Association of mammalian sterile twenty kinases, Mst1 and Mst2, with hSalvador via C-terminal coiled-coil domains, leads to its stabilization and phosphorylation

Authors

  • Bernard A. Callus,

    1.  The Walter and Eliza Hall Institute, Parkville, VC, Australia
    Search for more papers by this author
    • Present address
      Department of Biochemistry, La Trobe University, Plenty Road, Bundoora, VIC 3086, Australia

  • Anne M. Verhagen,

    1.  The Walter and Eliza Hall Institute, Parkville, VC, Australia
    Search for more papers by this author
  • David L. Vaux

    1.  The Walter and Eliza Hall Institute, Parkville, VC, Australia
    Search for more papers by this author
    • Present address
      Department of Biochemistry, La Trobe University, Plenty Road, Bundoora, VIC 3086, Australia


B. A. Callus, Department of Biochemistry, La Trobe University, Plenty Road, Bundoora, VIC 3086, Australia
Fax: +613 9479 2467
Tel: +613 9479 1669
E-mail: b.callus@latrobe.edu.au

Abstract

Genetic screens in Drosophila have revealed that the serine/threonine kinase Hippo (Hpo) and the scaffold protein Salvador participate in a pathway that controls cell proliferation and apoptosis. Hpo most closely resembles the pro-apoptotic mammalian sterile20 kinases 1 and 2 (Mst1 and 2), and Salvador (Sav) has a human orthologue hSav (also called hWW45). Here we show that Mst and hSav heterodimerize in an interaction requiring the conserved C-terminal coiled-coil domains of both proteins. hSav was also able to homodimerize, but this did not require its coiled-coil domain. Coexpression of Mst and hSav led to phosphorylation of hSav and also increased its abundance. In vitro phosphorylation experiments indicate that the phosphorylation of Sav by Mst is direct. The stabilizing effect of Mst was much greater on N-terminally truncated hSav mutants, as long as they retained the ability to bind Mst. Mst mutants that lacked the C-terminal coiled-coil domain and were unable to bind to hSav, also failed to stabilize or phosphorylate hSav, whereas catalytically inactive Mst mutants that retained the ability to bind to hSav were still able to increase its abundance, although they were no longer able to phosphorylate hSav. Together these results show that hSav can bind to, and be phosphorylated by, Mst, and that the stabilizing effect of Mst on hSav requires its interaction with hSav but is probably not due to phosphorylation of hSav by Mst.

Abbreviations
dIAP1

Drosophila inhibitor of apoptosis I

GCK

germinal center kinase

HA

hemagglutinin

Hpo

serine/threonine kinase Hippo

IL-2

interleukin-2

Lats

large tumour suppressor

MBP

myelin basic protein

Mst

mammalian sterile20 kinase

Nore1

novel Ras effector 1

PBST

NaCl/Pi-Tween 20

PPIA

peptidyl-prolyl cis-trans isomerase A

Rassf1

Ras suppressor factor 1

Sav

Salvador

SAPK

stress activated protein kinase

Ste20

Sterile Twenty

Wts

warts kinase

Yki

Yorkie

The mammalian serine/threonine kinases, Mst1 and Mst2, were originally identified by their similarity to yeast Sterile Twenty (Ste20) kinase [1,2]. Mst3 and Mst4 were subsequently identified the same way [3–5]. The four Mst kinases belong to a subfamily of Ste20-like germinal center kinases (GCKs) that is characterized by an N-terminal kinase domain (reviewed in [6]). Based on their similarity with each other, the Mst kinases can be further subdivided into two groups, Mst1 and Mst2 (GCKII) and Mst3 and Mst4 (GCKIII).

Mst1 has been widely studied and is the best characterized member of the family. In addition to its kinase domain, Mst1 contains an inhibitory domain, deletion of which results in increased kinase activity, and a predicted coiled-coil domain at the C-terminus that is essential for the formation of Mst1 dimers (multimers) [7]. Full-length Mst1 is mainly cytoplasmic, but can shuttle continuously between the cytoplasm and nucleus in a phosphorylation dependent manner [8–10]. Ectopic expression of Mst1 and Mst2 in certain cells types has been reported to induce cell death in a stress activated protein kinase (SAPK) dependent pathway [11–13]. In apoptotic cells, activated caspases can cleave Mst1 and Mst2 C-terminal to the kinase domain [11–13]. The proteolytic fragment encompassing the kinase domain accumulates in the nucleus and can phosphorylate histone H2B at Ser14, possibly triggering chromosomal condensation [9,11,14], in a positive feedback loop in cells undergoing apoptosis.

The physiological signals leading to activation of Mst1 and Mst2 are poorly understood. Mst1 has been reported to become activated if recruited to or artificially targeted to the plasma membrane [15,16] as well as in response to specific nonphysiological stress stimuli such as staurosporine, sodium arsenite, hyperosmotic concentrations of sucrose, and heat-shock [11,13,16], but Mst1 was not activated in HeLa cells in response to several cytokines nor affected by serum withdrawal or addition [16]. While cleavage of Mst1 has been observed in cells following CD95/Fas cross-linking or IL-2 withdrawal, this effect is apparently independent of activation of full-length Mst1 [11,12], and may be a late consequence of caspase activation.

Recently, several reports have revealed a role in Drosophila for the Mst1 and Mst2 homologue Hippo. Thesalvador gene (also known as shar-pei) was identified in flies in a screen to identify genes that imparted significant growth advantages in mutant versus normal tissue [17,18]. The Salvador protein has domains that permit protein–protein interaction, including a WW domain and a predicted coiled-coil motif in its C-terminus, suggesting it might function as a scaffold in a multimeric complex. Subsequently, the serine/threonine kinase Hippo (Hpo) was identified as a binding partner of Salvador [19–23]. Mutation of hpo and salvador yield identical phenotypes, characterized by increased cell proliferation and impaired apoptosis. These effects can be at least partly explained by the elevated levels of the cell cycle regulator, cyclin E, and the Drosophila inhibitor of apoptosis protein, dIAP1, in mutant tissue. The hpo and salvador mutant phenotypes also resemble those due to mutation of another serine/threonine kinase, warts (wts). Indeed, Wts can bind to the WW domains of Salvador [17], and was subsequently shown to complex with and be activated by Hpo in a Salvador dependent manner [19,20,23]. Furthermore, Wts was recently shown to phosphorylate and subsequently inactivate Yorkie (Yki), a Drosophila orthologue of the mammalian transcriptional coactivator Yes-associated protein, in a Hpo/Sav dependent manner [24]. Yki can transcriptionally up-regulate the genes for cyclin E and dIAP1. Therefore the failure to inactivate Yki in hpo/sav/wtsmutant tissue accounts for the elevated levels of cyclin E and dIAP1. Thus the Hpo/Salvador/Wts complex defines a novel pathway regulating cell growth and apoptosis in Drosophila in vivo, primarily through the regulation of Yki activity.

Hpo is a Drosophila orthologue of the Ste20-like kinases, and is most similar to mammalian Mst2. Mst1 and Mst2 have been shown to interact with a number or proteins, including the novel Ras effector 1, Nore1 [15,16], the putative tumour suppressor (Ras suppressor factor 1; Rassf1) [15,16,25], and most recently, Raf1 (with Mst2) [26]. The association of Mst with Nore1 or Rassf1 leads to an inhibition of Mst kinase activity, yet these complexes appear to mediate the pro-apoptotic activity of active Ras [15,25]. Raf1, on the other hand, directly inhibits Mst2 activation, thereby preventing apoptosis in cells following serum starvation [26]. These studies provide a possible link from Ras or Raf signalling to apoptosis through regulation of Mst activity. These findings are also consistent with the apoptotic effects of Hpo in flies and potentially link Mst/Hpo activity to upstream signalling events. Interestingly however, Salvador, also called WW45 in mammals [27], was not found in these complexes of Mst, suggesting that Mst may bind to Raf1/Rassf1/Nore1 or to Salvador, but not both at the same time. Furthermore, there appear to be no known orthologues of Nore1 and Rassf1 in flies, thus raising the possibility that complexes of Sav and Mst might not occur in mammals.

Here we report that hSalvador can tightly interact with the kinases Mst1 and Mst2, just as their counterparts, Salvador and Hpo interact in Drosophila.

Results

Mst kinase interacts with, and stabilizes, hSalvador protein

To determine whether hSalvador (referred to hereafter as Sav) can interact with Mst kinases, we generated isogenic stable cell lines that could inducibly express flag epitope-tagged Sav. Following treatment with doxycycline, two independent clones of cells efficiently expressed flag-Sav (Fig. 1A). Moreover, endogenous Mst1 was easily detected in antiflag immune complexes in cells that expressed flag-Sav. As a positive control for this experiment, in a separate cell line, induced myc-tagged Mst1 was also precipitated, albeit less efficiently, with antibody raised against Mst1. Interestingly, induction of Mst1 in these cells resulted in the appearance of two smaller proteins that corresponded in size to the caspase-cleaved forms of myc-tagged and endogenous Mst1. In contrast to Mst1, endogenous Mst2 was not detectable in these cells (data not shown).

Figure 1.

 Mst kinase interacts with hSalvador and increases its abundance. (A) Independent clones of Flp-In T-REx-293 cells (flag-SavB, flag-SavC and myc-Mst1) were cultured overnight with or without doxycycline as indicated. Cell lysates and immune complexes were separated by SDS/PAGE, transferred to membrane and sequentially immunoblotted as indicated on the left. (B–E) Flag-tagged Sav cDNA was cotransfected with or without myc-Mst1, myc-Mst2, HA-PPIA or green floiurescent protein cDNAs into 293T cells as indicated. Two days after transfection cell lysates were prepared. Immune complexes (B) or total cell lysates (C) were separated by SDS/PAGE, transferred to membrane and sequentially immunoblotted as indicated on the left. The position of relevant bands is indicated with arrows. The heavy and light immunoglobulin chains in antimyc immune complexes (B, bottom) are indicated (*). The panels shown in (C) are from identical duplicate gels of the same lysates. The blots shown in (D) are from a separate experiment to that shown in (B) and (C). In this experiment total cell lysates were separated on 10% denaturing gels prior to transfer and immunoblotting. Each experiment was performed at least twice with similar results.

To confirm and extend this observation, flag-Sav was coexpressed with myc-Mst1 or myc-Mst2 in 293T cells, and complexes were isolated by coimmunoprecipitation. As seen in Fig. 1B, both Mst1 and Mst2 were efficiently coimmunoprecipitated with flag-Sav. The efficiency of this coprecipitation was similar to that of the direct immunoprecipitation of Mst1 and Mst2 with anti-myc IgG, suggesting that most of the Mst kinases were in association with Sav. Interestingly, the coexpression of Mst kinases, especially Mst2, appeared to increase the abundance of Sav (Fig. 1C). To confirm this, we repeated the experiment, and again found that the presence of either Mst1 or Mst2 appeared to increase the abundance of Sav (Fig. 1D). Once again, despite similar expression levels themselves, Mst2 consistently had a greater stabilizing effect on Sav than Mst1. This effect was not due to differences in transfection efficiency because coexpression of Sav with green fluorescent protein or another protein that does not bind Sav (peptidyl-prolyl cis-trans isomerase A; PPIA) (see below), had no effect on Sav abundance (Fig. 1E).

Mst kinase and hSalvador interact via their C-terminal coiled-coil domains

Mst1, Mst2, and Sav all contain C-terminal coiled-coil domains (Fig. 2). Because coiled-coil domains mediate protein interactions, and Mst1 has previously been shown to homodimerize via its C-terminal coiled-coil domain [7], we hypothesized that Sav interacted with Mst kinases via these domains. To test this, we engineered C-terminally truncated mutants of Mst1 and Mst2 that lacked the coiled-coil domain, and determined whether they were capable of interacting with wild-type Sav. Consistent with earlier experiments, the full-length Mst kinases efficiently coprecipitated flag-Sav, but the truncated mutants of Mst1 and Mst2 did not. Similarly, in the reciprocal coimmunoprecipitations, flag-Sav was able to bring down full-length Mst1 and Mst2 but not Mst proteins that lacked their coiled-coil domains (Fig. 3A).

Figure 2.

 The structure of Mst kinase and hSalvador. Schematic illustrations of Mst kinase and hSalvador primary structures show the relative positions of their functional domains. The structure of Mst1 is given representatively for the very similar Mst1 and Mst2 kinases and shows the position of the kinase domain (light grey box), the inhibitory domain (medium grey box), C-terminal coiled-coil/dimerization domain (black box) and the caspase cleavage site (arrowhead). The Mst1 mutant lacking the coiled-coil domain, Δ433, analogous to that of Mst2 Δ437, is also shown. The structures of wild-type and mutant Sav constructs used in this study are shown indicating the location of the coiled-coil domain and the proline binding WW domains (dark grey box).

Figure 3.

 Mst kinase and hSalvador interact via their C-terminal coiled-coil domains. (A–D) WT, C-terminally or N-terminally truncated mutants of flag-Sav cDNA were cotransfected with or without WT or C-terminally truncated mutants of myc-tagged Mst1 or Mst2 cDNAs into 293T cells as indicated. Two days after transfection cell lysates were prepared. Immune complexes and total cell lysates were separated by SDS/PAGE, transferred to membrane and sequentially immunoblotted as indicated on the left. The migration of Mst and Sav are marked with arrows and arrowheads, respectively. The position of immunoglobulin heavy chain in antimyc immune complexes is indicated with an asterisk (*). Each experiment was performed at least twice with similar results.

To confirm that the C-terminal coiled-coil domain of Sav was also required for binding to Mst kinases, we generated a series of C-terminally truncated Sav mutants (Fig. 2), and examined their ability to interact with Mst1 and Mst2. As seen in Fig. 3B, full-length Mst1 and Mst2 were able to coprecipitate versions of flag-Sav that bore the coiled-coil domain, namely flag-Sav WT and Δ374, but not the smaller proteins, Δ344 and Δ321, that lacked the domain. Again, in the reciprocal coimmunoprecipitations, flag-Sav and Δ374 were able to efficiently bring down full-length Mst1 and Mst2. Thus the C-terminal coiled-coil domains of both Sav and the Mst kinases are essential for their interaction.

Once again we noted that in lysates from cells that coexpressed full-length Mst1 or Mst2 together with Sav, levels of Sav were elevated compared to extracts that expressed Sav alone (Fig. 3A). However, when the truncated versions of Mst1 and Mst2 that could not bind to Sav were coexpressed, levels of Sav were unaffected. Therefore it appears that the interaction of Mst with Sav is required, and might be sufficient, for it to increase levels of Sav.

The levels of N-terminally truncated mutants of Sav that retained the C-terminal coiled-coil domain, and thus were able to bind Mst, were increased even more dramatically than WT Sav. As shown in Fig. 3C (top), successive deletions of the Sav N-terminus strongly destabilized these proteins to such an extent that the Sav(268–383) and (321–383) constructs were expressed at or below the limit of detection in this system. Indeed, several attempts to detect Sav(321–383) when expressed alone were unsuccessful. However, coexpression of Mst2 with these truncation mutants dramatically enhanced their abundance, particularly Sav(268–383) and (321–383), such that they were readily detected (Fig. 3C, top). As expected these N-terminal mutants were all able to bind Mst2, as demonstrated by the presence of Mst2 in antiflag immune complexes (Fig. 3C, bottom). Notably, the Sav(321–383) fragment was able to efficiently coprecipitate Mst2, indicating that the Sav coiled-coil domain is not only essential but is also sufficient for binding. Furthermore, despite their greatly different abundances, the three Sav mutants were able to coprecipitate similar amounts of Mst2 compared to WT Sav, suggesting that in this system Mst is limiting. Alternatively, it is possible that the coiled-coil domain on its own is able to interact with Mst with higher efficiency than the full-length protein. If so, this could be because other parts of Sav reduce access to the coiled-coil domain or that regions, such as the WW domain, interact with other proteins that exclude the interaction of Mst.

Expression of Sav(321–383) was only detectable when coexpressed with either WT Mst1 or Mst2, but not with mutants of Mst that lacked their coiled-coil domain (Fig. 3D, bottom). Consistent with the earlier results, Sav(321–383) coprecipitated Mst2 but not the mutants that lacked their coiled-coil domains and unexpectedly, also failed to coprecipitate Mst1 (Fig. 3D, top). It is possible that the interaction between the Sav coiled-coil domain and Mst1 and Mst2 inside cells is sufficient to stabilize its abundance, but that Sav's interaction with Mst1 is significantly weaker than with Mst2, such that its interaction with Mst1 is disrupted upon cell lysis.

hSalvador can homodimerize/multimerize independently of its coiled-coil domain

Based on the above findings, and earlier observations that Mst1 can multimerize (dimerize) via its C-terminal coiled-coil domain [7], we hypothesized that Sav also homo-multimerized via its coiled-coil domain in a similar way. To test this we coexpressed full-length hemagglutinin (HA) tagged Sav together with C-terminally truncated Sav mutants tagged with the flag epitope. As predicted, full-length HA-Sav was indeed capable of coprecipitating full-length flag-Sav (Fig. 4A). Unexpectedly, however, HA-Sav was also efficiently coimmunoprecipitated with flag-Sav Δ344 and Δ321, two mutants that lacked the coiled-coil domain. These results indicate that Sav can homodimerize/multimerize, but it does not require its C-terminal coiled-coil domain to do so.

Figure 4.

 hSalvador can homo-multimerize independently of its C-terminal coiled-coil domain. (A,B) HA-tagged Sav cDNA was cotransfected with either WT or C-terminally truncated mutants of flag-Sav cDNA or with HA-PPIA or flag-CrmA-DQMD cDNAs into 293T cells as indicated. Two days after transfection the cells were washed and lysed. Immune complexes and total cell lysates were separated on denaturing gels, transferred to membrane and sequentially immunoblotted as indicated on the left. The migration of HA-Sav and PPIA (B) is indicated with arrows while the position of flag-Sav is marked with arrowheads. (B) The position of CrmA-DQMD (lane 6) is marked with an asterisk (*). Each experiment was performed at least twice with similar results.

Next we attempted to identify the region(s) that mediate Sav homo-multimerization. To do this we coexpressed C-terminally truncated flag-Sav mutants with full-length HA-Sav. As seen in Fig. 4B, all mutants we examined were able to coprecipitate full-length HA-Sav at levels comparable with that of WT flag-Sav. Importantly, this multimerization was specific to Sav because when Sav was coexpressed with two unrelated proteins, HA-PPIA and flag-cytokine response modifier A-DQMD mutant (CrmA-DQMD), they failed to coimmunoprecipitate with Sav (Fig. 4B, lanes 5 and 6).

hSalvador is phosphorylated by Mst kinase

To determine whether Sav was a phosphorylation substrate of Mst1 or Mst2, we coexpressed them with Sav and separated the lysates on a 10% linear gel. A mobility shift of WT flag-Sav that is only apparent on linear gels, suggestive of phosphorylation, was seen, but only in lanes that coexpressed Mst1 or Mst2 (Fig. 5A, top). An equivalent mobility shift was also detected with flag-Sav Δ374 in the presence of Mst. This suggests that Sav itself might be phosphorylated by Mst kinase. Again, consistent with earlier results, coexpression of Mst resulted in an increased abundance of Sav WT and Δ374 proteins. This increase was more apparent when the same samples were separated on gradient gels (Fig. 5A, middle).

Figure 5.

 Mst kinase phosphorylates hSalvador. (A) Myc-tagged Mst1 or Mst2 cDNAs were cotransfected with WT or Δ374 flag-Sav cDNA into 293T cells as indicated. (B) WT flag-Sav cDNA was cotransfected with either WT or mutant Mst1 or Mst2 cDNAs into 293T cells as indicated. Two days after transfection total cell lysates were prepared and separated by SDS/PAGE, transferred and sequentially immunoblotted as indicated on the left. The same cell lysates in (A) were separated either on a 10% linear gel (top) or on a 4–20% gradient gel (middle and bottom) prior to transfer. The migration of Mst and Sav is marked with arrows while the position of the slower migrating form of Sav is indicated with an arrowhead. (C) Flag-Sav cDNA was cotransfected with either WT or kinase-dead Mst2 (K56R) cDNAs into 293T cells as indicated. After two days cells were labelled in vivo with 32P-orthophosphate. Immune complexes were separated on a 10% denaturing gel, transferred to membrane, dried and exposed to film. Following autoradiography, membranes were sequentially immunoblotted as indicated on the left. (D) WT or kinase-dead Mst2 (K56R) cDNAs were transfected into 293T cells. After two days cells were lysed and Mst kinases were immunoprecipitated. Antimyc immune complexes were washed, equally divided three ways and incubated for 30 min at 30 °C in kinase buffer with [32P]ATP[γP] either alone, or with purified flag-Sav or MBP as indicated. Reactions were terminated and separated by SDS/PAGE, transferred to membrane, dried and exposed to film. Following autoradiography, membranes were sequentially immunoblotted as indicated on the left. The position of Mst and Sav is marked with arrows while the position of immunoglobulin chains from antimyc immune complexes in lanes 1–8 is indicated with an asterisk (*). Lanes 3, 6 and 9 are blank lanes. Each experiment was performed at least twice with similar results except that shown in (C) which was performed once.

To determine whether Mst kinase had to bind to Sav to induce this mobility shift, we coexpressed WT Sav with Mst1, Mst2 or Mst truncation mutants that are unable to interact with Sav (Fig. 3). As seen in Fig. 5B, the coexpression of Sav with WT Mst1 or Mst2 altered the mobility of Sav, as well as increasing its abundance. In contrast, coexpression of Sav with the truncated mutants of Mst that were unable to bind to Sav failed to induce a mobility shift. These results indicate that the mobility shift of Sav is not simply due to overexpressing Mst kinases, but is likely to be a direct consequence of Mst kinase interacting with, and directly phosphorylating, Sav.

To confirm that Mst phosphorylates Sav in vivo, we first generated catalytically inactive ‘kinase-dead’ mutants of both Mst1 and Mst2. The mutation of K59R in the ATP-binding region of the Mst1 kinase domain renders the kinase inactive [7,11,12,16], and by homology with Mst1 the analogous mutation of K56R in Mst2 should also render the kinase inactive.

Flag-Sav was coexpressed with WT Mst2 and the K56R mutant in cells, and labelled in vivo with 32P-orthophosphate. As shown in Fig. 5C, 32P-labelled Sav is clearly detected in antiflag immune complexes from cells expressing Sav and WT Mst2. However, the amount of 32P-labelled protein was much less when Sav was coexpressed with the mutant kinase. Consistent with this result, the amount of 32P-labelled Mst2 was also reduced in cells coexpressing the mutant kinase. 32P-labelling of Sav and Mst2-K56R in this sample was presumably due to endogenous kinases, most likely endogenous Mst. Thus, these results show that Sav is phosphorylated as a result of coexpression with Mst kinase.

To confirm these results and to address whether phosphorylation of Sav by Mst is direct, we performed an in vitro phosphorylation assay using purified Mst kinases and flag-Sav as substrate (see below). Incubation of WT Mst2 alone (without substrate) resulted in robust autophosphorylation of the kinase (Fig. 5D, top). In contrast, kinase-dead Mst2 was unable to autophosphorylate despite identical amounts of immunoprecipitated kinase being present (Fig. 5D, bottom). Because myelin basic protein (MBP) can serve as a pseudosubstrate for Mst1 [16], we reasoned that this is probably also the case for Mst2, and it would serve as a positive control for this assay. Indeed, MBP is well phosphorylated by WT Mst2 but not by the mutant kinase. Similarly, the incubation of flag-Sav with WT Mst2, but not the kinase-dead Mst2, resulted in the phosphorylation of a protein that was superimposable with that of flag-Sav (Fig. 5D, top and middle, lane 4). The incubation of flag-Sav alone in this assay yielded no radiolabelled proteins (Fig. 5D, lane 10) indicating that the phosphorylation of Sav is due to the addition of purified WT Mst2 rather than some other protein that had copurified with flag-Sav. This result provides strong evidence that Sav is directly phosphorylated by Mst2 kinase.

Having confirmed the ability of Mst to phosphorylate Sav, we then examined what role, if any, that phosphorylation played in the ability of Mst to increase the abundance of Sav. To do this we coexpressed both WT and kinase-dead mutants of Mst1 and Mst2 with flag-Sav, and determined the effect of these mutant Mst kinases on Sav stability. In contrast to WT Mst1 and Mst2, the kinase-dead mutants failed to induce a mobility shift in Sav (Fig. 6A). The failure of the kinase mutants to phosphorylate Sav was not due to an inability to bind to it, because both mutant kinases could be coimmunoprecipitated with flag-Sav (Fig. 6B). Despite being unable to phosphorylate Sav, both kinase mutants, particularly mutant Mst2, were nevertheless able to increase levels of Sav (Fig. 6A). Together, these results indicate that the stabilizing effect of Mst kinase on Sav is independent of its phosphorylation by Mst, but rather the association of Mst with Sav is required for stabilization of Sav.

Figure 6.

 The stabilization of hSalvador is independent of its phosphorylation by Mst kinase. (A,B) WT flag-Sav cDNA was cotransfected with either WT or kinase-dead Mst1 (K59R) or Mst2 (K56R) cDNAs into 293T cells as indicated. Two days after transfection cell lysates were prepared. Total cell lysates (A) or immune complexes (B) were separated by SDS/PAGE, transferred and sequentially immunoblotted as indicated on the left. Samples in (A) were separated on 9% a denaturing gel. The position of relevant bands is indicated with arrows except for the slower migrating form of Sav, which is marked with an arrowhead. Each experiment was performed at least twice with similar results.

Discussion

These results demonstrate that the mammalian scaffold protein, hSav (hWW45), can bind to the mammalian orthologues of Sterile Twenty kinase, Mst1 and Mst2. Recently, it was shown in yeast two-hybrid analyses that the C-terminal halves of the Mst2 and Sav proteins were required for interaction, but this study failed to define the region of interaction [28]. As we demonstrate here, this association is absolutely dependent upon their respective C-terminal coiled-coil domains (Fig. 3). A slightly larger region of similarity (≈ 50 amino acids) between Sav and Mst harbouring most of the coiled-coil domain, dubbed the Sarah domain, was previously predicted to be essential for interaction between the two proteins [29]. Consistent with this finding, the truncation mutants Mst1 Δ433, Mst2 Δ437 and Sav Δ321, that all lacked this region of similarity, all failed to heterodimerize. Furthermore, deletion of just the coiled-coil domain of Sav (Sav Δ344) was sufficient to abolish heterodimerization (Fig. 3B), indicating this domain is essential for interaction with Mst kinases. That the coiled-coil domain of Sav alone was sufficient to coprecipitate Mst2 (Fig. 3C,D) demonstrates that this domain is both necessary and sufficient to bind Mst kinase. These findings are consistent with studies in Drosophila that revealed the C-terminal coiled-coil domains of Sav and Hpo were also crucial and/or sufficient for their interaction [19,21,23]. These results indicate that the coiled-coil interaction between Mst and Sav has been evolutionarily conserved between flies and man.

The C-terminal coiled-coil domain in Mst1 is also required for it to form homodimers (multimers) [7]. Based on this finding, it seemed reasonable to predict that Sav might also homodimerize via its coiled-coil domain. We found that Sav could indeed specifically homodimerize (multimerize) but, surprisingly, this did not require its coiled-coil domain (Fig. 4). Because Salvador bears a WW domain, a motif that allows interaction with proline residues, we considered the possibility that this region might allow multimerization independently of the coiled-coil domain. Such a WW domain/proline interaction might explain why full-length HA-Sav, with its intact WW domains, was still able to interact with flag-Sav Δ199. Unfortunately, it was not possible to test this hypothesis, because the HA-Sav Δ199 construct, as well as a flag-tagged WW domain (residues 200–267) construct, was unstable. Therefore, although it is clear that Sav can homo-multimerize independently of its C-terminal coiled-coil domain, the region(s) involved remain to be determined.

It therefore seems probable that Sav, Mst1 and Mst2 exist in a state of equilibrium (competition) between Sav and Mst homodimers and the formation of Sav/Mst heterodimers, in an association dependent on their C-terminal coiled-coil domains (Fig. 7).

Figure 7.

 Possible activation model for the hSalvador/Mst kinase pathway. In the inactive state, Mst and Sav can coexist as functionally inactive homo- or heterodimers. Sav on its own is less stable than Sav bound to Mst. If activated by upstream signals active Mst kinase can interact with Sav via their coiled-coil domains. Alternatively, Mst might become activated while bound to Sav. The association of active Mst with Sav in itself induces a conformational change in Sav or leads to a phosphorylation-induced conformational change in Sav facilitating recruitment of unknown substrates (X) to the complex. Phosphorylation of substrates by Mst results in their activation or inactivation and subsequent downstream effects. Stabilization of Sav by Mst would enhance this process by potentially providing more scaffold ‘sites’ with which to recruit more substrates into the complex, thus strengthening the potential downstream effects.

The coexpression of Mst and Sav had two consequences. First, the abundance of Sav was increased in the presence of Mst, and second, Sav was phosphorylated by Mst (Figs 1, 3, 5 and 6). The effect of Mst2 on Sav phosphorylation and stability was almost always greater than that of Mst1, suggesting that Sav is a preferred partner/substrate of Mst2 compared to Mst1, which may reflect the fact that Drosophila Hpo is slightly more similar to Mst2 than to Mst1. Furthermore, the observation that Mst2 but not Mst1 could be coprecipitated with the Sav coiled-coil domain alone suggests that Sav binds Mst2 with a higher affinity than with Mst1 (Fig. 3D).

Both phosphorylation of Sav and its increase in abundance were dependent of the ability of Sav and Mst to interact. Deletion of the coiled-coil domain, and thus the ability of the two proteins to dimerize, abolished both the phosphorylation of Sav and the stabilizing effect of Mst on Sav abundance. We considered the possibility that the phosphorylation of Sav by Mst might account for the increased stability of Sav. However, while it is still possible that phosphorylation of Sav by Mst might further enhance its stability, the results in Fig. 6 show that association of kinase-dead mutants of Mst with Sav is sufficient to significantly enhance Sav abundance. A similar effect on Sav stability was also seen when a kinase-dead mutant of Hpo was coexpressed with Sav in Drosophila S2 cells [19], indicating the stabilizing effect of Mst on Sav expression is also conserved. Interestingly, N-terminal deletions of Sav rendered the mutant proteins less stable than the wild-type protein (Fig. 3C), however, when coexpressed with Mst2, a dramatic stabilizing effect was seen on the abundance of the smaller of these truncated proteins, namely Sav(268–383) and (321–383). Indeed, we have only ever been able to detect Sav(321–383) when coexpressed with either Mst1 or Mst2 (Fig. 3D). Sav/Mst heterodimers might be more stable than Sav homodimers because of conformational changes in Sav bound to Mst that render the protein more stable, or because Mst itself masks degradative signals in Sav. Alternatively, it may be that in its unbound state, the coiled-coil domain has a destabilizing influence on Sav. Thus, it seems that stability of Sav protein is increased by the presence of its N-terminal region as well its C-terminal coiled-coil domain due to its ability to bind Mst.

Phosphorylation substrates of Mst kinases have not been well characterized. Here we have provided strong evidence that Sav is indeed phosphorylated by Mst and that the phosphorylation is likely to be direct (Figs 5 and 6). The phosphorylation of Sav by Mst provides an additional means by which proteins may be recruited to the Mst/Sav complex, and in turn be phosphorylated by Mst (Fig. 7). Alternatively, phosphorylation of Sav might induce a conformational change in the protein that facilitates recruitment of substrates to the complex. The observation that Hpo can also phosphorylate Sav in S2 cells [19,20] suggests that this modification might be an important regulatory aspect of this complex.

In flies, the phenotypes of hpo and salvador mutants overlap with the phenotype of flies mutant for the serine/threonine kinase, warts[20–23]. Consistent with this, Hpo and Sav are capable of forming a complex with Wts that leads to its activation [20,22,23]. The WW domains of Sav mediate this interaction by binding to PPXY motifs in Wts [17]. It was shown recently that Mst2 could phosphorylate and activate the human orthologues of Wts, large tumour suppressor-1 and -2 (Lats1 and Lats2), both in vitro and in cells [28]. However, unlike the Hpo/Sav/Wts complex in flies, Lats1 was not detectable in a complex with Mst2 and Sav either in cells or using in vitro translated proteins, suggesting that the complex is either very unstable or that the activation of Lats by Mst2 might be indirect. Indeed, we also failed to detect endogenous Lats kinase in Sav/Mst immune complexes using a proteomics approach (data not shown), adding further support to the notion that Mst might indirectly activate Lats kinases in mammalian systems.

Mst1 and Mst2 are known to interact with several proteins. The growth inhibitory proteins, Rassf1 and Nore1 can form complexes with and inhibit Mst1 activity in an interaction involving their conserved C-termini [15,25]. Interestingly, while Nore1 and Rassf1 maintain Mst1 activity at low or basal levels it has been shown that Mst1 in complex with either Nore1 or with Rassf1 bound to the scaffold protein, connector enhancer of KSR1, CNK1, mediates the pro-apoptotic effects of a constitutively active Ras [15,25]. Furthermore, Nore1 appears to direct recruitment of Mst1 to Ras complexes following serum stimulation and the observation that artificially targeting Mst1 to the plasma membrane augments its pro-apoptotic activity has led to speculation that Nore1 and Rassf1 might direct Mst1 to sites of activation [15,16]. It is worth noting that endogenous Sav was not identified in these Mst-containing complexes. Moreover, in a proteomic screen using flag-Sav as bait we failed to detect the presence of endogenous Rassf-1 or Nore-1 in immune complexes (data not shown). Together these observations question the existence of proposed complexes such as Mst/Sav/Rassf1 [29]. Alternatively, the Mst/Sav interaction may be strong enough to prevent binding of other proteins to Mst, particularly when the interaction between Mst and Nore1/Rassf1 occurs through the same coiled-coil domain of Mst that we have shown binds Sav [15,16,25].

It was recently reported that in serum starved cells, Mst2 was sequestered into a complex with Raf1 that suppressed its activation and thereby prevented apoptosis from occurring [26]. Again, Sav was not detected in these Raf1 immune complexes, suggesting that complexes of Mst2/Sav and Mst2/Raf1 may also be mutually exclusive. For example, it may be that following mitogenic stimulation, Mst2 dissociates from Raf1, and is then free to bind Sav to induce its downstream effects. Nevertheless, more experiments will be required to exclude the possibility that Sav bound to Mst is also recruited into some of these regulatory complexes. Recent studies in Drosophila support a pro-apoptotic role for the Hpo/Sav/Wts pathway. Clearly the regulation of the Mst1 and Mst2 in mammals is more complex especially as orthologues of Nore1 and Rassf1 do not appear to exist in flies. How the Sav/Mst pathway contributes to apoptosis and/or coordinates with other regulatory pathways of Mst, if at all, remains to be determined.

Experimental procedures

Antibodies

Mouse monoclonal antiflag (M2) and agarose-conjugated antiflag (M2) beads were obtained from Sigma (Castle Hill, NSW, Australia). For immunoblotting, high affinity, rat monoclonal anti-HA (3F10) was purchased from Roche (Kew, VIC, Australia) and for immunoprecipitations, rabbit polyclonal anti-HA (HA.11) antisera was purchased from Covance (Berkeley, CA, USA). Mouse monoclonal antimyc (9E10) antibody was obtained from the monoclonal antibody facility, The Walter and Eliza Hall Institute (Bundoora, VIC, Australia). Mouse monoclonal antimyc (9B11) and rabbit polyclonal antibodies to Mst1 (#3682) and Mst2 (#3952) were purchased from Cell Signaling Technology (Genesearch, Arundel, QLD, Australia).

Plasmids and cDNAs

The mammalian expression plasmid, pcDNA3 (Invitrogen, Melbourne, VIC, Australia), containing N-terminally flag-tagged human Salvador (hSav) cDNA was a kind gift from D Haber (MGH Cancer Center, Charlestown, MA, USA). J Chernoff (Fox Chase Cancer Center, Philadelphia, PA, USA) generously provided the N-terminally myc-tagged cDNAs for mammalian sterile20 kinases (Mst1 and Mst2). Flag-CrmA-DQMD was described previously [30]. Peptidyl-prolyl cis-trans isomerase A cDNA was cloned into pcDNA3 with a C-terminal HA tag (HA-PPIA). Mst1 and Mst2 were subcloned into pcDNA3 prior to use in expression studies. All mutant cDNA constructs were generated by PCR using Pfu DNA polymerase and subcloned into either pcDNA3 or pcDNA5 FRT/TO (Invitrogen) expression plasmids. All constructs were sequenced for authenticity and purified using Qiagen Maxi prep kits (Qiagen, Clifton Hill, VIC, Australia).

Cell culture

293T cells were grown continuously in Dulbecco's Modified Eagle medium supplemented with 10% (v/v) foetal bovine serum (Gibco, Melbourne, VIC, Australia), penicillin G (50 U·mL−1), streptomycin (50 µg·mL−1) and l-glutamine (2 mm) in a humidified atmosphere of 10% CO2 at 37 °C. Cells were seeded at 10–15% confluency the day prior to transfection of cDNA constructs (1 µg plasmid DNA total per10 cm dish) using Effectene (Qiagen) according to the manufacturer's specifications. Flp-In™ T-REx™-293 cells (Invitrogen) were maintained as above but in the presence of blasticidin (9 µg·mL−1) and hygromycin (90 µg·mL−1). Isogenic, stable inducible cell lines were generated according to the manufacturer's guidelines. Protein expression was induced in these cells by overnight treatment with doxycycline (25 ng·mL−1).

Cell lysis, immunoprecipitation and immunoblotting

Cells were washed with phosphate-buffered saline (NaCl/Pi) and incubated for 1 h on ice in DISC lysis buffer [150 mm NaCl, 2 mm EDTA, 1% Triton X-100, 10% glycerol and 20 mm Tris pH 7.5 supplemented with complete protease inhibitor cocktail (Roche), 10 mm NaF, 2 mm Na pyrophosphate, 1 mm Na molybdate and 5 mmβ-glycerophosphate][31]. Total cell lysates were clarified by centrifugation before immunoprecipitation with antibodies raised against flag, Mst1, myc or HA. Anti-HA, -Mst1 and -myc immunoprecipitations were performed in the presence of protein G sepharose. Immune complexes were washed three times with DISC lysis buffer before being eluted with 100 mm glycine pH 3.0 and neutralized with 1 m Tris pH 8.0. Unless otherwise indicated, immune complexes or total cell lysates were separated by SDS/PAGE on 4–20% Tris-glycine gradient gels (Bio-Rad, Regent Park, VIC, Australia) and transferred to either poly(vinylidene diflouride) (Millipore, North Ryde, NSW, Australia) or Hybond C (Amersham, Castle Hill, NSW, Australia) membrane. Membranes were blocked in 20% horse serum (JRH Biosciences, Brooklyn, VIC, Australia)/NaCl/Pi containing 0.05% Tween-20 (PBST) before incubation with primary antibody. Membranes were washed with PBST, incubated with horseradish peroxidase-conjugated secondary antibody (Amersham) and washed before detection with enhanced chemiluminescence (Amersham).

In vivo labelling with 32P-orthophosphate

Two days after transfection, cells were incubated at 37 °C in phosphate-free media [Dulbecco's Modified Eagle without Na-phosphate (Gibco) supplemented with 10% (v/v) dialysed foetal bovine serum (Gibco), 1 mm Na-pyruvate and l-glutamine (2 mm)]. After 2 h, 2.5 mCi 32P-orthophosphate (Amersham) was added per 10 cm dish and incubated for a further 2 h at 37 °C. Cells were washed twice with ice-cold NaCl/Pi before being lysed in DISC lysis buffer and incubated for 1 h on ice. Immune complexes were prepared as described above and separated by SDS/PAGE before being transferred to poly(vinylidene diflouride) membrane, dried and exposed to film at −80 °C. Following autoradiography membranes were blocked and immunoblotted as necessary.

Purification of flag-Sav for in vitro phosphorylation

In our hands a bacterially expressed GST-Sav construct was insoluble. Therefore to generate sufficient quantity of Sav to use as a substrate in vitro we transiently transfected ten 15 cm plates of 293T cells with flag-Sav cDNA. Two days after transfection, cells were lysed and flag-Sav immunoprecipitated with antiflag beads as described above. Immune complexes were eluted with 100 mm glycine pH 3.0, neutralized with 1 m Tris pH 8.0 and dialysed against kinase buffer (see below) and stored at −80 °C. A Coomassie stainable amount of purified flag-Sav protein was verified by SDS/PAGE and used as substrate in subsequent in vitro phosphorylation assays.

In vitro phosphorylation assay

Two days after transfection, 293T cells were lysed and myc-tagged WT or kinase-dead (K56R) Mst2 was immunoprecipitated with antimyc (9B11) and EZ-view Red protein A affinity gel (Sigma) as described above. Immune complexes were washed twice with DISC lysis buffer and once with NaCl/Pi. Complexes were equally divided three ways before being pre-equilibrated in kinase buffer at 4 °C (50 mm Hepes pH 7.4, 10 mm MgCl2, 1 mm dithiothreitol, 10% glycerol, 1 mm EDTA, 1 mm EGTA, 100 mm NaCl, 1 mm NaF, 5 mmβ-glycerophosphate, 1 mm Na molybdate, 100 µm ATP and protease inhibitor cocktail). Five microcuries of [32P]ATP[γP] (PerkinElmer, Rowville, VIC, Australia) was added to each sample before being incubated for 30 min at 30 °C either alone or in the presence of purified flag-Sav (see above) or 2.5 µg of MBP (Sigma). Reactions were terminated by the addition of 5× SDS-sample buffer and samples separated by SDS/PAGE, transferred to membrane, dried and exposed to film at −80 °C. Following autoradiography membranes were blocked and immunoblotted as necessary.

Acknowledgements

We sincerely thank Jonathon Chernoff and Dan Haber for generously providing cDNAs used in this study. We thank members of the Vaux lab for helpful discussions. DLV is an ARC Federation Fellow. AMV is an ARC Queen Elizabeth II Research Fellow. This work was supported by funds from NHMRC Program Grant (257502) and Leukemia and Lymphoma Society Center Grant.

Ancillary