From the realization that cell number homoeostasis is fundamental to the biology of all metazoans, and that deregulation of this process leads to human diseases, enormous interest has been devoted over the last two decades to map the requirements of cell death and cell survival. This effort has led to tangible progress, and we can now chart with reasonable accuracy complex signalling circuitries controlling cell-fate decisions. Some of this knowledge has translated into novel therapeutics, and the outcome of these strategies, especially in cancer, is eagerly awaited. However, the function of cell-death modifiers have considerably broadened over the last few years, and these molecules are increasingly recognized as arbiters of cellular homoeostasis, from cell division, to intracellular signalling to cellular adaptation. This panoply of functions is best exemplified by members of the IAP (inhibitor of apoptosis) gene family, molecules originally narrowly defined as endogenous caspase inhibitors, but now firmly positioned at the crossroads of multiple normal and transformed cellular responses.
Although there are various morphologically distinct and biochemically separate forms of cell death, only apoptosis embodies an orderly genetic programme of cellular suicide . This process is designed to sculpt the developing organism  and maintain the cell number homoeostasis of tissues and organs throughout adult life . Deregulation of apoptosis is a pathogenic factor of many human diseases, and aberrantly increased cell survival is a hallmark of virtually every human tumour .
Extensive studies over the last two decades have identified two main pathways by which mammalian cells commit suicide . An extrinsic apoptotic pathway is activated by ligand binding to so-called death receptors at the cell surface, molecules structurally reminiscent of the TNFR [TNF (tumour necrosis factor) receptor]. This results in the assembly of a multiprotein complex associated with the cytosolic tail of the receptor, and culminates with the activation of upstream caspase 8 . An intrinsic pathway of apoptosis is activated by genotoxic, metabolic and other stimuli, and is centred on a sudden loss of mitochondrial integrity . Dubbed ‘mitochondrial permeability transition’ , this process ultimately leads to the rupture of the organelle outer membrane with discharge of apoptogenic proteins normally stored in the mitochondrial intermembrane space, in particular cytochrome c . Once released into the cytosol, cytochrome c assembles in a large supramolecular complex called an apoptosome that promotes the activation of initiator caspase 9 via induced proximity . Regardless of the triggering stimulus, active initiator caspases promote the downstream processing of executioner caspases, which dismantle a cell's architecture imparting the classical morphological features of apoptosis . There is extensive cross-talk between the two apoptotic pathways, and mechanisms for signal amplification in selected cell types have been described .
Among the regulators of cell death, the Bcl-2 gene family comprises both apoptosis inducers and apoptosis inhibitors . These molecules are structurally diverse, and form heteromeric complexes to control mitochondrial integrity, especially at the level of outer membrane permeability . In contrast, the IAP (inhibitor of apoptosis) family of proteins were originally characterized as physical inhibitors of caspases , providing a cytoprotective step downstream of death receptor or mitochondrial apoptosis. However, studies over the last few years have uncovered a far more complex biology of IAPs with broadened roles in various facets of cellular homoeostasis [12,13]. The review of these multiple IAP functions is the main theme of the present article. Excellent contributions covering virtually every aspect of cell-death regulation, including mechanisms of death receptor activation , mitochondrial permeability transition [7,8], apoptosis modifiers [10,13] or caspases , have been published in the literature, and the reader is directed to those articles for a more in-depth perspective.
THE BIOCHEMISTRY OF IAPs: THE ‘OLD’ CASPASE INHIBITORS
IAPs are recognized by the presence of a ~70 amino acid BIR (baculovirus IAP repeat), a zinc-finger-fold present at least once in each family member  (Figure 1). The eight IAPs in humans contain one to three BIRs, typically arranged at the N-terminus of the protein. Several mammalian IAPs, for instance c-IAP1 (cellular IAP1), c-IAP2 (cellular IAP2) and XIAP (X-linked IAP) contain additional structural domains, including a C-terminus RING, which functions as an E3 ubiquitin ligase, a ubiquitin-associated domain implicated in binding to ubiquitinated proteins, and a caspase-recruitment domain (CARD in c-IAP1 and c-IAP2), of less clear function (Figure 1). There is extensive modularity in the assembly of these domains, and different IAPs can variously display BIRs as well as other protein domains (Figure 1).
Schematic diagram of domain structure in representative IAPs
Compared with other IAPs, survivin is structurally unique. At 142 amino acids, survivin is the smallest mammalian IAP, containing a single BIR and a long C-terminal α-helix, but no other identifiable protein domain. Structural data suggest that survivin forms a stable homodimer in solution , but definitive evidence that this organization is required for function(s) is still lacking. Conversely, certain aspects of survivin nucleo-cytoplasmic trafficking , and key protein recognition, for instance binding to the chromosomal passenger protein borealin [17,18], appear to require the monomeric protein.
BIRs mediate protein recognition and protein–protein interactions . Accordingly, a deep peptide-binding groove in the BIRs of XIAP, c-IAP1 and c-IAP2 serves as a hydrophobic recognition site for proteins containing an IBM (IAP-binding motif). The IBM is a tetrapeptide region with an invariant N-terminal alanine residue and other conserved residues found in initiator (caspase 9) and effector (caspase 7) caspases , as well as in certain apoptosis inducers, for instance Smac (second mitochondrial-derived activator of caspase)/DIABLO (direct IAP-binding protein with low pI) . Not all BIRs contain a ‘canonical’ IBM-recognition motif . For instance, BIR1 in XIAP does not bind IBM-containing proteins, but recognizes molecules implicated in NF-κB (nuclear factor-κB) activation (see below) [22,23]. Similarly, the BIRs in survivin and some of its likely orthologues in yeast, Caenorhabditis elegans or Drosophila do not appear to contain an IBM-binding motif. However, this is clearly not a rigid rule, as survivin binds IBM-containing Smac/DIABLO  in a complex that resembles the Smac interaction with XIAP BIR3 .
One of the most studied IBM-dependent complexes is the interaction between IAPs and caspases , which obliterates their enzymatic activity. Historically, this has been the role proposed for all IAPs , expanding a cytoprotective function first observed with the viral orthologues of these proteins . However, we now know that only one mammalian IAP, XIAP, is truly a physiological inhibitor of caspases in vivo . Other IAPs, for instance c-IAP1 and c-IAP2, bind caspases in vitro, but these interactions are unlikely to be physiologically meaningful in vivo. Conversely, XIAP associates with executioner caspases 3 and 7, as well as initiator caspase 9, with high affinity, shutting off their cell killing ability. The structural requirements of these interactions have been worked out in detail . With respect to executioner caspases, it is the XIAP linker region upstream of BIR2 that inserts into the catalytic cleft of the enzymes, preventing substrate accessibility and thus blocking activity [29–31]. Instead, XIAP binds caspase 9 through its BIR3, associating with the homodimerization domain of the enzyme, and preventing the conformational change that is necessary for activity .
In addition, XIAP contains a RING domain (Figure 1) involved in cell-death regulation . How this happens, however, has not been conclusively elucidated. Early work with IAPs orthologues in Drosophila suggested that the E3 ligase activity of the RING catalysed a non-degradative ubiquitination step of bound caspase , blocking substrate access to their catalytic sites . A similar paradigm has been proposed for mammalian IAPs, but in this case RING-mediated poly-ubiquitination of caspases 3 and 7 was degradative, and resulted in proteasomal destruction of the modified caspase . Recent evidence reinforced the role of the RING domain in cytoprotection, as mice expressing a BIR-only form of XIAP, thus deleted in the RING domain, exhibited higher caspase activity, and increased cell death, in vivo .
Similar to all other IAPs, except XIAP, survivin does not directly bind caspases . Instead, a prevailing model is that survivin inhibits apoptosis via co-operative interactions with other partners in vivo. An example of these interactions is an IAP–IAP complex between survivin and XIAP . The structure of this recognition is not yet available, but biochemical data suggest that survivin BIR residues 15–38  associate with discontinuous sites in XIAP BIR1 and BIR3 . IAP–IAP complexes may provide a general mechanism to expand the functional repertoire of these molecules, as survivin also interacts with the large IAP BRUCE , as well as c-IAP1 , in the control of cytokinesis and the mitotic spindle checkpoint .
The biological implications of a survivin–XIAP interaction are complex (Figure 2). Current evidence suggests that only a pool of survivin compartmentalized in mitochondria and released in the cytosol in response to cell-death stimuli  has the ability to associate with XIAP, and this recognition is inhibited by survivin phosphorylation on Ser20 by protein kinase A (Figure 2) . Functionally, a survivin–XIAP complex enhances XIAP stability against ubiquitin-dependent degradation, synergistically increases the activity of XIAP for caspase inhibition [37,38], promotes tumour growth in vivo  and directly participates in XIAP-mediated intracellular signalling, in particular NF-κB activation (see below)  (Figure 2). This IAP–IAP complex may also reciprocally control survivin stability, as an XIAP-associated molecule, XAF-1, promotes RING-mediated poyubiquitination and proteasomal destruction of survivin . Other mechanisms of survivin cytoprotection have been proposed, including the ability of mitochondria-localized survivin to sequester pro-apoptotic Smac/DIABLO away from XIAP , or altogether prevent its release from mitochondria , although the functional implications of this pathway have not been clearly defined .
Survivin cytoprotection involves a pathway of cytoplasmic–mitochondrial shuttling and intermolecular co-operation with XIAP
MORE THAN CASPASE INHIBITION: OTHER IAP FUNCTIONS
The idea that IAPs could have functions beyond the control of cell death was first inferred from work with survivin  as it became clear that, in addition to cytoprotection, the molecule had additional roles in cellular homoeostasis. Characterized by a sharp cell-cycle-regulated expression that peaked at mitosis, and subcellular localization to various compartments of the mitotic apparatus , survivin is now unanimously recognized as an indispensable regulator of cell division [48,49]. Differently from all other IAPs, except BRUCE , homozygous deletion of the survivin gene caused early embryonic lethality  and, similarly, conditional deletion of survivin in adult tissues triggered mitotic defects, cell death and tissue involution [51,52]. Evidence collected in other model systems supports this scenario, as putative survivin orthologues in C. elegans [53,54], and yeast  have key roles in mitosis, especially with respect to chromosomal segregation and cytokinesis.
However, teasing out how survivin controls mitosis proved challenging [48,49]. A unifying, albeit not completely satisfying, model for this pathway is that independent pools of survivin localized to various aspects of the mitotic apparatus orchestrate different phases of cell division. As an essential member of the chromosomal passenger complex , survivin physically interacts with Aurora B, borealin and INCENP (inner centromere protein)  to regulate chromosomal alignment, chromatin-associated spindle assembly and cytokinesis . A second pool of survivin has been implicated in stabilization of the mitotic spindle , by binding to polymerized microtubules via its C-terminal α-helices (Figure 1), and actively repressing microtubule dynamics . Independent evidence suggests that this pool of survivin may also participate in the spindle assembly checkpoint and kinetochore–microtubule attachment . How the multiple pools of survivin work together in a seamless continuum at mitosis is not entirely clear, but post-translational modifications play an important role in this pathway. Accordingly, mono-ubiquitination of survivin by both Lys48 and Lys63 regulates its mitotic trafficking in the context of the chromosomal passenger complex , whereas phosphorylation of survivin by mitotic kinases, including p34cdc2/Cdk1 (cyclin-dependent kinase 1) [60,61], Aurora B [62,63] and Polo-like kinase-1 , controls protein stability, subcellular localization, association with protein partners and cytoprotection during the cell cycle.
Another emerging function of IAPs is in the cellular stress response (Figure 3). So far, this has been studied in some detail only for survivin, and whether a similar function applies to other IAPs remains to be explored. With respect to survivin, biochemical studies combined with proteomics screenings identified at least three molecular chaperones, Hsp90 (heat-shock protein 90) , Hsp60  and AIP (aryl hydrocarbon receptor-interacting protein) , that physically interact with survivin in vivo (Figure 3). Based on initial mapping studies, survivin may simultaneously accommodate the binding of at least two of these chaperones, Hsp90  and AIP , as they engage spatially distinct sites, but the cellular implications of a potential survivin–multi-chaperone complex have not yet been established. Functionally, these interactions preserve survivin stability against proteasomal degradation, and inhibit mitochondrial apoptosis [65–67]. However, it is also possible that chaperones help localize survivin to specific subcellular compartments (Figure 3), including mitochondria, as both AIP  and Hsp90  have been implicated in organelle preprotein import.
Role of survivin in the cellular stress response
IAPs AS INTRACELLULAR SIGNAL TRANSDUCERS AND SIGNAL INTEGRATORS
Building on pioneering work that linked XIAP to various intracellular signalling pathways [70,71], it is now clear that IAPs have diverse functions in signal transduction, independently of caspase inhibition . Much emphasis has focused on the role of IAPs as modulators of NF-κB, a pleiotropic gene expression programme , which is pivotal for inflammation, immunity and cell survival [74,75].
Similar to model organisms, for instance Drosophila, where IAP orthologues activate NF-κB , mammalian XIAP is also now recognized as a physiological activator of NF-κB. This pathway is centred on a non-IBM BIR1-dependent recruitment of an activator complex comprising TAK1 [TGFβ (transforming growth factor β)-activating kinase 1] and its adapter protein, TAB1 (TAK1-binding protein) . In turn, this complex facilitates dimerization and activation of TAK1 with subsequent phosphorylation-dependent ubiquitination and proteasomal degradation of the NF-κB inhibitor, IκBα (inhibitor of NF-κB α) . There is also a postulated role for the XIAP RING domain in NF-κB activation, potentially via a non-degradative ubiquitination step , but this activity has not been characterized in detail. Because NF-κB triggers the transcriptional up-regulation of the same IAPs , as well as survivin , this pathway functions as an amplification loop ideally suited to enhance cell survival , especially in cancer, where high NF-κB activity correlates with aggressive disease . In addition, recent evidence has suggested that IAP-mediated NF-κB activation may directly contribute to tumour progression, in particular metastasis . Accordingly, assembly of a survivin–XIAP complex in tumour cells functions as a better activator of NF-κB than XIAP alone, resulting in NF-κB-dependent transcription of the extracellular matrix protein fibronectin . In turn, the newly produced fibronectin engages β1 integrins at the cell surface, with activation of cell motility kinases, Src and FAK (focal adhesion kinase), and dramatically increased tumour cell migration, invasion and metastatic dissemination in vivo, independently of cytoprotection .
Further studies on the role of c-IAPs in NF-κB regulation have uncovered an even greater degree of complexity, with implications for tumour cell survival and novel cancer therapeutics. It had been known that c-IAP1 and c-IAP2 form a complex with TNFR1, and promote TNFα-induced NF-κB activation [81,82] via ubiquitin-dependent stabilization of RIP-1 (receptor-interacting protein-1) kinase . Functionally, this pathway protects cells from the noxious effects of TNFα, as loss of both c-IAPs attenuated TNFα-mediated NF-κB activation [81,82], but also unhindered the assembly of a pro-apoptotic caspase 8-activating complex in the cytosol .
However, it was the more recent characterization of so-called ‘Smac mimetics’ that unravelled a second function of c-IAPs in NF-κB signalling. Smac mimetics are a class of small molecules that reproduce the physical competition of Smac/DIABLO for the caspase-binding site(s) of XIAP, thus eliminating its anti-apoptotic function . Unexpectedly, a brief exposure of tumour cells to these compounds caused sudden degradation of c-IAP1 and c-IAP2 [86,87], with concomitant loss of RIP-1 ubiquitination [81,83]. In turn, this activated NF-κB via the non-canonical pathway [86,87], a mechanism used by certain TNFR family members that involves stabilization of NIK (NF-κB-inducing kinase) . When induced by Smac mimetics in certain tumour cells, non-canonical NF-κB activity enhances the production of TNFα , causing TNFR1- and caspase-8-dependent apoptosis [86,87]. Such a response is attractive for cancer therapy, as production of TNFα confined to the tumour cells may avoid systemic toxicity in vivo. Unfortunately, at least in vitro, only a minority of tumour cells produce TNFα in response to Smac mimetics, and the so-called ‘resistant’ cells do not die unless challenged with exogenous TNFα . Therefore, unexpectedly, c-IAP1 and c-IAP2 act as both activators and repressors of canonical and non-canonical NF-κB signaling respectively, and the balance between these two activities probably controls a broad survival threshold in tumour cells.
IAPs IN CANCER
Given their role in cellular homoeostasis, it is not surprising that deregulated IAP expression or function is frequently associated with human diseases, most notably cancer. In this context, the survivin locus on 17q25 is often amplified in neuroblastoma , whereas the c-IAP1 and c-IAP2 locus on 11q22 is amplified in several epithelial malignancies . Aside from copy number increase, the expression of IAPs is deregulated in many types of cancer, with aberrantly increased protein levels in transformed cells. In this context, survivin is a striking cancer gene, overexpressed in virtually every human tumour examined, whereas largely undetectable or expressed at very low levels in normal tissues . The sharp differential distribution of survivin is unique among IAPs, which are typically found in normal tissues as well, and occasionally further up-regulated in cancer .
The basis for such ‘cancer-specific’ expression of survivin is not completely understood. There is compelling evidence that this reflects transcriptional changes, and several oncogenic pathways have been identified that independently turn on survivin gene expression . Conversely, many tumour suppressor networks have also been shown to exert the opposite effect, and actively silence transcription of the survivin gene, by various mechanisms . It is possible that this finely tuned balance maintains survivin levels low in normal tissues, where tumour suppression mechanisms dominate , whereas transformed cells characterized by oncogene activation and/or loss of tumour suppression may exhibit early deregulation, i.e. induction of survivin gene expression in vivo . Non-transcriptional mechanisms that deregulate survivin expression in cancer have also been described, for instance stabilization of survivin mRNA in an mTOR (mammalian target of rapamycin)-mediated pathway in prostate cancer . Once overexpressed in tumours, retrospective analysis of patient series and genome-wide microarray studies have consistently identified survivin as a risk-associated gene for resistance to therapy, disseminated disease and overall unfavourable disease outcome .
Although there may be one function of survivin pivotally important for disease progression, a more likely scenario is that tumours globally exploit the multifaceted biology of the protein for the broadest advantage in cell proliferation, survival and adaptation. Consistent with this model, deregulation of survivin profoundly affects mitotic transitions in tumour cells, maintaining viability of aneuploid cells , bypassing cell-cycle checkpoints , promoting resistance to microtubule-targeting agents  and co-operating with oncogenes, i.e. Myc, for disease progression . The link between survivin and molecular chaperones (Figure 3) may similarly be important to preserve cell proliferation and cell survival in the face of the highly unfavourable environments characteristic of tumour growth in vivo . This concept is further reinforced by the overexpression of Hsp60 in tumours compared with normal tissues , and the differential subcellular recruitment of Hsp90 to mitochondria of transformed cells . And, finally, there is evidence from transgenic animals that survivin up-regulation during tumour progression in vivo may also occur independently of the cell cycle [100,101], suggesting that the non-mitotic functions of survivin in blocking apoptosis in interphase cells may be also prominently exploited in vivo.
Although these findings reinforce the model that survivin and XIAP confer a broad advantage for tumour growth, the situation for other IAPs, in particular c-IAP1 and c-IAP2, is seemingly more complex. In particular, genomic deletions of the c-IAP1 and c-IAP2 locus have been observed in some types of cancer, for instance multiple myeloma, a condition that would be expected to produce unbridled non-canonical NF-κB activation . Although it is too soon to conclude that c-IAPs contribute to a yet-to-be-elucidated tumour suppression pathway, it is intriguing that unrestrained non-canonical NF-κB activation is observed in other tumours in vivo , suggesting a role for this response in disease progression.
Over the last decade and half, unravelling the biology of IAPs has produced important insights into disparate cellular circuitries of cell survival, adaptation, mitosis and intracellular signalling. Although considered at first to be somewhat redundant endogenous caspase inhibitors, it is now clear that IAPs serve unique and cornerstone functions in cellular homoeostasis. In a little over 10 years, significant progress has also been made in exploiting IAP biology for novel cancer therapeutics [104,105]: no small feat when one considers the excruciatingly long timeline for bringing new agents to the clinic. However, it is also clear that important questions about IAP function remains, for instance how these molecules intersect other signalling pathways, participate in adaptation or regulate the cell cycle, just to name a few. Given the fast pace of IAP research, the answer to some of these questions is undoubtedly forthcoming, helping frame new more rationally grounded strategies for targeting IAPs in human diseases, especially cancer.
aryl hydrocarbon receptor-interacting protein
baculovirus IAP repeat
cyclin-dependent kinase 1
direct IAP-binding protein with low pI
inhibitor of apoptosis
second mitochondrial-derived activator of caspase
TGFβ (transforming growth factor β)-activating kinase 1
tumour necrosis factor
I apologize to all of the colleagues whose work could not be cited for reasons of space constraints.
Work of the authors was supported by the National Institutes of Health [grant numbers CA118005, CA90917, CA78810, HL54131].