New wirings in the survivin networks

Abstract

A little over 10 years after its discovery in 1997, the small inhibitor of apoptosis (IAP) protein, survivin, continues to generate intense interest and keen attention from disparate segments of basic and disease-related research. Part of this interest reflects the intricate biology of this multifunctional protein that intersects fundamental networks of cellular homeostasis. Part is because of the role of survivin as a cancer gene, which touches nearly every aspect of the disease, fromonset to outcome. And part is due to the potential value of survivin for novel cancer diagnostics and therapeutics, which have already reached the clinic, and with some promise. Grappling with emerging new signaling circuits in survivin biology, and their implications in cancer, will further our understanding of this nodal protein, and open fresh opportunities for translational oncology research.

Ten years of survivin

Survivin is the smallest member of the inhibitor of apoptosis (IAP) gene family in mammalian cells (Srinivasula and Ashwell, 2008). Since the cloning of the survivin gene in 1997, there has been mounting interest in the biology and functions of this pathway in several areas of biomedical research, and peer-reviewed publications dealing with various facets of survivin biology have continued to appear at an exponential rate. Survivin is a structural homodimer, containing a single baculovirus IAP repeat, which is the hallmark of all IAPs, and a -COOH terminus coiled coil (Verdecia et al., 2000), with no other identifiable protein domain(s) (Srinivasula and Ashwell, 2008). The regulation of the survivin gene is complex, involving multiple pathways of transcriptional and post-transcriptional control. Located at the tip of chromosome 17 in humans (17q25), a single survivin gene is extensively alternatively spliced to generate several protein isoforms (Sampath and Pelus, 2007), and is transcriptionally controlled in a sharp cell cycle-dependent manner, with peak expression at mitosis (Altieri, 2006; Lens et al., 2006). Examples of survivin gene expression independent of cell cycle periodicity have also been described, especially in response to growth factor and cytokine stimulation (Aoki et al., 2003). There are also mechanisms of post-transcriptional modulation of survivin levels, especially in tumor cells. One example involves stimulation with insulin growth factor-1, which results in stabilization and rapid translation of a pool of survivin mRNA, in a pathway dependent on the mammalian target of rapamycin (Vaira et al., 2007; Oh et al., 2008).

A newly generated survivin protein is post-translationally regulated by degradative (Zhao et al., 2000) and non-degradative cycles (Vong et al., 2005) of ubiquitylation and de-ubiquitylation, as well as multiple phosphorylation events by kinases, including Cdk1 (O'Connor et al., 2002), Aurora B (Wheatley et al., 2007) and protein kinase A (Dohi et al., 2007). These post-translational mechanisms have been largely implicated in survivin protein stability, and in controlling protein trafficking among various subcellular compartments. Survivin (Uren et al., 2000) and BRUCE (Ren et al., 2005), another single-baculovirus IAP repeat IAP protein, are the only two essential members of this gene family, in that their homozygous deletion in mice results in early embryonic lethality, and is incompatible with tissue viability (Okada et al., 2004). Most likely, orthologs of survivin are found in lower organisms, such as yeast, worms and flies, underscoring the evolutionary conservation of this pathway in cellular homeostasis.

Soon after its discovery, there was uncertainty as to the true function(s) of survivin in mammalian cells. Overexpression of survivin in various cellular systems was clearly associated with inhibition of cell death, and, conversely, targeting survivin function and/or expression, especially in tumor cells, caused spontaneous cell death and enhanced the effect of other apoptotic stimuli (Altieri, 2003). However, it was confusing that survivin did not appear to function as a direct caspase inhibitor (Verdecia et al., 2000), which was thought of as the main mechanism of IAP-mediated cytoprotection (Salvesen and Duckett, 2002). Similarly, survivin knockdown in worms, a model organism extensively used to study apoptosis, did not uncover defects in cell death, rather a phenotype of aberrant mitosis (Speliotes et al., 2000). This controversy has been solved over the years with the realization that survivin is a bona fide multifunctional protein. Indeed, survivin is an essential mitotic gene, localized to multiple aspects of the mitotic apparatus, and indispensable for several steps in cell division. This pathway, at least some of its aspects, is highly evolutionarily conserved among various model organisms (Altieri, 2006; Lens et al., 2006). In addition, survivin functions as a cell death inhibitor: this second pathway is operative in both interphase and mitotic cells, relies on intermolecular cooperation with other cofactor molecules (Dohi et al., 2007) and is evolutionarily conserved in certain model organisms, for instance flies (Jones et al., 2000) and yeast (Walter et al., 2006). With respect to the inability of survivin to suppress caspase activity, this turned out to be the rule rather than the exception, as most, perhaps all IAP family members except XIAP, are not physiologic antagonists of caspases in vivo (Eckelman et al., 2006).

In addition to this complex biology, survivin has generated considerable interest as a `cancer gene' (Altieri, 2003). Abundantly found in virtually every human tumor tested, including those with low mitotic index, survivin is conversely expressed at low to undetectable levels in most normal tissues (Fukuda and Pelus, 2006). This sharp differential distribution comes predominantly from the deregulation of survivin gene expression in transformed cells. Although no mutations or polymorphisms have been identified that selectively induce survivin gene transcription in tumor cells, loss of tumor suppressor genes, most notably p53, or the expression of oncogenes, for instance mutated Ras, results in aberrantly increased survivin promoter activity (Altieri, 2003). Although still retaining cell cycle periodicity at mitosis, this causes a much increased expression of survivin in interphase cells, which may be relevant for anti-apoptotic mechanisms in developing tumors, as suggested by transgenic mouse models in vivo (Xia and Altieri, 2006). By molecular profiling, survivin has been consistently identified as a risk-associated gene in various malignancies, carrying unfavorable implications for cancer prognosis, disease recurrence and abbreviated survival. Together with other gene signatures, this information is now being used in the clinic for the risk assessment of breast cancer patients (van 't Veer et al., 2002; Paik et al., 2004). Because of its essential functions, and differential expression in cancer, survivin has been actively pursued as a cancer therapeutic target (Fesik, 2005). Although the portfolio of survivin antagonists currently in the clinic is not particularly large, including antisense sequences, small-molecule transcriptional repressors and inhibitors of protein-protein recognition, encouraging responses were seen in early-phase clinical trials with some of these regimens, warranting further exploration of this strategy for anticancer therapy (Altieri, 2008).

The present contribution will not deal with the specific aspects of survivin biology: several excellent reviews covering these topics have appeared, and the reader is directed to those articles for a more in-depth perspective (Kotwaliwale and Biggins, 2006; Lens et al., 2006; Li and Brattain, 2006; Stauber et al., 2007). Efforts to reconcile the different mechanistic functions of survivin in cell death, cell division and cellular adaptation, and provide a unified framework for how these may participate in disease, have also been reported (Altieri, 2008). Instead, this contribution will focus on three emerging themes of survivin biology, they are: (i) the compartmentalization of survivin signaling pathways in specialized subcellular microenvironments; (ii) the assembly of `cancer-specific' survivin networks; and (iii) the participation of the survivin pathway in the maintenance of `cancer stem cells.' Upcoming research along these lines of investigation is likely to continue to blossom. The new information will undoubtedly uncover new clues for how survivin participates in cellular homeostasis, contributes to cancer onset and progression and can be targeted for novel molecular cancer therapeutics.

Compartmentalization of survivin networks in specialized subcellular `microenvironments'?

A fresh approach to recapitulate the complexity of cellular signaling pathways, especially in cancer, is to use systems biology tools (van der Greef and McBurney, 2005) and build connectivity maps that link together multiple functional networks. This approach (Lamb, 2007) has been used to represent the `nodal' properties of survivin, and tentatively define its position at the crossroads of disparate molecular networks of cell division, subcellular trafficking, transcriptional responses, intracellular signaling and cell death (Altieri, 2008).

But how are these signaling networks regulated so that they can execute clearly distinct cellular functions? One distinguishing feature of the biology of survivin is not only the ability to associate with multiple protein partners but also to localize to disparate subcellular compartments. One of these survivin pools localizes to nuclei of interphase cells. The functional role(s) of nuclear survivin, or its prognostic role in cancer, has not been completely elucidated. However, recent data suggest that this fraction of the molecule is immunochemically and post-translationally unique (Fortugno et al., 2002) and may lack anti-apoptotic properties (Connell et al., 2008), and that it undergoes export from the nucleus by binding to Crm1, an effector molecule of Ran-GTP signaling (Knauer et al., 2007). Consistent with an emerging concept of how survivin may recognize some of its protein partners in vivo (Jeyaprakash et al., 2007), Crm1-dependent nuclear export of survivin involves the monomeric protein, whereas dimeric survivin is apparently not efficiently exported fromnuclei (Engelsma et al., 2007). Another abundant pool of survivin accumulates during cell division and localizes to multiple aspects of the mitotic apparatus. This recognition involves multiple protein-protein interactions, including a direct association of survivin with polymerized microtubules (Altieri, 2006; Lens et al., 2006). Third, survivin localizes to mitochondria, especially in tumor cells (Dohi et al., 2007), accumulating in both the mitochondrial intermembrane space (Dohi et al., 2004a) and the organelle matrix. Finally, there is an abundant fraction of cytosolic survivin in tumor cells, in equilibriumwith trafficking pathways controlling survivin nuclear export (Engelsma et al., 2007) or stimulus-dependent release from the mitochondria (Dohi et al., 2004a). The perspective presented below is to consider these multiple subcellular localizations of survivin as semiautonomous, independently regulated, compartmentalized subnetworks and devoted to unique, specialized functions.

There is experimental evidence in support of this hypothesis. For instance, the mitochondrial pool of survivin, especially in its location in the intermembrane space (Dohi et al., 2004a), appears specifically earmarked to inhibit apoptosis. How this process is regulated may illustrate the concept of a functionally compartmentalized pathway, which is controlled by diverse intracellular signals and protein-protein interactions. The first subnetwork in this pathway deals with how survivin localizes to the mitochondria. Although lacking a recognizable, cleavable mitochondrial targeting sequence, survivin is actively imported to the mitochondria (Ghosh et al., 2008), and this mechanism may involve its physical association with cytosolic chaperones (Fortugno et al., 2003; Kang and Altieri, 2006), which is known to contribute to preprotein import to mitochondria. These include heat shock protein-90 (Hsp90), which docks with the mitochondrial Tom70 receptor import system (Young et al., 2003), or the immunophilin aryl-hydrocarbon receptor-interacting protein, which delivers preproteins to mitochondria via a Tom20 receptor-dependent recognition (Yano et al., 2003). Once in the mitochondria, survivin interacts with another chaperone, Hsp60, and this complex may be required for proper re-folding of survivin after transfer across the mitochondrial membrane (Ghosh et al., 2008). Following an as-yet uncharacterized sorting pathway to the intermembrane space, properly folded mitochondrial survivin has been shown to bind the pro-apoptotic mediator Smac (Sun et al., 2005). Although the role of this complex in apoptosis inhibition has not been completely elucidated, it has been proposed that mitochondrial survivin may help sequestering Smac away from XIAP (Song et al., 2003), or altogether prevent its release from the mitochondria (Ceballos-Cancino et al., 2007). Conversely, disparate cell death stimuli have been shown to induce the release of survivin from the mitochondria into the cytosol (Dohi et al., 2004a). The dynamics of this process, and whether it follows the same requirements as the release of other mitochondrial intermembrane proteins during apoptosis (Green and Kroemer, 2004), have not been elucidated. However, once in the cytosol, mitochondria-derived survivin assembles in an intermolecular complex with its related molecule, XIAP. The formation of this complex promotes increased XIAP stability against proteasomal degradation and cooperative inhibition of caspase activity (Dohi et al., 2004b). The heightened anti-apoptotic environment afforded by this pathway is functionally important, and a survivin-XIAP complex has been associated with accelerated tumor growth in vivo (Dohi et al., 2004a).

The assembly of a functional IAP-IAP complex is not without precedent, as it has already been speculated that these molecules may operate in homo- and hetero-multimeric complexes in vivo. In this context, survivin has been recently shown to associate with the IAP family protein BRUCE in regulating cytokinesis (Pohl and Jentsch, 2008), and several IAPs are found in heteromeric complexes that may be important for the inhibition of apoptosis in vivo (Rajalingam et al., 2006). In the case of a survivin-XIAP complex, this recognition is negatively regulated by a compartmentalized process of phosphorylation by protein kinase A, which occurs in the cytosol but not in the mitochondria. The basis for this differential subcellular phosphorylation is unknown, but may be due to the presence in mitochondria of protein phosphatase 2A, which efficiently dephosphorylate protein kinase A-phosphorylated survivin on Ser20 (Dohi et al., 2007). Regardless of these dynamics, protein kinase A phosphorylation of survivin is expected to alter a discrete binding interface for XIAP, which is comprised between residues 15 and 38. In turn, this results in the loss of XIAP binding, lack of protection against apoptosis and inefficient tumor growth in vivo (Dohi et al., 2007).

Similar to the paradigm of mitochondrial survivin, it may be possible to model the function of survivin at mitosis as a compartmentalized subcellular network, which is regulated by multiple molecular interactions and signaling pathways. One critical step in this network is the proper targeting of chromosomal passenger proteins, INCENP, Borealin, and Aurora B to kinetochores (Lens et al., 2006), a process that depends on survivin (Klein et al., 2006; Vader et al., 2006). The chromosomal passenger complex is essential for chromatid bi-allelic attachment and orientation (Sandall et al., 2006), bipolar spindle formation (Sampath et al., 2004; Tulu et al., 2006), activation of the spindle assembly checkpoint (Carvalho et al., 2003) and completion of cytokinesis (Uren et al., 2000).

Several signaling pathways and protein-protein interactions centered on survivin ensure the proper recruitment of the chromosomal passenger complex in vivo. First, survivin coordinates a unique structural arrangement of the chromosomal passenger complex with the organization of a single functional unit (Jeyaprakash et al., 2007). The first mechanism that directs centromeric targeting of this complex involves sequential cycles of survivin ubiquitylation/deubiquitylation. These modifications do not affect protein stability or promote degradation but control subcellular trafficking, with survivin deubiquitylation by the enzyme hFAM being a pivotal step to direct the chromosomal passenger complex to mitotic chromatin (Vong et al., 2005). A second pathway of regulation involves multiple phosphorylation steps of survivin, potentially with opposite and balancing effects. Phosphorylation of survivin on Thr34 by the mitotic kinase Cdk1 has a positive effect, stabilizes survivin at prometaphase and metaphase against proteasomal degradation and heightens an anti-apoptotic threshold in cells traversing mitosis (O'Connor et al., 2002). On the other hand, phosphorylation of survivin on Thr117 by Aurora B kinase (Wheatley et al., 2004) has a negative effect, lowering its affinity for the interaction with centromeric chromatin (Delacour-Larose et al., 2007; Wheatley et al., 2007). Although it is unclear how survivin becomes dephosphorylated on Thr117, this may be critical to maintain the localization of the chromosomal passenger complex on kinetochores and to ensure progression through anaphase. Last, this pathway involves a physical interaction between survivin and the Ran-GTP effector molecule Crm1 (Quimby and Dasso, 2003). This requires a leucine-rich nuclear export sequence in survivin (Stauber et al., 2007), which, intriguingly, overlaps with the binding interface for other chromosomal passenger proteins identified by X-ray crystallography (Jeyaprakash et al., 2007). The dynamics of these mutual recognitions in vivo are unknown, but it has been proposed that a survivin-Crm1 interaction also participates in proper targeting of the chromosomal passenger complex to kinetochores (Knauer et al., 2006).

Altogether, the two examples presented above underscore how a dedicated set of signaling pathways converge to modulate the function of survivin in the mitochondria and during cell division. The fact that these pathways are physically (mitochondria) or spatial-temporally (cell division) coordinated in specialized subcellular microenvironments appears ideal to selectively regulate these pools of survivin, without affecting other fractions of the molecule in different subcellular compartments. This model may confer a unique `multi-functionality' of survivin, enabling a single gene product to carry out disparate tasks in cellular homeostasis.

Are the survivin networks `tumor-specific'?

There is little ambiguity that survivin is a `cancer gene.' Its sharp differential expression in nearly every human tumor, as compared with normal tissues, underscores how this pathway is broadly exploited for tumor maintenance, and perhaps multiple aspects of tumorigenesis, from initiation to metastasis (Altieri, 2003). As anticipated above, this is also likely important for disease outcome, being survivin expression often linked to resistance to therapy, aggressive tumor behavior and shortened survival. Taken together, this has heightened the interest in survivin as a novel and somewhat unique molecular target for cancer therapy, which is required for tumor maintenance but is at the same time largely undetectable in normal tissues. However, transformed cells are not the only tissue where survivin is abundant. A comparable high level of survivin is practically ubiquitous during embryonic and fetal development, indicating that this pathway is equally important for homeostasis of normal tissues (Adida et al., 1998). This conclusion is reinforced by the phenotypes of various conditional knockout mice: whether dominated by exaggerated apoptosis (Jiang et al., 2005; Leung et al., 2007), catastrophic mitotic defects (Xing et al., 2004) or both (Okada et al., 2004), deletion of survivin is invariably incompatible with tissue viability. Altogether, this gives reason to pause in thinking about survivin-based cancer therapies, given their potential risk for normal tissues (Fukuda and Pelus, 2006).

So, why is it that molecular antagonists of survivin, at least the ones tested so far, appear relatively well tolerated in vivo? (Altieri, 2008) Laboratory reagents, including antisense, small-interfering RNA or dominant-negative constructs, have shown efficacy on tumor cells, shutting off cell proliferation, causing mitotic defects often incompatible with cell cycle re-entry, and triggering, depending on the cell type, spontaneous apoptosis. In contrast, these agents are for the most part harmless to normal cells, and relatively well tolerated when given systemically to mice as experimental anticancer regimens (Altieri, 2003). A similar paradigm appears to hold true, at least until now, for cancer patients. For instance, a small-molecule transcriptional repressor of survivin, YM155 (Nakahara et al., 2007), produced encouraging responses in heavily pretreated cancer patients, and this was associated with only minor and potentially treatment-unrelated side effects, like pyrexia, arthralgia, nausea, fatigue and mucosal inflammation (Tolcher et al., 2006).

A possible model to explain this conundrum may be to consider the survivin networks (Altieri, 2008) as `qualitatively' different in tumor cells, as compared with normal tissues. Following this reasoning, it is intriguing that many of the protein partners that survivin interacts with are, in fact, themselves `cancer genes,' that is molecules that are differentially expressed, deregulated or otherwise functionally exploited in tumor cells. As an example, molecular chaperones Hsp90 (Fortugno et al., 2003) and Hsp60 (Ghosh et al., 2008), which are critical players in the survivin cell death/mitochondrial network, are clearly `cancer genes.' Hsp90 is broadly overexpressed in most tumors, where it exhibits a much higher ATPase activity in vivo (Whitesell and Lindquist, 2005). In addition, Hsp90 is recruited to the mitochondria of tumor cells, where it antagonizes the process of mitochondrial permeability transition and preserves organelle integrity against cell death (Kang et al., 2007). Conversely, mitochondria of most normal tissues do not contain Hsp90 chaperones in vivo (Kang et al., 2007). Similarly, Hsp60-directed mitochondrial protein folding is also exploited in tumors, as reflected by the dramatic overexpression of this chaperone in various cancers in vivo (Ghosh et al., 2008).

If these were truly `cancer-specific' survivin networks, then one would expect that their disabling or therapeutic targeting would have catastrophic effects for tumor cells, while leaving most normal cells unaffected. Initial experimental evidence seems consistent with this model. Accordingly, disabling Hsp90-directed protein folding in the mitochondria using organelle-specific ATPase antagonists resulted in mitochondrial dysfunction with massive tumor cell death, whereas normal cells were not affected (Kang et al., 2007). Similarly, acute silencing of Hsp60 in tumor cells resulted in p53-dependent apoptosis but had no effect on the viability of normal cell types (Ghosh et al., 2008). Altogether, these data suggest that it may be the arrangement of multiple cancer genes in a network, rather than individual gene expression, to confer selectivity of functions in tumor versus normal cells.

Similar considerations may apply to the survivin network at mitosis. Several key survivin-associated molecules in this pathway, including Aurora and Cdk1 kinases, are often exploited in cancer, and have been pursued for novel antimitotic therapeutics in cancer. However, it is the intersection between survivin and the Ran-GTP pathway that again may illustrate a qualitative difference of this signaling network in tumor versus normal cells. As indicated above, survivin intersects Ran-GTP signaling either indirectly, by binding to its effector molecule, Crm1 (Knauer et al., 2007), or, directly, by physically associating with Ran-GTP in S phase and again at mitosis (Xia, F, Canovas, P, Guadagno, T and Altieri, DC, unpublished observations). A survivin-Crm1 complex participates in kinetochore trafficking, whereas a survivin-Ran-GTP interaction is required for proper spindle assembly, by increasing the local concentration of the spindle assembly factor TPX2 to microtubules (Xia, F, Canovas, P, Guadagno, T and Altieri, DC, unpublished observations). The reason why this may be important for `tumor-specific' mitotic signaling is because spindle formation mediated by Ran-GTP relies on downstream effector molecules (Hetzer et al., 2002), which themselves behave as `cancer genes.' Accordingly, Aurora A is upregulated in tumors and has transforming potential (Giet et al., 2005), the microtubule-associated protein HURP is highly expressed in certain human cancers (Koffa et al., 2006) and the checkpoint protein complex BRCA1/BARD1 is critical for DNA damage/repair mechanisms of tumor suppression (Joukov et al., 2006). Some of these molecules are also important for tumor progression, as Aurora A and TPX2 are part of a 70-gene signature predictive of chromosomal instability and poor outcome in various cancers (Carter et al., 2006). Because of its broad oncogenic properties, the identification of survivin as a direct Ran-GTP effector molecule fits well with the paradigm of a `cancer-specific' signaling network. Consistent with this model, acute ablation of Ran-GTP in tumor cell types resulted in loss of survivin levels, extensive mitotic defects, especially at the level of spindle formation, and p53- or Bcl2-independent apoptosis (Xia et al., 2008). Conversely, Ran-GTP silencing in normal cells did not impair mitosis and had no effect on cell viability (Xia et al., 2008). These results are consistent with other data, in which a small-interfering RNA screen identified the Ran-GTP effector molecule TPX2 as a critical regulator of tumor cell viability, especially in cells carrying activating mutations of K-Ras (Morgan-Lappe et al., 2007). The differential sensitivity of tumor versus normal cells to Ran-GTP targeting is not an isolated curiosity. There are numerous examples in which disabling fundamental mitotic regulators is incompatible with tumor cell homeostasis but it has surprisingly little to no effect on normal cells. For instance, loss or inhibition of cell division molecules, including Cdk4 (Landis et al., 2006), Polo kinase (Liu et al., 2006), Aurora B kinase (Harrington et al., 2004) or the kinesin motor Eg5 (Duhl and Renhowe, 2005), resulted in catastrophic defects of mitosis in tumor cells, whereas normal cells were largely unaffected.

Therefore, a unifying model can be formulated in which at least some of the survivin subnetworks assemble through the recruitment of regulatory molecules that are themselves `cancer genes,' which are aberrantly expressed in tumors as compared with normal cells (Altieri, 2008). By extension, one can postulate that such signaling networks are either not assembled or assembled in `qualitatively' different ways in normal tissues by recruiting different effector molecules. The differential and exquisite sensitivities of tumor versus normal cells to the targeting of such `cancer-specific' networks suggest that they may have an essential role in tumor maintenance. This model may fit well with current concepts of tumor cell dependence or `addiction' to selected signaling pathways (Weinstein and Joe, 2006). The extent to which such `oncogene addiction' maintains the malignant phenotype, or its role in tumor cell survival versus differentiation in vivo, has not been fully defined (Jonkers and Berns, 2004). However, therapeutic disabling of certain signaling networks has generated spectacular responses in subsets of cancer patients (Sharma et al., 2007), lending further credibility to the idea that these pathways may provide an Achilles' heel for tumor maintenance. Whether tumor cells become addicted to at least some of the survivin `cancer-specific' networks is currently unknown. However, such possibility may explain why survivin antagonists have limited toxicity for normal tissues, despite the fundamental roles of this pathway in cellular homeostasis. If validated by further experimental evidence, such model would further support the safety of survivin-based molecular antagonists for novel cancer therapeutics (Altieri, 2008).

Does survivin contribute to stemcell maintenance?

There is considerable interest in the therapeutic potential of progenitor/stemcells in tissue regeneration or, conversely, as targets for cancer treatment. The idea that rare `cancer stemcells' may function as a reservoir of the disease (Al-Hajj et al., 2004) and reconstitute the tumor, while defying conventional and perhaps even molecular therapies (Dean et al., 2005), is not without controversy (Kelly et al., 2007). On the other hand, a role of cancer stem cells in the pathogenesis and maintenance of certain malignancies, for instance colon cancer (Radtke and Clevers, 2005) and adult acute myeloid leukemia (Krause and Van Etten, 2007), is amply validated, and is being intensely investigated for other types of tumors (Dick, 2003; Kim et al., 2005). Because of its nodal properties (Altieri, 2008), the idea has been put forward that survivin may contribute to stem cell homeostasis (Fukuda and Pelus, 2006). Accordingly, earlier studies suggested that survivin was required to preserve the viability of hematopoietic progenitor cells (Fukuda and Pelus, 2002), a model further reinforced by the results of conditional knockout studies, in which heterozygous deletion of survivin produced lethal bone marrow ablation with the loss of hematopoietic progenitors (Leung et al., 2007). This pathway may have a direct link to human disease, as bone marrow depletion in myelodysplastic syndrome is associated with lineage-specific methylation and silencing of the survivin gene, and reduced survivin expression in hematopoietic progenitor cells (Hopfer et al., 2007). A role of survivin in stem cell maintenance may extend beyond the hematopoietic compartment. Accordingly, survivin expression has been associated with `stemness' gene signatures of mesenchymal (Taubert et al., 2007), neuronal (Pennartz et al., 2004) and skin (Marconi et al., 2007) progenitor cells.

If survivin helps maintaining progenitor/stem cell homeostasis, what controls its expression in this cellular compartment? There is evidence that multiple gene expression pathways control this response, including developmental signaling mechanisms. In this context, survivin has been identified as a direct transcriptional target of Wnt/β-catenin (Fodde and Brabletz, 2007), which involves the recognition of discrete T-cell factor-4 (TCF-4)-binding elements in the survivin promoter. Functionally, forced expression of non-destructible β-catenin readily increases survivin levels and supports survivin-mediated cytoprotection (Kim et al., 2003). Conversely, Wnt inhibition suppressed survivin expression and produced anticancer activity (You et al., 2004). There is evidence that this gene expression pathway may be important in vivo, as intestinal crypt cells fromTCF-4 knockout embryos are completely devoid of endogenous survivin (Kim et al., 2003), and survivin levels in intestinal crypt cells appear to inversely correlate with the expression of the Wnt negative regulator adenomatous polyposis coli protein (Zhang et al., 2001). We now know that Wnt/β-catenin provides a broad homeostatic role not only in intestinal crypt cells but also in hematopoietic progenitors (Reya et al., 2003). This involves the control of neutrophil granulocytopoiesis (Skokowa and Welte, 2007) as well as survival mechanisms such as inhibition of ionizing radiation-induced apoptosis (Chen et al., 2007; Woodward et al., 2007).

Another gene expression network that may control survivin levels in stem cells involves active Stat-3 (Aoki et al., 2003). This is a pathway that contributes to survivin cytoprotection in various tumor cell types and may favor aggressive tumor behavior in vivo (Nam et al., 2005; Siddiquee et al., 2007). In addition, Stat-3 signaling increases survivin levels in CD34⁺ hematopoietic progenitor cells (Gu et al., 2007) as well as in embryonic stem cells in response to cellular stress (Guo et al., 2008).

Is a similar survivin pathway also operative in cancer stem cells? This question has not yet been experimentally addressed. However, it is of interest that another developmental gene expression pathway exploited in cancer stem cells, that is Notch, targets survivin for upregulation (Lee et al., 2008). Similar to the paradigm of Wnt signaling (Kim et al., 2003), recent studies have shown that survivin is a direct transcriptional target of Notch-dependent gene expression, and this involves the recognition of RPB-Jκ-binding sites in the survivin promoter (Lee et al., 2008). In addition to its developmental role, Notch signaling is often deregulated in cancer (Weng et al., 2004), contributing to cellular transformation (Stylianou et al., 2006), by the expression of disease-causing oncogenes (Sharma et al., 2006), disruption of cell cycle dynamics (Ronchini and Capobianco, 2001), angiogenesis (Keith and Simon, 2007) and emergence of anti-apoptotic mechanisms (Beverly et al., 2005). There is evidence, especially in breast cancer (Liu et al., 2005), that this pathway may be important for the maintenance of cancer stem cells (van Es et al., 2005), in particular for the pathogenesis of ductal carcinoma in situ (Farnie et al., 2007) and in general for the link to high-risk disease with unfavorable outcome (Reedijk et al., 2005). Consistent with this paradigm, Notch upregulation of survivin occurred preferentially, if not exclusively, in estrogen receptor (ER)-negative, but not in ER-positive, breast cancer cell types, and resulted in the inhibition of apoptosis and accelerated mitotic transitions (Lee et al., 2008). Conversely, Notch inhibition by the antagonists of γ-secretase downregulated survivin levels in ER-negative breast cancer cells, caused apoptosis and suppressed tumor growth in mice, with no detectable side effects (Lee et al., 2008). These data may be part of a broader context in which Notch signaling affects the cytoprotective environment mediated by IAP proteins. Accordingly, the transactivation domain of Notch has been recently shown to associate with the RING domain of XIAP, thus blocking binding of the E2 ligase and preventing XIAP's own ubiquitin-dependent proteasomal destruction (Liu et al., 2007).

Although this evidence points to a role of transcriptional induction of survivin in multiple progenitor/stem cell compartments, it is too early to say whether this pathway contributes to the maintenance of cancer stem cells, or, conversely, whether targeting survivin could provide a means to kill these cells in vivo. Should this be the case, the new data would clearly strengthen the rationale of developing survivin antagonists for novel cancer therapeutics (Altieri, 2008). Upcoming research on this topic is likely to conclusively resolve this issue and define potential `cancer-specific' survivin networks selectively operative in the cancer stem cell compartment (Fukuda and Pelus, 2006).

Concluding remarks and future directions

Over the past 10 years, much has been learned about the biology of survivin. Through the efforts of many laboratories, we can now trace with reasonable accuracy some of the signaling requirements of this multifunctional protein, its impact on organ and tissue maintenance and its exploitation in tumor onset and progression. Emerging evidence now suggests that the multiple roles of survivin in cell death inhibition, regulation of mitosis and control of cellular adaptation, which have been difficult to ascribe mechanistically to the same protein, may derive from the regulated function of separate pools of survivin. The model proposed here is that survivin pools are compartmentalized in specialized subcellular microenvironments and recruit other `cancer-specific' signaling molecules for their regulation, and it awaits further experimental validation. This is most likely forthcoming given the fast pace of survivin research, and the new clues will undoubtedly help in further expanding a mechanistic foundation for the development of survivin antagonists as anticancer agents in humans.