to the high divergence rate reported for nematode genes in general (Aguinaldo et al. 1997; Mushegian et al. 1998). The analysis of Bcl-2 genes from molluscs or other annelids might help to settle this question. In all cases, our results suggest that emergence of multidomain members of the Bcl-2 family predates the divergence of bilaterians and nonbilaterians and is thus of ancient origin in the natural history of Metazoa. Later in the evolution, the Bcl-2 multidomain ancestor or the primitive paralogous lineages found in Cnidaria and Ascidiacea underwent extensive duplications throughout the evolution of Bilateria, especially during vertebrate evolution (see tables 1–3), that lead to the definition of 10 orthology groups in birds and amphibians and 13 orthology groups in mammals. Overall, mammalian subfamily sizes for multidomain members are consistently larger in fishes, mouse, and humans than in Ciona, and very little difference is seen between humans and rodents, suggesting that the expansion in the vertebrate lineage has been followed by a stasis in mammals. The vertebrate set displays a high degree of duplication and divergence relative to both Drosophila multidomain proteins and C. elegans CED-9. We identified various multidomain members of the Bcl-2 family in fishes including clear homologues of Bcl2l10 and Mcl-1. Interestingly, fish orthologues for the recently cloned Bfk protein (Coultas et al. 2003), which is preferentially expressed in the mammary gland during pregnancy, were not found. Curiously, we failed to find a Bax protein orthologue in chicken, in spite of a Bax-like sequence in the bird Taeniopygia guttata. Moreover, sequences displaying some similarity to mammalian Bid could be retrieved in a number of fish species, with conserved BH3 domain and cleavage site (LQTD in mammals and Oncorhynchus mykiss, Salmo salar, Anguilla japonica, and Ictalurus punctatus and IETD in Haplochromis), but it is unclear if they represent true orthologues of Bid. Absence of putative Bid orthologues in ascidians is puzzling because C. intestinalis possesses numerous apoptosis-related genes otherwise, including tumor necrosis factor-like receptors and caspases (Terajima et al. 2003).

    [in this window]

    [in a new window]

    FIG. 2.— Bootstrapped neighbor-joining tree of multidomain Bcl-2–related proteins. Unrooted neighbor-joining tree that depicts the phylogenetic relationships of the multidomain members of the Bcl-2 family (see text for subfamily definition). The tree was inferred based on the alignment of the 68 most conserved amino acid sites between BH3 or equivalent helix to BH2 ("GGW" motif). Numbers indicate percentage support in bootstrap analyses (1,000 replicates). Branch lengths are proportional to distances between sequences. A ML tree inferred using Phyml was largely congruent (see Supplementary file 1, Supplementary Material online). Nomenclature: sequence names contain abbreviations of the taxonomic group, the genus and species. Species abbreviations are mentioned in Materials and Methods. As orthology groups are sometimes difficult to assign, "L" means "-like" (e.g., BakL, BokL, or BaxL). "Bcl" means that the corresponding member could not be clearly assigned to any defined orthology group. When needed, letters and/or numbers after the sequence names are abbreviations of the accession records as figured in tables 1–3 and Supplementary file 4 (Supplementary Material online). BclW Xl and BclX Xl correspond to formerly reported XR11 and XR11 family members in frog, respectively.

    We performed a phylogenetic analysis for representative members of the "helical bundle" clade. As low levels of sequence similarity between full-length subfamily members made completely automated alignments impractical, only the most conserved region lying from BH3 to BH2 (or end of the protein in the case of Bid) of the diverse multidomain Bcl-2 members was compared (fig. 2). If the phylogenetic tree shows some clustering of pro- and antiapoptotic members, a group of multidomain members with poorly defined BH domains (i.e., Bcl-rambo, Bcl-G, BPR, Bfk, and Bid) is clearly visible (see table 1 for motif conservation patterns). As a C. intestinalis sequence segregating with Bax with a 95% bootstrap support displays all the BH motifs, it is likely that these proteins with hardly recognizable motifs are rather newly evolved members. On the contrary, the Nr-13/Bcl2l10 orthology group, which segregates with sequences from sponge, Ciona, and nematodes, shows remarkably long branch lengths, suggesting that its members are rather ancient and could have undergone some functional divergence in vertebrates, especially in mammals. In support to this, we noticed that the structurally important predicted turn between 5 and 6 helices bears an additional dozen amino acids in Bcl2l10 and Diva/Boo which are not present in chicken Nr-13. These supplementary amino acids may have some consequences on protein insertion in membranes and on activity of this member toward apoptosis modulation. Interestingly, although anti- or proapoptotic activity of Bcl2l10 has been reported (Aouacheria et al. 2001; Ke, Godzik, and Reed 2001; Lee et al. 2001; Zhang, Holzgreve, and De Geyter 2001), genetic knockout of mouse Diva/Boo did not cause any severe phenotype in line with apoptosis (Russell et al. 2002). Therefore, it would be of great interest to address potential extraroles beyond cell death regulation for these proteins, and as a general rule, for all multidomain members as the classical tissue-specific expression argument might not be the unique reason accounting for the apparent redundancy within the Bcl-2 family.

    BH3-Only Proteins with Unrelated Secondary Structure

    BH3-only proteins with secondary structures unrelated to multidomain members or BNip proteins have been identified in various vertebrates but not in invertebrates except the nematode C. elegans. Notably, we were not able to identify any Egl-1–related proapoptotic BH3-only proteins in arthropoda, especially in the fly genome. In this particular species, one potential functional substitute for the BH3 domain may be the amphipathic -helical "GH3" motif recently identified in the proapoptotic Grim protein (Claveria et al. 2002). We also failed to identify many BH3-only Bcl-2 family members, such as Harakiri, Bik, and Puma, in the fish genomes. It could be because of lack of statistically significant motif in the BH3-only proapoptotic proteins that would allow reliable sequence-based predictions. Indeed, because most BH3-only proteins are rather short and given that these sequences are known to be quite divergent in distant species, it is possible that the search parameters used may not be optimal for detecting their orthologues. Alternatively, these genes are either lost in the fish lineage or arose independently after the divergence with the mammalian lineage. In all cases, it is noteworthy that out of the two p53 target genes Puma and Noxa involved in the regulation of apoptosis in mammals, only a Noxa-like sequence could be retrieved in fishes, although data mining of the Fugu genome has otherwise revealed a good level of conservation for the p53 pathway and p53 responsive elements (Le Bras, Bensaad, and Soussi 2003). An incongruence reminiscent of the one reported above for Bax-like sequence in birds occurred with BH3-only protein Bad in amphibians, which seems to be present in A. tigrinum but apparently absent in other amphibians including X. laevis.

    A phylogenetic analysis has been performed on the single BH3 domain, and the unrooted phylogenetic tree is shown in figure 3. Obviously, the genes encoding the structurally unrelated BH3-only proteins have an even more complex evolutionary history than the multidomain members, as revealed by the unresolved star-like tree topology. Nonetheless, we collected a number of lines of evidences that could shed some light on the origin and early evolution of BH3-only proteins. First, in rodents, the Noxa/APR protein consists of two tandem repeats of the BH3 domain (fig. 1), suggesting that recombination processes involving this motif could have occurred during the course of evolution. Furthermore, while searching for fish orthologues of Noxa, we found that the extremely short open reading frame (128 bp long; accession number: BE557856) encoding the probable Noxa homologue in D. rerio was bearing Alu and MER35 repetitive elements, suggesting that the Noxa-like BH3 domain could actually correspond to an exonization of transposable elements. Exaptation of repetitive elements into genes encoding cellular proteins has been described in primates as a mechanism enriching transcriptome diversity, with examples available inside the Bcl-2 family itself (e.g., Alu-like cassette insertions in Bim ?3 and Bcl-rambo ?) (U et al. 2001; Yi et al. 2003). Thus, transposon origin for Noxa BH3 domain is theoretically conceivable because mobile DNA elements might have a profound influence in creating novel protein diversity. Second, the Bik/Nbk protein displays unusually high resemblance (for a BH3-only protein) to the multidomain member Bak in an 60 amino acid region encompassing the BH3 domains (e.g., 36% identity and 58% similarity between chicken Bik and murine Bak), raising the possibility of some evolutionary relationships. Because alternatively spliced forms of multidomain members that bear only the BH3 domain and lack BH1/BH2/TM domains (e.g., Bcl-XS, Mcl-1-S; Bax-, Bcl-GS) have been depicted, it is tempting to speculate that some BH3-only proteins are derived from common ancestry with multi-BH members through gene duplication and exon loss. Lastly, when an [L-X-X-X-G-D-D/E] pattern designed to identify BH3-only proteins is used to search the UNIPROT database (http://npsa-pbil.ibcp.fr/cgi-bin/npsa_automat.pl?page=/NPSA/npsa_pattinprot.html), as many as 1,994 hits (present on 1,966 proteins) are retrieved. Because the database contains 155,576 proteins, this output corresponds to a fraction of 0.01, with the randomized probability for this pattern being P = 4 x 10–5. When the same experiment is done with the [W-X-X-X-X-G-G-W] and [N-W-G-R] patterns roughly corresponding to the BH2 and BH1 domains, respectively, only 78 and 169 proteins match (which correspond to random probabilities of P = 7.045 x 10–7 and P = 1.932 x 10–6, respectively). These observations raise the possibility that a number of putative BH3 domains or so-called "BH3-like" domains actually arose either randomly or by a process of convergent evolution. The recently reported Mcf toxin encoded by the "makes caterpillars floppy" gene of the Gram-negative bacteria Photorhabdus luminescens provides a good example of such possible convergent evolution. Indeed, this rather long insecticidal protein (2,929 amino acids) has no primary or secondary structure similarity to any other known protein sequence except in a short, degenerated BH3-like stretch of residues, which is still awaiting further characterization (Daborn et al. 2002). On the other hand, despite lack of any sequence similarity with BH3 domains, a so-called Siva-1 amphipathic helical region on another proapoptotic protein, Siva-1, is also of helical structure and sufficient to mediate the interaction with antiapoptotic members of the Bcl-2 family such as Bcl-2 and Bcl-XL (Chu et al. 2004), suggesting that Siva-1 is formed and behaves as a BH3-only protein. Thus, structural/functional equivalents of the BH3 domain might exist in absence of clear sequence similarity (e.g., in Grim, Mcf, and Siva-1), but the corresponding proteins should best be termed structural analogs of BH3-only members.

    [in this window]

    [in a new window]

    FIG. 3.— Bootstrapped neighbor-joining tree of BH3-only proteins with unrelated secondary structures. Unrooted tree is shown because common ancestry is unlikely. Tree has been calculated from amino acid sequence alignment of BH3 domains plus a few flanking amino acids (total residues in ungapped alignment = 37). Only bootstrap values 50 are shown. A ML tree from Phyml led to roughly similar results (see Supplementary file 2, Supplementary Material online).

    BNip Proteins

    BNip1, 2, and 3 proteins were first identified as adenoviral E1B 19K interacting proteins present in mammalian cells (Boyd et al. 1994). We found BNip-related sequences in a wide range of eukaryotes, from unicellular (e.g., BNip2) to plants (e.g., BNip1), and simple multicellular animals. For the sake of report clarity, we decided to restrict phylogenetic analyses to BNip sequences present in multicellular animals.

    BNip1.

    Human BNip1 protein contains a BH3-like domain followed by a region of similarity to Sec20 (a membrane glycoprotein associated with secretory pathway), including two ankyrin repeats and a putative C-terminal TM domain. Various BNip1-related sequences were found in multicellular animals (tables 1–3). Alignments of BNip1 subgroup members reveal unexpectedly poor conservation between fishes and other vertebrates of the BH3-like domains compared to adjacent regions and compared to the Sec20 homology region. In addition, several fungi and flowering plant sequences (e.g., unnamed protein product BAB02931 from Arabidopsis thaliana) were found to cluster with BNip1 proteins, and careful examination indicates that the similarity extends to a region encompassing the supposed BH3 domain, the two ankyrin repeats, and the TM domain. The Sec20 homology region (PFAM03908) present in BNip1 is also retrieved in Sec20 proteins from Candida albicans (CAB76200), Saccharomyces cerevisiae (P28791), and Schizosaccharomyces pombe (CAA16989). These results suggest that the BH3-like domain of BNip1 proteins rather corresponds to a divergent amino acid stretch most probably included in an unrelated functional region. Unlike arthropods (fig. 4A), worms and platyhelminthes do not seem to possess any true BNip1 orthologue at all (only a weak similarity could be detected between the N-terminal moiety of C. elegans ankyrin protein and the C-terminal part of BNip1 proteins). In contrast, a putative homolog sharing 66% similarity with chicken BNip1 is present in C. intestinalis (and also H. roretzi). Therefore, it seems that the BNip1 subfamily has been established more recently in metazoan evolution, most probably after the protostome/deuterostome divergence. It has been previously proposed that sponge death domain-containing protein DD2 evolved from ankyrin building blocks by gene duplication followed by divergence (Wiens et al. 2000). Based on this example, the evolution of BNip1 in multicellular animals could also have involved the accretion of a similar ankyrin building block with an unrelated C-terminal domain (corresponding to PFAM-B 63610 and 53542 domains).

    [in this window]

    [in a new window]

    FIG. 4.— Bootstrapped ML trees of BNip proteins. (A), (B), and (C): Rooted neighbor-joining tree of BNip1, BNip2, and BNip3 proteins, respectively. For the BNip1 clade, phylogenetic calculations were computed based on alignment of a region encompassing the BH3-like domain, the ankyrin repeats, and the TM domain, and the tree was rooted using the nematode ankyrin sequence. The carboxyl terminus part of BNip2-like proteins with homology with Sec14/GAP and comprising the PFAM-B 7813, CRAL-TRIO, and BCH domains was used for BNip2 subfamily analysis; the tree was rooted with the Ciona sequence. A region encompassing the BH3 and TM domains was retained for the BNip3 phylogenetic analysis, and the tree was rooted with the Strongylocentrotus sequence. Number of examined sites is 142, 206, and 75 for BNip1, BNip2, and BNip3, respectively. Similar tree topologies were obtained with both neighbor-joining and ML methods (see Supplementary file 3, Supplementary Material online).

    BNip2.

    From N- to C-terminus, BNip2 proteins have been reported to contain a Ca2+-binding motif, a BH3-like domain, a CRAL-TRIO motif (Cellular retinaldehyde–binding/Triple function domain) common to proteins that bind small lipophilic molecules, and a region termed the BCH domain that shares similarity to the noncatalytic domain of Cdc42GAP ("Bnip2 and Cdc42GAP homology domain") (Zhang et al. 2003). Contrary to other BNip proteins, BNip2 is devoid of a TM hydrophobic domain. Overall, sequence comparisons show that the N-terminal region of BNip2 relatives including the putative calcium-binding and BH3-like sites is only poorly conserved. Proposed BH3 domain for human BNip2 protein corresponds to the amino acid stretch 88PSENSDEFE97 (Zhang et al. 2003) included in the PFAM-B 7813 domain. When compared to other well-known BH3 signatures, the BH3-like motif present in BNip2 proteins exhibits a Pro residue in position 1 rather than the highly conserved Leu residue and never bears the highly frequent triplet GDE in position 6-8 or the frequent triplet EFE in position 8-10. In addition, the supposed BH3-like site is not predicted to adopt a -helical conformation (fig. 1). Taken together, these observations suggest that BNip2, as BNip1, may not exert its function primarily owing to this motif. In sharp contrast, the C-terminal part of BNip2 including the BCH domain and the CRAL-TRIO motifs (PFAM-B 1538) is much more conserved, confirming the crucial role of this region as reported by functional characterization (Zhou et al. 2002). Several sequences from unicellular (e.g., Dictyostelium discoideum C91038 [GenBank] ), plant (e.g., A. thaliana BX834778 [GenBank] ), tunicates (Oikopleura dioica), molluscs (C. virginica), and platyhelminthes (S. mansoni) sharing some similarity with this C-terminal part of BNip2 proteins were also found. As most of these sequences are partial, it remains unclear if they also share similarity outside the Sec14/GAP homology region. The phyletic distribution of BNip2-like sequences roughly resembles to that of the BNip1 subgroup. Indeed, while putative orthologues were detected in fly and C. intestinalis genomes, we failed to identify any worm orthologue of BNip2. The subfamily size differences among organisms suggest that this subclass experienced a period of stasis until the separation of the vertebrate lineage from Ciona (fig. 4B). The number of different BNip2 members is one in fly and Ciona and 4–5 in fishes, primates, and rodents and hence may reveal the occurrence of two whole-genome duplications (2R hypothesis) in early vertebrate evolution. In mouse and rat, the presence of an intronless gene suggests a lineage-specific retro-pseudogenization event. Interestingly, fishes have a species-specific duplication of one particular BNip2 isoform called caytaxin, possibly reflecting some ecological and physiological relevance of this gene in this particular lineage. In mammals, this novel gene encodes a neuron-specific protein whose mutations are responsible for severe disorders in humans (Cayman ataxia) and mouse (hesitant and jittery phenotypes) (Bomar et al. 2003).

    BNip3.

    BNip3 and its variants have been reported to contain a BH3 homology region necessary for interaction with other Bcl-2 family members and a C-terminal TM domain important for mitochondrial localization and homodimerization (Boyd et al. 1994; Yasuda et al. 1998a, 1998b; Ray et al. 2000). BNip3-related sequences were found in all major lineages of Bilateria, from Echinoidea and Ascidiacea to mammals. Our data bank searches also identified several proteins significantly similar to BNip3 in arthropoda and worms, including homologues of C. elegans proapoptotic ceBNip3 protein in a dozen of other nematodes (Yasuda et al. 1998a), and a fly candidate that might be useful in future genetic studies. Although overall similarity between ceBNip3, BNip3 Drosophila homologue, and human BNip3 is weak (between 30% and 40%), BNip3 may effectively bear a bona fide (yet atypical, see below) BH3 domain as "threading" programs such as LIBRA (http://www.ddbj.nig.ac.jp/search/libra_i-e.html) and 3D-PSSM (http://www.sbg.bio.ic.ac.uk/3dpssm.index2.html) successfully align the BH3 domain of human BNip3 with that of BH3-only protein Bid as compatible structures in forward-folding searches (not shown). This conclusion is largely supported by experimental evidences (Chen et al. 1997; Matsushima et al. 1998; Yasuda et al. 1998a, 1998b; Chen et al. 1999; Bruick 2000). However, when more sequences are added to the alignments, one can observe that the conserved residues are rather unusual for a BH3 domain. Indeed, instead of the [L/V/M1-X2-X3-X4-G5-D6-D/E7-F8-E9-R10] "canonical" pattern of amino acid conservation commonly found for BH3 domains in other Bcl-2 family members, BNip3 proteins display at least two specific and highly conserved Trp residues in positions 7 and 11. BNip3 has been shown to homodimerize but, in contrast to other Bcl-2 relatives, BNip3 homodimerization does not depend on the BH3-like domain but rather on the C-terminus hydrophobic TM domain (Chen et al. 1997). Therefore, the unusual conservative pattern observed for the BNip3 BH3-like domain could be an indication of specific constraints for interaction with Bcl-2 multidomain family members or other unrelated proteins. Another proapoptotic homologue of BNip3 called Nix (or BNip3L for BNip3-like) has been previously reported (Chen et al. 1999). Our phylogenetic analysis generated trees supporting the notion that this member of the BNip3 subgroup originated from a duplication of the ancestral BNip3 gene present in nematodes, the duplication event occurring early in the chordate evolution, most probably before the gnathostome radiation (fig. 4C). Increase of duplication patterns were found for BNip3 relatives in fishes and mammalians compared to the C. elegans lineage, particularly in zebrafish and rodents. As we previously mentioned, the increase of gene duplication during this time period is regarded as evidence for the 2R hypothesis by some investigators (Wolfe 2001; Hokamp, McLysaght, and Wolfe 2003). As we also found a tandem duplication of the BNip3 gene in zebrafish, another explanation for the observed increase in duplication events is that multiple, independent segmental duplications of chromosomes occurred within a relatively short time frame. Lastly, we noticed that several degenerated copies of BNip3-related genes with high number of stop codons and point mutations are present in the human and mouse/rat genomes. As we assumed that these copies are likely to represent pseudogenes, they were not included in the present analysis.

    Taken together, these results indicate that BNip1–3 proteins form three monophyletic subgroups inside the Bcl-2 family. Yet, classification inside the Bcl-2 family remains doubtful (at least for BNip1 and BNip2 proteins) based on domain composition and poor conservation of the BH3-like motif and TM domain (which is a low-complexity hydrophobic region). Moreover, it has been reported that interaction with other Bcl-2 family members and proapoptotic activity might be BH3 independent for these proteins (Ray et al. 2000; Zhang et al. 2003), so experimental confirmations are required to assess precise structure-function relationships for all three BNip subgroups.

    Analyses of Genomic Features

    Owing to the high number of genes within the updated Bcl-2 family (fig. 5), gene structure/location analysis could not be described here in detail for the whole set of Bcl-2 members. We chose to present and discuss here the results for some of the genes encoding multidomain members sharing the helical bundle structural fold as it appeared to us that this well-characterized and rather homogenous group is the more suitable for phylogenetic inference. Complete transcript structures could be retrieved in Supplementary file 5 (Supplementary Material online).

    [in this window]

    [in a new window]

    FIG. 5.— Phylogeny of Bcl-2–like proteins overlaid upon the general taxonomy of Metazoa. The diagram illustrates the extent of gene lineage sorting and inferred diversification events (in bold). Circles indicate the presence in a particular lineage of at least one member of a specific Bcl-2 family subclass. Approximate number of members in the different lineages for which whole genomes are available is represented (legend at the top).

    Intron-Exon Structures

    First, as shown in figure 6, some kind of clustering between anti- and proapoptotic genes could be observed on the basis of predicted intron-exon structures: the coding region of "antiapoptotic" genes (e.g., Bcl-2, Bcl-XL, Bcl-W, Bfl1/A1, Mcl-1, and sponge BHP2) is spread over two or three coding exons, whereas "proapoptotic" genes always display more coding exons, from five to seven in general (e.g., Bax, Bak, Bcl-G, and Bok). Nonetheless, a common feature among both prosurvival and prodeath members (minus Bid) is the presence of a conserved intron splitting the C-terminal BH2 domain just after the "GGW" signature. Because this conserved intron within the BH2 domain is common to anti- and proapoptotic family members, it has been hypothesized that both sides of the family may have arisen from a primordial member, which contained a single intron in the coding region and which underwent subsequent gene duplication and intron gain in proapoptotic members (Hatakeyama et al. 1998; Herberg et al. 1998). CED-9, the sole worm multidomain protein, could represent a living molecular fossil of such putative bifunctional primordial member. Indeed, the fact that antiapoptotic CED-9 has also been reported to promote apoptosis in nematodes with a certain "genetic makeup" (Hengartner and Horvitz 1994; Xue and Horvitz 1997), suggesting that it could assume both prosurvival and proapoptotic roles, argues in favor to this hypothesis. Next, position of the additional upstream introns is not precisely conserved among proapoptotic members, suggesting that these additional introns were added after the divergence of the primordial gene. However, the reason for specific addition of introns upstream of the conserved BH2 intron/exon junction during the evolution of proapoptotic family members and not of antiapoptotic members is not clear. One possibility is that there might be a direct link between intron acquisition and the emergence of proapoptotic members. In this respect, the complex intron-exon structures observed in prodeath member genes might have arisen as a mechanism to finely regulate the toxic activities of their encoded products, e.g., via alternative splicing, as exemplified by the numerous spliced products of the bax gene (Apte, Mattei, and Olsen 1995; Zhou et al. 1998; Shi et al. 1999; Schmitt et al. 2000; Cartron et al. 2002). Furthermore, such a mechanism could have constituted a mean to generate novel biochemical functions and/or tissue-specific expression patterns without further, potentially harmful, gene duplications of prodeath members. Notably, presence of introns is thought to speed up the formation of new combinations between exons in a process called exon shuffling (Gilbert, de Souza, and Long 1997; Kolkman and Stemmer 2001). It is noteworthy that within the Bak, Bax, Bcl-rambo, Mcl-1, Bfk, Bcl-G, BPR, and Bid genes, the BH3 domain is encoded by a single exon bearing no other domain (BH4, loop, BH1, BH2, or TM). Likewise, we noticed that, in all BH3-only genes reported to date, the BH3 domain is a structural and functional subunit encoded by a single short exon. These remarks raise the possibility that the BH3-encoding exons of proapoptotic multidomain members could represent a target for exon shuffling, involved in the formation of new gene combinations evolving into genes with novel functions, for instance BH3-only genes.

    [in this window]

    [in a new window]

    FIG. 6.— Schematic representation of transcript structures inside the Bcl-2 family. Transcripts are grouped by subfamily. Only coding exons are presented (gray horizontal lines). The length of the line is not proportional to the sequence length. Arrows denote intron position. BH motifs are schematized on top of the transcripts. Coding is the same as in figure 1 (see legend at the top). For Bcl-rambo, only the BH moiety was depicted. Figure not drawn to scale.

    Comparative Mapping Study

    The most striking observation that appeared from our analysis is that Bcl-rambo and Bid correspond to a conserved linkage group in Fugu, chicken, rodents, chimpanzee, and humans, with preservation of both gene order and orientation. In all these species, Bcl-rambo and BH3-only Bid form two distally adjacent genes on opposite strands. In addition, both members exhibit very similar phyletic distribution (tables 1 and 2). Bcl-rambo contains BH1–BH4 domains and a TM anchor and shows overall predicted structural similarity to the multidomain member of the Bcl-2 family (fig. 1). However, the Bcl-rambo protein constitutes somehow a unique case in the Bcl-2 family because its TM domain is preceded by a 250–amino acid insertion (the so-called "BHNo" region) that displays no similarity to any other known protein (Kataoka et al. 2001; Yi et al. 2003). In addition, although the effect of the BH domain has not been addressed yet, it is noteworthy that Bcl-rambo has proapoptotic activity that is induced by this terminal extension and not by the BH moiety. On the other hand, the C-terminal part of the BH2 domain, the BHNo region, and the TM domain are encoded by a single exon, suggesting that the extra region has been brought by an in-frame insertion of noncoding sequences that intervened just after the conserved 3' intron. Because Bid precisely lacks the BH2 domain but not a C-terminal conserved intron, one intriguing possibility is that Bid originated from the Bcl-rambo gene which is located in its close vicinity. In support to this hypothesis is the fact that, as previously mentioned, Bcl-rambo has undergone loss-of-function divergence or at least neofunctionalization in its BH half, while Bid has gained a unique role in signaling of apoptosis because it links the death receptor pathway to the mitochondrial pathway mediated by the Bcl-2 proteins (Li et al. 1998). It is now widely believed that neo- or subfunctionalization processes explain at least in part the retention of duplicated genes (Prince and Pickett 2002). We were not able to gain any clue from direct analysis of nucleotide sequences or cis-regulatory elements (not shown), suggesting that the duplication event, if any, should be ancient. Although supported by a low bootstrap value (58%), putative common evolutionary origin for Bid and Bcl-rambo is apparent on the multidomain Bcl-2 family phylogenetic tree (fig. 2). Of note, Bcl-rambo and Bid genes are close enough in vertebrate genomes to raise the possibility of some overlapping transcripts or even antisense regulation, the two hypotheses now warranting further experimental evidence.

    Establishment of Domain Configuration in the Multidomain Bcl-2 Family

    For the multidomain subfamily, domain architectures are made of up to four amino acid typical signatures, namely, BH1–4. In this part, Bcl-2 family members were compared among organisms to uncover the pattern of conservation or divergence in domain organization, and a model for domain recruitment events is proposed.

    BH1 and BH2 Domains

    The BH1 and BH2 domains are well recognizable in BHP1 and BHP2 sponge sequences, indicating that these motifs were already present in the common ancestor of Cnidaria and Bilateria, i.e., very early in the evolution of animals. 3D structures of Bcl-2 family members have revealed that these two domains are surrounding a -hairpin structural motif. This helical hairpin is formed by the helices 5–6 and has been shown to be crucial for pore formation in mitochondrial membranes (Schendel et al. 1997; Matsuyama et al. 1998; Heimlich et al. 2004). Although necessary for Bcl-2 function, the 5–6 region appears to be insufficient for promoting cell survival as replacing the 5–6 helices of Bax with this segment of Bcl-2 did not convert Bax to a cytoprotective protein. Likewise, although the 5–6 region of Bax was necessary for its cytotoxic activity in yeast, engineering these -helices into the Bcl-2 protein was insufficient for switching its phenotype (Matsuyama et al. 1998). Thus, both pro- and antiapoptotic Bcl-2 family proteins can regulate membrane efflux at the level of mitochondria (Kuwana et al. 2002), and BH1–2 and 5–6 region may not be responsible for the opposite activity of Bcl-2 pro- and antiapoptotic members. Instead, this region might have constituted the first elementary unit with biological activity in the Bcl-2 family ancestor, i.e., a toxic module responsible for pore formation. If true, this hypothesis implies that the ancestor and its antiapoptotic descendants should have developed some means to inhibit the latent toxic activity of this toxic domain, in order to prevent inappropriate cell death induction (see below). Conversely, the ancestor and its proapoptotic counterparts should somehow have evolved a specificity allowing the controlled activation of this intrinsic prodeath function.

    BH4 Domain and Flexible Loop

    In contrast to the C-terminal moiety, the N-terminal moiety is poorly conserved among Bcl-2 family members, particularly in proapoptotic relatives (Adams and Cory 1998; Bonnefoy-Berard et al. 2004). This N-terminal region is formed by the BH4 domain (corresponding to an amphipathic helix) linked to the rest of the protein by a flexible loop of undefined structure and various lengths. This region present in nematode CED-9 is characteristic of antiapoptotic proteins of the Bcl-2 family and is supposed to confer specificity on each individual member of the family, for instance through interaction with specific partners and posttranslational modifications such as phosphorylation or deamidation (Breitschopf et al. 2000; Blagosklonny 2001; Deverman et al. 2002). Analysis of BHP1 and BHP2 proteins shows that a -helix presumably corresponding to BH4 location plus a loop upstream of the 5–6 helices were already coexisting in these living molecular fossils. The BH4 domain and the loop have been shown to display a negative regulatory function on the activity of antiapoptotic multidomain members (Reed 1997). Indeed, deletion of the BH4 domain results in a total inactivation of Bcl-2 or Bcl-XL (Huang et al. 1997). Moreover, in some cases, deletions or mutations of these regions could induce cell death (e.g., Bcl-XS, a proapoptotic variant of Bcl-XL) and have the same consequences as a cleavage releasing the C-terminal pore-forming domain: they convert an antiapoptotic member into a proapoptotic one (e.g., cleavage of Bcl-2 or Bcl-XL) (Cheng et al. 1997; Kirsch et al. 1999; Basanez et al. 2001; Liang et al. 2002). Therefore, pro- and antiapoptotic multidomain members of the Bcl-2 family presumably contain a same "elementary toxic module" required for pore formation, one difference being that antiapoptotic multidomain members have developed specific N-terminal regulatory regions (BH4 plus loop) to neutralize their own toxicity. Equivalent N-terminal sites of negative autoregulation may exist for Bid as well, as K. O. Tan, K. M. Tan, and Yu (1999) identified a "BH3B" domain important for intramolecular inhibition of proapoptotic activity.

    BH3 Domain

    The BH3 domain is required for the heterodimerization of the Bcl-2–like multidomain proteins and for the binding with other members such as BH3-only proteins. In contrast to the BH1–2 domains, no recognizable BH3 domain signature could be observed in either sponge BHP1 or BHP2 sequences, despite -helical–predicted secondary structures at the expected location. On the other hand, clear typical signatures of the BH3 domain could be identified in at least two sequences from both Ciona and Hydra, suggesting that this domain evolved after the parazoan-eumetazoan split, i.e., at the time of establishment of the initial family of paralogous genes that probably included both pro- and antiapoptotic members (see in fig. 2 the C. intestinalis sequences segregating with Bax/Bok and BHP2/CED-9). It is thus tempting to speculate that emergence of the BH3 domain may have permitted the appearance of heterodimers between paralogous prosurvival and prodeath Bcl-2 family members. However, CED-9 exhibits a bona fide BH3 domain in absence of any Bax-like counterpart, which is in apparent contradiction with this speculation. The most parsimonious possibility is that one of the duplicate genes might have been lost in the C. elegans lineage (contrary to T. spiralis and L. rumbellus lineages). In this model, CED-9 might have been enforced to assume both the prosurvival and the proapoptotic roles in the nematode cells, probably owing to a newly evolved capacity to adopt different conformations depending on the cellular context.

    In conclusion, the opposing phenotypes of Bcl-2–like survival proteins and Bax-like death factors presumably require both the 5–6 region and additional domains such as BH3 and BH4. A possible evolutionary scenario is that the basic function of Bcl-2 ancestral proteins was primarily pore formation via their 5–6 helices, the BH4 domain, and the loop acting as an internal switch and that the BH3 domain has evolved because of dimerization constraints linked to the pairing of antagonist duet effectors.

    Conclusion

    Bcl-2 Family Diversification and Evolution of Organism Complexity

    Because Bcl-2–related proteins have been discovered in Porifera, the origin of the Bcl-2 family must have predated the Cambrian explosion (Valentine, Erwin, and Jablonski 1996), the time of main divergence of metazoan phyla. Long after the Porifera-Eumetazoa split, extensive duplication (probably corresponding to tissue-specific isoforms and possibly neofunctionalized paralogues) occurred in the first half-period of chordate evolution, between the protostome/deuterostome split and the fish-tetrapod split, by which subfamily members expanded (fig. 5). Noticeably, the divergence pattern of multidomain family members shows a close resemblance to that of the Pax family, G proteins, collagens, integrins alpha family, and tyrosine kinases, i.e., gene families related to the evolution of multicellularity (Muller 1997; Suga et al. 1997; Hoshiyama et al. 1998; Breitling and Gerber 2000; Hughes 2001; Aouacheria et al. 2004). Moreover, multidomain Bcl-2 family members have been identified in all multicellular animals studied so far and, conversely, no multidomain Bcl-2 members have ever been identified outside the metazoan phyla. Lastly, Bcl-2–like genes experienced serial duplication accompanying the emergence and continuing evolution of complex developmental processes after the origin of Metazoa, as exemplified by knockout studies in mice (Chao and Korsmeyer 1998; Sorenson 2004). Therefore, we wish to propose that acquisition and subsequent diversification of multidomain Bcl-2 family proteins might be linked to the acquisition and development of multicellularity in the animal kingdom. Restricted heterodimerization between pro- and antisurvival multidomain members, and emergence of multiple BH3-only proteins, may have evolved later to regulate independent death and survival pathways in response to distinct stimuli. Lastly, alternative splicing, posttranslational modifications such as proteolytic cleavage and phosphorylation, and even exonization of Alu sequences in primates (Yi et al. 2003) may have constituted diverse devices to further increase variability and multifunctionality of Bcl-2 family members.

    The results of the present study provide a basis to further study the origin and evolution of the Bcl-2 family. In particular, the fish models are expected to bridge the gap between the nematode/Drosophila and mouse/human models. Indeed, the genomic architecture of fishes could provide an excellent genetic tool to allow rapid identification of important conserved regulatory regions and investigate the molecular mechanisms of apoptosis during development and pathogenesis. From a more evolutionary perspective, the striking similarities in the 3D structures and pore-forming abilities of Bcl-2 proteins and a subset of bacteriocins (e.g., colicins and diphtheria toxin) raise the possibility that Bcl-2 family proteins and these bacterial toxins may share a common ancestral origin. As it is difficult at present to evaluate whether this common structural and functional analogy represents an example of parallel or convergent evolution, explaining the acquisition of a pore-forming multidomain Bcl-2 member by early multicellular animals is challenging. Increasing evolutionary diversity for comparative genomics to early-emerging multicellular animals (e.g., Porifera, Placozoa, Cnidaria, Ctenophora) may help to decipher animal origins.

    Supplementary Material

    Supplementary files 1–5 are available at Molecular Biology and Evolution online (http://www.mbe.oxfordjournals.org/).

    Supplementary file 1. Maximum-likelihood tree for multidomain Bcl-2 family members.

    Supplementary file 2. Maximum-likelihood tree for BH3-only proteins with unrelated secondary structures.

    Supplementary file 3. Neighbor-joining trees for BNip proteins.

    Supplementary file 4. List of Bcl-2 family sequences extracted from EST databases.

    Supplementary file 5. Genomic organization of Bcl-2 family genes. Exon-intron features are reproduced with authorization from the Ensembl Web site.

    Acknowledgements

    We thank G. Gillet at the Institute de Biologie et Chimie des Protéincs for insightful critique of the manuscript. We are also grateful to C. Gautier for his continuous support and to A. Khelifi for pseudogene detection. A.A. is a recipient of a fellowship from the Association pour la Recherche sur le Cancer.

    References

    Adams, J. M., and S. Cory. 1998. The Bcl-2 protein family: arbiters of cell survival. Science 281:1322–1326.

    ———. 2001. Life-or-death decisions by the Bcl-2 protein family. Trends Biochem. Sci. 26:61–66.

    Aguinaldo, A. M., J. M. Turbeville, L. S. Linford, M. C. Rivera, J. R. Garey, R. A. Raff, and J. A. Lake. 1997. Evidence for a clade of nematodes, arthropods and other moulting animals. Nature 387:489–493.

    Altschul, S. F., T. L. Madden, A. A. Schaffer, J. Zhang, Z. Zhang, W. Miller, and D. J. Lipman. 1997. Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res. 25:3389–3402.

    Aouacheria, A., E. Arnaud, S. Venet, P. Lalle, M. Gouy, D. Rigal, and G. Gillet. 2001. Nrh, a human homologue of Nr-13 associates with Bcl-Xs and is an inhibitor of apoptosis. Oncogene 20:5846–5855.

    Aouacheria, A., C. Cluzel, C. Lethias, M. Gouy, R. Garrone, and J. Y. Exposito. 2004. Invertebrate data predict an early emergence of vertebrate fibrillar collagen clades and an anti-incest model. J. Biol. Chem. 279:47711–47719.

    Aparicio, S., J. Chapman, E. Stupka et al. (41 co-authors). 2002. Whole-genome shotgun assembly and analysis of the genome of Fugu rubripes. Science 297:1301–1310.

    Apte, S. S., M. G. Mattei, and B. R. Olsen. 1995. Mapping of the human BAX gene to chromosome 19q13.3-q13.4 and isolation of a novel alternatively spliced transcript, BAX delta. Genomics 26:592–594.

    Basanez, G., J. Zhang, B. N. Chau, G. I. Maksaev, V. A. Frolov, T. A. Brandt, J. Burch, J. M. Hardwick, and J. Zimmerberg. 2001. Pro-apoptotic cleavage products of Bcl-xL form cytochrome c-conducting pores in pure lipid membranes. J. Biol. Chem. 276:31083–31091.

    Blagosklonny, M. V. 2001. Unwinding the loop of Bcl-2 phosphorylation. Leukemia 15:869–874.

    Bomar, J. M., P. J. Benke, E. L. Slattery et al. (13 co-authors). 2003. Mutations in a novel gene encoding a CRAL-TRIO domain cause human Cayman ataxia and ataxia/dystonia in the jittery mouse. Nat. Genet. 35:264–269.

    Bonnefoy-Berard, N., A. Aouacheria, C. Verschelde, L. Quemeneur, A. Marcais, and J. Marvel. 2004. Control of proliferation by Bcl-2 family members. Biochim. Biophys. Acta 1644:159–168.

    Borner, C. 2003. The Bcl-2 protein family: sensors and checkpoints for life-or-death decisions. Mol. Immunol. 39:615–647.

    Boyd, J. M., S. Malstrom, T. Subramanian, L. K. Venkatesh, U. Schaeper, B. Elangovan, C. D'Sa-Eipper, and G. Chinnadurai. 1994. Adenovirus E1B 19 kDa and Bcl-2 proteins interact with a common set of cellular proteins. Cell 79:341–351.

    Brachmann, C. B., O. W. Jassim, B. D. Wachsmuth, and R. L. Cagan. 2000. The Drosophila bcl-2 family member dBorg-1 functions in the apoptotic response to UV-irradiation. Curr. Biol. 10:547–550.

    Breitling, R., and J. K. Gerber. 2000. Origin of the paired domain. Dev. Genes Evol. 210:644–650.

    Breitschopf, K., J. Haendeler, P. Malchow, A. M. Zeiher, and S. Dimmeler. 2000. Posttranslational modification of Bcl-2 facilitates its proteasome-dependent degradation: molecular characterization of the involved signaling pathway. Mol. Cell. Biol. 20:1886–1896.

    Bruick, R. K. 2000. Expression of the gene encoding the proapoptotic Nip3 protein is induced by hypoxia. Proc. Natl. Acad. Sci. USA 97:9082–9087.

    Cartron, P. F., L. Oliver, S. Martin, C. Moreau, M. T. LeCabellec, P. Jezequel, K. Meflah, and F. M. Vallette. 2002. The expression of a new variant of the pro-apoptotic molecule Bax, Baxpsi, is correlated with an increased survival of glioblastoma multiforme patients. Hum. Mol. Genet. 11:675–687.

    Chao, D. T., and S. J. Korsmeyer. 1998. BCL-2 family: regulators of cell death. Annu. Rev. Immunol. 16:395–419.

    Chen, G., J. Cizeau, C. Vande Velde, J. H. Park, G. Bozek, J. Bolton, L. Shi, D. Dubik, and A. Greenberg. 1999. Nix and Nip3 form a subfamily of pro-apoptotic mitochondrial proteins. J. Biol. Chem. 274:7–10.

    Chen, G., R. Ray, D. Dubik, L. Shi, J. Cizeau, R. C. Bleackley, S. Saxena, R. D. Gietz, and A. H. Greenberg. 1997. The E1B 19K/Bcl-2-binding protein Nip3 is a dimeric mitochondrial protein that activates apoptosis. J. Exp. Med. 186:1975–1983.

    Cheng, E. H., D. G. Kirsch, R. J. Clem, R. Ravi, M. B. Kastan, A. Bedi, K. Ueno, and J. M. Hardwick. 1997. Conversion of Bcl-2 to a Bax-like death effector by caspases. Science 278:1966–1968.

    Chou, J. J., H. Li, G. S. Salvesen, J. Yuan, and G. Wagner. 1999. Solution structure of BID, an intracellular amplifier of apoptotic signaling. Cell 96:615–624.

    Chu, F., A. Borthakur, X. Sun, J. Barkinge, R. Gudi, S. Hawkins, and K. V. Prasad. 2004. The Siva-1 putative amphipathic helical region (SAH) is sufficient to bind to BCL-XL and sensitize cells to UV radiation induced apoptosis. Apoptosis 9:83–95.

    Claveria, C., E. Caminero, A. C. Martinez, S. Campuzano, and M. Torres. 2002. GH3, a novel proapoptotic domain in Drosophila Grim, promotes a mitochondrial death pathway. EMBO J. 21:3327–3336.

    Colussi, P. A., L. M. Quinn, D. C. Huang, M. Coombe, S. H. Read, H. Richardson, and S. Kumar. 2000. Debcl, a proapoptotic Bcl-2 homologue, is a component of the Drosophila melanogaster cell death machinery. J. Cell. Biol. 148:703–714.

    Cory, S., and J. M. Adams. 2002. The Bcl2 family: regulators of the cellular life-or-death switch. Nat. Rev. Cancer 2:647–656.

    Cory, S., D. C. Huang, and J. M. Adams. 2003. The Bcl-2 family: roles in cell survival and oncogenesis. Oncogene 22:8590–8607.

    Coultas, L., M. Pellegrini, J. E. Visvader, G. J. Lindeman, L. Chen, J. M. Adams, D. C. Huang, and A. Strasser. 2003. Bfk: a novel weakly proapoptotic member of the Bcl-2 protein family with a BH3 and a BH2 region. Cell Death Differ. 10:185–192.

    Coultas, L., and A. Strasser. 2003. The role of the Bcl-2 protein family in cancer. Semin. Cancer Biol. 13:115–123.

    Daborn, P. J., N. Waterfield, C. P. Silva, C. P. Au, S. Sharma, and R. H. Ffrench-Constant. 2002. A single Photorhabdus gene, makes caterpillars floppy (mcf), allows Escherichia coli to persist within and kill insects. Proc. Natl. Acad. Sci. USA 99:10742–10747.

    Denisov, A. Y., M. S. Madiraju, G. Chen, A. Khadir, P. Beauparlant, G. Attardo, G. C. Shore, and K. Gehring. 2003. Solution structure of human BCL-w: modulation of ligand binding by the C-terminal helix. J. Biol. Chem. 278:21124–21128.

    Deverman, B. E., B. L. Cook, S. R. Manson et al. (12 co-authors). 2002. Bcl-xL deamidation is a critical switch in the regulation of the response to DNA damage. Cell 111:51–62.

    Felsenstein, J. 1996. Inferring phylogenies from protein sequences by parsimony, distance, and likelihood methods. Methods Enzymol. 266:418–427.

    Galtier, N., M. Gouy, and C. Gautier. 1996. SEAVIEW and PHYLO_WIN: two graphic tools for sequence alignment and molecular phylogeny. Comput. Appl. Biosci. 12:543–548.

    Gilbert, W., S. J. de Souza, and M. Long. 1997. Origin of genes. Proc. Natl. Acad. Sci. USA 94:7698–7703.

    Green, D. R., and G. I. Evan. 2002. A matter of life and death. Cancer Cells 1:19–30.

    Guindon, S., and O. Gascuel. 2003. A simple, fast, and accurate algorithm to estimate large phylogenies by maximum likelihood. Syst. Biol. 52:696–704.

    Hatakeyama, S., A. Hamasaki, I. Negishi, D. Y. Loh, F. Sendo, and K. Nakayama. 1998. Multiple gene duplication and expression of mouse bcl-2-related genes, A1. Int. Immunol. 10:631–637.

    Heimlich, G., A. D. McKinnon, K. Bernardo, D. Brdiczka, J. C. Reed, R. Kain, M. Kronke, and J. M. Jurgensmeier. 2004. Bax-induced cytochrome c release from mitochondria depends on alpha-helices-5 and -6. Biochem. J. 378:247–255.

    Hengartner, M. O., and H. R. Horvitz. 1994. Activation of C. elegans cell death protein CED-9 by an amino-acid substitution in a domain conserved in Bcl-2. Nature 369:318–320.

    Herberg, J. A., S. Phillips, S. Beck, T. Jones, D. Sheer, J. J. Wu, V. Prochazka, P. J. Barr, M. C. Kiefer, and J. Trowsdale. 1998. Genomic structure and domain organisation of the human Bak gene. Gene 211:87–94.

    Hokamp, K., A. McLysaght, and K. H. Wolfe. 2003. The 2R hypothesis and the human genome sequence. J. Struct. Funct. Genomics 3:95–110.

    Hoshiyama, D., H. Suga, N. Iwabe, M. Koyanagi, N. Nikoh, K. Kuma, F. Matsuda, T. Honjo, and T. Miyata. 1998. Sponge Pax cDNA related to Pax-2/5/8 and ancient gene duplications in the Pax family. J. Mol. Evol. 47:640–648.

    Huang, D. C., L. A. O'Reilly, A. Strasser, and S. Cory. 1997. The anti-apoptosis function of Bcl-2 can be genetically separated from its inhibitory effect on cell cycle entry. EMBO J. 16:4628–4638.

    Huang, Q., A. M. Petros, H. W. Virgin, S. W. Fesik, and E. T. Olejniczak. 2002. Solution structure of a Bcl-2 homolog from Kaposi sarcoma virus. Proc. Natl. Acad. Sci. USA 99:3428–3433.

    ———. 2003. Solution structure of the BHRF1 protein from Epstein-Barr virus, a homolog of human Bcl-2. J. Mol. Biol. 332:1123–1130.

    Hughes, A. L. 2001. Evolution of the integrin alpha and beta protein families. J. Mol. Evol. 52:63–72.

    Igaki, T., H. Kanuka, N. Inohara, K. Sawamoto, G. Nunez, H. Okano, and M. Miura. 2000. Drob-1, a Drosophila member of the Bcl-2/CED-9 family that promotes cell death. Proc. Natl. Acad. Sci. USA 97:662–667.

    Igaki, T., and M. Miura. 2004. Role of Bcl-2 family members in invertebrates. Biochim. Biophys. Acta 1644:73–81.

    Jaillon, O., J. M. Aury, F. Brunet et al. (61 co-authors). 2004. Genome duplication in the teleost fish Tetraodon nigroviridis reveals the early vertebrate proto-karyotype. Nature 431:946–957.

    Kataoka, T., N. Holler, O. Micheau, F. Martinon, A. Tinel, K. Hofmann, and J. Tschopp. 2001. Bcl-rambo, a novel Bcl-2 homologue that induces apoptosis via its unique C-terminal extension. J. Biol. Chem. 276:19548–19554.

    Ke, N., A. Godzik, and J. C. Reed. 2001. Bcl-B, a novel Bcl-2 family member that differentially binds and regulates Bax and Bak. J. Biol. Chem. 276:12481–12484.

    Kirkin, V., S. Joos, and M. Zornig. 2004. The role of Bcl-2 family members in tumorigenesis. Biochim. Biophys. Acta 1644:229–249.

    Kirsch, D. G., A. Doseff, B. N. Chau, D. S. Lim, N. C. de Souza-Pinto, R. Hansford, M. B. Kastan, Y. A. Lazebnik, and J. M. Hardwick. 1999. Caspase-3-dependent cleavage of Bcl-2 promotes release of cytochrome c. J. Biol. Chem. 274:21155–21161.

    Kolkman, J. A., and W. P. Stemmer. 2001. Directed evolution of proteins by exon shuffling. Nat. Biotechnol. 19:423–428.

    Koonin, E. V., and L. Aravind. 2002. Origin and evolution of eukaryotic apoptosis: the bacterial connection. Cell Death Differ. 9:394–404.

    Kuwana, T., M. R. Mackey, G. Perkins, M. H. Ellisman, M. Latterich, R. Schneiter, D. R. Green, and D. D. Newmeyer. 2002. Bid, Bax, and lipids cooperate to form supramolecular openings in the outer mitochondrial membrane. Cell 111:331–342.

    Kuwana, T., and D. D. Newmeyer. 2003. Bcl-2-family proteins and the role of mitochondria in apoptosis. Curr. Opin. Cell Biol. 15:691–699.

    Lanave, C., M. Santamaria, and C. Saccone. 2004. Comparative genomics: the evolutionary history of the Bcl-2 family. Gene 333:71–79.

    Le Bras, M., K. Bensaad, and T. Soussi. 2003. Data mining the p53 pathway in the Fugu genome: evidence for strong conservation of the apoptotic pathway. Oncogene 22:5082–5090.

    Lee, R., J. Chen, C. P. Matthews, J. K. McDougall, and P. E. Neiman. 2001. Characterization of NR13-related human cell death regulator, Boo/Diva, in normal and cancer tissues. Biochim. Biophys. Acta 1520:187–194.

    Li, H., H. Zhu, C. J. Xu, and J. Yuan. 1998. Cleavage of BID by caspase 8 mediates the mitochondrial damage in the Fas pathway of apoptosis. Cell 94:491–501.

    Liang, Y., K. D. Nylander, C. Yan, and N. F. Schor. 2002. Role of caspase 3-dependent Bcl-2 cleavage in potentiation of apoptosis by Bcl-2. Mol. Pharmacol. 61:142–149.

    Matsushima, M., T. Fujiwara, E. Takahashi, T. Minaguchi, Y. Eguchi, Y. Tsujimoto, K. Suzumori, and Y. Nakamura. 1998. Isolation, mapping, and functional analysis of a novel human cDNA (BNIP3L) encoding a protein homologous to human NIP3. Genes Chromosomes Cancer 21:230–235.

    Matsuyama, S., S. L. Schendel, Z. Xie, and J. C. Reed. 1998. Cytoprotection by Bcl-2 requires the pore-forming alpha5 and alpha6 helices. J. Biol. Chem. 273:30995–31001.

    McDonnell, J. M., D. Fushman, C. L. Milliman, S. J. Korsmeyer, and D. Cowburn. 1999. Solution structure of the proapoptotic molecule BID: a structural basis for apoptotic agonists and antagonists. Cell 96:625–634.

    Muchmore, S. W., M. Sattler, H. Liang et al. (12 co-authors). 1996. X-ray and NMR structure of human Bcl-xL, an inhibitor of programmed cell death. Nature 381:335–341.

    Muller, W. E. 1997. Origin of metazoan adhesion molecules and adhesion receptors as deduced from cDNA analyses in the marine sponge Geodia cydonium: a review. Cell Tissue Res. 289:383–395.

    Mushegian, A. R., J. R. Garey, J. Martin, and L. X. Liu. 1998. Large-scale taxonomic profiling of eukaryotic model organisms: a comparison of orthologous proteins encoded by the human, fly, nematode, and yeast genomes. Genome Res. 8:590–598.

    Opferman, J. T., and S. J. Korsmeyer. 2003. Apoptosis in the development and maintenance of the immune system. Nat. Immunol. 4:410–415.

    Page, R. D. 1996. TreeView: an application to display phylogenetic trees on personal computers. Comput. Appl. Biosci. 12:357–358.

    Petros, A. M., A. Medek, D. G. Nettesheim, D. H. Kim, H. S. Yoon, K. Swift, E. D. Matayoshi, T. Oltersdorf, and S. W. Fesik. 2001. Solution structure of the antiapoptotic protein bcl-2. Proc. Natl. Acad. Sci. USA 98:3012–3017.

    Prince, V. E., and F. B. Pickett. 2002. Splitting pairs: the diverging fates of duplicated genes. Nat. Rev. Genet. 3:827–837.

    Quinn, L., M. Coombe, K. Mills, T. Daish, P. Colussi, S. Kumar, and H. Richardson. 2003. Buffy, a Drosophila Bcl-2 protein, has anti-apoptotic and cell cycle inhibitory functions. EMBO J. 22:3568–3579.

    Ray, R., G. Chen, C. Vande Velde, J. Cizeau, J. H. Park, J. C. Reed, R. D. Gietz, and A. H. Greenberg. 2000. BNIP3 heterodimerizes with Bcl-2/Bcl-X(L) and induces cell death independent of a Bcl-2 homology 3 (BH3) domain at both mitochondrial and nonmitochondrial sites. J. Biol. Chem. 275:1439–1448.

    Reed, J. C. 1997. Double identity for proteins of the Bcl-2 family. Nature 387:773–776.

    ———. 2003. Apoptosis-targeted therapies for cancer. Cancer Cells 3:17–22.

    Reed, J. C., K. S. Doctor, and A. Godzik. 2004. The domains of apoptosis: a genomics perspective. Sci. STKE 2004:re9.

    Russell, H. R., Y. Lee, H. L. Miller, J. Zhao, and P. J. McKinnon. 2002. Murine ovarian development is not affected by inactivation of the bcl-2 family member diva. Mol. Cell. Biol. 22:6866–6870.

    Sattler, M., H. Liang, D. Nettesheim et al. (12 co-authors). 1997. Structure of Bcl-xL-Bak peptide complex: recognition between regulators of apoptosis. Science 275:983–986.

    Schendel, S. L., Z. Xie, M. O. Montal, S. Matsuyama, M. Montal, and J. C. Reed. 1997. Channel formation by antiapoptotic protein Bcl-2. Proc. Natl. Acad. Sci. USA 94:5113–5118.

    Schmitt, E., C. Paquet, M. Beauchemin, J. Dever-Bertrand, and R. Bertrand. 2000. Characterization of Bax-sigma, a cell death-inducing isoform of Bax. Biochem. Biophys. Res. Commun. 270:868–879.

    Shi, B., D. Triebe, S. Kajiji, K. K. Iwata, A. Bruskin, and J. Mahajna. 1999. Identification and characterization of baxepsilon, a novel bax variant missing the BH2 and the transmembrane domains. Biochem. Biophys. Res. Commun. 254:779–785.

    Sorenson, C. M. 2004. Bcl-2 family members and disease. Biochim. Biophys. Acta 1644:169–177.

    Suga, H., K. Kuma, N. Iwabe, N. Nikoh, K. Ono, M. Koyanagi, D. Hoshiyama, and T. Miyata. 1997. Intermittent divergence of the protein tyrosine kinase family during animal evolution. FEBS Lett. 412:540–546.

    Suzuki, M., R. J. Youle, and N. Tjandra. 2000. Structure of Bax: coregulation of dimer formation and intracellular localization. Cell 103:645–654.

    Tan, K. O., K. M. Tan, and V. C. Yu. 1999. A novel BH3-like domain in BID is required for intramolecular interaction and autoinhibition of pro-apoptotic activity. J. Biol. Chem. 274:23687–23690.

    Terajima, D., K. Shida, N. Takada, A. Kasuya, D. Rokhsar, N. Satoh, M. Satake, and H. G. Wang. 2003. Identification of candidate genes encoding the core components of the cell death machinery in the Ciona intestinalis genome. Cell Death Differ. 10:749–753.

    U, M., T. Miyashita, Y. Shikama, K. Tadokoro, and M. Yamada. 2001. Molecular cloning and characterization of six novel isoforms of human Bim, a member of the proapoptotic Bcl-2 family. FEBS Lett. 509:135–141.

    Valentine, J. W., D. H. Erwin, and D. Jablonski. 1996. Developmental evolution of metazoan bodyplans: the fossil evidence. Dev. Biol. 173:373–381.

    Verjovski-Almeida, S., R. DeMarco, E. A. Martins et al. (37 co-authors). 2003. Transcriptome analysis of the acoelomate human parasite Schistosoma mansoni. Nat. Genet. 35:148–157.

    Wiens, M., B. Diehl-Seifert, and W. E. Muller. 2001. Sponge Bcl-2 homologous protein (BHP2-GC) confers distinct stress resistance to human HEK-293 cells. Cell Death Differ. 8:887–898.

    Wiens, M., A. Krasko, C. I. Muller, and W. E. Muller. 2000. Molecular evolution of apoptotic pathways: cloning of key domains from sponges (Bcl-2 homology domains and death domains) and their phylogenetic relationships. J. Mol. Evol. 50:520–531.

    Wiens, M., S. Perovic-Ottstadt, I. M. Muller, and W. E. Muller. 2004. Allograft rejection in the mixed cell reaction system of the demosponge Suberites domuncula is controlled by differential expression of apoptotic genes. Immunogenetics 56:597–610.

    Wolfe, K. H. 2001. Yesterday's polyploids and the mystery of diploidization. Nat. Rev. Genet. 2:333–341.

    Woo, J. S., J. S. Jung, N. C. Ha, J. Shin, K. H. Kim, W. Lee, and B. H. Oh. 2003. Unique structural features of a BCL-2 family protein CED-9 and biophysical characterization of CED-9/EGL-1 interactions. Cell Death Differ. 10:1310–1319.

    Xue, D., and H. R. Horvitz. 1997. Caenorhabditis elegans CED-9 protein is a bifunctional cell-death inhibitor. Nature 390:305–308.

    Yasuda, M., C. D'Sa-Eipper, X. L. Gong, and G. Chinnadurai. 1998a. Regulation of apoptosis by a Caenorhabditis elegans BNIP3 homolog. Oncogene 17:2525–2530.

    Yasuda, M., P. Theodorakis, T. Subramanian, and G. Chinnadurai. 1998b. Adenovirus E1B-19K/BCL-2 interacting protein BNIP3 contains a BH3 domain and a mitochondrial targeting sequence. J. Biol. Chem. 273:12415–12421.

    Yi, P., W. Zhang, Z. Zhai, L. Miao, Y. Wang, and M. Wu. 2003. Bcl-rambo beta, a special splicing variant with an insertion of an Alu-like cassette, promotes etoposide- and taxol-induced cell death. FEBS Lett. 534:61–68.

    Zhang, H., W. Holzgreve, and C. De Geyter. 2001. Bcl2-L-10, a novel anti-apoptotic member of the Bcl-2 family, blocks apoptosis in the mitochondria death pathway but not in the death receptor pathway. Hum. Mol. Genet. 10:2329–2339.

    Zhang, H. M., P. Cheung, B. Yanagawa, B. M. McManus, and D. C. Yang. 2003. BNips: a group of pro-apoptotic proteins in the Bcl-2 family. Apoptosis 8:229–236.

    Zhou, M., S. D. Demo, T. N. McClure, R. Crea, and C. M. Bitler. 1998. A novel splice variant of the cell death-promoting protein BAX. J. Biol. Chem. 273:11930–11936.

    Zhou, Y. T., U. J. Soh, X. Shang, G. R. Guy, and B. C. Low. 2002. The BNIP-2 and Cdc42GAP homology/Sec14p-like domain of BNIP-Salpha is a novel apoptosis-inducing sequence. J. Biol. Chem. 277:7483–7492.(Abdel Aouacheria*, Frédér)

    http://www.100md.com/html/DirDu/2006/10/17/25/96/21.htm