2.1.1. Upstream activator
Many stimuli, including calf serum, heat shock, and high salt, have been examined for effects on MST1 kinase activity;
however, none of these stimuli affect endogenous MST1 kinase activity. Interestingly, EGF stimulation has been observed by in-gel kinase assay to cause a transient decrease of MST1 kinase activity; the kinase activity decreases up to two-fold within 1 min and returns to normal level 30 min after EGF treatment [9]. The first indication of MST1 upstream stimulus comes from the observation that MST1 is specifically cleaved by a
caspase 3-like activity during apoptosis, induced either by cross-linking CD95/Fas [20], anticancer drugs [27], or by bipho-sphonates that induce apoptosis in osteoclasts [28]. The in-duction of MST1 activation cleavage is not restricted to specific types of inducers. For example, addition of staurosporine or withdrawal of both serum and M-CSF also induced activation cleavage of MST1[28]. The role of MST1 in apoptosis may vary in different cell types. Caspase cleavage and activation of MST1 correlates with eosinophil apoptosis, but not with neutrophil apoptosis[29].
2.1.2. MST1, JNK, and histone 2B
Overexpression of MST1 activates JNK and p38 MAPK, but not ERK1 in 293T cells[20]; additionally, activation of JNK can be blocked by cotransfection with kinase-dead MEKK1 [30]
and dominant-negative JNK inhibits MST1-induced apoptosis.
In contrast, neither p38 inhibitor nor dominant-negative p38 affects MST1-induced apoptosis [31]. The appearance of cleaved active MST1 fragment was usually 2 to 6 h after the addition of an apoptosis-inducing drug. However, the first appearance of JNK activation (approximately 30 min after stimulus) was earlier than MST1 during apoptosis induced by the anticancer drug cytotrienin A [27], suggesting that early phase activation of JNK was not mediated by MST1. MST1 induces DNA fragmentation through caspase activation, but induces chromatin condensation and membrane blebbing independent of caspase activation. One important issue is how MST1 activates chromatin condensation. It has been demon-strated that expression of a dominant-negative JNK mutant inhibited MST1-induced chromatin condensation [31].
Enhancement of MST1 nuclear translocation by mutation of the nuclear exporting sequence in the C-terminal of MST1 increased its ability to induce chromatin condensation, suggesting that nuclear translocation of MST1 was important for chromatin condensation [32]. Allis's group has demon-strated that the mechanism of triggering chromatin condensa-tion was mediated by phosphorylating histone 2B in vertebrate and yeast[33,34]. Deletion of ste20 rendered yeast resistant to apoptosis and loss of histone 2B phosphorylation[34]. MST1 phosphorylated histone 2B in vitro and the kinetics of histone 2B phosphorylation was similar to that of MST1 cleavage in HL-60 cells. In addition, expression of a C-terminal truncated MST1 could phosphorylate histone 2B in Hela cells; however, the requirement of MST1 in inducing histone 2B phosphoryla-tion has not been demonstrated in vivo[33]. The role of MST1 in chromatin condensation has been questioned by showing that depletion of MST1 further enhanced histone 2B phosphoryla-tion during apoptosis induced by apoptotic chromatin con-densation inducer in the nucleus (acinus). Furthermore, PKC delta may mediate H2B phosphorylation downstream of acinus [35]. It was interesting to note that the active form of acinus binded to MST1 during apoptosis and enhanced its kinase activity, and that the binding of acinus to MST1 was reduced by nerve growth factor treatment. Recently, activation of the c-Jun N-terminal kinase pathway has been shown to be essential and sufficient for induction of chromatin condensation by over-expression of MST1[36]. The discrepancy in these reports may
result from interactions of several parameters. The first is that other MST members, such as MST2, may replace the function of MST1 on chromatin condensation or histone 2B phosphoryla-tion during apoptosis in certain cellular condiphosphoryla-tions. Therefore, depletion of MST1 has minimal effects in some experiments.
The second is that histone 2B phosphorylation may be a good marker for chromatin condensation, but not necessarily for all types of chromatin condensation. The third is that MST1 exists in different types of complexes, including acinus, RASSF1, NOR1, and death-associated protein 4 (DAP4). The equilibrium and balancing of these complexes may determine the final consequence of MST1. The interacting partner DAP4 was identified by affinity co-purification with FLAG-tagged MST1.
Co-expression of DAP4 and a sub-optimal level of MST1 enhanced MST1-induced apoptosis. However, DAP4 was not phosphorylated by MST1, and did not enhance MST1 kinase activity. DAP4 may enhance the apoptotic effect through promoting the caspase-catalyzed cleavage of MST1. DAP4 was originally identified by its ability to block apoptosis induced by interferon-gamma (IFN-γ) in Hela cells; however, MST1 was not activated by IFN-γ in Hela cells. The biological implication of this interaction requires further investigation[37].
2.1.3. MST1, ras, Nore1, RASSF1A, and CNK1
In search of the potential biological function of a Ras effector, Nore1/Rassf5, MST1 was shown to directly interact with Nore1/
RASSF5. The complex between Nore1 and MST1 was insensitive to serum withdrawal. On the other hand, the complex between endogenous Ras and MST1 and Nore1 only occurred in the presence of serum. Interestingly, the complex of Ras and Nore1 with MST1 did not enhance MST1 kinase activity, suggesting that Ras controls MST1 signaling primarily by its recruitment of MST1 to a site regulating MST1 action, probably a membrane fraction[38].The complex may involve other RASSF members, such as RASSF1A. RASSF1A is frequently inactivated in lung cancer cells due to hypermethylation of a CpG island in its promoter[39]. RASSF1A can form homodimers or heterodimers with Nore1 [40]. The complex appears to contain many other proteins, e.g., human scaffold protein CNK1, a c-Raf1 binding protein. Cell death induced by CNK1 required the participation of MST1 or MST2, which was mediated indirectly through association of RASSF1 polypeptides [41]. Recombinant RASSF1A inhibited MST1 kinase activity in vitro, but appeared to increase MST1 kinase activity in vivo. It seems that the interaction is more than a simple association. Depletion of RASSF1A by RNA interference reduced MST1 activation and Fas ligation-induced apoptosis [42]. Therefore, the interaction between RASSF1A and MST1 indeed plays a role in mediating apoptosis. Treatment of K-ras-transformed cells with bortezomib resulted in nuclear translocation of MST1 and an increase of phosphorylated histone H2B. Knockdown of MST1 expression by RNA interference reduced bortezomib-induced apoptosis.
These results suggest that MST1 is an essential mediator in apoptosis of K-ras transformed cells. Although Ras-Nore1-MST1 is known to be involved in mediating Ras-induced apoptosis, bortezomib has little effect on NORE1, suggesting that other molecular interactions in this complex may be altered[43].
2.1.4. MST1, hSav, Wts, and cyclin E
Drosophila MST1/2 homologue Hippo (hpo) binds to and phosphorylates a tumor suppressor protein Salvador (Sav), which is known to interact with Warts (Wts) protein kinase.
Loss of Hippo results in increased expression of cyclin E and cell-death inhibitor diap1[44–48]. The human orthologue hSav (also called hWW45) can bind to and be phosphorylated by MST1 and 2. MST2 has a stronger stabilizing effect on hSav, and may play a more prominent role in this complex[49].
2.1.5. MST1, FOXO3, and AKT
MST1 phosphorylates FOXO3 and disrupts its interaction with 14-3-3 proteins, promotes FOXO3 nuclear translocation, and induces apoptosis in neuronal cells. Knockdown of C. elegans MST1 ortholog CST-1 shortened life span and accelerated tissue aging, while overexpression of CST-1 promoted life span and delays aging[50]. The effect of MST1 is opposite to that of AKT. AKT phosphorylates FOXO3 and promotes its association with cytoplasmic 14-3-3 protein, preventing its transcriptional activity. Therefore, FOXO tran-scription factor can undergo inhibitory phosphorylation by AKT, and activating phosphorylation by MST1[51]. The signal transduction appears to be interconnected. AKT could also directly phosphorylate MST1 and inhibited its kinase activity towards FOXO3, but enhanced its kinase activity on the other substrate histone 2B [52]. In transgenic mice with constitu-tionally active p110-alpha subunit of PI3K, MST1 expression was elevated and the elevation could be blocked by the PI3K-specific inhibitor LY294002 [53]. These results suggest that AKT phosphorylates MST1 and prevents its action from promoting apoptosis. Furthermore, MST1/2 may be an important branch pathway of PI3K/AKT based on the observation that MST1 and AKT1 are localized to identical subcellular compartments in human prostate tumors. Both MST1 and its active cleavage form physically interact with AKT and act as direct inhibitors of AKT. Depletion of MST1 or MST2 with siRNA increased AKT activity, and depletion of both proteins further enhanced AKT activity [54]. The MST complex appears to contain many protein components, and the association or dissociation of one component in response to outside apoptotic stimulus may alter the activity of other components of this complex. A simplified scheme of MST1 signal transduction is shown inFig. 1.
2.2. Mst2
MST2 shares at least two similar pathways with MST1. The first is that MST1 and MST2 can phosphorylate AKT and inhibit AKT activation. In addition, they can cause activating phosphorylation of FOXO3. The second is that both proteins can be detected in a complex containing hSav, although the role of MST2 in this complex has been more clearly demonstrated.
During apoptosis, MST2 was cleaved and undergone irreversible autophosphorylation, which was resistant to phosphatase [17]. Thyroid transcription factor-1 could be phosphorylated by rat MST2 in thyroid cells [55]; however, the role of phosphorylation in the regulation of thyroid
tran-scription factor-1 function was not investigated afterwards.
Constitutionally active Ras can inhibit thyroid transcription factor-1 and thyroid differentiation [56]. MST2 also interacts with several Ras effector complexes including Raf-1 and AKT.
Therefore, MST2 may play a role in mediating inhibitory effects of Ras on thyroid differentiation. MST2 can be inhibited by Raf-1 by forming a complex. Raf-1 inhibits MST2 independent of its kinase activity, probably by sequestering MST2 from its activation site[57].
The complex of hSav and MST2 also contains RASSF1A, Nore1 and LATS1. MST2 can be co-precipitated with LATS1 only in the presence of hSav, and furthermore, RASSF1A and hSav promote LATS phosphorylation by MST2 through this recruitment [58]. RASSF proteins have been suggested to participate in the Hpo-Sav-Wts pathway in mammalian cells in a manner dependent on a protein–protein interaction domain, SARAH, that is shared by Sav, RASSF, and Hpo[57].
MST2 is in the tumor suppressor pathway Hippo/Salvador/
Lats; MST2 direct phosphorylates LATS1 and LATS2 at their C-terminal catalytic domain. MST1 can perform similar kinase activity, but at a lower efficiency. In contrast, MST4 has almost no activity[59]. The interaction between raf-1 and MST2 was further extended to RASSF1A. RASSF1A causes the disruption of the inhibitory Raf1-MST2 complex, and enhances the released MST2 to phosphorylate its substrate, LATS1. The phosphorylated LATS1 releases the cytosol-sequestered transcription factor YAP1. YAP1 is the human homologue of Drosophila Yorkie, which connects the LATS1 Drosophila homologue Warts to the transcription of cyclin E gene[60]. The transcriptional regulator YAP1 translocates to the nucleus and associates with p73, resulting in transcription of the proapoptotic target gene, such as Puma[61].
An outline of MST2 signal pathways is illustrated in Fig.2.
Fig. 1. MST1 kinase signal transduction. MST1 has several interacting partners, including DAP4, acinus, and hSav/Nore1/RASSF1A. MST1 can be cleaved by caspase to remove inhibitory domain, and induce activation of JNK and p38 during apoptosis. The MST1 substrates include histone 2B, AKT, FOXO3, and hSav. MST1 causes activating phosphorylation of FOXO3 transcription factor, but introduces an inhibitory phosphorylation of AKT. AKT can also cause inhibitory phosphorylation of MST1. Interacting proteins are indicated by a rectangle; kinase substrates are indicated by a hexagon.
2.3. Mst3
MST3 was unable to activate ERK, JNK, and p38 MAPK kinase activity[12]. However, one report indicated that MST3, but not the isoform MST3b, showed activation of ERK in HEK 293 cells [15]. Overexpression of MST3 did not lead to activation of ERK in MDCK cells [unpublished observation].
Interestingly, protein kinase A appears to phosphorylate Thr-18 of MST3b isoform. Mutation of Thr-18 to alanine of MST3b enhances its activation of ERK. The role of ERK in wild-type MST3 downstream signaling remains to be investigated.
The first identified natural substrate for MST3 is NDR protein kinase[62].The nuclear Dbf2-related (NDR) family of Ser/Thr kinases consists of NDR and large tumor suppressor (LATS) kinase. Members of the NDR family regulate cell proliferation, tubulin cytoskeleton organization, cell spreading, morphogen-esis, polarized growth in C. elegans and D. elanogaster[63].
LATS/WARTS kinase is a tumor suppressor, negatively regulating cell cycles by cyclin-dependent kinase CDK1 and reduces cell proliferation and survival leading to growth arrest in the G2/M phase[63]. The function of Trc, the drosophila NDR kinase is required for the normal morphogenesis of polarized growth, including epidermal hairs, bristles, arista laterals and dendrites [64]. Casnorhabditis elegans NDR homolog SAX1 regulates neuronal cell shape and is involved in cell spreading, neurite initiation and dendritic tiling[65]. Genetic studies with yeast, C. elegans, and drosophila showed that many of the NDR members are highly conserved and phosphorylated by Ste20 family kinases [62]. From this genetic evidence, MST3 was identified to be able to phosphorylate and activate NDR protein kinase. NDR kinase was directly phosphorylated at Thr442/444
by MST3 when cells were treated with okadaic acid, a serine/
threonine phosphatase inhibitor 2A [62]. Therefore, the activated MST3 locates upstream of the NDR kinase signal pathway and controls cell shape and cell cycle. This pathway appears to be parallel to MST2 on LATS. The Ste20-like kinase, Hippo, phosphorylates WARTS/LATS kinase in S. cerevisiae.
MST2 in mammals phosphorylates LATS1 and is involved in tumor suppressors pathways in mammals[57]. Cdc7 and Sid1, two other Ste20-like kinases, are upstream of NDR kinase homologue Sid2 in S. pombe, and are involved in cytokinesis [66]. Recently, the svkA gene encoding severin kinase, a homolog of the human MST3, MST4 and YSK1 kinases, was knockout in Dictyostelium discoideum. SvkA-knockout cells showed dramatic defects in cytokinesis, leading to multi-nucleated cells[67]. MST3, MST4 and YSK1 share very high homology in the kinase domain. YSK1 was originally identified by the activation by oxidative stress and named as Ste20/oxidant stress response kinase-1 (SOK-1) [68,69]. The expression of MST3 was elevated by the oxidative stress, but not by the hormones released during labor such as prostaglandin E, in human term placenta. The hydrogen peroxide-induced apopto-sis of trophoblast cell was suppressed by overexpression of kinase-dead Mst3 or by knockdown of endogenous Mst3 with siRNA[70]. This result suggested that oxidative stress might promote apoptosis in part by upregulating MST3 expression.
Downregulation of endogenous MST3 with RNA interfer-ence was shown to enhance cellular migration in MCF-7 breast cancer cells. Aberrant protrusions were reduced by reconstitu-tion of RNAi-resistant MST3 in MST3-knockdown MCF7 cells [18]. Overexpression of wild-type MST3 in MCF7 and MDCK cells leaded to tyrosine phosphorylation of Tyr-31 and Tyr-118 of paxillin. MST3 could keep paxillin tyrosine phosphory-lated to inhibit protrusion and migration in cells plating on fibronectin[18](Fig. 3).
MST3, a Ser/Thr kinase, can not act directly on the tyrosine residue of paxillin to control migration. Protein tyrosine
Fig. 2. MST2 kinase signal transduction. MST2 substrates include FOXO3, AKT, LATS1, and thyroid transcription factor-1. The interacting partner Raf-1 inhibits MST2 function. MST2 reacts similarly with MST1 in two aspects:
(1) interaction with AKT and FOXO3, (2) activation by caspase cleavage. MST2 interacts with RASSF1A and hSav to phosphorylate LATS and regulate p73-target genes. The biological function of MST2 is inhibited by Raf-1 binding.
Interacting proteins are indicated by a rectangle; kinase substrates are indicated by a hexagon.
Fig. 3. MST3 kinase signal transduction. MST3 can be cleaved by caspase to promote apoptosis. MST3 can phosphorylate NDR to regulate cell cycle and PTP-PEST to regulate migration. Kinase substrates are indicated by a hexagon.
phosphatase PTP-PEST binds with paxillin C-terminal LIM 3-4 domains as well as the N-terminal LD4 motif to phosphorylate tyrosine 31 and tyrosine 118 of paxillin[71]. The optimal level, localization or activity of PTP-PEST are important regulators that maintain the balance of promoting control cell motility in different contexts[72]. MST3 was shown to directly phosphor-ylate PTP-PEST but not PTP1B. The phosphorylation inactivated PTP-PEST and inhibited migration through tyr-31 and tyr-118 of paxillin[18]. The potential mechanism of PTP-PEST-dependent inhibition of migration by MST3 may involve other proteins such as Rac1. PTP-PEST can suppress the activation of Rac1 and impair membrane protrusion, cell spreading and hepatotaxis in CHOK1 cells [72]. It is possible that MST3 increases phosphorylation of paxillin and hence decreases Rac1 activity at the cadherin-containing junctions. At these sites the cadherin assembling increases and leads to reduction of cell migration.
PTP-PEST is not only regulating cell migration, but also controlling cell growth and apoptosis. Expression of PTP-PEST sensitized cells to TNF-alpha-induced apoptosis, and PTP-PEST was cleaved by caspase 3 during apoptosis[73]. Therefore, MST3 may phosphorylate PTP-PEST and alter the scaffold function of PTP-PEST during apoptosis.
2.4. Mst4
MST4 was cloned by screening Raf-1 with yeast two-hybrid;
however, Raf-1 did not interact with MST4 in mammalian cells.
Furthermore, the C-terminal domain enhanced MST4 kinase activity in contrast to the inhibitory non-kinase domain of MST1[14]. MST4 did not activate ERK, JNK, or p38 in 293 cells, but MST4 could activate ERK via MEKK, but not the Ras/Raf pathway in Pheonix cells[14,74]. In contrast to other MST family kinases, MST4 did not activate JNK or p38.
Overexpression of MST4 further enhanced transformation of phoenix cells as demonstrated by growth in soft agar. One conflicting report has suggested that overexpression of MST4 induced apoptosis in MCF-7 cells and 293 cells, and that kinase-dead MST4 did not have pro-apoptotic effects which indicating kinase activity was required[74]. The programmed cell-death 10 (PDCD10) gene has been implicated in mutations associated with Cerebral Cavernous Malformations (CCM).
MST4 was shown to interact with PDCD10 with yeast two-hybrid and co-immunoprecipitation assay. Overexpression of either PDCD10 or MST4 enhances ERK kinase activity, and co-expression further enhanced ERK activity and cellular trans-formation[75](Fig. 4).
The orientation of Golgi matrix positions at wound edges, and polarizing the edge, coordinates the signal of the cytoskeleton and Rho family proteins. Activation of YSK1 or MST4 involves autophosphorylation at Thr174 in YSK1 and Thr178 in MST4. It may stabilize the binding of YSK1 and MST4 to GM130, a cis-Golgi protein, and restrict YSK1 and MST4 at the Golgi matrix to control cell migration. Interfering with YSK1 and MST4 function by siRNA disturbed the ordered localization of the Golgi matrix in the perinuclear region, and the Golgi was dispersed into the cell periphery. However, overexpressing YSK1 but not MST4 induced migration of cells
on collagen [76]. One possibility for the difference seen with YSK1 and MST4 is that they may have different substrates.
YSK1, but not MST4, specifically phosphorylates 14-3-3ζ, which is localized in the Golgi matrix. Although MST3 is highly homologous to YSK1 and MST4, MST3 showed cyto-plasmic distribution and has no interaction with G130. Knock-down of MST3, unlike dominant negative YSK1, did not
YSK1, but not MST4, specifically phosphorylates 14-3-3ζ, which is localized in the Golgi matrix. Although MST3 is highly homologous to YSK1 and MST4, MST3 showed cyto-plasmic distribution and has no interaction with G130. Knock-down of MST3, unlike dominant negative YSK1, did not