The Domains of Apoptosis: A Genomics Perspective

See allHide authors and affiliations

Science's STKE  29 Jun 2004:
Vol. 2004, Issue 239, pp. re9
DOI: 10.1126/stke.2392004re9


Apoptosis plays important roles in many facets of normal physiology in animal species, including programmed cell death associated with fetal development or metamorphosis, tissue homeostasis, immune cell education, and some aspects of aging. Defects in the regulation of apoptosis contribute to multiple diseases associated with either inappropriate cell loss or pathological cell accumulation. Host-pathogen interactions have additionally provided evolutionary pressure for apoptosis as a defense mechanism against viruses and microbes, sometimes linking apoptosis mechanisms with inflammatory responses. To a large extent, the apoptosis machinery can be viewed as a network, with different nodes connected by physical interactions of evolutionarily conserved domains. These domains can serve as signatures for identification of proteins involved in the network. In particular, the caspase recruitment domains (CARDs); death effector domains (DEDs); death domains (DDs); BIR (baculovirus IAP repeat) domains of inhibitor of apoptosis proteins (IAPs); Bcl-2 family proteins; caspase protease domains; and endonuclease-associated CIDE (cell death–inducing DFF45-like effector) domains are found in common in proteins involved in apoptosis. In the genomes of mammals, genes encoding proteins that carry one or more of these signature domains are often present in multiple copies, making up diverse gene families that permit tissue-specific and highly regulated control of cell life and death decisions through combinations of stimulus-specific gene expression and complex protein interaction networks. In this Review, we organize the repertoire of apoptosis proteins of humans into domain families, drawing comparisons with homologs in other vertebrate and invertebrate animal species, and discuss some of the functional implications of these findings.


Apoptosis is a type of cell death that plays important roles in many facets of normal physiology in animal species (1). Defects in apoptosis regulation are implicated in the pathogenesis of multiple diseases, perhaps explaining why the study of apoptosis has emerged as one of the fastest growing areas of biomedical research in recent years (24). Already, knowledge of apoptosis mechanisms has resulted in new strategies for treating certain illnesses, several of which are currently in clinical testing [reviewed in (57)].

Strictly defined, apoptosis is determined by its morphological features. As viewed with the light (or preferably electron) microscope, the characteristics of the apoptotic cell include chromatin condensation and nuclear fragmentation (pyknosis), plasma membrane blebbing, and cell shrinkage. Eventually, the cell breaks into small, membrane-surrounded fragments (apoptotic bodies), which are cleared by phagocytosis without inciting an inflammatory response. The release of apoptotic bodies inspired the term "apoptosis," from the Greek word meaning "to fall away from," and conjuring notions of the falling of leaves in the autumn from deciduous trees (8). Apoptosis is of recognized importance in a wide variety of both vertebrate and invertebrate animal species. Although programmed cell death with features resembling apoptosis has also been documented in plants, the underlying molecular mechanisms and core cell death machinery of plants remain to be defined, and thus the similarity to animal cell death regulation is currently uncertain [reviewed in (9)]. Also, in mammals and certain other animal species, cell death can be induced by mechanisms that entirely lack the constellation of morphological changes associated with apoptosis, or that retain only portions of the program, and thus that do not strictly qualify to be called apoptosis. Such cell death is particularly evident in certain pathological conditions, such as injury caused by ischemia, but also can occur in developmental contexts as well (10, 11).

Although defined by its morphological features, at the biochemical level, apoptosis can be attributed to the activation in cells of a family of evolutionarily conserved proteases, known as caspases. These proteases are present as inactive zymogens in essentially all animal cells, but can be triggered to assume active states, generally involving their proteolytic processing at conserved aspartic acid (Asp) residues. Upon activation, members of this family of intracellular cysteine proteases cleave their substrates at aspartic acid residues, thus giving rise to the moniker "caspases" for cysteine aspartyl-specific proteases (12). Based on the observations that (i) caspases cleave their substrates at Asp residues and that (ii) caspases are activated by proteolytic processing at Asp residues, it has become clear that these proteases collaborate in proteolytic cascades in which caspases activate themselves and each other (13).

The human genome encodes 11 or 12 caspases (14), depending on certain polymorphisms (15), whereas the mouse genome contains 10 such proteases. Caenorhabditis elegans and Drosophilia encode five and seven caspase-like proteins in their genomes, respectively. The caspases can be subgrouped according to either their protease specificity or their amino acid sequence similarities (1618). However, from a functional perspective, it is probably most germane to view the caspases as either upstream "initiator" caspases or downstream "effector" caspases.

The substrates of effector caspases include protein kinases (in which caspases often separate the autorepressing regulatory domains from catalytic domains) and other signal transduction proteins, cytoskeletal and nuclear matrix proteins, chromatin-modifying enzymes [for example, poly(ADP ribose) polymerase (PARP)], DNA repair proteins, and inhibitory subunits of certain endonucleases [CIDE (cell death–inducing DFF45-like effector) family proteins] that are responsible in part for genome digestion during apoptosis [reviewed in (18)]. These proteolytic cleavage events irreversibly commit the cell to death. Cross-talk between caspases and other protease networks, such as those made up of calpains, can occur in some pathological conditions, creating confusion about upstream, initiating versus downstream, effector mechanisms of cell death [reviewed in (19)].

Apoptosis and Inflammation

Apoptosis is an important defense mechanism against pathogens. For example, cell suicide can provide a mechanism for depriving viruses of host cells for replication, thus limiting viral spread (20). Also, some of the families of proteins involved in apoptosis regulation participate in inflammatory responses to microbial pathogens. For instance, caspase family proteases are critical effectors of the apoptotic program, but some of these proteases also cleave and activate proinflammatory cytokines such as pro-interleukin-1β (pro-IL-1β) and pro-IL-18 (21). Similarly, some proteins involved in caspase activation participate in triggering induction of nuclear factor (NF)-κB family transcription factors, which regulate expression of numerous genes important for inflammatory responses and for innate and acquired immunity (22). NF-κB also regulates the expression of several genes involved in apoptosis control [reviewed in (23)]. Thus, the worlds of apoptosis and inflammation are often closely intertwined.

Apoptosis Domains

Proteins involved in apoptosis form a tightly regulated network, in which most interactions carried out by evolutionarily conserved domains can serve as signatures for identification, including (i) the caspase-associated recruitment domain (CARD); (ii) the death effector domain (DED); (iii) the death domain (DD); (iv) the BIR domain of inhibitor of apoptosis (IAP) proteins; and (v) the Bcl-2 homology (BH) domains of Bcl-2 family proteins. The output of this network of protein interactions determines whether certain effectors of apoptosis become activated in sufficient quantities to commit a cell to death. Most prominent in cell death decisions are (vi) the caspase family cell death proteases mentioned above, but other effectors such as (vii) the CIDE domain–containing apoptotic endonucleases also have roles. Higher organisms such as humans and mice have gene families that encode multiple proteins that contain one or more of these signature domains. Despite the availability of draft genomes, the precise number of copies of some apoptosis-relevant proteins is still in dispute. This confusion results from at least two factors: (i) Genes encoding apoptosis proteins appear to have undergone expansion in recent evolution and are sometimes located at the most variable genomic regions that are the last to be sequenced; and (ii) some apoptosis genes have been diverging rapidly during mammalian evolution, with some of the more distant homologs, such as para- and meta-caspases and the pyrin domain families having been recognized only recently. Thus, in many cases, we simply do not know how many distant homologs might be present in various genomes. This caveat should be considered when interpreting the summary of apoptosis domains presented here.

In general, the larger the genome, the larger the gene family (Table 1). Presumably, the amplification and diversification of genes encoding apoptosis proteins in more complex organisms afford tissue-specific mechanisms for precisely regulating life and death decisions in the cell, whereby certain members of these gene families are expressed at appropiate times in specific cell types during development and differentiation. Complex networks of interactions among various proteins encoded by these gene families, and a plethora of possible posttranslational modifications, create additional levels of control, allowing for exquisite fine-tuning of cell life-span regulation.

In addition to these seven domain families that constitute core components of the apoptosis machinery, several families of proteins containing other types of domains have been implicated either in the regulation of the core apoptotic machinery or in the control of closely linked inflammatory response pathways. These protein families minimally include (i) tumor necrosis factor (TNF) family ligands; (ii) TNF receptors (TNFRs); (iii) TIR (toll and IL-1 receptor) domains; (iv) pyrin domains [also called PAAD, for pyrin, AIM (absent-in-melanoma), ASC (apoptosis-associated speck-like protein containing a CARD [caspase-recruitment domain]) and DD-like]; Pyk; or DAPIN (domain in apoptosis and interferon response); (v) TRAFs (TNF receptor–associated factors); (vi) REL (NF-κB) and IκB family proteins; and (vii) BAG (Bcl-2–associated athanogene) domains. These domain families will be addressed here only superficially. Other genes and their encoded proteins that play critical or contributory roles in apoptosis regulation or execution will be introduced here only briefly. Much of this information about apoptosis domain families has recently been assembled into a database (, providing a convenient site for initial inquiries and links to additional information (24).

Here we focus on the apoptosis gene families of humans, with occasional reference or comparison to mice, Drosophila melanogaster, or C. elegans. First, we summarize basic characteristic and functions of the domain families that create networks of protein interactions that dictate whether or not the cell commits to activation of the cell death program. Then we address the effectors of apoptosis, the caspases and certain endonucleases implicated in apoptotic cell destruction.

Caspase recruitment domains (CARDs)

The CARD is a protein interaction module composed of a bundle of six α helices (25, 26) (Fig. 1). This domain is commonly implicated in regulation of caspases that contain CARDs in their N-terminal prodomains, including human Caspases 1, 2, 4, 5, and 9; mouse Caspases 1, 2, 9, 11, and 12; insect caspase Dronc in Drosophila; and nematode caspase Ced3 in C. elegans (Fig. 2). However, CARDs are also commonly involved in the regulation of NF-κB, coinciding with the close links between apoptosis, inflammation, and host-defense mechanisms. CARDs associate with other CARD domains, forming dimers with complementary surfaces that dictate specificity (Fig. 3). The CARD-encoding genes are spread throughout the genome, with the exception of a cluster of CARDs on the long arm of chromosome 11.

Fig. 1.

The CARD fold. The fold of a prototypical CARD domain showing the six α helices.

Fig. 2.

CARD-containing caspases. The caspase family members containing CARD domains are depicted for human Caspases-1, 2, 4, 5, and 9; mouse Caspase-11 and 12; bovine Caspase-13; worm Ced3; and fly Dronc.

Fig. 3.

CARD surfaces. The three-dimensional (3D) structures of the CARD domain of selected apoptosis proteins are shown in space-filling mode, with basic residues in blue, acidic residues in red, and hydrophobic residues in yellow. Complementary surface patches are thought to account for the selective pairing of specific CARDs (top to bottom). [Figure courtesy of F. Pio]

CARD-containing proteins are remarkably diverse in terms of the domains that accompany the CARD and their arrangement. Humans have 22 CARD-encoding genes (Fig. 4), in addition to the six CARD-containing caspases. Mice have at least 19 CARD-encoding genes, in addition to five genes encoding CARD-containing caspases. Drosophila melanogaster has two and C. elegans has one, in addition to their genes encoding CARD-containing caspases (Table 1). These CARD family proteins can be simple in their structural arrangement, consisting essentially only of a CARD [for example, Iceberg or COP (CARD-only protein; also called Pseudo-ICE)], or very complex, where the CARD is combined with as many as four additional types of domains, including nucleotide-binding NACHT or NB-ARC (nucleotide-binding Apaf-1, R-gene, and Ced4) domains, pyrin domains, leucine-rich repeats (LRRs), WD repeats, SH3 (Src-homology 3), PDZ, membrane-associated guanylate kinase (MAGUK), BIR, Ring, kinase, and death domains(DDs) (Fig. 4).

Fig. 4.

Human CARD family proteins. The domain architectures of the human CARD family proteins are shown, excluding CARD-containing caspases.

In general, the mechanisms by which CARDs participate in activation of caspases or NF-κB involve assembly of multiprotein complexes. The complexes can facilitate dimerization or serve as a scaffold on which proteinases or protein kinases are brought together and activated through the so-called induced proximity mechanism (27). Some complexes assemble spontaneously, but others require endogenous or exogenous inducers. For example, caspase activation resulting from mitochondrial release of cytochrome c (Cyt-c) into the cytosol results in the assembly of a multiprotein caspase-activating complex, referred to as the apoptosome [reviewed in (2830)]. The central component of the apoptosome is Apaf1, a caspase-activating protein that oligomerizes into a hepatmer through its nucleotide-binding NB-ARC domain after it binds Cyt-c through a series of WD repeat domains. The CARD of Apaf1 binds a complementary CARD in the prodomain of pro-Caspase-9, in a 1:1 stoichiometric arrangement (30) (Fig. 5), thus triggering Caspase-9 activation by the induced-proximity mechanism. An analogous system exists in Drosophila, involving the Apaf1 homolog Dark (also known as Hac1 and Dapaf1) and the Caspase-9 homolog Dronc (31). In contrast, the Apaf1 homolog of C. elegans, Ced4, lacks the WD repeats seen in the human and fly proteins and is constitutively active, not requiring a stimulatory ligand such as cytochrome c (Fig. 6). Instead, worms rely on binding of Ced4 by an antiapoptotic protein called Ced9 to prevent spontaneous initiation of the death program (32, 33).

Fig. 5.

Apaf1 family protease-activating proteins. The domain topologies of homologous caspase-activating proteins from human (bottom), fly (middle), and worm (top) proteins are depicted. Note that human Apaf1 and fly Dark proteins contain WD repeats at their C-termini, whereas worm Ced4 does not.

Fig. 6.

Caspase activation by induced proximity. Mechanism of Ced4-mediated activation of the caspase Ced3. The nucleotide-binding (NB) domains of Ced4 are thought to oligomerize in a ATP- or deoxyadenosine triphosphate (dATP)–dependent manner. For worm Ced4, no additional stimulus is required, whereas human protein Apaf1 requires cytochrome c and the fly protein Dark require mitochondrial-derived cofactors (not shown). The oligomerized scaffold recruits pro-Ced3 through CARD-CARD interactions. This brings the protease zymogens into close proximity and promotes caspase activation and release of activated Ced3 protease consisting of a heterotetramer with two large (L) and two small subunits (S). In the case of Apaf1, the CARD-containing prodomain is not cleaved from the protein and the activated protease remains bound to oligomerized Apaf1, forming a holoenzyme complex called the "apoptosome" (not shown).

The CARD-protein Nod1 uses a mechanism analagous to that of Apaf1 to induce activation of NF-κB . In response to ligands displayed on intracellular bacteria that bind the LRRs in Nod1, the Nod1 protein is thought to oligomerize through its central nucleotide-binding NACHT domain. Nod1’s CARD binds to an adapter protein RIP2 (receptor-interacting protein 2, also called Cardiak), which in turn binds the IκB kinase γ (IKKγ) subunit of the IKK complex, (34, 35) (Fig. 7). Presumably, bringing multiple IKK complexes into close proximity on the scaffold created by Nod1 results in kinase activation and consequent phosphorylation of the NF-κB inhibitory protein IκBα, which is thus targeted for ubiquitination and proteasome-dependent destruction. RIP2 also can bind the CARD of pro-Caspase-1 in addition to IKKγ (36), possibly explaining why Nod1 can induce both cleavage of pro-IL-1β and activation of NF-κB (37). Several CARD proteins can operate as trans-dominant inhibitors of CARD family proteins such as Apaf1 and Nod1, interfering with assembly of these proteins into complexes, and thereby creating a network of checks and balances that prevent unscheduled activation of caspases or NF-κB.

Fig. 7.

Model for IKK activation by Nod1 and Nod2 (35). Ligands derived from microbes trigger the Nod1 and Nod2 proteins to oligomerize via their nucleotide-binding NACHT domains. The CARDs of Nod1 and Nod2 bind the CARD of RIP2, which in turn binds to the IKKγ subunit of the IKK complex. Presumably, clustering of IKK complexes promotes kinase activation through intersubunit phosphorylation, which results in phosphorylation of IkBα on serines that allow it to be recognized by an E3 complex. Proteasome-dependent destruction of IkBα releases p50-p65 NF-κB for entry into the nucleus. Many aspects of the proposed model remain speculative.

The human CARD-containing proteins (excluding caspases) include (names are followed by synonyms in parentheses) Apaf1; Arc (CARD2, Myp); ASC (apoptosis-associated speck-like protein containing a CARD) (TMS-1; PyCARD); Bcl-10 (CIPER, CARMEN, mE10, cE10, CLAP); Bimp1 (Bcl10-interacting MAGUK protein–1) (CARD10, Carma3); Bimp2 (CARD14, Carma2); Bimp3 (CARD11; Carma1); Cardiak (RIP2, RICK); CARD6; CARD9; CARP (Caspases 8– and 10–associated RING proteins); CLAN (CARD, LRR, and NACHT) (Ipaf; CARD12); COP (Pseudo-ICE); cIAP1 (HIAP1, MIHB); cIAP2 (HIAP2, MIHC); Helicard (Mda5); Iceberg; Nac (nucleotide binding domain and CARD) (NALP1, DEFCAP); Nod1 (nucleotide-binding oligomerization domain1) (CARD4); Nod2 (CARD15); RAIDD (RIP-associated ICH-1 or Ced3-homologous protein with death domain) (CRADD); and TUCAN (tumor-up-regulated CARD-containing antagonist of Caspase-9) (Cardinal, CARD8, NDPP, Dakar) (Table 2). CD-CIITA, an alternative splice variant of CIITA (class II transactivator), has a domain with weak similarity to members of the combined CARD-DED-DD structural superfamily, as determined with FFAS (Fold and Function Alignment System). It has been interpreted to be a CARD domain (38). Among these CARD-containing proteins, ASC, Cardiak, CLAN, COP, Iceberg, NAC, Nod1, and Nod2 directly or indirectly associate with and regulate the activation of pro-Caspase-1 in either a positive (ASC; Cardiak; CLAN; NAC; Nod1; Nod2) or negative (ASC; COP; Iceberg) manner (36, 37, 3948). RAIDD binds pro-Caspase-2 through its CARD, and facilitates activation of this protease (49, 50). NAC (NALP1) associates via its CARD with pro-Caspase-5 (48). Apaf1 and TUCAN bind pro-Caspase-9, inducing or blocking its activation, respectively (51, 52). NF-κB-regulating activity has been reported for ASC, Bcl10, Bimp1, Bimp2, Bimp3, Cardiak, CARD6, cIAP1, cIAP2, Nod1, Nod2, and TUCAN (5362). Some of these CARD-carrying proteins have been reported to control both caspases and NF-κB.

Within mammals, some interesting differences in CARD family genes are found. For example, the mouse genome appears to lack genes encoding the Caspase-1 antagonists COP (Pseudo-ICE) and Iceberg (43, 44, 46), implying that regulation of Caspase-1–activation networks may be more complex in humans than in mice. Mice also appear to lack CARD 9, whose function is unknown, and TUCAN (Cardinal, CARD8, NDPP, Dakar), a CARD-carrying protein reported to either suppress pro-Caspase-9 activation (52) or to modulate activation of NF-κB through interactions with IKKγ (NEMO) (61). Gene ablation studies in mice have been accomplished for Apaf1, Bcl10, Cardiak, (6365), Nod1, and Nod2, leaving much work still undone in terms of delineating the in vivo roles of these CARD-carrying genes. Flies and worms contain apparent orthologs only of Apaf1 and Caspase-9 . Genetic analysis of the fly and worm versions of Apaf1 and Caspase-9 has demonstrated critical roles for these genes in programmed cell death in these lower organisms [reviewed in (31, 66)].

Some CARD-containing proteins contain other protein interaction domains that allow them to interact with members of other large families of proteins. For example, the bipartite adapter protein, ASC, contains a CARD in combination with a pyrin domain (67). The CARD domain of ASC binds the CARD of pro-Caspase-1 and selected other CARD family members, and the pyrin domain binds several other proteins that contain complementary pyrin domains. Thus, ASC acts as a molecular bridge or adapter between members of the CARD and pyrin domain families [reviewed in (14, 39)].

The pyrin domain comprises a bundle of five or six α helices (6874). It represents another branch of the death domain superfamily (Fig. 8), along with the DDs, DEDs, and CARDs. Pyrin domains function as homotypic protein interaction motifs, mediating interactions among proteins involved in the activation of inflammatory caspases (for example, Caspase-1) and NF-κB (39, 40, 54, 7577). Although the structural basis for interaction of pairs of pyrin domains remains to be formally demonstrated, molecular modeling studies suggest a prominent role for electrostatic interactions, involving acidic residues in helices α1 and α4, complementing basic residues in helices α2 and α3 (73). Some of the pyrin family proteins known to bind ASC activate pro-Caspase-1 and have topologies highly similar to those of Nod1, Nod2, and CLAN described above, with NACHT and LRR domains but substituting a pyrin domain for the CARDs (Fig. 9). These proteins presumably form oligomeric assemblies through the NACHT domains, using their pyrin domains to link indirectly to pro-Caspase-1 through ASC (Fig. 10). Hereditary mutations in some genes encoding pyrin family proteins have been implicated in autoimmune and hyperinflammatory syndromes, further supporting a role for these proteins in regulating inflammatory responses (78, 79).

Fig. 8.

Pyrin domain structure. The structures of the pyrin domains of ASC (A and C) and NAC (protein with NB domain and CARD) (B and D) are compared in ribbon (A and B) or space-filling (C and D) modes. ASC has six α helices with an unstructured loop in the location where α3 of other death domain folds is found (73), whereas the pyrin domain of NAC has five α helices (74).

Fig. 9.

Pyrin family proteins. Domain topologies are shown for proteins containing pyrin domains. Unless otherwise specified, the protein names used refer to molecules of human origin. MNDA, myeloid cell nuclear differentiation antigen.

Fig. 10.

Caspase-1 activation by PAN family proteins. Proposed model for activation of pro-Caspase-1 by PAN family proteins that interact with bipartite adapter ASC. In response to unidentified ligands (probably derived from pathogens), PANs oligomerize through their nucleotide-binding (NB) NACHT domains. The pyrin domain of oligomerized PANs binds the pyrin domain of ASC. The CARD of ASC binds the CARD of pro-Caspase-1. Clustering of pro-Caspase-1 molecules results in protease activation. Active Caspase-1 cleaves proinflammatory cytokine pro-IL-1β, resulting in secretion of active IL-1β. LRR, leucine-rich repeats.

In humans, 19 candidate pyrin family genes have been identified, including ASC (PyCARD), NAC (NALP1; PyPAF1; DEFCAP), Cryopyrin (PYPAF1; NALP3), pyrin (Marenostrin), AIM2 (absent in melanoma 2), IFI16 (interferon-inducible gene 16), POP1 (PAAD-only protein-1) (ASC2), POP2, and PAN1 through PAN11 (PAAD and NACHT-containing proteins; also known as NALPs, PYPAFs, and CATERPILLER proteins) (14, 39, 80). In some cases, the pyrin domains are present as the only motifs within these proteins (as in POP1 and POP2), but in most instances, the pyrin domain is combined with various other domains, including CARDs (ASC; NAC), nucleotide-binding NACHT domains (NAC; Cyropyrin; PAN1 through PAN11), LRRs (NAC; Cyropyrin; PAN1 through PAN11), HIN-200 (hemopoietic interferon-inducible nuclear protein 200–amino acid motif), B-Box zinc fingers, or SPRY (sprouty) domains (68). Mice appear to contain substantially fewer pyrin family genes (about 12). Drosophila and C. elegans lack PAAD/PYRIN domains, suggesting that this domain may have evolved recently. The presence of pyrin domains in zebra fish and fugu fish, however, indicates that other vertebrates beside mammals have acquired this domain (81). In fact, caspases exist in fish that contain a pyrin domain within the N-terminal prodomain (68). Thus, according to the "rosetta stone" hypothesis (which holds that two polypeptides from one organism are likely to interact if their homologs are expressed as a single polypeptide in some other organism), pyrin family proteins are expected to be involved in caspase activation. Members of the pox family of animal viruses have acquired pyrin domain–encoding sequences, suggesting a role in ablating host immune responses (68, 81).

Death effector domains (DEDs)

The DED is another homotypic protein interaction module similar to the CARD, also made up of a six–α-helical bundle (82) (Fig. 11). Dimerization of DEDs, like that of other members of the death domain superfamily, is mediated primarily by electrostatic interactions. This typically involves surface residues on α1 and α4 associating with complementary charged residues on α2 and α3 (26). DED-containing proteins have been implicated in apoptosis regulation through interactions with DED-containing caspases (Caspases 8 and 10 in human; Caspase-8 in mouse; Dredd in Drosophila).

Fig. 11.

DED structure. The 3D structure of the DED of FADD is presented in (A) ribbon and (B) space-filling modes (82). Colored surface residues correspond to acidic (red) and basic (blue) residues.

cDNAs representative of 11 DED family genes were identified in humans (not counting DED-containing caspases) (Fig. 12), and orthologous sequences are found for 10 of these in mice (Tables 1 and 3). In addition to the caspase, Dredd, Drosophila contains two proteins proposed to have a DED-like domain: dFADD, representing an apparent ortholog of mammalian death domain–containing adapter protein FADD, and the caspase Dredd (83).

Fig. 12.

Human DED family proteins. The human DED family proteins are depicted, showing their domain architectures. The activity of the proteins is shown at right, indicating whether the proteins induce or block apoptosis. Bap31, BAR, HIP, and HIPPI (Hip-1 protein interactor) contain "variant" DEDs, representing a distinct branch of the DED family. Dap3 and FLASH also contain nonclassical DED-like (DED-L) domains. NLS, nuclear localization sequence; NB, nucleotide binding (GTP-binding) domain; TM, transmembrane domain; Talin-L, Talin-like domain; MLD, myosin-like domain.

The DED can be found either in isolation or in combination with other protein domains. A tandem pair of DEDs is found within the prodomain region of pro-Caspases 8 and 10 of humans, thus providing the basis for a network of DED-DED interactions that either suppress or promote activation of these proteases involved in apoptosis. Among the domains associated with DED-containing proteins in humans are (i) caspase catalytic domains (in pro-Caspases 8 and 10); (ii) death domains (in FADD); (iii) nuclear localization sequences (NLSs) [in DEDD (83a) and DEDD2 (83b)]; (iv) transmembrane (TM) domains (Bap31; Bar); (v) guanosine 5­′-triphosphate or guanosine 5­′-diphosphate nucleotide-binding (NB) domains (in Dap3); (vi) coiled-coil (CC) domains (in Hip and Hippi); (vii) SAM domains (in Bar); (viii) E2-binding RING domains (in Bar); (ix) Talin-like domains (in Hip); and (x) a myosin-like domain (MLD) (in Hippi).

The mechanism for activating DED-containing caspases was revealed in studies of apoptosis induction by the tumor necrosis factor (TNF) family receptor Fas (also called APO1 or CD95). Ligation of this receptor at the cell surface results in receptor clustering and recruitment to the cytosolic domain of DED-containing pro-caspases, forming a "death-inducing signaling complex" (DISC) that triggers caspase activation and leads to apoptosis [reviewed in (84, 85)]. The DED in pro-Caspases 8 and 10 bind a complementary DED in FADD, a bipartite adapter protein containing a DED and a structurally related domain that links FADD directly or indirectly to the cytosolic tails of certain TNF family death receptors. Cells derived from Caspase-8 knockout mice fail to undergo apoptosis in response to ligands or antibodies that activate TNF family death receptors, demonstrating an essential role for this caspase in this pathway (86). It is important to recognize, however, that mice lack the highly similar protease, Caspase-10, which is found in humans, having arisen from an apparent gene duplication on chromosome 2 (14). Thus, both Caspases 8 and 10 would potentially need to be ablated to achieve the same effect in human cells (87).

Hints that additional multiprotein complexes may create "DISCs" that cause activation of Caspase-8 have recently surfaced. For example, the protein Hip (Huntingtin-interacting protein) was discovered through its interaction with Huntingtin, the causative protein of Huntington’s disease, in which an expanded poly-glutamine (polyQ) track in Huntingtin is associated with neurodegeneration of basal ganglia structures (88). Hip was reported to contain a variant DED, similar to domains seen in the Hip-interacting protein Hippi and in the endoplasmic reticulum (ER)–associated transmembrane proteins BAR (bifunctional apoptosis regulator) and Bap31 (B cell receptor–associated protein of 31 kD) (8992). However, our analysis suggests that the DED-like domains in Hip, Hippi, and other proteins are simply coiled-coil regions, and thus we refer to them as variant DEDs. Through interactions of these domains, Hip and Hippi appear to form a complex that can trigger Caspase-8 activation. Wild-type Huntingtin may restrain this death mechanism through its interactions with Hip (91). PolyQ expansions in Huntingtin are thought to loosen its grip on Hip, thus favoring apoptosis.

Multiple DED-containing modulators of apoptosis have been identified, some of which enhance and others of which inhibit Caspase-8 activation [reviewed in (4)]. Among these, the DED-containing protein FLIP (also known as Flame, CASH, Clarp, MRIT, Casper, I-Flice, or Usurpin) has emerged as an important modulator of Caspase-8 activation and apoptosis. FLIP shares extensive amino acid sequence similarity with pro-Caspases 8 and 10, containing two N-terminal DEDs followed by a pseudo-caspase domain that lacks critical residues required for protease activity, including the catalytic cysteine [reviewed in (93, 94)]. In humans, the FLIP gene is located on chromosome 2, in a tandem array with the CASPASE-8 and CASPASE-10 genes, indicative of gene duplication. In most cases, FLIP association with pro-Caspase-8 squelches death receptor signaling by competing with pro-Caspases 8 and 10 for binding to FADD. Occasionally, however, FLIP may enhance Caspase-8 activation, acting like a chaperone to favor conformational changes that create a catalytically active protease (95). Several animal viruses harbor genes encoding FLIP homologs. These viral proteins typically contain only the DEDs, and act as suppressors of Caspase-8 activation by TNF family receptors. Presumably this would help allow virus-infected cells to survive attack by Fas ligand (FasL)–expressing cytotoxic T lymphocytes (CTLs) [reviewed in (94)].

Knockout mice lacking DED-encoding genes for caspase-8, fadd, or flip have been described (86, 96, 97). Mice lacking either caspase-8 or flip die during fetal development because of cardiac problems, suggesting a critical role of these proteins in organogenesis of the heart.

Death domains (DDs)

The DD is another protein interaction module belonging to the same superfamily that includes the CARDs and DEDs (26) (Fig. 13). DDs bind each other, forming oligomers of unknown stoichiometry (possibly hexamers). Specificity for partner selection among DDs is dictated by differences in surface residues. The DD is commonly implicated in activation of NF-κB or caspases and typically involves interactions with members of the TNF family of cytokine receptors (98).

Fig. 13.

Structure of DD. The 3D structure of the DD of FAS is presented in (A) ribbon and (B) space-filling modes (274). Colored surface residues in (B) correspond to acidic (red) and basic (blue) residues.

Mammals have numerous and diverse DD-containing proteins, with 33 DD-encoding genes recognized in humans and mice (Tables 1 and 4). Proteins have been identified in mammals that combine the DD with a diversity of other types of domains, including Ankryin repeats, CARDs, caspase-like folds (para-caspase), DEDs, immunoglobulin (Ig)–like folds, kinase domains, leucine-zippers, LRRs, Rel-homology domains (RHDs), TIR domains, Thm (thrombospondin type 1) repeats, TNFR extracellular domains, and ZU5 domains (Fig. 14). In addition, an apoptosis-inducing protein that binds certain DD family members has been described in Chlamydia species, obligate intracellular bacteria that induce apoptosis of host cells (99). This protein has regions of predicted α-helical structure, but its overall similarity to DDs remains to be defined.

Fig. 14.

DD family proteins. The DD family proteins of humans showing their domain architectures. EC, extracellular cysteine-rich domain; A, ankyrin domain; repeated domains are shown in parentheses; LRR, leucine-rich repeat; PRO/SER, proline/serine-rich region; Ig, immunoglobulin-like fold; TM, transmembrane domain; ZU5 domain; LZ, leucine zipper; Thm, Thrombospondin type-1 domain. Figure not drawn to scale.

Eight of the ~30 known members of the TNF receptor family in humans contain a DD in their cytosolic tail (100). The TNF receptors contain a conserved cysteine-rich domain (CRD) present in one to four copies, typically preceded by a hydrophobic leader peptide sequence and followed by a transmembrane (TM) domain, typical of type I transmembrane proteins that are sorted to the cell surface (100). Several of these DD-containing TNF family receptors use caspase activation as a signaling mechanism, including TNFR1 (CD120a); Fas (APO1, CD95); DR3 (Apo2, Weasle); DR4 (TRAIL-R1); DR5 (TRAIL-R2); and DR6. The p75 nerve growth factor receptor (p75-NGFR) also contains a modified ("type II") DD (101), and induces apoptosis under some circumstances, but its links to caspases are tenuous [reviewed in (102)]. Mice are similar to humans with respect to their TNF family death receptors, but lack the gene encoding TRAIL-R2 (DR5). The tandem arrangement of the genes encoding TRAILl Receptors, DR4 and DR5 on human chromosome 8p21 suggests that the extra death receptor found in humans may have arisen by gene duplication during mammalian speciation (Tables 1 and 4).

The TNF family receptors TNFR1, DR3, and DR6 bind an adapter protein TRADD, through its DD (103106). The DD of TRADD binds certain other DD-containing proteins, including the adapter protein FADD. As mentioned above, the FADD (Mort1) protein contains two protein interaction modules: a DD and a DED. Moreover, FADD is the only protein in the human and mouse genomes to carry both a DD and a DED, serving as a critical bridge between these two large families of domains. FADD links TNF family death receptors to caspases, using its DD to bind TRADD or to interact directly with the cytosolic DD of the TNFR family member Fas, and binding through its DED to DED-containing caspases (Fig. 15). FADD is present within the receptor complexes of all known DD-containing members of the TNF family, except p75-NGFR. Thus, this protein plays a central role in linking caspases to TNF family death receptors, as confirmed in gene ablation studies in mice (97, 107109).

Fig. 15.

Linking TNF family death receptors to DED family caspases through the bipartite adapter FADD. The DD of FADD binds either directly to the DD in the cytosolic tails of TNF family receptors, or indirectly through other DD-containing adapter proteins (not shown). The DED of FADD binds the DEDs of pro-Caspases 8 and 10.

RAIDD (Cradd) is the only protein in the human and mouse genomes that contains both a DD and a CARD. This protein reportedly binds the DD of TRADD and the CARD of pro-Caspase-2, and thus may serve as an adapter protein for linking pro-Caspase-2 to TNF family receptors (49, 50). There is scant evidence to support an important role for Caspase-2 in signaling by TNF family death receptors; however, other roles for Raidd may await discovery.

In addition to the DD-containing TNF family, receptors and adapters such as FADD and RAIDD have also been implicated in apoptosis. DAP (death-associated protein) kinase, for example, reportedly modulates apoptosis induction by TNF family death receptors through unclear mechanisms [reviewed in (110)]. DAP kinase contains a DD, several ankryin-repeats, and a serine kinase domain that is highly similar to domains found in a family of kinases implicated indirectly in apoptosis that include Zip kinase (ZipK), DAPK2 (DRP1), DRAK1 (DAP kinase-related apoptosis-inducing protein kinase), and DRAK2 (110). The DAP kinase (DAPK) has been implicated in suppression of metastasis. Several cytoskeleton-associated ankryin family proteins contain DDs, but their relevance to apoptosis remains uncertain. However, Caspase-8 activation is triggered when adherent epithelial cells are released from the substrate and cultured in suspension (111, 112), an event that disrupts the cytoskeleton. Thus, Caspase-8 might function in the phenomenon of "anoikis" (that is, apoptosis induced by depriving cells of integrin-mediated attachments to the extracellular matrix). Avoidance of anoikis represents an important aspect of tumor invasion, metastasis, and angiogenesis [reviewed in (113)]. It is also fundamental to correct positioning of cells during development.

DDs are not always involved in caspase activation, and some DDs indirectly suppress apoptosis through effects on NF-κB, a family of transcription factors that have important roles in host defense and cell survival [reviewed in (22, 114)]. For example, the RIP protein binds TRADD and activates kinases that induce degradation of IκB, thus releasing NF-κB so that it can translocate to the nucleus and function as a transcription factor (115117). Several antiapoptotic genes are among those induced by NF-κB, including the Caspase-8 antagonist FLIP, as well as certain members of the IAP family and Bcl-2 family (118123). The dual function of TRADD, as a partner for both caspase-activator FADD and NF-κB activator RIP, causes many of the TNF family receptors to counteract their own apoptosis-inducing activity. Thus, TNFR1, DR3, and DR6 are uncertain inducers of apoptosis unless NF-κB induction is inhibited, in which case they typically elicit robust apoptotic responses (124, 125). In contrast, Fas and the Trail receptors DR4 and DR5 only rarely activate NF-κB, probably because these receptor complexes contain FADD but not TRADD (108, 109, 126, 127).

Close ties of DD proteins to NF-κB regulation are also implied by the presence of two genes in humans and mice, NFkB-1 and NFkB-2, encoding proteins with both a DD and an RHD. The RHD is an evolutionarily conserved domain spanning ~300 amino acids that is found in all NF-κB family transcription factors and that is involved in DNA binding and dimerization (114, 128). The NF-κB family transcription factors comprise homo- and heterodimeric pairs of Rel family proteins. Although regulation of these transcription factors is complex, in general, their activity is controlled by a counteracting family of suppressors, the IκB family, that sequesters these transcription factors in the cytosol. IκB family proteins contain conserved Ankyrin-repeat structures, which bind RHDs. Release of NF-κB typically entails degradation of IκB family proteins, which results from a mechanism involving phosphorylation by IκB-kinases (IKKs or IκB-Ks), followed by ubiquitin-dependent proteasome-mediated destruction. The DD-containing NF-κB1 and NF-κB2 proteins, however, appear to participate in an alternative activation pathway in which these proteins are phosphorylated and then digested from p105 and p100 precursors to produce p50 and p52 polypeptides that retain the RHD but lack the DD and other structures that retain them in the cytosol (114).

DD-containing proteins also have links to innate immunity, communicating with Toll family receptors (TLRs) through bipartite adapter proteins such as MyD88 (129). The MyD88 protein is the only protein encoded in the human and mouse genomes that contains both a DD and a TIR domain. The TIR domain represents a ~130–amino acid fold related to flavodoxin, consisting of five-stranded parallel β sheets surrounded by two layers of parallel α helices (130). This domain is found uniformly in the cytosolic (intracellular) tails of TLRs, as well as in some plant proteins involved in host-pathogen responses (131). In mammals, MyD88 connects TLRs to the IL-1 receptor-associated kinases (IRAKs), kinases that contain a DD and a serine-kinase domain (IRAK1, IRAK2, IRAK-M, and IRAK4). The IRAKs then confer signals that result in activation of NF-κB and of stress-related kinases [for example, c-Jun N-terminal kinase (JNK)] (132). Several elements of this signaling network are conserved in Drosophila [reviewed in (133)], but not in C. elegans. The DD-containing protein, para-caspase (also called MALT), is also conserved in mammals and flies, and participates in a NF-κB–signaling pathway relevant to host-pathogen responses. Gene ablation studies in mice indicate a nonredundent role for para-caspase in linking T and B cell antigen receptors to activaton of NF-κB in lymphocytes and a role in B cell responses to LPS (lipopolysaccharide) (134135).

In addition to apoptosis and inflammation, DDs may also participate in other aspects of biology. For example, DDs are found in the cytosolic tails of UNC5 family proteins, a family of putative Netrin receptors implicated in migration of neural progenitor cells and in axon patterning (136138). UNC5 was first discovered in C. elegans, where it participates in axon guidance. Mice and humans contain five UNC5 homologs: UNC5H1a, UNC5H1b, UNC5H2, UNC5H3, and UNC5H4. These proteins typically contain extracellular domains comprising an Ig-like fold and tandem Thm domains (Fig. 14). Drosophila has only one UNC5-like gene.

Knockouts of genes encoding DD-containing proteins in mice have been accomplished thus far for tnfr1, fadd, trail-receptor 2 (DR5), irak1, irak2, irak4, nfκb1, nfκb2, and para-Caspase. Hereditary loss-of-function mutations in Fas (CD95) have also been identified in mice and humans, providing insights into the important role of this DD protein in lymphocyte homeostasis [reviewed in (100)].

BIR domains of inhibitor of apoptosis proteins

A zinc-binding fold, termed the BIR (baculovirus IAP repeat) domain (139), is present in at least one copy in the inhibitor of apoptosis proteins (IAPs), a family of evolutionarily conserved apoptosis suppressors (140). Some IAPs contain as many as three BIR domains in a tandem array (Fig. 16). Unlike the DD-DED-CARD-PRYIN superfamily of domains, which function as homotypic interaction motifs, BIR domains are involved in protein-protein interactions, but they interact principally with other types of proteins, most prominently the caspases and a diverse family of IAP antagonists described below.

Fig. 16.

The BIR domain. The 3D structure of the BIR2 domain of XIAP is presented in ribbon diagram mode, showing the zinc atom that is coordinated by C200, C203, C227, and H220 (275).

The human genome has eight genes that encode BIR-containing proteins, including NAIP, C-IAP1, C-IAP2, XIAP, SURVIVIN, APOLLON, ML-IAP (also called Livin or K-IAP), and ILP2 (also called TsIAP). The mouse genome encodes orthologs of seven of these BIR family members, apparently lacking ILP2 (Table 5). Human naip resides in the SMA (spinal muscular atrophy) locus on chromosome 5. The locus is a complex patchwork of recently (in primates) duplicated genomic segments (141). Two incomplete duplications of naip are present along with two complete and indistinguishable protein and mRNA sequences. Murine naip-related genes are present in multiple tandem copies on chromosome 13 (142, 143). Mice appear to express at least three versions of the NAIP protein from distinct genes located in this genetic locus, which has been associated with strain-dependent differences in responses to Legionella. The genomes of Drosophila and C. elegans encode four and two BIR-containing proteins, respectively. BIR domains can be found in association with several other types of protein domains, including CARD, NACHT, LRR, RING, and ubiquitin-conjugating enzyme (Ubc) domains (Fig. 17).

Fig. 17.

BIR family proteins of humans. The domain architectures of the human BIR family proteins are shown. The sequence of ILP2 (TsIAP) is most similar to the BIR3 domains of XIAP, cIAP1, and cIAP2, as signified by the alignment of domain in the figure. Ubc, ubiquitin-conjugating enzyme; LRR, leucine-rich repeats.

Several IAPs directly bind to and suppress activity of caspases (140), thus serving as endogenous antagonists of the cell death proteases. In humans, XIAP, cIAP1, cIAP2, and ILP2 suppress activity of certain caspases (144146), but evidence that other BIR-containing proteins are direct regulators of caspases is less compelling. The Drosophila proteins DIAP1 and DIAP2 are also caspase suppressors [reviewed in (31)]. Some IAPs use specific BIRs to inhibit particular caspases. XIAP, for example, binds the downstream effector proteases Caspase-3 and Caspase-7 through its BIR2 domain (requiring both BIR2 and portions of a flanking segment of the protein located between BIR1 and BIR2), whereas it binds the upstream initiator protease, Caspase-9 through its BIR3 domain (147149). IAPs are selective caspase inhibitors, inasmuch as they lack activity against many members of the caspase family of cell death proteases. Thus, overexpression of IAPs may block some apoptosis pathways but not others. This stands in marked contrast to the properties of the baculovirus p35 protein, an apoptosis suppressor that displays broad activity against caspases, but for which no cellular homolog has been identified (150).

Not all BIR-containing proteins are suppressors of apoptosis. For example, C. elegans contains two BIR-encoding genes (bir1, bir2) that are similar to proteins in budding and fission yeast. These BIR-containing proteins seem to regulate chromosome segregation and cytokinesis during mitosis or meiosis, rather than apoptosis (151, 152). The BIR-containing protein known as Survivin has a similar function in regulating cell division in mammals [reviewed in (153)]. Indeed, the Survivin protein is associated with chromosomes during mitosis and with the mitotic spindle apparatus and midbody during anaphase, where it plays a critical role in ensuring proper chromosome segregation and in cytokinesis. Suppression of the gene encoding Survivin by antisense methods or interfering with Survivin protein function with dominant-inhibitory mutants results in polyploidy, aneuploidy, and apoptosis (154). Although the promoter of the SURVIVIN gene is normally cell cycle–regulated and activated specifically in late G- to M-phase of the cell cycle (155), many cancers continuously express this protein in a deregulated fashion. The Survivin protein in many tumors is also mislocalized, found throughout the cell interior, including in the cytosol, where it blocks apoptosis. Survivin appears to bind the zymogen-form of Caspase-9 in combination with another protein, HBXIP (hepatitis B X-interacting protein), and to suppress the activation of this caspase in a mitochondrial pathway leading to apoptosis (see below) (156). Many details about the mechanisms by which Survivin interferes with apoptosis remain unresolved, however, and the relative importance of its antiapoptotic action in the division of normal (nontransformed) cells remains to be defined.

In addition to suppressing activity of caspases, some BIR family proteins have other biochemical functions that remain only partly resolved. For example, of the eight human BIR family proteins, four contain RING domains (Fig. 17). The RINGs of IAPs are implicated in interactions with the cellular components of the ubiquitination machinery (157), thus controlling turnover of these RING-containing proteins and of other proteins with which they associate. Also, the BIR-containing protein, Apollon (also called Bruce), contains a domain with ubiquitin-conjugating enzyme (E2) activity, further suggesting links of BIR family proteins to the cellular ubiquitination machinery (158). The targets of the E3 activity of IAPs remain to be fully elucidated. IAPs can control their own degradation through self-ubiquitination (157, 159), as do other E3 proteins. IAPs can also promote the degradation of the caspases to which they bind (160, 161). Finally, IAPs bind to and induce ubiquitination (and subsequent degradation) of endogenous IAP-antagonist proteins, both in insect and in mammalian cells (162166). Thus, RING-containing IAPs can destroy death-inducing proteins that block their antiapoptotic actions, suggesting that such mutual antagonism might determine the balance between cell life and death. In addition, some BIR family proteins also modulate signaling pathways controlling activation of NF-κB or JNK [reviewed in (140, 167)]. The RING of cIAP1 is important for inducing degradation of the signaling protein TRAF2 to which it binds (62). The relevance of these signal-transducing functions to apoptosis suppression is unresolved, but the findings raise the possibility that some BIR family proteins function both as caspase inhibitors and as signal-transduction modulators.

Endogenous antagonists of the IAPs help to keep these apoptosis suppressors in check and promote apoptosis (Fig. 18). In Drosophila, five endogenous IAP antagonists have been identified, including Reaper, Hid, Grim, Sickle, and Jafra2 (168171). The IAP-binding domain of Reaper, Hid, Grim, and Sickle maps to their N-termini, revealing a conserved peptidyl motif that is necessary and sufficient for binding DIAP1. It is thought that the N-terminal methionine of these proteins is removed by an exoprotease, to reveal this IAP-binding motif, which has the consensus sequence A (V or I or W) (P or A) (I or S or Y or F), where alanine represents the second amino acid in the native protein. In contrast, the N terminus of Jafrac2 is removed by proteolysis during its sorting to the ER lumen, exposing the IAP-binding motif (170). Analogous IAP antagonists have been identified in humans and mice, including SMAC (second mitochondria-derived activator of caspases; also called Diablo) and the serine protease Omi (also called Htra2). The IAP antagonists SMAC and Omi are sequestered inside mitochondria and are released into the cytosol during apoptosis (172, 173). The N-terminal leader sequences of these proteins are removed by proteolysis upon import into mitochondria, exposing the tetrapeptide IAP-binding motif. These IAP antagonists compete with caspases for binding to IAPs, thus freeing caspases from the grip of the IAPs and promoting apoptosis. Synthetic peptides that mimic SMAC and HtrA2 induce apoptosis or sensitize tumor cell lines to apoptosis (174176). Structural information derived from high-field nuclear magnetic resonance (NMR) and x-ray crystallography indicates that a tetrapeptide motif is sufficient to bind the BIRs of IAPs. This suggests a path to drug discovery in which small-molecule chemical compounds that occupy the same tetrapeptide-binding motif on BIRs might act as apoptosis-sensitizing agents (Fig. 19). Other IAP antagonists have been described, such as the mammalian XAF [X-linked inhibitor of apoptosis protein (XIAP)–associated factor] protein, which may inhibit certain IAP family members through an alternative mechanism that remains to be defined (177). Also, given that Drosophila possess only three BIR-containing genes but has at least five IAP antagonists, one would presume that additional endogenous antagonists of the mammalian IAP await discovery.

Fig. 18.

IAP antagonists. The domain architectures of IAP antagonists are presented. Reaper, Hid, Grim, Sickle, and Jafrac2 are Drosophila proteins. Smac, HtrA2, and XAF-1 are human proteins that have homologs in mice (not shown).

Fig. 19.

SMAC peptide bound to BIR3 of XIAP. The 3D structure of a tetrapeptide from SMAC (shown as a stick diagram) bound to the BIR3 domain of XIAP (shown as a space-filling model) is based on molecular modeling (276, 277). [Figure courtesy of G. Salvesen]

Bcl-2 family proteins

Proteins of the Bcl-2 family are critical regulators of apoptosis, whose functions included governing mitochondrial-dependent steps in cell death pathways (29, 178). Many proteins from this family share a common fold of a helical bundle, reminiscent of the diphtheria toxin structure (Fig. 20), but other proteins commonly listed among Bcl-2 family members have diverse structures and share only a short sequence motif, known as the Bcl-2 homology-3 (BH3) domain, that allows them to interact with other members of the family.

Fig. 20.

Three-dimensional structure of Bcl-XL. (A) The 3D structure of Bcl-XL (lacking its C-terminal TM domain) is shown in ribbon representation, alongside (B) the pore-forming domain of diphtheria toxin (196). The central pair of hydrophobic α helices (red) and surrounding amphipathic α helices (yellow and blue) are indicated. [Figure courtesy of S. Fesik]

Humans and mice contain 25 and 28 recognized Bcl-2 family proteins, respectively, whereas Drosophila and C. elegans each have two (Table 1 and Table 6). The Bcl-2 family proteins derive their name from the founding member of the family, Bcl-2, which was initially discovered because of its role in B cell lymphomas (179). Bcl-2 family proteins are usually partitioned into different groups based on the presence or absence of certain sequence motifs, Bcl-2 homology (BH) domains. However, recent structural studies suggest that the Bcl-2 family may consist of a "proper" family of proteins that share a common three-dimensional fold but may lack classical BH domains and a second more diverse set of proteins, including short peptides, that share only limited sequence similarity in the BH-3 motif. The BH domains are difficult to assign by genomic database searches.

In mammals, the Bcl-2 family proteins include (i) the "multidomain" members of the family, which contain BH1, BH2, BH3, and (sometimes) BH4 domains [Bcl-2; BclX; Mcl1; Bcl-W, Bf11 (also called A1), Bcl-B, Diva (also called Boo), Bax, Bak, Bok (also called Mtd), and Bcl-Rambo (also called Mil); (ii) Bcl-GL and Bfk, which have BH2 and BH3 domains (180, 181); (iii) several proteins that contain only the BH3 domain ("BH3-only" proteins) [Bad; Bid; Bim (also called Bod); Spike; Bmf; Bik (also called Blk); Noxa (aslo calledAPR); Puma; Hrk (also called Dp5); and Spike] (182, 183); (iv) proteins with BH3-like domains [Nip3 (Bnip3); Nix (Nip3L); Map1; and p193] (184); and (v) a protein containing a putative BH2 domain (Bcl2L12) (185) (Fig. 21). In mice, the Bfl1 gene has undergone amplification, with four copies of genes encoding Bfl1-like proteins present, compared to only one such gene in humans.

Fig. 21.

Bcl-2 family proteins. The domain architecture of the Bcl-2 family of proteins is presented, showing Bcl-2 homology (BH) domains, and transmembrane (TM) regions. Most proteins shown are mammalian, except Ced9, Egl-1, and CeBnip, which are from C. elegans; Drob1 (also called Dbok), which is of fly origin; and Nr13, which is from chicken. Family members from amphibians, fish, and other species are not shown, nor are viral homologs. Nip3, Nix, and C. elegans Bnip have BH3-like domains (BH3L). The human ortholog of mouse Noxa has only one BH3 domain.

Bcl-2, Bcl-XL, Mcl-1, Bak, and several other members of the Bcl-2 family have a hydrophobic stretch of amino acids near their C terminus that anchors them in internal cell membranes, predominantly the outer mitochondrial membrane and ER (29, 186, 187). In contrast, other Bcl-2 family members such as Bid, Bim, and Bad lack these membrane-anchoring domains, but are translocated to mitochondria in response to specific stimuli (188191). Still others have the membrane-anchoring domain but keep it latched against the body of the protein until stimulated to expose it (such as Bax or Bcl-W) (192). Endogenous regulators that control translocation of Bcl-2 family proteins have been identified, and more are likely to be revealed in the future (193195).

A picture of how Bcl-2 family proteins regulate life and death decisions in the cell has begun to emerge, although many details remain unclear. The multidomain members of the family, as well as some BH3-only members, have been documented to or are predicted to share structural similarity with the α-helical pore-forming domains of certain bacterial toxins (26, 196201) (Fig. 20). These Bcl-2 family proteins are thought to directly regulate the permeability of mitochondrial membranes, either permitting or inhibiting efflux of apoptogenic proteins from these organelles. The mammalian α-helical porelike proteins include both antiapoptotic proteins (Bcl-2, Bcl-XL, Mcl-1, Bfl-1, Bcl-W, and possibly Bcl-B) and proapoptotic proteins (Bax, Bak, Bok, and Bid). Although most of the proteins in this subcategory contain the BH1, BH2, BH3, and sometimes BH4 domains, the Bid protein contains only a BH3 domain as shown by sequence analysis, but has nevertheless been determined to share the same overall protein-fold with Bcl-XL, Bcl-2, Bcl-W, and Bax (197, 198). Where tested to date, these proteins all form ion-conducting channels in synthetic membranes in vitro, including Bcl-2, Bcl-XL, Bax, and Bid (202206).

Precisely how multidomain Bcl-2 family proteins control the permeability of mitochondrial membranes remains enigmatic. Unresolved questions include whether the proapoptotic members such as Bax and Bak are sufficient to create pores by themselves or require additional accessory proteins [leading candidates include the voltage-dependent anion channel (VDAC) and adenine nucleotide translocator (ANT)] (178). Bax and Bak oligermize in mitochondrial membranes when activated, and this oligomerization is associated with the release of cytochrome c and other proteins from these organelles that contribute to caspase activation and other cell death mechanisms (190). In contrast, antiapoptotic proteins such as Bcl-2 and Bcl-XL block oligomerization of Bax and Bak, preserving cell viability.

Another aspect of the biology of Bcl-2 family proteins that might relate to their similarity to pore-forming molecules concerns the effects of overexpressing Bcl-2 or Bcl-XL on regulation of Ca2+ in the ER. These and some other antiapoptotic proteins of the Bcl-2 family insert not only in mitochondrial membranes but also in the membranes of the ER (207). Overexpression of Bcl-2 appears to cause perturbations in the amount of free Ca2+ sequestered in the ER and reduces the amounts of Ca2+ liberated from the ER in response to some types of cell death stimuli and drugs (208210). Knocking out the bax and bak genes in murine cells has the same effect [reviewed in (211)]. It remains to be formally demonstrated, however, whether the ability of multidomain Bcl-2 family proteins to function as ion channels explains these phenomena. Also unclear is why some members of the family block cell death, whereas others promote it, given the extensive structural similarlity between the anti- and proapoptotic multidomain members of the Bcl-2 family (196199).

The BH3 domain mediates dimerization among Bcl-2 family proteins. This domain consists of an amphipathic α helix ~16 amino acids in length that inserts into a hydrophobic crevice on the surface of antiapoptotic proteins such as Bcl-2 and Bcl-XL (212) (Fig. 22). Members of the BH3-only branch of the family in mammals (182) are uniformly proapoptotic. These proteins include minimally Bad, Bik, Bid, Bim, Hrk, Bfm, Bcl-GS, p193, Noxa, Puma, and Spike. Some of the Drosophila IAP antagonists also may contain BH3-like domains, in addition to their IAP-binding motifs (213), suggesting dual functions for these death-inducing proteins. The cell death–inducing activity of most BH3 domain proteins depends on their ability to dimerize with antiapoptotic Bcl-2 family members and thus to function as trans-dominant inhibitors of proteins such as Bcl-2 and Bcl-XL [reviewed in (182, 214)]. Alternatively, certain BH3-only proteins, in particular Bid and Bim, can either bind proapoptotic multidomain proteins (such as Bax or Bak) and function as agonists of the killers or dimerize with antiapoptotic family members (such as Bcl-2 or Bcl-XL) and function as apoptosis antagonists (215217) (Fig. 23). Binding of Bid to Bax or Bak promotes their insertion into membranes and their oligomerization, thus resulting in changes in mitochondrial membrane permeability and release of apoptogenic proteins into the cytosol (190, 218). Mutations in the BH3 domains of proapoptotic BH3-only proteins that abolish their ability to bind other Bcl-2 family members abrogate their proapoptotic activity.

Fig. 22.

Structure of Bcl-2 with BH3 peptide. A model of the 3D structure of Bcl-2 is presented [adapted from (278)] in space-filling representation, showing a BH3 peptide bound to a hydrophobic surface crevice.

Fig. 23.

Hierarchies of Bcl-2 family proteins. A proposed model for hierarchical relations among Bcl-2 family proteins [adapted from (190)]. Bax and Bak (multidomain proapoptotic) proteins are envisioned as essential downstream mediators of cell death. Bcl-2, Bcl-XL, and similar antiapoptotic multidomain proteins suppress Bax and Bak. BH3-only proteins, through dimerization with multidomain proteins, can act either as antagonists of antiapoptotic Bcl-2, Bcl-XL, and their related proteins or as agonists of Bax and Bak, or both.

The BH3-only proteins link various environmental stimuli to the mitochondrial pathway for apoptosis (Fig. 24). For example, the transcription factor p53 directly induces expression of BH3-only proteins Noxa, Puma, and Bid (219221), thus linking p53 to the death machinery. The BH3-only protein Bfm is associated with the actin cytoskeleton (222). Conditions that disrupt the cytoskeleton, such as cell detachment from extracellular matrix, release Bfm, allowing it to dimerize with antiapoptotic proteins on the surface of mitochondria and induce apoptosis. Bid is cleaved by Caspase-8 in the context of signaling by apoptosis-inducing TNF family receptors, removing the N-terminal 52 amino acids and exposing both the BH3 dimerization domain, as well as generating a novel N terminus that becomes myristyolated, facilitating targeting of the Bid protein into membranes (206, 223225). This Caspase-8–mediated activation of Bid represents an important mechanism accounting for cross-talk between the death receptor (extrinsic) and mitochondrial (intrinsic) pathways that lead to apoptosis.

Fig. 24.

Signal transduction and BH3-only proteins. Many BH3-only proteins are linked to pathways for inducing apoptosis [reviewed in (182)]. For example, p53 induces transcription of Noxa and Puma genes. Bim associates indirectly with microtubules and, where released, translocates to mitochondria to interact with antiapoptotic Bcl-2 family proteins. Bfm is associated with the actin cytoskeleton and is released upon disruption of the cytoskeleton by release of adherent cells into suspension [a condition that triggers a from of apoptosis known as "anoikis" (279)]. Transcription of Hrk is induced by withdrawal of neurotrophins from neurons. Bid is cleaved and activated by Caspase-8 during signaling by TNF family death receptors. Proapoptotic short forms of Bcl-X and Bcl-G are produced by alternative mRNA splicing. The ability of Bad to dimerize with antiapoptotic Bcl-2 family proteins is regulated by phosphorylation. Several kinases inactivate and some phosphatases activate this proapoptotic protein.

Isoforms of the BH3-only protein Bim associate with microtubules through direct binding to microtubule-associated dynein light-chain (DLC) (191). Disruption of these protein interactions frees the microtubule-associated isoforms of Bim, allowing them to dimerize through their BH3 domains with antiapoptotic Bcl-2 family proteins on the surface of mitochondria. The BH3-only protein Bad translocates between the cytosol and mitochondria, depending on whether it is phosphorylated. Several protein kinases, including Akt [also called protein kinase B (PKB)], PKA (cyclic AMP-dependent protein kinase), Raf1, Rsk1, and Pak1 (p21-activated kinase 1) have been reported to phosphorylate Bad, thus inactivating the protein so that it cannot dimerize with and antagonize Bcl-2 or Bcl-XL [reviewed in (226, 227)]. Finally, expression of some proapoptotic Bcl-2 family proteins is controlled at the level of mRNA splicing. Specific signals switch mRNA splicing patterns to produce versions of the proteins with proapoptotic activity (e.g., Bcl-XS; Bcl-GS) (180, 228). Perturbations in the expression and posttranslation modification of Bcl-2 family proteins have been associated with many disease states, and causative relations have been established between specific members of the family and either pathological cell loss or cell accumulation.

Nip3 (Bnip3) and Nix (Bnip3a) appear to be exceptions to the rule where BH3-only proteins are concerned. Mutagenesis of the BH3 domains of these proteins has failed to demonstrate a faithful correlation between ability to bind other Bcl-2 family protein and proapoptotic function (229). These proteins contain a transmembrane domain that anchors them in mitochondrial membranes and is critical for cell death–inducing activity. The details of how Bnip3 and Nix regulate mitochondrial-dependent steps in cell death pathways require further elucidation. Drosophila and C. elegans also have homologs of Nip3 and Nix, although little is known about their functions (230). Spike is a proapoptotic protein with a BH3-only motif (231) The apparent BH3 motif is embedded within a region of strong similarity to sequences from the SNF7 family of coiled-coil proteins involved in yeast multivesicular body function and formation (232).

Several viruses contain genes encoding Bcl-2–related proteins [reviewed in (233)]. The Herpes family viruses encode structural homologs of Bcl-2 that suppress apoptosis, including the BHFR protein of Epstein-Barr virus and the Kbcl-2 protein of Kaposi sarcoma virus (234236). Adenovirus also produces an antiapoptotic protein, E1b-19 kD, which shares very little sequence similarity with Bcl-2 family proteins, but nevertheless binds Bax and suppresses apoptosis (237).

Caspases and Other Proteases with Roles in Apoptosis

Caspases are a family of evolutionarily conserved intracellular cysteine proteases that either induce apoptosis or that are required for proteolytic processing of certain proinflammatory cytokines [reviewed in (21)]. The cysteine protease fold that constitutes the caspase domain is typically composed of ~20-kD large and ~10-kD small catalytic subunits. The subunits are generated upon proteolytic cleavage from a proprotein precursor and assemble into a symmetrical heterotetramer with two active sites formed at the junction of the large and small subunits [reviewed in (26)] (Fig. 25). However, variations on this theme have been described, such as Caspase-9, which forms an asymmetrical dimer with only one active site per molecule (238).

Fig. 25.

Structure of Caspase-3. The 3D structure of Caspase-3 is presented, based on x-ray crystallographic data. The protease was cocrystallized with a peptide inhibitor in both active sites of the heterotetramer. [Adapted from (280)]

In addition to caspases, other proteases have been described that have some relevance to apoptosis, including granzymes. These apoptosis-relevant proteases are also briefly addressed here.

Domain organization and functions of caspases and other proteases

Caspases are encoded as polypeptides encompassing an N-terminal prodomain followed by a conserved C-terminal catalytic domain. The proforms of caspases contain N-terminal prodomains of variable length, with the upstream initiator proteases generally having larger prodomains than do downstream effector proteases. The larger prodomains serve as protein interaction modules for controlling caspase activation through the "induced proximity" mechanism (239) or by promoting caspase dimerization (27).

The genomes of most humans contain 11 functional caspase-encoding genes (Fig. 26) (Tables 1 and 7), which can be divided into four subgroups on the basis of sequence comparisons. Caspases with short or essentially nonexistent N-terminal prodomains include Caspases 3, 6, 7, and 14. These proteases are generally implicated as downstream effectors of apoptosis and are typically dependent on upstream initiator caspases for their proteolytic cleavage and activation [reviewed in (18, 21)]. Caspases with similarly short N-terminal domains are found in the genomes of mice. Among model organisms, mice appear to have 10 such orthologous caspases, Drosophila has 7, and C. elegans has 5 (Table 1).

Fig. 26.

Caspase family cell death proteases. The domain architecture of the human caspases is presented. The prodomains of the upstream initiator caspases are absent in the downstream effector caspases. Aspartyl residues at which proteolytic processing occurs and the active-site cysteine are indicated. The prodomain of Caspase-2 can be removed by cleavage at an aspartic acid residue (not shown), whereas the prodomain of pro-Caspase-9 remains intact. [Figure courtesy of G. Salvesen, Burnham Institute]

Two caspases whose prodomains contain a tandem pair of DEDs are found in humans: Caspases 8 and 10. The mouse genome encodes a Caspase-8 ortholog, but is lacking Caspase-10. In humans, the Caspase-8 and Caspase-10 genes are located adjacent to each other on chromosome 2, possibly as a result of a recent gene-duplication event. Caspases 8 and 10 typically become activated in the context of signal transduction by TNF family cytokines. Although the genome of Drosophila also encodes a caspase with a DED, called Dredd (240), there are no recognizable TNF family receptors in the fly. However, a candidate DED-containing adapter protein for Dredd has been identified in Drosophila, which shows sequence similarity to analogous proteins in mammals (83). The gene encoding Dredd is required for resistance to Gram-negative bacterial infection in flies (241).

Five human caspases contain N-terminal prodomains that comprise a CARD (caspase recruitment domain) (25). Two of these, Caspase-2 and Caspase-9, are implicated in regulation of apoptosis. Caspase-2 is putatively involved in linking stress in the Golgi to caspase activation (242), and Caspase-9 functions in linking mitochondria to the caspase protease network (51). However, three of the caspases that have CARD domains are better known for their roles in cleaving and activating proinflammatory cytokines (pro-IL-1 and pro-IL-18) than for their role in apoptosis. In humans, these include Caspases 1, 4, and 5, which appear to have arisen by gene amplification on chromosome 11, where they exist in a tandem array. Human Caspases 4 and 5 both appear to be orthologs of murine Caspase-11. Thus, although humans have three Caspase-1–like proteases (Caspases 1, 4, and 5), mice have only two (caspases 1 and 11). Various CARD-containing proteins have been identified that participate in either inducing or suppressing activation of Caspase-1 and related proteases. Although Caspase-1 does not appear to be required for physiological cell death on the basis of gene ablation studies in mice (243), it may be relevant to the induction of apoptosis during certain pathological conditions, such as ischemia-reperfusion injury in the brain (stroke), and during certain host-pathogen interactions in which bacterial toxins trigger apoptosis of macrophages or intestinal epithelial cells (244, 245). Phylogeny analysis indicates that, in mice and humans, Caspases 2 and 9 constitute a branch of the caspase family distinct from that containing Caspases 1, 4, 5, and 11.

The same region of human chromosome 11 where the genes encoding CARD-containing proteases Caspase-1, Caspase-4, and Caspase-5 are located contains multiple caspase-like pseudogenes, including a transcribed gene that would produce a human ortholog of Caspase-12 were it not for the presence of a termination codon early in the predicted open reading frame, in the region corresponding to the CARD (15). The genomes of mice, rats, and even some primates contain orthologous genes that lack the termination codon and which therefore encode the full-length protein, Caspase-12. Some humans of African descent have alleles of Caspase-12 that lack the termination codon, suggesting that the loss of this caspase family member represents a very recent event (246). The Caspase-12 protein in mice purportedly associates with the ER, and mediates a pathway linking ER stress to caspase activation and apoptosis (247). It remains to be determined whether this orthologous gene retained in some human populations functions in an analogous manner.

Gene ablation (knockout) analysis of caspases in mice has been accomplished thus far for Caspases 1, 2, 3, 8, 9, 11, and 12, providing insights into circumstances in which these proteases are either essential or redundant for regulation of apoptosis or processing of proinflammatory cytokines (86, 243, 247252). Genetic analysis of CARD-containing caspases Dronc and Ced3 in Drosophila and C. elegans, respectively, indicates essential roles for these proteases in developmental cell death (253, 254).

Caspase-like proteins

In addition to the caspases implicated in apoptosis, several genes encoding proteins with a caspase-like fold have been identified, and are known as para-caspases, meta-caspases, and separases (57). None of these proteins has been firmly linked to apoptosis regulation, and most have not even formally been demonstrated to be proteases. Humans, mice, and flies have a single gene encoding a para-caspase. The human para-caspase protein has been implicated in a pathway for activation of NF-κB (57). Although mechanistic details are lacking, it has been proposed that the principal function of para-caspase is as an E3 ubiquitin ligase, rather than as a protease (255). Mice lacking para-caspase have defects in T and B cell antigen receptor signaling (134, 135).

Granzymes and other proteases

Another type of protease implicated in apoptosis is Granzyme-B, a serine protease that cleaves substrate targets at Asp residues, as do caspases (256). Cytotoxic T lymphocytes and natural killer cells inject apoptosis-inducing Granzyme B into target cells through a mechanism involving the mannose-6-phosphate receptors and pore-forming perforins (256258). A single Granzyme B gene is found in the genomes of mice and humans, although closely related serine proteases with no proven involvement in apoptosis, such as Granzyme C and H, are also recognizable. Granzyme B can cleave and activate multiple caspases and some caspase substrates (259). Endogenous and viral inhibitors of Granzyme B have been identified that promote resistance to this inducer of apoptosis (150, 260, 261).

Finally, a serine protease sequestered in mitochondria has been implicated in apoptosis regulation. Omi is a highly conserved protein that contains an N-terminal leader peptide for mitochondrial import. The normal function of Omi in mitochondria is unclear, but when released into the cytosol by proapoptotic stimuli, Omi encourages cell death by both a caspase-dependent mechanism (involving inhibitory effects on IAPs) and through its proteolytic activity, which reportedly contributes to caspase-independent cell death (262264).

CIDE domains and apoptotic endonucleases

DNA fragmentation is often considered a hallmark of apoptosis, reflecting the activation of endonucleases that cleave DNA between nucleosomes (265). A domain is found in the apoptotic endonuclease CAD (DFF40) and its homologs (266268), called the CIDE or CIDE-N domain (Fig. 27). The CIDE-N domain represents a ~75–amino acid fold consisting of a twisted five-stranded β sheet with two α helices arranged in an α/β roll (269). The CIDE-containing endonuclease CAD is held in an inactive state by a specific chaperone protein, ICAD (DFF45), that also contains the CIDE domain and that associates via CIDE-CIDE interactions (270). The endonuclease CAD is released upon cleavage of chaperone ICAD by caspases, thus linking endonuclease activation to activation of the cell death proteases. Five CIDE family members are evident in the transcriptomes of humans and mice (Fig. 28) (Table 8).

Fig. 27.

Structure of CIDE domain. The 3D structure of the CIDE domain of DFF40 is presented, showing ribbon (A) and space-filling (B) representations of the domain (269, 270). Blue and red indicate basic and acidic residues, respectively.

Fig. 28.

CIDE family proteins. The domain architecture of the CIDE family proteins of humans is presented.

Table 1.

Apoptosis proteins. Data represent unique genes within each organism’s genome. Genes for which sequence similarity is noted in the literature, but is not substantiated by either PFAM or unambiguous pairwise sequence similarity to known domains, were excluded. CARD-containing caspases were excluded from the CARD group.

Table 2.

CARDs. The human genes with CARD domains are listed along with genomic positions, identifiers, and annotated protein products. CARD-containing caspases listed in Table 7 are excluded here.

Table 3.

DEDs. The human genes encoding DED domains are listed along with genomic positions, identifiers, and annotated protein products. The DED-like (DED-L) proteins do not show similarity to the PFAM DED domain (expectation value < 10.0), nor do they show pairwise sequence similarity to any known PFAM-based DED domain. The variant DEDs (vDED) of BAR, Bap31, Hip, and Hippi represent a cluster of domains distinct from the other DED and DED-L domains.

Table 4.

DDs. The human genes encoding death domains are listed along with genomic positions, identifiers, and annotated protein products. The MADD protein does not have structural similarity to death domains on the basis of a PFAM profile.

Table 5.

BIR-containing proteins (IAPs). The human genes encoding BIR domains are listed along with genomic positions, identifiers, and annotated protein products.

Table 6.

Bcl-2 family. The human genes encoding Bcl-2 homology domains are listed along with genomic positions, identifiers, and annotated protein products.

Table 7.

Caspases. The human genes encoding caspases are listed along with genomic positions, identifiers, and annotated protein products. Also included are para-Caspase, which has predicted structural similarity to caspases, and other structurally unrelated proteases of relevance to apoptosis (HtrA2; Granzyme B).

Table 8.

CIDE N-terminal. The human genes encoding CIDE N-terminal domains are listed along with genomic positions, identifiers, and annotated protein products.

In addition to CIDE family endonucleases, two proteins have been identified that are sequestered in mitochondria and can contribute directly or indirectly to DNA digestion: EndoG, an evolutionarily conserved endonuclease (271, 272); and AIF, a flavoprotein that promotes large-scale cleavage of genomic DNA during apoptosis (273). Orthologs of AIF and EndoG are found in both Drosophila and C. elegans, but their overall importance to cell death requires further investigation (274).


The domains associated with the core apoptotic machinery are conserved throughout animal species, presumably reflecting the evolutionary pressure favoring a mechanism for altruistic cell suicide that benefits the multicellular organism while sacrificing some of its cellular members. Host-pathogen interactions represent one of the contexts in which a cell suicide mechanism can be beneficial. This may explain the close links, in many instances, between apoptosis and inflammatory responses mechanisms. We have introduced the major domain families associated with apoptosis and have provided a sketch of their mechanisms and interrelations, focusing on Homo sapiens as a frame of reference and making comparisons with selected other animal species. The network of protein interactions that defines the full panoply of apoptosis regulatory mechanisms is, however, far broader than the subject matter reviewed here. Consequently, the domain families reviewed here should be viewed as a foundation upon which to build a deeper understanding of the molecular mechanisms responsible for regulating apoptosis.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 47.
  48. 48.
  49. 49.
  50. 50.
  51. 51.
  52. 52.
  53. 53.
  54. 54.
  55. 55.
  56. 56.
  57. 57.
  58. 58.
  59. 59.
  60. 60.
  61. 61.
  62. 62.
  63. 63.
  64. 64.
  65. 65.
  66. 66.
  67. 67.
  68. 68.
  69. 69.
  70. 70.
  71. 71.
  72. 72.
  73. 73.
  74. 74.
  75. 75.
  76. 76.
  77. 77.
  78. 78.
  79. 79.
  80. 80.
  81. 81.
  82. 82.
  83. 83.
  84. 83a.
  85. 83b.
  86. 84.
  87. 85.
  88. 86.
  89. 87.
  90. 88.
  91. 89.
  92. 90.
  93. 91.
  94. 92.
  95. 93.
  96. 94.
  97. 95.
  98. 96.
  99. 97.
  100. 98.
  101. 99.
  102. 100.
  103. 101.
  104. 102.
  105. 103.
  106. 104.
  107. 105.
  108. 106.
  109. 107.
  110. 108.
  111. 109.
  112. 110.
  113. 111.
  114. 112.
  115. 113.
  116. 114.
  117. 115.
  118. 116.
  119. 117.
  120. 118.
  121. 119.
  122. 120.
  123. 121.
  124. 122.
  125. 123.
  126. 124.
  127. 125.
  128. 126.
  129. 127.
  130. 128.
  131. 129.
  132. 130.
  133. 131.
  134. 132.
  135. 133.
  136. 134.
  137. 135.
  138. 136.
  139. 137.
  140. 138.
  141. 139.
  142. 140.
  143. 141.
  144. 142.
  145. 143.
  146. 144.
  147. 145.
  148. 146.
  149. 147.
  150. 148.
  151. 149.
  152. 150.
  153. 151.
  154. 152.
  155. 153.
  156. 154.
  157. 155.
  158. 156.
  159. 157.
  160. 158.
  161. 159.
  162. 160.
  163. 161.
  164. 162.
  165. 163.
  166. 164.
  167. 165.
  168. 166.
  169. 167.
  170. 168.
  171. 169.
  172. 170.
  173. 171.
  174. 172.
  175. 173.
  176. 174.
  177. 175.
  178. 176.
  179. 177.
  180. 178.
  181. 179.
  182. 180.
  183. 181.
  184. 182.
  185. 183.
  186. 184.
  187. 185.
  188. 186.
  189. 187.
  190. 188.
  191. 189.
  192. 190.
  193. 191.
  194. 192.
  195. 193.
  196. 194.
  197. 195.
  198. 196.
  199. 197.
  200. 198.
  201. 199.
  202. 200.
  203. 201.
  204. 202.
  205. 203.
  206. 204.
  207. 205.
  208. 206.
  209. 207.
  210. 208.
  211. 209.
  212. 210.
  213. 211.
  214. 212.
  215. 213.
  216. 214.
  217. 215.
  218. 216.
  219. 217.
  220. 218.
  221. 219.
  222. 220.
  223. 221.
  224. 222.
  225. 223.
  226. 224.
  227. 225.
  228. 226.
  229. 227.
  230. 228.
  231. 229.
  232. 230.
  233. 231.
  234. 232.
  235. 233.
  236. 234.
  237. 235.
  238. 236.
  239. 237.
  240. 238.
  241. 239.
  242. 240.
  243. 241.
  244. 242.
  245. 243.
  246. 244.
  247. 245.
  248. 246.
  249. 247.
  250. 248.
  251. 249.
  252. 250.
  253. 251.
  254. 252.
  255. 253.
  256. 254.
  257. 255.
  258. 256.
  259. 257.
  260. 258.
  261. 259.
  262. 260.
  263. 261.
  264. 262.
  265. 263.
  266. 264.
  267. 265.
  268. 266.
  269. 267.
  270. 268.
  271. 269.
  272. 270.
  273. 271.
  274. 272.
  275. 273.
  276. 274.
  277. 275.
  278. 276.
  279. 277.
  280. 278.
  281. 279.
  282. 280.
  283. 281.

Stay Connected to Science Signaling

Navigate This Article