刨花水

袖箭

砗磲

砗磲是稀有的有机宝石、白皙如玉，亦是佛教圣物。 砗磲、珍珠、珊瑚、琥珀在西方被誉为四大有机宝石，在中国佛教与金、银、琉璃、玛瑙、珊瑚、珍珠一起被尊为七宝。

Tridacninae

Eros

Yam

Manuka

Durian

Laevis

Poplar

Metis

Pineapple

Raspberry

PTEN signaling in brain: neuropathology and tumorigenesis

Abstract Phosphatase and tensin homolog deleted on chromosome 10 (PTEN) is a tumor suppressor that antagonizes the phosphatidylinositol-3-kinase (PI3K)/Akt/mTOR pathway by functioning as a lipid phosphatase. This ubiquitous and evolutionarily conserved signaling cascade influences numerous functions including cell growth, survival, proliferation, migration and metabolism. Inherited mutations in PTEN cause pleiotropic effects including cancer predisposition as well as a range of neurological abnormalities revealing specialized roles for PTEN in nervous system development and maintenance. Somatic mutations in PTEN occur frequently as late events in sporadic brain tumors. Mouse models based on Pten deletion in the brain have provided insights into the normal functions of Pten in the nervous system as well as the initiation and progression of gliomas. Compromised PTEN function may contribute to gliomagenesis through disrupted regulation of proliferation, migration, invasion, angiogenesis, stem cell self-renewal and regulation of other tumor suppressor pathways such as p53. Clinical findings in high-grade glioma suggest that PTEN gene alterations are associated with poor prognosis and may influence response to specific therapies. Emerging research using specific pharmacological inhibitors of the PI3K pathway may provide novel therapeutic options for the treatment of PTEN-deficient tumors. Introduction PTEN (phosphatase and tensin homolog deleted on chromosome 10), also known as MMAC (mutated in multiple advanced cancers) (Steck et al., 1997) or TEP1 (transforming growth factor β-regulated and epithelial cell enriched phosphatase 1) (Li and Sun, 1997) was originally identified in 1997 as a tumor suppressor gene that was mutated in prostate, breast and brain tumors, including glioblastoma multiforme (GBM) (Li et al., 1997). Loss of heterozygosity (LOH) of chromosome 10q23.3, the chromosomal region containing PTEN, was known to occur frequently in sporadic tumors, and subsequently PTEN somatic mutations were identified in several additional types of human tumors (Simpson and Parsons, 2001). Furthermore, germline mutations in PTEN are associated with a group of inherited disorders collectively known as the PTEN hamartoma tumor syndromes (PHTS) (Eng, 2003). Inherited mutation of PTEN is associated with a wide range of phenotypes with variable penetrance including hamartomas in multiple tissues, and a high risk of breast, thyroid and endometrial cancer, as well as multiple neurological features. As its name suggests, PTEN has homology to protein phosphatases and early studies showed PTEN dephosphorylated serine, threonine and tyrosine residues in peptide substrates in vitro, with particular preference for extremely acidic substrates (Myers et al., 1997). Subsequently, PTEN was shown to act as a phosphatase for the lipid signaling intermediate phosphatidylinositol-3,4,5-trisphosphate (PIP3), removing the phosphate from the three position of the inositol ring (Maehama and Dixon, 1998), to create phosphatidylinositol-4,5-bisphosphate (PIP2), thereby directly antagonizing signaling through the phosphatidylinositol-3-kinase (PI3K) pathway (Figure 1). This evolutionarily conserved pathway regulates diverse cellular processes, including growth, proliferation, survival, apoptosis, metabolism and cell migration as well as specialized functions in the development and maintenance of the nervous system. To date, the lipid phosphatase activity of PTEN remains its most well-characterized physiological role. Figure 1

Simplified representation of the PI3K pathway. A multitude of growth factors bind and activate receptor tyrosine kinases (RTKs) that in turn recruit Class IA PI3Ks directly or through adaptor proteins stimulating the production of PIP3 from PIP2. The phosphatase activity of PTEN reverses this event and thereby directly counteracts PI3K signals. The serine/threonine kinases AKT and PDK1 are recruited to the membrane through binding of their PH domains to PIP3. Following recruitment to the membrane, AKT is activated though its phosphorylation by PDK1 and mTORC2. AKT phosphorylates numerous targets to transduce signals for growth, proliferation and survival. Shown here are two AKT substrates FOXO and TSC2 that are evolutionarily conserved regulators of PI3K signaling. AKT phosphorylates TSC2 and thereby relieves the inhibitory activity of the TSC1/TSC2 complex towards Rheb. This allows active Rheb to accumulate and activate the mTORC1 complex, a central growth regulator. mTORC1 phosphorylates 4E-BP1 and p70S6K to influence ribosome biogenesis and translation efficiency as well as numerous other processes. FOXO proteins inhibit cell cycle transition by activating the transcription of cell cycle inhibitors such as p27Kip1 and the retinoblastoma protein, p130. AKT phosphorylation inhibits FOXO by stimulating its nuclear export and sequestration in the cytoplasm by 14-3-3 proteins. This also promotes cell survival by blocking FOXO-mediated transcription of proapoptotic proteins such as FAS-L, BIM and TRAIL. Furthermore, AKT can phosphorylate the proapoptotic protein BAD causing its inactivation by 14-3-3 binding. In mammals, multiple family members exist for many components of the pathway including three isoforms for each of the Class IA PI3K subunits, three AKT isoforms, three FOXO family members and two S6K isoforms. Full size image The PI3K pathway Phosphatidylinositol-3-kinases are lipid kinases that phosphorylate the three position of the inositol ring of phosphatidylinositols and phosphoinositides (Engelman et al., 2006). This family of kinases can be subdivided into three classes (I–III) according to their substrate preference and sequence homology. Class IA and IB PI3Ks are the primary enzymes that produce the second messenger, PIP3, in response to activation of growth factor receptor tyrosine kinases or G-protein coupled receptors, respectively (Engelman et al., 2006). The major downstream effector of PI3K signaling is the serine-threonine kinase AKT. As with PI3K, there are multiple AKT isoforms in mammals encoded by three separate genes and designated AKT1, AKT2 and AKT3 (also known as PKBα, PKBβ and PKBγ, respectively). All three enzymes (collectively referred to as AKT) respond similarly to PI3K signaling, but have some differences in tissue-specific expression patterns. Mouse knockout models indicate that the three Akt family members play distinct physiological roles in vivo, with some overlapping function and/or compensation (Chen et al., 2001; Cho et al., 2001a, 2001b; Easton et al., 2005; Tschopp et al., 2005). AKT and the serine/threonine kinase PDK1 (phosphoinositide-dependent kinase 1) are recruited to the plasma membrane by binding of their pleckstrin homology (PH) domains to PIP3. Here, AKT is activated by phosphorylation of its activation loop by PDK1 (Alessi et al., 1997), and phosphorylation of its hydrophobic motif by the rapamycin-insensitive mTOR complex (mTORC2), which changes the protein conformation rendering AKT fully active (Sarbassov et al., 2005). In turn, AKT phosphorylates many target proteins to regulate a broad range of cellular processes. AKT signaling can be regulated either by cellular PIP3 pools or by dephosphorylation by the PHLPP phosphatases (Brognard et al., 2007). Downstream of AKT The specific substrates for AKT vary depending on physiological stimuli and cell context, and each AKT isoform has both unique and overlapping targets, although the factors that determine these specificities are not fully understood. This signaling cascade is highly conserved evolutionarily, with FOXO and TSC2 (tuberous sclerosis complex 2) identified as AKT substrates and key downstream effectors by genetic and biochemical studies in nematodes, flies and mammals (Engelman et al., 2006). In mammals, FOXO is a family of transcription factors that includes FOXO1, FOXO3a and FOXO4 (also known as FKHR, FKHRL1 and AFX, respectively), all three of which are expressed in brain (Hoekman et al., 2006; Lein et al., 2007). AKT phosphorylation of the FOXO (forkhead) family of transcription factors negatively regulates their activity by inducing their release from DNA and translocation to the cytoplasm where they are retained by binding to 14-3-3 proteins (Brunet et al., 1999). Inhibition of FOXO-mediated transcription promotes cell survival and increases proliferation by preventing activation of target genes including the proapoptotic genes Fas-L (Fas ligand) (Brunet et al., 1999), TRAIL (tumor necrosis factor-related apoptosis inducing ligand) (Modur et al., 2002) and Bim (Bcl-2 interacting mediator of cell death) (Dijkers et al., 2002), the cell cycle inhibitor p27Kip1 (Medema et al., 2000) and the retinoblastoma (Rb)-related protein, p130 (Kops et al., 2002). The tuberous sclerosis complex (TSC) is composed of TSC1/hamartin, a cytoskeletal protein, and a GTPase-activating protein, TSC2/tuberin. The TSC1/TSC2 complex acts to negatively regulate the activity of the small G protein RHEB (Ras homolog enriched in brain), but phosphorylation of TSC2 by AKT inhibits this activity (Inoki et al., 2002). This allows GTP-bound RHEB to accumulate and activate the rapamycin-sensitive mTOR complex (mTORC1). As part of mTORC1, the kinase activity of mTOR (mammalian target of rapamycin) controls cell growth by regulating many cellular processes, including protein synthesis, ribosome biogenesis, translation efficiency and transcription of downstream target genes among others. This is mediated in part by phosphorylation of 4EBP1 (eukaryotic translation initiation factor 4E binding protein 1) and ribosomal S6 kinase (S6K) (Shaw and Cantley, 2006). mTOR-mediated phosphorylation of S6K, of which two family members exist in mammals, subsequently allows further phosphorylation and activation by PDK1 (the same kinase that activates AKT) (Balendran et al., 1999). In addition to PI3K activation, available amino-acid pools can regulate mTORC1 activity directly and signaling through LKB1/STK11 (serine/threonine kinase 11) and AMPK (AMP-activated protein kinase) can regulate TSC1/TSC2 activity resulting in the inhibition of mTORC1 signaling under energy-poor conditions (Shaw et al., 2004). PTEN is a unique phosphatase Unlike several components of the pathway for which multiple family members exist, there are no PTEN-related proteins that compensate for its loss of function. Therefore, as the sole central negative regulator of PI3K signaling, it is perhaps not surprising that loss of PTEN function has a substantial impact in tumorigenesis and normal development. In addition to lipid phosphatase activity, the protein phosphatase activity of PTEN appears important for regulating its own activity by autodephosphorylation of amino acids in its C-terminal tail (Leslie et al., 2007). The identification of PTEN mutants that have defects in either lipid phosphatase activity, protein phosphatase activity or both have made it possible to define the importance of each on cellular function. For example, the C124S mutation inactivates both lipid and protein phosphatase activity and expression in PTEN-null cell lines suggests that it is unable to function in cell cycle regulation (Maehama and Dixon, 1998). On the other hand, the G129E mutation disrupts the lipid but not protein phosphatase activity of PTEN (Myers et al., 1998). The presence of this mutation in patients with PHTS and sporadic human tumors clearly indicates that loss of the lipid phosphatase activity is sufficient to cause these clinical phenotypes. However, there may be physiological settings where the protein phosphatase activity plays important roles as well, for example in cell migration (Raftopoulou et al., 2004). Germline mutations of PTEN cause neurological abnormalities PHTS are a group of rare clinical syndromes that result from inheritance of a mutated PTEN allele and include Cowden's syndrome, Bannayan-Riley-Ruvalcada syndrome and Lhermitte–Duclos disease (LDD). The diagnostic criteria for each of these disorders have overlapping features, such as the presence of hamartomas, although only Cowden's syndrome includes cancer predisposition (Gustafson et al., 2007). Comprehensive analysis of germline mutations in PTEN reveals missense, nonsense, frameshift and splice site mutations distributed throughout the gene, as well as small deletions or insertions and larger deletions encompassing whole exons, or in some cases whole gene deletions or translocation (Figure 2). The fact that many mutations disrupt the phosphatase domain or result in production of an inactive protein by truncation or instability supports the finding that loss of function of PTEN is the explanation for the clinical features of PHTS. Although the phenotypic features associated with germline PTEN mutation show highly variable penetrance, no consistent genotype–phenotype relationship has been identified. Indeed, the same mutation can cause either Cowden's syndrome or Bannayan-Riley-Ruvalcada syndrome within the same family (Zori et al., 1998) strongly suggesting that the different PTEN-mutation disorders are in fact one disease with a broad and highly variable spectrum of clinical features, as is the case with other autosomal dominant tumor suppressor disorders such as neurofibromatosis type I (Lachlan et al., 2007). Figure 2

Graphic representation of PTEN mutations in central nervous system (CNS) tumors and PHTS. PTEN contains 9 exons (grey) and encodes a 403 amino-acid protein. The domains within PTEN include a PIP2-binding region (PBR, red), a phosphatase domain (green), a C2 domain (yellow), with a C-terminal tail containing two PEST domains for degradation (orange) and a PDZ-interaction motif (blue). Amino acid substitutions (missense mutations) are shown as circles, nonsense mutations are stars and mutations involving DNA insertions or deletions, both in frame or causing frameshifts are squares. Interestingly, many inherited mutations in PTEN (red) are different to those mutations found in sporadic human CNS tumors (blue), although some mutations are overlapping (purple). Not represented here is the frequency of occurrence of each mutation; however, the most frequently observed mutations in CNS tumors are amino acid substitutions at arginine 173 and nonsense mutation at arginine 130 (black outline, arrows). Note that the mutations are distributed along the entire gene, although a clustering of mutations is seen within exons encoding the phosphatase domain. Not included here are more complex mutations including translocations or splice site mutations. Somatic mutations were compiled from the COSMIC database (http://www.sanger.ac.uk/genetics/CGP/cosmic/) and germline mutations were from the Human Gene Mutation Database (http://www.hgmd.cf.ac.uk/). Full size image Lhermitte–Duclos disease is described clinically as a benign overgrowth of neurons in the cerebellum that causes increased intracranial pressure, ataxia and seizure (Zhou et al., 2003). Although patients have inherited one mutant copy of PTEN, the dysplastic cells have either complete loss of PTEN expression or express only the mutant allele, which is accompanied by elevated phosphorylated AKT (Iida et al., 1998; Zhou et al., 2003; Abel et al., 2005). The features of LDD were recapitulated in mouse models where Pten was conditionally deleted late in development of granule neurons of the cerebellum resulting in a cell-autonomous loss of size regulation (Backman et al., 2001; Kwon et al., 2001). In the mouse, the size of Pten-deficient granule neurons progressively increases without evidence of abnormal proliferation. Similarly, the focal lesions in LDD rarely contain proliferative cells, as distinct from malignancies in which the second allele of PTEN has been inactivated by mutation. Further, some dysplastic neurons are ectopically placed in the molecular layer in LDD, similar to ectopically positioned granule neurons resulting from a neuronal migration defect in the mouse model (Backman et al., 2001; Kwon et al., 2001; Abel et al., 2005). Indeed, deletion of Pten in multiple neuronal types during development results in profound defects in migration and patterning in brain (Backman et al., 2001; Groszer et al., 2001; Kwon et al., 2001; Marino et al., 2002; Yue et al., 2005). Finally, abnormalities in synaptic structure have been identified in LDD patients, as well as Pten conditional knockout mice (Kwon et al., 2006; Fraser et al., 2008). Thus, the abnormalities observed in LDD can be attributed to key roles of PTEN in neuronal migration, size regulation and specialized subcellular structure. Although not malignant, this nonproliferative disease resulting from PTEN dysregulation is often associated with substantial morbidity. Other neurological manifestations of inherited PTEN mutation include decreased or delayed learning, such as mental retardation (IQ <75), which is a minor criterion for diagnosis of Cowden's syndrome (Gustafson et al., 2007). Macrocephaly is common and several studies have identified autism or autistic behaviors in PHTS patients with this feature (Goffin et al., 2001; Butler et al., 2005). It is unclear why the neurological phenotypes associated with PHTS are so highly variable. The development of LDD is clearly associated with a second hit that inactivates the wild-type allele of PTEN in the cerebellar lesions. It is possible that other neurological abnormalities, such as macrocephaly, mental retardation and autism, are also associated with second hits that occur stochastically during development. In this case, the timing during development and specific cell populations in which PTEN function is lost would determine the severity and specific neurological deficit. Interestingly, a mouse model in which Pten was deleted selectively in subsets of differentiated neurons in the cerebral cortex and hippocampus showed anxiety-like behavior and decreased learning, perhaps showing some similarities to autistic features of some PHTS patients (Kwon et al., 2006). Alternatively, it is possible that individual variation in other regulators of PI3K signaling may cause neurological dysfunction in the context of PTEN haploinsufficiency. Experimental systems have shown that loss of Pten function can have profound consequences on patterning in the brain, proper neuronal morphogenesis and function, as well as defects in myelination (Fraser et al., 2008; van Diepen and Eickholt, 2008), showing remarkably specialized functions for a pathway that is ubiquitously expressed and evolutionarily conserved. Interestingly, inherited mutation of several other genes involved in PI3K and mTOR signaling can result in familial syndromes also characterized by hamartomas and cancer predisposition. Germline mutations of LKB1/STK11 cause the autosomal dominant Peutz–Jeghers syndrome associated with intestinal hamartomas and increased cancer susceptibility in the gastrointestinal tract, breast, testis and ovary (Hemminki et al., 1998; Katajisto et al., 2007). Also, mutation in either TSC1 or TSC2 results in tuberous sclerosis, an autosomal dominant disorder characterized by hamartomas, susceptibility to renal cell carcinoma and giant cell astrocytomas, and neurological problems such as cortical tubers, autism, seizure and learning disabilities (Young and Povey, 1998). Furthermore, learning deficits were also observed in Tsc1 conditional knockout mice (Zeng et al., 2007). Intriguingly, despite the connections of these genes within converging signaling pathways (Figure 1) and the shared association with hamartomas and tumor predisposition, the specific phenotypic features of the familial syndromes for each gene are quite distinct. Only PTEN is frequently targeted by mutation in sporadic cancer, indicating that disruption of different effectors in the pathway clearly have different physiological outcomes and can contribute differently to neuropathological conditions as well as tumorigenesis. Somatic mutation of PTEN in brain tumors Clearly the PI3K pathway has important functional roles in the nervous system, but what is the principal contribution of PTEN loss to tumorigenesis? PHTS is not associated with an increased incidence of brain tumors, and Pten+/− mice fail to develop brain tumors despite predisposition to multiple other tumor types. Brain tumors were also not observed in conditional knockouts targeting Pten deletion in the brain, indicating that cooperating mutations in other genes are required for the neoplastic process (Backman et al., 2001; Groszer et al., 2001; Kwon et al., 2001; Marino et al., 2002; Fraser et al., 2004). Within the central nervous system (CNS), tumors are classified based on their predominant cell type as determined by morphological and immunohistological criteria (Louis et al., 2007). In the adult, most primary malignant tumors are of glial origin. Astrocytomas are the most common CNS neoplasm and are classified according to a WHO grading system (I–IV). Grade IV astrocytomas are termed GBM, and are the most common malignant brain tumor and one of the most aggressive human cancers, with a mean survival time of less than 1 year after diagnosis (Louis et al., 2007). Loss of 10q, including PTEN, is the most common alteration associated with GBM (70% incidence) (Ohgaki et al., 2004). Clinically, GBM can be separated into two subgroups. Primary GBM arises de novo, more often in older patients and without any pre-existing low-grade lesion, whereas secondary GBM develops progressively from lower-grade astrocytoma in patients who are generally younger at diagnosis (Louis et al., 2007). PTEN mutation occurs more frequently in primary tumors (25%) compared to secondary tumors (5%), although the rate of LOH at 10q is similar between the two groups (Tohma et al., 1998). The spectrum of PTEN mutations in GBM overlaps with that observed in PHTS patients, as well as other tumor types (Figure 2). Interestingly, although PTEN mutations are observed frequently in GBM, they are rare in lower-grade astrocytic tumors (Louis et al., 2007). Although the sequence of mutational events in primary GBM is unknown, in secondary GBM, TP53 mutation is usually an early event occurring frequently in lower-grade astrocytic tumors (Bigner and Vogelstein, 1990); however, PTEN inactivation is not detected frequently until the transition to GBM. This suggests that PTEN loss does not confer a selective growth advantage early in astrocytic tumor development (Figure 3). Figure 3

Genetic pathways involved in glioblastoma multiforme (GBM) development. Analyses of primary tumor samples have shown several pathways are frequently disrupted in the initiation and progression of GBM. Loss of p53 regulation (through mutation, MDM2 amplification or ARF deletion) and perturbation in the RB pathway (such as CDKN2A (INK4a) deletion or RB1 loss) are frequently observed. Activation of growth factor signaling is also common; however, this is predominantly though platelet-derived growth factor signaling in secondary GBM compared with EGF signaling in primary GBM. In secondary GBM it is clear that p53 mutation is involved in tumor initiation, whereas PTEN mutation occurs at a later stage and seems to be involved in conferring an aggressive and invasive phenotype. Note that 10q LOH occurs frequently in both primary and secondary GBM (70 and 65% respectively). Activating mutations in PIK3CA, encoding the p110α catalytic subunit of PI3K, are also observed in GBM. Full size image The frequency of 10q LOH (70%) is much greater than the incidence of PTEN mutation (25%) in GBM, raising the question of whether haploinsufficiency of PTEN may dysregulate PI3K signaling in a substantial proportion of tumors. Alternatively, another tumor suppressor gene also located on 10q could be the target for deletion in tumors that retain one wild-type copy of PTEN. Interestingly, evidence of PI3K pathway activation such as increased phosphorylation of AKT is observed in more than 50% of GBM, suggesting involvement of the pathway in tumors that lack PTEN mutations (Chakravarti et al., 2001). This may be explained in part by an underestimate of PTEN inactivation by incomplete sequencing for mutations in the PTEN promoter or intronic sequences, or by epigenetic silencing of PTEN expression, which has been reported in GBM as well as low-grade gliomas (Baeza et al., 2003; Wiencke et al., 2007). Importantly, the relationship between PTEN promoter methylation and regulation of PTEN expression is inconsistent, indicating that methylation differences must be interpreted with caution (Baeza et al., 2003). Although biallelic inactivation of PTEN function in at least 25% of GBM provides strong evidence that loss of the second allele is required to overcome tumor suppression in that subset of tumors, it is possible that haploinsufficiency of PTEN may provide a selective growth advantage to some tumors. This may be particularly relevant in contexts where other mutations elevate the extent of signaling through the PI3K pathway. PI3K pathway involvement in brain tumors In many cases, it is assumed that loss of PTEN primarily functions through AKT to promote tumorigenesis. However, the different AKT isoforms do not necessarily play redundant roles in oncogenic signaling. For example, Akt2 deletion unexpectedly accelerates tumor formation, whereas Akt1 deletion inhibits tumor formation in two mouse models of breast cancer (Maroulakou et al., 2007). In Pten heterozygous mice, Akt1 deficiency largely prevents development of endometrial and prostate neoplasia, and to a lesser extent thyroid and adrenal neoplasia (Chen et al., 2006). These studies suggest that selective inhibition of AKT1 may be an effective therapy for cancer. However, in the postnatal brain, AKT3 appears to be the critical isoform, as deletion in mice results in decreased brain size (Easton et al., 2005), whereas in humans, haploinsufficiency of AKT3 has been associated with postnatal microcephaly (Boland et al., 2007). It remains unclear, which AKT isoform is the critical effector downstream of deregulated PI3K signaling in brain tumors. Indeed, it is possible that AKT1, AKT3 or both isoforms may play crucial roles in brain tumorigenesis given their shared contribution to embryonic development of the CNS (Yang et al., 2005). Although elevation of AKT activity is seen more frequently than PTEN mutation in gliomas and medulloblastomas (Schlegel et al., 2002; Hartmann et al., 2006), mutations causing increased activity of any AKT isoform have not been observed frequently. AKT1 amplification has been reported in various human cancers, including a single case of gliosarcoma (Knobbe and Reifenberger, 2003) and AKT3 mutation has been reported in a single case of glioma (Hunter et al., 2006); however, the consequence of this missense mutation (G171R) on the activity of the enzyme is unknown. While AKT2 amplification has been observed frequently in head and neck tumors, as well as pancreatic, ovarian and breast cancers (Bellacosa et al., 1995; Cheng et al., 1996; Pedrero et al., 2005; Nakayama et al., 2006), it has not yet been described in primary human brain tumors, although studies indicate it may be overexpressed and drive tumorigenicity in some glioma cell lines (Pu et al., 2006). The increased AKT phosphorylation in many tumors suggests that other events resulting in PI3K pathway activation must also be a contributing factor to neoplasia. In fact, activating mutations in the gene PIK3CA, which encodes the p110α catalytic subunit of PI3K, have been observed in anaplastic oligodendrogliomas, anaplastic astrocytomas, GBM and medulloblastomas, as well as other common malignancies such as prostate, breast and colon carcinoma (Broderick et al., 2004; Samuels et al., 2004). Another study investigating all three genes that encode the catalytic subunits of PI3K (PIK3CA, PIK3CD and PIK3C2B) did not find PIK3CA mutations but did observe PIK3C2B amplification and PIK3CD mRNA overexpression in GBM (Knobbe and Reifenberger, 2003). Interestingly, PIK3CA and PTEN mutations have been observed to occur simultaneously in endometrial tumors and GBM indicating a potential additive effect of both mutations on pathway activation (Broderick et al., 2004; Oda et al., 2005; Hartmann et al., 2005a; Hayes et al., 2006). Intriguingly, 100% (five of five) of GBM with PIK3CA mutation also had 10q LOH (Hartmann et al., 2005a), suggesting the possibility that PTEN may be haploinsufficient when other mutations serve to upregulate signaling through the PI3K pathway, although the small number of tumors involved in this study do not allow a clear conclusion. Other oncogenic signals that occur in GBM such as abnormal receptor tyrosine kinase signaling and RAS activation can also engage the pathway to drive neoplastic transformation (Zhu and Parada, 2002). For example, PTEN mutation often occurs together with aberrant growth factor receptor activity in tumors, such as amplification of the epidermal growth factor receptor (EGFR) gene and/or activating mutations in the receptor. Increased signaling in response to platelet-derived growth factor has also been observed. Other mutations that are involved in glioma development and progression include those that effect p53 function, including TP53 gene mutation, MDM2 amplification/overexpression or CDKN2A deletion targeting the p14ARF tumor suppressor protein; or those that influence cell-cycle progression, such as RB1 alteration, CDK4 overexpression or CDKN2A deletion resulting in loss of the CDK inhibitor p16INK4A (Chow and Baker, 2006) (Figure 3). While PTEN mutations in glioma have been intensely investigated, the importance of PI3K pathway activation in medulloblastoma is more recently becoming better understood. Medulloblastomas represent the most frequent solid malignancy in children with an incidence of five cases per million (Louis et al., 2007). Large scale analysis of genomic gains and loss from over 200 medulloblastoma cases found that loss of 10q is a frequent event (Rickert and Paulus, 2004). Despite this, only rare cases of PTEN mutation have been identified in these tumors (Rasheed et al., 1997), instead observing lower levels of PTEN mRNA expression, potentially caused by epigenetic modifications (Hartmann et al., 2006). An important role of PI3K pathway activation in medulloblastoma was recently suggested for tumors driven by mutations in the Shh pathway. Shh secretion by Purkinje cells drives proliferation of granule neuron precursors (CGNPs) during development of the cerebellum, and activating mutations in this pathway (for example by inactivating the Shh-antagonist PATCHED1) are associated with the desmoplastic subtype of medulloblastoma. Activating mutations in PIK3CA have been observed in medulloblastoma (Broderick et al., 2004) and PI3K/Akt signaling enhanced the proliferative effects of Shh in these cells (Hartmann et al., 2005b), in addition to increasing tumor incidence in mouse models (Rao et al., 2004). Moreover, a recent report showed loss of Pten in CGNPs, together with Shh expression, also increased medulloblastoma incidence in mice, confirming that the PI3K pathway can contribute to tumorigenesis in this pediatric malignant brain tumor (Hambardzumyan et al., 2008). Prognostic significance of PTEN mutation Although some clinical studies have suggested that PTEN mutation in glioma has no correlation with survival (Kraus et al., 2000; Smith et al., 2001; Rich et al., 2005), elevated AKT activity has been associated with poor prognosis (Ermoian et al., 2002). In support of the latter, multiple studies have generally associated loss of function of PTEN with a more adverse outcome. Pediatric patients harboring PTEN mutation in tumors have poorer prognosis (Raffel et al., 1999; Phillips et al., 2006), whereas in adults several groups have shown that LOH at 10q has negative correlation with survival (Lin et al., 1998; Tada et al., 2001; Homma et al., 2006). Additionally, decreased PTEN mRNA expression or low PTEN protein levels are also linked with decreased survival (Sano et al., 1999; Ermoian et al., 2002; Phillips et al., 2006). Unfortunately, many of these studies lack the sample size or thorough evaluation of PTEN genetic alterations to make concrete conclusions. To precisely evaluate the genuine prognostic significance of PTEN function in CNS malignancies, comprehensive analysis of tumors at the genetic, epigenetic, transcript and/or protein levels on a large number of morphologically well-documented cases is required. Specific functions of Pten in brain tumors The diverse phenotypes that are observed upon PTEN loss in different tissues illustrate that it regulates many critical processes depending on the cellular environment; therefore the contribution of PTEN in tumorigenesis is likely also context dependent. There is no specific evidence to suggest any one of these activities alone is sufficient for tumorigenesis. Instead, it is likely that PTEN is frequently targeted in neoplastic transformation because it impinges on multiple critical pathways. Proliferation In contrast to the effects of PTEN deletion on neuronal cell size, deletion in astrocytes results in increased proliferation, without influencing cell size in vitro (Fraser et al., 2004). This effect on proliferation rather than cell size is also seen in other cell types, such as hepatocytes, B cells, T cells and keratinocytes (Suzuki et al., 2001, 2003a, 2003b; Horie et al., 2004). Interestingly, Tsc1-deficient astrocytes display increased cell size without substantial changes in proliferation, suggesting that Akt activation by Pten deletion has additional effects in astrocytes separate from mTORC1 activation (Uhlmann et al., 2004). At a simple level, it has been suggested that Pten will influence cell size in terminally differentiated cells that are unable to re-enter the cell cycle, despite an aberrant growth signal, whereas if deleted in cycling cells, it will accelerate proliferation without influencing size. However, there are examples in the brain where Pten deletion influences both size and proliferation in the same cell type, for example in neural progenitor cells (Groszer et al., 2001; Marino et al., 2002). The regulation of cell size is naturally coupled with cell division, as cells need to reach a critical size before entering the cell cycle. The coordination of these events ensures that adequate nutrients are available and is achieved by the intersection of the nutrient-sensing pathway with the PI3K pathway, which both signal through mTOR. Therefore, signals increasing growth may drive cells more rapidly through the cell cycle or even drive quiescent cells into cycle if there are no major restrictions within the cell to cell-cycle progression. Indeed, deletion of Pten in neural progenitor cells decreased the number of cells in the G0 (quiescent) phase of the cell cycle, resulting in an increase in the number of cells cycling at a given time (Groszer et al., 2006). Several studies have reported that re-expression of PTEN in PTEN-deficient glioblastoma cell lines suppresses proliferation in vitro. This growth-suppressive activity is due to arrest of cells in the G1 phase of the cell cycle (Furnari et al., 1998; Gottschalk et al., 2001). The limitation of these studies is that the level of PTEN expression is often not physiological, making conclusions difficult. Activation of the PI3K pathway has been shown to regulate cell-cycle progression directly through AKT-mediated phosphorylation of cell cycle inhibitors, such as p27Kip1 resulting in its cytoplasmic retention (Liang et al., 2002; Shin et al., 2002; Viglietto et al., 2002) or through indirect means, like FOXO-mediated transcriptional regulation (Medema et al., 2000; Stahl et al., 2002). Indeed, p27Kip1 appears to be a key regulator of the effects of PTEN loss in vivo, as simultaneous inactivation of one Pten allele and one p27Kip1 allele increases tumor incidence in a mouse model of prostate cancer (Di Cristofano et al., 2001). Whether this is also the case in brain tumors is unknown. Migration and invasion A distinctive feature of GBM is their invasive growth that leads to diffuse infiltration of adjacent normal brain, making complete surgical removal impossible. There are several lines of evidence implicating PTEN in the regulation of cell migration and invasion. Pten deletion in early neural precursors in vivo caused profound neuronal migration defects (Backman et al., 2001; Kwon et al., 2001; Marino et al., 2002; Fraser et al., 2004; Yue et al., 2005). In mammalian fibroblasts, Pten influences migration by regulation of Rac1 and Cdc42 in a lipid phosphatase-dependent manner (Liliental et al., 2000). However, many studies suggest that the effects of PTEN on migration are independent of lipid phosphatase activity, with one study indicating that the C2 domain was required (Raftopoulou et al., 2004) and another implicating the protein phosphatase activity of PTEN (Leslie et al., 2007). However, many of these studies were performed in vitro under conditions that lack the complexity of extracellular cues that are found within a structure such as the brain. It is possible that the effects of PTEN on cell migration in vitro are dependent on experimental conditions. Indeed, studies suggesting that the lipid phosphatase activity of PTEN regulates migration by RAC1 inhibition were performed on fibronectin and mediated by α5β1 integrin (Liliental et al., 2000); however, Dey et al. (2008) studied glioma cell migration on vitronectin, which binds αvβ3 integrin, and showed that PTEN's protein phosphatase activity negatively regulated RAC1 indirectly by regulating the activity of the SRC-family kinase, FYN. It has also been suggested that PTEN may regulate cell migration by directly dephosphorylating FAK in the DBTRG-05MG glioblastoma cell line (Tamura et al., 1998); however, other studies have shown that PTEN does not influence FAK phosphorylation in other cell lines (Maier et al., 1999; Jones et al., 2001). The process of tumor cell invasion and metastasis involves more than the mechanics of cell motility. In vitro studies with glioma cell lines deficient in PTEN expression showed that inhibitors of PI3K (LY294002 and wortmannin), or overexpression of PTEN, reduced cell invasiveness in vitro, and this was associated with reduced activity of matrix metalloproteinases including MMP2 and MMP9 (Kubiatowski et al., 2001; Koul et al., 2001). It is not clear whether PTEN genuinely regulates cell migration, tumor invasiveness and metastasis in vivo using the mechanisms and pathways defined by in vitro systems. Importantly, if PTEN protein phosphatase activity proves to be significant in controlling the invasiveness of tumor cells in vivo, this function will not be affected by inhibitors of PI3K signaling, such as rapamycin, that target the downstream effectors of the pathway. Angiogenesis A histological characteristic of GBM is the presence of aberrant microvascular proliferation (Louis et al., 2007). Tumors greater than 1–2 mm in diameter require neoangiogenesis to sustain their expansion (Folkman, 2006). Recent evidence suggests that PTEN is potentially important in the control of angiogenesis within brain tumors. Reintroduction of PTEN into the PTEN-deficient U87MG glioblastoma cell line caused dramatically decreased tumor growth when implanted subcutaneously into mice without having a significant effect on the proliferation of these cells in vitro. This study suggested that this was due to decreased recruitment of blood vessels to the tumor and showed that these cells had increased expression of a negative regulator of angiogenesis, thrombospondin 1 (Wen et al., 2001). Tumor angiogenesis can also be initiated by hypoxia, which stabilizes hypoxia inducible transcription factor 1α (HIF1α), which in turn forms a dimer with HIF1β and upregulates certain growth factors, such as vascular endothelial growth factor and its receptors (Folkman, 2006). In addition to hypoxia, increased PI3K signaling, such as through PTEN loss or EGFR overexpression, can also facilitate glioma growth by inducing HIF1 activity (Maity et al., 2000). Interestingly, even under nonhypoxic conditions, activation of AKT can induce transcription of HIF1-target genes (Zundel et al., 2000), whereas expression of PTEN in glioblastoma cell lines can inhibit them (Pore et al., 2006). Included among HIF1-target genes are glycolytic enzymes that enable cells to adapt to hypoxia by switching from oxidative to glycolytic metabolism. Therefore, in addition to influencing hypoxic growth and angiogenesis, the effects of PTEN loss on HIF1 activity may also regulate metabolism to facilitate neoplastic transformation. Stem cell regulation Evidence suggests that brain tumors, similar to several other malignancies, contain a cancer cell hierarchy reminiscent of cells in the normal developing brain (Singh et al., 2004). These cancer stem cells (CSCs) have the ability to self-renew while producing progeny that make up the bulk of the tumor but have limited replication potential. This hypothesis may provide some explanation for recurrence as, even if treatment is able to eliminate the majority of a tumor, failure to eliminate CSCs sets the stage for the regrowth of the tumor following cessation of chemotherapy. Indeed, brain tumor stem cells isolated by expression of the stem cell marker CD133 have been shown to be relatively resistant to radiation by preferential activation of the DNA damage response (Bao et al., 2006). Pten loss has been shown to enhance the number of tumor initiating cells in a mouse model of leukemia (Yilmaz et al., 2006) and in solid tumors such as breast cancer (Zhou et al., 2007). As deletion of Pten has been shown to regulate neural stem cell self-renewal (Groszer et al., 2001) and proliferation (Li et al., 2002; Groszer et al., 2006), it is possible that PTEN loss may also influence the proportion of CSCs within brain tumors. Alternatively, animal models suggest that progenitor cells are more susceptible than mature astrocytes to transformation by some oncogenic stimuli (Holland et al., 2000). Thus, deletion of PTEN may promote tumor initiation by increasing the available pool of target cells with the greatest capacity for tumorigenesis. p53 regulation PTEN and TP53 are the two most frequently mutated genes in human cancer and several studies suggest that they are functionally connected. p53 can activate transcription of PTEN (Stambolic et al., 2001). In some cells, loss of Pten can increase the expression of Mdm2, a negative regulator of p53 (Chang et al., 2004). Further, MDM2 is a substrate of AKT, thus activation of AKT upon PTEN loss results in MDM2 phosphorylation and increased nuclear import to enhance p53 degradation (Mayo et al., 2002). This regulation is further complicated by the fact that PTEN can physically associate with p53, thereby enhancing its DNA binding ability (Freeman et al., 2003). These data indicate that loss of PTEN would reduce p53 activity. This influence of PTEN dose on p53 levels is supported by mouse models in which lymphoma development is similar in Pten+/−; Tp53+/− and Tp53−/− mice (Freeman et al., 2003). In direct contrast, Pten loss has been suggested to be detrimental to some cell types by increasing p53 protein levels and inducing senescence, the complete and irreversible cessation of cell division (Chen et al., 2005). This phenomenon was observed in a mouse model of prostatic intraepithelial neoplasia where combined deletion of Pten and Tp53 was synergistic, causing much more aggressive tumors with decreased latency. The significant progression of tumors deficient in both genes suggests that loss of Pten and p53 are required for maximal tumorigenicity. Certainly, both genes are often concomitantly deleted in other tumor types, including glioma, supporting this notion. The finding that PTEN loss is most often seen as a late stage event in the progression of grade III astrocytoma to GBM may lend weight to the hypothesis that the timing of PTEN loss is critical for tumor progression. It could be inferred that if PTEN function were lost before p53 is deleted, then the consequence would be increased p53 activity and senescence. It is likely that additional cell type-specific signaling pathways influence these phenomena, so additional experiments are needed to clarify this connection. Mouse brain tumor models Much of our understanding of brain tumor pathways has come from cell culture experiments, which are often the first step in identifying aberrant signals that drive tumor growth. However, because in vitro conditions do not fully recapitulate the nervous system environment, in vivo evolution of a tumor may impose different selection pressures, which may also vary depending on tumor grade. Tumor cell lines cultured for decades differ greatly from the tumors that gave rise to them. Even cells directly cultured from biopsies rarely resemble their original tumors, although altered culture conditions may serve to maintain the genotype, gene expression profile and biology of parental primary tumors (Lee et al., 2006). Nevertheless, in vivo modeling provides a more accurate means of studying human cancer. Most commonly established in mice, this can include xenograft tumor models or models of spontaneous tumor formation in genetically engineered animals. Most models of spontaneous brain tumors have been made by targeting expression to specific cells of genes that mimic the effects of common mutations in human brain tumors. For glioma models, genetic alterations have been targeted to progenitor cells that express nestin, or cells belonging to the astrocyte lineage, which may express the markers Gfap (glial fibrillary acidic protein) or S100β. These experimental systems have provided several clues for the role of Pten in gliomagenesis. Conditional astrocytic expression of a truncated SV40 T antigen, which inactivates the Rb family of proteins, caused high-grade astrocytoma development in adult mice. Despite high levels of proliferation, these tumors exhibited significant apoptosis, but also had areas of better cell survival when phosphorylated Akt was present (Xiao et al., 2002). Furthermore, reducing Pten expression in this system using cre-mediated deletion accelerated tumor incidence and resulted in gliomas that were more invasive, had increased angiogenesis and decreased apoptosis without a substantial impact on proliferation (Xiao et al., 2005). Interestingly, Pten deletion accelerated tumor onset and increased tumor burden through alternative mechanisms, causing increased proliferation with no decrease in apoptosis, in an oligodendroglioma model driven by v-erbB expression combined with Pten deletion (Chow LM, Weiss W and Baker SJ, in preparation). In an alternative model, expression of a mutant activated Ras protein (Hras) in astrocytes, using the GFAP promoter, generated mice which developed low-grade astrocytomas that progressed to higher-grade tumors later in life (Ding et al., 2001). Interestingly, it was found that Pten loss was associated with the higher-grade tumors, whereas expression was retained in the low-grade tumors, very similar to the timing of PTEN deletion in human astrocytoma. Other mutations associated with human astrocytoma progression, such as Cdkn2a (Ink4a) loss or EGFR overexpression, were also seen (Shannon et al., 2005). The role of Pten, or activation of EGFR signaling, for increasing the severity of Ras-driven tumors was elegantly confirmed in vivo by targeted deletion of Pten using adenovirus expressing cre recombinase, or expression of a constitutively active EGFR (EGFRvIII) (Wei et al., 2006). Similarly, the combined activation of Ras and Akt (using KRas and myristylated Akt) also led to the formation of GBM in mice. Interestingly this system, which utilizes the RCAS/tv-a system of gene delivery, showed that, while infection of nestin-expressing cells produced tumors in a proportion of mice, if the same genes are targeted to Gfap-expressing cells, this was not tumorigenic (Holland et al., 2000). This may provide some clues as to the relative susceptibility of particular cell types to neoplastic transformation by specific oncogenic signals. Furthermore, it was shown that while inhibition of mTor signaling using the inhibitor CCI-779 induced apoptosis within the tumor, this was specific to the astrocytic cells within the tumor, resulting in a switch from astrocytoma to oligodendroglioma (Hu et al., 2005). Thus, activation of the PI3K pathway in different cell types has specific outcomes, as does inhibition, highlighting the complexities of the pathway in vivo and illustrating that selective inhibition of specific effectors in this pathway may cause unexpected results in cancer therapy. The limitations of several animal models is that the genes driving tumorigenesis, such as mutated Ras, SV40 T antigen and myristylated Akt, are not the same mutations found in human tumors, and this may alter the feedback and cross-talk of PI3K pathway activation and regulation. Preclinical and clinical studies of PI3K inhibition In addition to providing clues for the role of Pten in tumorigenesis, mouse models are integral in drug development. Historically, cancer cell lines such as U87 have been used to study the biology of brain tumors and employed as preclinical models for screening potential therapeutic agents. Although the Food and Drug Administration does not require study of spontaneous tumor models before the initiation of clinical trials, experimental artifacts can arise using xenograft models for preclinical testing of novel drugs. This may partly be because the surrounding environment of CNS-derived tumor cells implanted subcutaneously, can change tumor properties. Also the delivery and effective concentrations of chemotherapeutic compounds to a subcutaneous xenograft may be different than to tumors arising in brain. The phenotypic and genotypic differences of primary tumors and cell lines also likely contributes to the fact that xenograft models based on cell lines have been poorly predictive for identifying clinically useful therapeutic agents for GBM. The selective pressures that drive tumorigenic progression in brain will likely influence the outcome of selectively inhibiting key signaling effectors. Since in vitro and in vivo studies have clearly established a role of PI3K signal activation in the development of tumors such as GBM, then it seems feasible that restoration of PTEN function in mutant tumor cells would be advantageous for treatment. Indeed, re-expression of PTEN in glioma cell lines can sensitize them to DNA damaging agents such as etoposide (Mayo et al., 2002). Although inhibitors of PI3K, such as LY294002 or wortmannin, are extensively used as tools in research, several pharmacological characteristics such as insolubility or instability in vivo limit their potential for clinical application. The most extensively used drug that targets this pathway is the macrolide fungicide rapamycin. In addition, analogs of rapamycin, CCI-779 and RAD001 have been synthesized with improved pharmacological properties. Originally used as immunosuppressive agents, these drugs have been shown to have antitumor activity and function by forming a complex with FK506 binding protein 12, which inhibits the mTORC1 complex (Bjornsti and Houghton, 2004). Numerous studies suggest that mTOR inhibition can induce cell cycle arrest or apoptosis in PTEN-deficient cell lines in vitro or in xenograft models. However, the usefulness of these inhibitors in clinical studies has been difficult to demonstrate using conventional criteria for clinical response. This may be because the action of mTOR inhibition is mostly cytostatic, thus not affecting tumor burden. Studies have also been limited by not selecting patients that are more likely to respond based on the genetics of their tumor, such as patients harboring tumors with PTEN mutations. A recent phase I clinical trial carefully selected GBM patients based on the absence of PTEN protein using immunohistochemistry. Using pre- and post-treatment tissue sampling they showed that treatment with rapamycin could inhibit mTORC1 activity within tumors of a subset of patients and this correlated with reduced proliferation (Cloughesy et al., 2008). However, in half of the patients, rapamycin treatment led to increased levels of phosphorylated AKT and shorter time to tumor progression. Inhibition of mTORC1 likely relieved feedback inhibition within the PI3K pathway, paradoxically causing increased signaling through AKT, which lies upstream of mTORC1 (Um et al., 2004; O'Reilly et al., 2006). Therefore, inhibitors of the pathway that act further upstream may have improved outcomes. Indeed, a compound that inhibits both mTORC1 and PI3K (PI-103) has been shown to be more effective than rapamycin on the inhibition of glioma cell line growth in xenograft models (Fan et al., 2006; Raynaud et al., 2007). Clinical and preclinical studies have highlighted the value of combination therapy to target multiple pathways simultaneously. Determining which other pathways should be targeted will ultimately depend on the genetic makeup of an individual tumor for specifically tailored therapy. In GBM, increased receptor tyrosine kinase activity and PI3K pathway activation are common. EGFR is amplified, mutated or both in a number of neoplasms including glioblastoma. Furthermore, a subset of patients responds favorably when treated with small molecule inhibitors of the receptor, such as erlotinib or gefitinib (Rich et al., 2004; Prados et al., 2006). Many patients, however, do not respond to treatment and a recent study suggested that poor response in GBM patients was associated with PTEN-deficient tumors (Mellinghoff et al., 2005). Precedence for this had already been established in breast cancer cell lines (Bianco et al., 2003) and PTEN loss was associated with resistance to ErbB2 inhibitors in breast cancer patients (Nagata et al., 2004). It was therefore hypothesized that resistance to receptor inhibitors was due to uncoupling of EGFR signaling from the PI3K pathway downstream and in vitro studies showed that combined EGFR/mTOR inhibition was effective at inhibiting cell growth independently of PTEN status (Wang et al., 2006). Similar results were found by combining RAD001 treatment with a drug that inhibits both EGFR and vascular endothelial growth factor receptor 2 (Goudar et al., 2005), as did the combination of PI-103 with erlotinib (Fan et al., 2007). A molecular explanation for this was recently elucidated in GBM cell lines. She et al. (2005) found that signaling downstream of EGFR and PI3K results in phosphorylation and inhibition of the proapoptotic protein BAD either directly by AKT or by MAPK (mitogen-activated protein kinase) signaling. As this occurs on separate residues, phosphorylation at either site could inhibit BAD by 14-3-3 binding, and only by inhibiting both pathways simultaneously could BAD-mediated cell death occur. Conclusions Since the discovery of PTEN as a tumor suppressor gene over a decade ago, the importance of PI3K signaling in normal development and cancer has been highlighted through experimental systems and the study of human disease. Despite the ubiquitous expression of many proteins within this pathway, the various outcomes resulting from its dysregulation are highly dependent on cellular context, with PTEN having specific functions in nervous system development and maintenance. Questions remain regarding how different physiological stimuli that activate PI3K signaling cause selective and diverse effects. A greater understanding of the key players that dictate these responses is required to design effective therapeutic intervention for brain tumors and neurological abnormalities resulting from aberrant PTEN/PI3K pathway regulation.

Kazal domain

The Kazal domain is an evolutionary conserved protein domain usually indicative of serine protease inhibitors.

Tissue-Specific Splicing of Omi Stress-Regulated Endoprotease Leads to an Inactive Protease with a Modified PDZ Motif

Omi is a human serine protease whose catalytic domain is homologous to a bacterial heat shock endoprotease (HtrA), a protein indispensable to the survival of bacteria at elevated temperatures. Omi is expressed ubiquitously, and its protein product is predominantly localized in the endoplasmic reticulum of mammalian cells. Here we present the genomic structure of Omi, consisting of eight exons located on human chromosome 2p12–p13.

The mitochondrial serine protease HtrA2/Omi

The HtrA family refers to a group of related oligomeric serine proteases that combine a trypsin-like protease domain with at least one PDZ interaction domain. Mammals encode four HtrA proteases, named HtrA1–4. The protease activity of the HtrA member HtrA2/Omi is required for mitochondrial homeostasis in mice and humans and inactivating mutations associated with neurodegenerative disorders such as Parkinson's disease. Moreover, HtrA2/Omi is released in the cytosol, where it contributes to apoptosis through both caspase-dependent and -independent pathways. Here, we review the current knowledge of HtrA2/Omi biology and discuss the signaling pathways that underlie its mitochondrial and apoptotic functions from an evolutionary perspective. The evolutionarily conserved high-temperature requirement (HtrA) family of oligomeric serine proteases has been classified in family S1B of the PA protease clan in the MEROPS protease database (http://merops.sanger.ac.uk), and its members are characterized by the combined presence of a trypsin-like protease domain and one or two C-terminal PDZ domains (Figure 1a). The PDZ domain functions as a protein–protein interaction motif that preferentially binds C-terminal peptides of the target protein to stabilize interactions and modulate the proteolytic activity of the trypsin-like protease domain.1 The bacterial HtrA family members have been implicated in stress tolerance and pathogenicity.2 Although the functions of their eukaryotic homologs have been less well studied, it has become apparent in recent years that the human HtrA member HtrA2/Omi executes essential roles in the mitochondria and contributes to apoptosis through caspase-dependent and -independent mechanisms. Here, we review the current knowledge of HtrA2/Omi biology and discuss its mitochondrial and apoptotic functions from an evolutionary perspective.

Domain organization and phylogenetic analysis of human HtrA2/Omi. (a) Full-length HtrA2/Omi consists of five functional domains and motifs: the N-terminal mitochondrial localization signal (MLS), the transmembrane (TM) segment, the IAP-binding motif (IBM), the serine protease domain and the C-terminal PDZ domain. Amino-acid substitutions associated with Parkinson's disease and the Parkinsonian phenotype of the Mnd2 mice are indicated below the functional domains in italic and bold, respectively. The catalytic triad residues are depicted above the functional domains. Phylogenetic Analysis of HtrA2/Omi and its Homologs Members of the HtrA family are present in nearly all bacterial and eukaryotic genomes, with no less than eight paralogs identified in the α-proteobacterial species Mesorhizobium loti.In contrast to the phylogenetic domains of Eukaryota and bacteria, HtrA homologs are absent from nearly all archaean genomes.In line with Margulis's endosymbiosis theory,these findings support a monophyletic origin of eukaryotic HtrA proteases in a mitochondrial ancestor of the α-proteobacterial lineage.3 The apparent absence of HtrA proteases from the bacterial class Mollicutes, to which the human parasites Mycoplasma pneumonia and M. genitalium belong, and the presence of many HtrA homologs in the related phylogenetic classes Clostridia and Bacilli strongly suggests that Mycoplasma species lost their HtrA-encoding genes after their diversification from the remaining classes of the Firmicutes. In the animal kingdom, HtrA-like genes are absent from all sequenced genomes of the phylum Nematoda, including that of the well-studied model organism Caenorhabditis elegans. These findings are in marked contrast with a previous report that suggested the presence of six nematode HtrA genes, but failed to provide further information.Our studies indicate that nematodes lack genes encoding trypsin-like protease domains, although PDZ-containing proteins are present. A possible explanation for the apparent discrepancy is that PDZ-encoding genes without trypsin-like domains were selected in the study of Koonin and Aravind.As mutations in the genes encoding HtrA proteins correlate with decreased bacterial fitness,perinatal lethality in mice5 and human Parkinson's disease,the apparent lack of HtrA homologs in mycobacteria and nematodes suggests the functional convergence of structurally unrelated proteins in the latter organisms. In contrast to nematodes, the arthropod Drosophila melanogaster and the amphibian model organism Xenopus tropicalis encode an HtrA homolog in their respective genomes. Animals of the vertebrate lineage have expanded their repertoire of HtrA homologs, with four paralogs present in humans and mice. Whereas HtrA2/Omi resides in the mitochondrial intermembrane space (IMS), its paralogs HtrA1, 3 and 4 are most likely targeted to the secretory pathway. Indeed, whereas the HtrA2/Omi precursor contains a mitochondrial localization signal (MLS) in its N terminus, HtrA1, 3 and 4 all harbor secretion signals in addition to insulin-like growth factor binding motifs and KAZAL domains in their N terminus. Interestingly, the HtrA2/Omi orthologs from human, rhesus monkey, dog, cow, mouse and rat also segregate phylogenetically from the cluster harboring HtrA1, 3 and 4 (Figure 1b). Moreover, the identified HtrA homolog in the fruitfly represents an ortholog of HtrA2/Omi (Figure 1b), in accordance with a recent report describing its cloning and functional characterization.7, 8 In contrast, the frog most likely expresses an HtrA1 ortholog. Although HtrA2/Omi segregates phylogenetically from the other metazoan HtrA proteins, the eukaryotic HtrA proteins, nevertheless, relate more to each other than to their bacterial homologs DegP, DegQ and DegS (Figure 1b). Escherichia coli HtrA/DegP functions as a chaperonin at normal temperatures, but relies on its proteolytic activity to prevent the accumulation of misfolded proteins in the periplasmic space at higher temperatures.In line with this function, its protease activity displays only limited substrate selectivity.In contrast, the anti-σ factor RseA is the only known target of bacterial DegS, which cleaves its substrate to initiate the transcription of stress-responsive genes when misfolded outer membrane proteins bind to its C-terminal PDZ domain.11 The physiological function of bacterial DegQ is less well understood, although it is believed to fulfill roles redundant with those of DegP and DegS as many bacteria lack DegP and DegS, but encode DegQ genes.12 Additionally, its protease activity displays a substrate specificity profile resembling that of DegP10 and it may functionally substitute for DegP when overexpressed in E. coli. Mitochondrial HtrA2/Omi and Neurodegenerative Disorders HtrA2/Omi is expressed as a 49-kDa proenzyme that is targeted to the mitochondrial IMS, although a fraction of the endogenous HtrA2/Omi pool has been detected in the nucleus of resting cells.15, 16, 17 The transmembrane anchor behind the N-terminal MLS most likely attaches the precursor protein into the mitochondrial inner membrane, where it undergoes proteolytic maturation. The fully processed protein is devoid of the first 133 amino acids encompassing the MLS and the transmembrane anchor, thus exposing an N-terminal inhibitor of apoptosis protein (IAP)-binding motif (IBM) related to those found in the Drosophila IAP inhibitors Reaper, Hid and Grim, and the mammalian IAP antagonist Smac/DIABLO.14, 15, 18, 19, 20 Although it is evident that the HtrA2/Omi proenzyme undergoes proteolytic maturation within the IMS, the mechanism involved requires further analysis. Autocatalysis is suggested by the observation that purified recombinant HtrA2/Omi undergoes autoprocessing at Ala133 in vitro, whereas the enzymatically inactive S306A mutant remains uncleaved.21, 22 However, HtrA2/Omi appears to be correctly processed in cells derived from Mnd2 mice (motor neuron degeneration 2), which are homozygous for a naturally occurring Ser276Cys mutation in the HtrA2/Omi protease domain that greatly reduces its catalytic activity.5 The latter observation suggests that the HtrA2/Omi zymogen may be cleaved by another protease in the IMS, although it cannot be ruled out that residual HtrA2/Omi activity is responsible for the unaffected HtrA2/Omi maturation observed in Mnd2 mice.5 Studies on the maturation of HtrA2/Omi zymogens containing mutations in the residues of the catalytic triad that are performed in a HtrA2/Omi-deficient background may clarify this important issue. Although studies in Mnd2 mice were not conclusive enough to elucidate the mechanism involved in HtrA2/Omi maturation, the striking Parkinsonian phenotype displayed by these mice clearly demonstrated the essential role of HtrA2/Omi in vivo.5 In addition to this neurodegenerative phenotype, Mnd2 mice failed to gain weight, and organs such as the heart, thymus and spleen were dramatically smaller when compared to wild-type littermates.5 The reduced body weight and progressive loss of neurons in the striatum of the basal ganglia were also evident in mice with a targeted deletion of the HtrA2/Omi gene,23 hence confirming the results obtained in Mnd2 mice. Before the neuronal cell loss became lethal approximately 30 days after birth, HtrA2/Omi−/− mice displayed a lack of coordination, decreased mobility and tremor,23 resembling the clinical manifestations of Parkinson's disease. Indeed, two single nucleotide polymorphisms in the HtrA2/Omi gene that cause missense mutations (A141S and G399S; Figure 1a) and affect the enzymatic activity of the protease have been associated with the development of Parkinson's disease in humans (Table 1).6 A recent study demonstrated the phosphorylation of HtrA2/Omi at a residue adjacent to a position found mutated in patients with Parkinson's disease.35 HtrA2/Omi phosphorylation depended on the cytosolic MAP kinase p38 and required the putative mitochondrial protein kinase PTEN-induced putative kinase 1 (PINK1), a known susceptibility factor for early-onset Parkinson's disease.39 Interestingly, lower HtrA2/Omi phosphorylation levels were detected in brains of patients with Parkinson's disease carrying mutations in PINK1.35 These findings suggest that PINK1-dependent phosphorylation of HtrA2/Omi might modulate HtrA2/Omi protease activity. Mutations in HtrA2/Omi or PINK1 that affect HtrA2/Omi phosphorylation might abolish the induction of HtrA2/Omi protease activity in patients with Parkinson's disease, possibly causing an increased susceptibility to mitochondrial stress and neuronal cell death. Table 1 Diseases associated with human HtrA homologs Full size table Albeit less well established, some studies have suggested a link between HtrA2/Omi and Alzheimer's disease. The precursor of the β-amyloid protein that forms the plaques associated with Alzheimer disease undergoes post-translational processing by the mutually exclusive α- and β/γ-secretase pathways.40 Cathepsin B was identified as the α-secretase,41 whereas γ-secretase-mediated cleavage of amyloid precursor protein (APP) requires presenilin-1.42 Interestingly, one study identified HtrA2/Omi as a presenilin-1-interacting factor in a yeast two-hybrid screen.16 The association of endogenous HtrA2/Omi with presenilin-1 was later confirmed in cell lysates of untreated 293T cells.36 Notably, presenilin-1 localizes primarily to the plasma membrane, endoplasmic reticulum, Golgi and nucleus,43, 44 whereas HtrA2/Omi resides mostly inside mitochondria, questioning the physiological context in which the interaction between HtrA2/Omi and presenilin-1 might occur. However, a fraction of the endogenous HtrA2/Omi pool may be targeted to the nucleus15, 16, 17 and presenilin-1 may also traffic to mitochondrial membranes,36, 45 thus providing possible cellular niches for interaction. Alternatively, presenilin-1 and HtrA2/Omi may interact in the cytosol of apoptotic cells. Indeed, presenilin-1-derived peptides that bind to the PDZ domain of HtrA2/Omi induce apoptosis by upregulating its enzymatic activity.36 In addition to its interaction with γ-secretase factor presenilin-1,16, 36 HtrA2/Omi was reported to generate a 28-kDa APP fragment in vitro and upon ectopic expression in 293T cells.37 In accordance with APP-processing activity, the occurrence of this APP fragment was greatly reduced in brain extracts of mnd2 mice carrying the Ser276Cys missense mutation in HtrA2/Omi,37 suggesting a role for HtrA2/Omi in the turnover of APP that is targeted to the mitochondria by virtue of an N-terminal signal sequence.46 It would be interesting to determine the fate of this 28-kDa APP fragment in the brains of mice with deficiencies in α-, β- and γ-secretase activities. Clearly, studies addressing the in vivo context in which endogenous HtrA2/Omi interacts with presenilin-1 and cleaves APP would greatly improve our understanding of these links to Alzheimer's disease. Is HtrA2/Omi a Mitochondrial Chaperone? The neurodegenerative phenotype of mice entirely lacking HtrA2/Omi or expressing the enzymatically inactive Mnd2 mutant indicates that the protease activity of HtrA2/Omi fulfills a protective role in the mitochondria of neuronal cells.5, 23 Although the mechanism by which HtrA2/Omi exerts its protective effect is not clear, a role for HtrA2/Omi in the regulation of mitochondrial energy metabolism is excluded because the activity of the mitochondrial electron transport chain complexes was not affected in HtrA2/Omi-deficient cells.23 It is tempting to speculate that HtrA2/Omi monitors and controls protein folding in the mitochondria, similar to the role of its homolog DegP in the bacterial periplasmic space. In this respect, HtrA2/Omi protein levels were shown to be upregulated several fold when the unfolded protein response was triggered by tunicamycin or heat shock.16 Additionally, elevated HtrA2/Omi expression occurred following activation of the p53 stress pathway with etoposide.47 Similar to the bacterial chaperone DegP, a transient exposure to elevated temperatures augments the protease activity of human HtrA2/Omi.48 Moreover, the serine protease domains of both HtrA2/Omi and DegP favor the aliphatic residues Val or Ile in the P1 position.10, 48, 49 In spite of these similarities between HtrA2/Omi and bacterial DegP, some important functional and structural characteristics of HtrA2/Omi are shared with DegS, but not with DegP, thus arguing against a DegP-like chaperone function for HtrA2/Omi and suggesting a role closer to that of DegS. Particularly, HtrA2/Omi displays protease activity at room temperature,21 whereas DegP requires elevated temperatures to become activated.9 These functional differences are mirrored by extensive structural dissimilarities. For instance, DegP possesses two PDZ domains and oligomerizes into a hexameric cage in which the inner cavity is occupied by the protease domains with the side walls being constructed by the 12 PDZ domains.50 In contrast, HtrA2/Omi and DegS form a trimeric pyramid-like structure with the N termini on top and the three PDZ domains at the bottom.51, 52, 53 The overall stability of the HtrA2/Omi complex is ensured by extensive van der Waals interactions involving residues of the protease domains and a large hydrophobic interface formed by aromatic residues in the N-terminal segment of each monomer, which is referred to as the trimerization motif (Figure 2).51 Second, both HtrA2/Omi and DegS lack the extended LA loop that ensures the dimerization of two DegP homotrimers and controls the height of the large central cavity of the hexamer to prevent properly folded proteins from entering the proteolytic sites.1, 54 In marked contrast to DegP, the short LA loop found in HtrA2/Omi and DegS does not interfere with proper formation of the active site.1, 12, 53 Indeed, the relative orientation of the PDZ domain was suggested to modulate proteolytic activity in these proteases.51, 52, 53 In the inactive state, the PDZ domain is directed back to the body of the protease domain through non-canonical interactions, trapping both the active site of the protease and its own peptide binding site in a distorted state.51, 52 Upon ligand binding, the PDZ domain would release the flexible L3 loop in the vicinity of the active site, thus triggering conformational reorganizations that result in the formation of a functional and accessible active site.52 This mechanistic model is supported by the observation that the proteolytic activity of HtrA2/Omi is significantly augmented in the absence of its PDZ domain.51 In addition, the protease activity of both HtrA2/Omi and DegS are upregulated in the presence of peptides that bind to their PDZ domains.11, 48 Notably, the PDZ domains of HtrA2/Omi and DegS have similar ligand specification, with both displaying a high affinity for peptides ending with the hydrophobic peptide sequence YYF(V) in the C terminus.11, 48 In addition to recognizing C-terminal peptide sequences, the HtrA2/Omi PDZ domain was recently reported to recognize internal stretches of extended, hydrophobic polypeptides as well.55 In conclusion, arguments have been provided in favor of a role for HtrA2/Omi as a mitochondrial chaperone, as well as against it. Therefore, studies targeting the capacity of HtrA2/Omi to resolve the aggregation of unfolded proteins are required and may reveal more clues to its putative role in protein quality control. Figure 2

Three-dimensional structure of HtrA2/Omi and primary structure of the IBM and trimerization motifs. (a) Schematic representation of the HtrA2/Omi monomer. The serine protease and PDZ domains are indicated in green and blue, respectively. (b) A close-up view of the catalytic triad. The catalytic residues His198 (blue), Asp228 (yellow) and Ser306 (red) are shown in space fill. (c) The amino-acid sequences of the conserved trimerization motifs in human HtrA2/Omi and its orthologs are indicated. The amino-acid sequences of confirmed and putative IBM motifs are underlined. Bt, Bos taurus; Cf, Canis familiaris; Dm, Drosophila melanogaster; Hs, Homo sapiens; Mm, Mus musculus; Mu, Macaca mulatta; Rn, Rattus norvegicus Full size image Role of HtrA2/Omi in Apoptosis As discussed above, proteolytic maturation of the HtrA2/Omi zymogen in mitochondria exposes an N-terminal IBM homologous to those of the Drosophila IAP inhibitors Reaper, Hid and Grim and the mammalian IAP antagonist Smac/DIABLO.14, 15, 18, 19, 20 Nuclear DNA damage, death receptor activation and numerous other apoptotic insults trigger the translocation of the mature protease into the cytosol, where it contributes to apoptosis through both caspase-dependent and -independent mechanisms. In this respect, antisense- and RNAi-mediated knockdown of HtrA2/Omi increases the resistance of multiple cell lines against apoptotic stimuli, such as anoikis, staurosporine, cisplatin, UV irradiation, anti-Fas and TRAIL.Moreover, the synthetic HtrA2/Omi inhibitor Ucf-10161 partially protects from cell death induced by cisplatin,56, 62 TNF-α63 and staurosporine.64 HtrA2/Omi unleashes caspase activity in a biphasic process that frees the active forms of caspases-3, -7 and -9 by proteolytically removing their natural inhibitors.58, 59 Indeed, RNAi-mediated downregulation of HtrA2/Omi diminished the degradation of XIAP and cIAP1 in cells undergoing apoptosis in response to etoposide, staurosporine and TRAIL.57, 58, 59 Mechanistically, the Reaper-like IBM sequesters IAP proteins in a first step.14, 15, 18, 19, 20 The protease activity of HtrA2/Omi may then steer the reaction into the favorable thermodynamic direction by actively degrading bound IAP proteins.58, 59 In line with the two-step model for the degradation of IAP proteins by HtrA2/Omi, IBM-deficient mutants of HtrA2/Omi cleaved recombinant cIAP1 10 times less efficiently than the wild-type protease.58 In addition, blocking the proteolytic activity of HtrA2/Omi attenuated post-ischemic myocardial apoptosis in vivo by preventing HtrA2/Omi-mediated XIAP degradation and subsequent caspase activity.65 The HtrA2/Omi inhibitor Ucf-101 did not alter caspase-3-like activity in Mnd2 MEF cells that underwent hypoxia/reoxygenation, in line with the proposal that Ucf-101 exerted its cardioprotective role by specifically inhibiting the protease activity of HtrA2/Omi.65 A recent report confirmed the evolutionary conservation of this function by demonstrating that the Drosophila HtrA2/Omi ortholog harbors two IBM motifs to recruit DIAP1, easing its removal by the serine protease activity.8 IBM motifs closely resembling that of human HtrA2/Omi have been retained in the rhesus monkey and rodent orthologs of this protease (Figure 2c). Serine substitutes for the N-terminal alanine in the putative IBM motif of bovine and canine HtrA2/Omi (Figure 2c). As noted for the IBM motifs of caspase-7 and glutamate dehydrogenase,66 a serine residue may be tolerated in interactions with the BIR2 domain of IAP proteins, thus suggesting that the IAP-binding capacity of HtrA2/Omi is conserved in mammals. As discussed above, the genomes of nematodes, such as the model organism C. elegans, lack HtrA2/Omi homologs. Interestingly, the two IAP-like proteins in C. elegans do not seem to be implicated in the regulation of apoptosis,67, 68 suggesting that IAP proteins and the concomitant emergence of IAP-antagonistic proteins such as HtrA2/Omi and Drosophila Reaper, Hid and Grim represent more recent additions to the repertoire of apoptotic molecules. Although multiple IAP family members were shown to be targeted and degraded by human HtrA2/Omi and its evolutionary paralogs, a recent study demonstrated that XIAP is the only bona fide inhibitor of caspases-3, -7 and -9.69 Indeed, the BIR2 and BIR3 domains of cIAP1, cIAP2 and XIAP all bind the IBM motifs in the N terminus of the small catalytic subunits of active caspases-3, -7 and -9, but only XIAP engages a second interaction surface that allows potent inhibition of caspases.70 The lack of this key feature in cIAP1 and 2 results in a catalytic inhibition that is 100- to 1000-fold less efficient.69 Nevertheless, cIAP1 and 2, as well as XIAP, may prevent caspase activation by targeting bound caspases for ubiquitin-mediated proteasomal degradation,71 providing an explanation why HtrA2/Omi targets all three IAP members. In marked contrast to its effect on caspases, XIAP binding enhanced the proteolytic activity of HtrA2/Omi.48 Although IBM-mediated catalytic inhibition of caspases is a current focus for therapeutic exploitation in cancer,72, 73 XIAP-deficient mice lacked a clear apoptotic phenotype.74, 75 The only apoptotic phenotype is that sympathetic neurons from these mice are more sensitive to cytochrome c injection.76 The absence of a major apoptotic phenotype may point to a limited physiological role for IAP proteins in the control of apoptosis or may be explained by the functional redundancy of XIAP with cIAP1 and 2. Double and triple knockout mice lacking these IAP members may reveal the extent to which these proteins protect against apoptosis. Similarly, the endogenous role of HtrA2/Omi in the sequestration of IAP proteins during apoptosis may be masked by its functional redundancy with IAP-binding proteins such as Smac/DIABLO77, 78 and the endoplasmic reticulum-associated protein GSPT1.79 Moreover, a number of caspase substrates80 and mitochondrial proteins that are released into the cytosol during apoptosis66 have been proposed to contain XIAP-antagonizing IBM motifs. In addition, a detailed study of the physiological role of HtrA2/Omi in apoptosis has been hampered by the concomitant loss of its mitochondrial function in knockout mice.23 A knock-in strategy that introduces a functionally defective IBM motif may circumvent this caveat by preserving its intramitochondrial function. The retinoic acid/IFNβ-induced cell death activator Grim-1981 was proposed to interact with the PDZ domain of HtrA2/Omi to enhance the proteolytic degradation of XIAP.82 Grim-19 physically associates with the PDZ domain of HtrA2/Omi, and their interaction is enhanced by the combination of retinoic acid and IFNβ.82 Grim-19 augmented HtrA2/Omi activity in vitro, thus providing an explanation for the reduced cell death and impaired HtrA2/Omi-mediated degradation of XIAP in antisense Grim-19-expressing MCF7 cells.82 Similar to the results obtained with Grim-19, antisense-mediated downregulation of HtrA2/Omi conferred resistance to retinoic acid/IFNβ-induced cell death.82 Grim-19 was previously shown to contribute to the activity of mitochondrial complex I83 and homozygous deletion of Grim-19 caused embryonic death.84 Therefore, both Grim-19 and HtrA2/Omi exert life-essential functions in mitochondria, but interact in the cytosol of apoptotic cells to promote HtrA2/Omi-mediated degradation of XIAP. In contrast to Grim-19, the death effector domain-containing protein Ped/Pea-15 was identified as a substrate of recombinant HtrA2/Omi in vitro and the HtrA2/Omi inhibitor Ucf-101 prevented Ped/Pea-15 degradation in UV-irradiated 293T and HeLa cells.85 As Ped/Pea-15 interfered with XIAP binding on HtrA2/Omi and prevented UV-induced caspase-3 activity, cellular Ped/Pea-15 levels were proposed to modulate the ability of HtrA2/Omi to relieve XIAP-mediated inhibition of caspases.85 Similar to the results obtained with Ped/Pea-15, siRNA-mediated knockdown of the tumor suppressor and mitotic regulator WARTS protected HeLa cells against HtrA2/Omi-induced cellular toxicity and XIAP degradation.57 Interestingly, the kinase activity of WARTS was required for its association with the PDZ domain of HtrA2/Omi and the consequent increases in HtrA2/Omi protease activity and apoptosis.57 It remains to be seen whether WARTS can phosphorylate HtrA2/Omi to modulate its protease activity as has been demonstrated for the serine/threonine kinases Akt1 and Akt2.86 Indeed, phosphorylation of HtrA2/Omi on Ser212 attenuated its protease activity in vitro and impaired its pro-apoptotic function in vivo.86 The authors demonstrated that phosphorylated HtrA2/Omi failed to cleave XIAP, although its binding with HtrA2/Omi was not affected.86 These findings extend the anti-apoptotic role of Akt1 beyond the transcriptional upregulation of the anti-apoptotic proteins Bcl-2 and Mcl-187, 88, 89, 90 and the inactivating phosphorylation of caspase-991 and the pro-apoptotic Bcl-2 members Bad and Bax. In addition to antagonizing IAP proteins to augment caspase activity, HtrA2/Omi contributes to apoptosis independently of its IBM. Indeed, siRNA-mediated knockdown of HtrA2/Omi combined with the pan-caspase inhibitor zVAD-fmk almost completely protected HeLa cells from undergoing staurosporine-induced cell death, whereas caspase inhibition alone was significantly less effective.57 Additional evidence came from the observation that unlike HtrA2/Omi, caspase-9 and the IAP antagonist Smac/DIABLO were dispensable for detachment-induced anoikis in the epithelial cell line IEC-18.Overexpression of IBM-defective HtrA2/Omi, but not the enzymatically inactive S306A mutant, induced morphological features of anoikis such as cell rounding and shrinkage.This HtrA2/Omi-induced morphology persisted in Apaf-1−/− and caspase-9−/− cells as well as in the presence of the caspase inhibitors XIAP and zVAD-fmk, pointing to a caspase-independent mechanism.A comprehensive proteome-wide analysis of Jurkat cell lysates led to the identification of 15 potential HtrA2/Omi substrates, 10 of which were validated in vitro. Interestingly, this group included the cytoskeleton-associated proteins actin, α- and β-tubulin and vimentin, providing a likely explanation for the anoikis-like phenotype that occurs upon overexpression of HtrA2/Omi. In addition to these structural proteins, eIF-4G1 and EF-1α were identified as putative HtrA2/Omi targets.The cleavage of these components of the translation machinery may contribute to the abrogation of de novo protein synthesis during apoptosis, next to the caspase-mediated cleavage of eIF-4G1.The list of identified HtrA2/Omi substrates also included KIAA1967 and KIAA0251, two proteins that have recently been associated with apoptosis. Another study demonstrated the mitochondrial anti-apoptotic protein HS1-associated protein X-1 (HAX-1) to be a HtrA2/Omi substrate.62 Interestingly, experiments in Mnd2 MEF cells reconstituted with wild-type HtrA2/Omi demonstrated that HtrA2/Omi-mediated degradation of HAX-1 correlated with extensive cell death in response to etoposide, cisplatin and H2O2.62 In contrast to HtrA2/Omi, HAX-1 protein levels were significantly reduced but remained associated with mitochondria of cisplatin- and H2O2-treated 293T cells, suggesting that HtrA2/Omi degrades HAX-1 early in apoptosis.62 Although more studies are required to decipher the physiological contribution of the identified HtrA2/Omi substrates, the results discussed above (summarized in Table 2) substantiate a role for HtrA2/Omi in apoptosis apart from antagonizing IAP proteins to induce caspase activity.