The results of this study help define the architecture and regulation of the striatin family complexes. A model for the organization of the striatin complex is presented in Figure . The creation of striatin point mutants deficient in PP2A association facilitated a set of experiments clearly identifying striatin-associated PP2A as the phosphatase negatively regulating Mst3 activation in striatin family complexes. Moreover, mutational analysis of Mst3 autophosphorylation sites and studies with the PP2A inhibitor, okadaic acid, provided new insight on Mst3 activation.
Figure 10 Model of human wild-type striatin complex, based on results from previous studies and the current study. Schematic of wild-type human striatin, drawn to scale, showing the locations of its previously published protein-interaction motifs/domains and residues (more ...)
The A and C subunits of PP2A were previously shown to bind to striatin family complexes but the region of striatin to which the A/C heterodimer bound was unknown [8
]. Our current data show that the coiled-coil domain of striatin mediates the formation of this PP2A heterotrimer (Figure ). Although we cannot rule out a direct contribution of the caveolin-binding motif, the fact that double or triple point mutations in the middle of the coiled-coil/oligomerization domain of striatin almost completely disrupt PP2A association suggests that the primary determinants for PP2A binding are near the middle of the coiled-coil domain (residues 84-105).
Our data also suggest that PP2A association with striatin is dependent on oligomerization of striatin complexes. Of the charged coiled-coil residues that were mutated to disrupt PP2A binding to striatin, only arginine 88 is predicted to be at the dimerization interface when either NCOILS [33
] or Paircoil2 [34
] prediction programs are used, while the other residues are predicted to be more available to potentially interact with PP2A. However, arginine 101, mutated in the R100S/R101E mutant, is part of a signature in striatin called a trimerization motif [35
], and could have an effect on topology as well. Of the hydrophobic coiled-coil mutations, leucine 84, isoleucine 102, and leucine 105 are predicted to be at the dimerization interface, while leucine 94 may be more accessible. Thus, it is interesting that the striatin mutant L84A/L94A/L105A is significantly more defective in PP2A binding than the mutant L84A/L94A/I102A (p = 0.029), because the only difference between these two mutants is in amino acids predicted to be at the interface of the helices. These results are consistent with the idea that association of the PP2A A/C heterodimer is sensitive to local alterations in the dimerization interface of striatin. We hypothesize that this sensitivity results from PP2A binding asymmetrically across the dimerization interface to more than one striatin family coiled-coil domain. Further support for this model comes from the fact that deletion of the caveolin-binding motif, which leaves the residues mutated above intact, abolishes oligomerization of striatin and its association with PP2A. Thus, although a precise understanding of the effects of the different striatin mutants on PP2A association may await crystallization studies, our current data are consistent with a model in which the PP2A A/C heterodimer binds to the coiled-coil domain of striatin family members in an oligomerization-dependent manner. Because we have shown that Mst3 binding to striatin is not dependent on striatin family oligomerization, this model provides an attractive feature of restricting the regulation of striatin-associated Mst3 by PP2A until oligomerization occurs.
A surprising result from this study was the complete loss of oligomerization caused by deletion of the caveolin-binding motif within striatin (mutant Δ(53-66)). A previous study [15
] reported that a fusion protein consisting of residues 81-131 of mouse SG2NA (corresponding to striatin amino acids 65-115) fused to the C-terminus of GFP did not oligomerize for unknown reasons. One possible explanation for both of these results is that striatin family oligomerization requires binding of caveolin. This would have to be a transient requirement, however, because we could find no associated caveolin even in wild-type striatin immunoprecipitates (data not shown) and no caveolin was found in striatin family complexes in a recent proteomics study [9
]. Alternatively, loss of oligomerization could result from one or more of the deleted residues being directly required for oligomerization. Analyses using NCOILS [33
] and Paircoil2 [34
] prediction programs suggest that the N- and C-terminal limits previously assigned to the coiled-coil domain of striatin (70-116; [11
]) may need to be extended. Using a 21 residue window with NCOILS, which is recommended for locating the ends of coil-coils, and an unweighted analysis (heptad residues a-g given equal weight), probabilities of almost 1 are obtained for residues 61-69 being part of the coiled-coil of human striatin. Weighted analysis (residues a and d given 2.5-fold more weight than the other heptad residues) using the same window still assigns these residues high probability (0.621 for residues 61-63 and 0.956 for residues 64-69) of being part of the coiled-coil. Results from similar analyses of human SG2NA suggest that SG2NA residues corresponding to striatin aa61-69 (unweighted) or aa64-69 (weighted) have a high probability of being part of the coiled-coil. Finally, for human zinedin, NCOILS predicts unweighted and weighted N-terminal ends corresponding to striatin aa56 and aa61, respectively. For striatin, SG2NA, and zinedin, these analyses also suggest that the C-terminal end of the coiled-coil may be around aa120-121. Paircoil2 analysis of these proteins predicts a similar end to the coiled-coil but an even earlier start to it, ranging from striatin aa56-61, potentially including much of the caveolin-binding motif, residues 53-63. The results of these analyses are consistent with the idea that our caveolin deletion mutant and the GFP-SG2NA fusion protein published by others may have removed amino acids critical to the formation of striatin coiled-coils. Since the missing residues common to both of these constructs are residues 53-64, it is possible that these are the critical amino acids. In fact, the complete loss of coiled-coil formation caused by our small caveolin-binding motif deletion is reminiscent of "trigger" sequences, sequences absolutely required for coiled-coil formation [36
]. Additional analyses will be needed to firmly define the striatin coiled-coil and resolve these possibilities.
There are at least two distinct domains within striatin (aa1-269 and aa310-780) that interact directly or indirectly with Mob (Figure ). The C-terminal striatin residues important for Mob3 binding were further narrowed down by the fact that deletion of striatin residues 270-344 had no significant effect on Mob3 binding (Figure ). This result suggested that the most important C-terminal determinants of striatin for Mob binding likely occur after residue 344 and include the WD domain. However, because of the larger standard deviation of Mob3 binding to the Δ(270-344) mutant, a contribution of striatin residues 310-344 cannot be completely ruled out. In the N-terminus of striatin, Mob3 binding seems to be largely restricted to the coiled-coil domain because deletion mutants collectively spanning residues 115-269 and a mutant deleting the caveolin-binding motif had no significant effect on Mob3 binding, while deletion of the coiled-coil domain had the same reduction of Mob3 binding as the Δ(3-309) striatin mutant. The fact that Mob3 binds much better to the caveolin-binding motif deletion mutant than to the coiled-coil deletion mutant and close to wild-type levels also supports the idea that Mob3 binding to the striatin N-terminus is not highly dependent on oligomerization. This result suggests that Mob3 binding to the striatin N-terminus is not solely due to indirect binding of Mob3 through oligomerization of endogenous striatin family members with mutants. However, additional studies are necessary to strengthen this conclusion. It has not been firmly established whether Mob3 binds directly or indirectly to striatin. A previous study identified Mob3 (phocein) by its binding to striatin in a two-hybrid assay but no additional data was presented to rule out the possibility that the interaction was mediated by a conserved protein in yeast (e.g., PP2A A and C subunits) bridging the association. Mob3 binds striatin independently of PP2A C subunit, Mst3, and CCM3 because when close to 85% of the latter associations were disrupted by various mutations little to no reduction in Mob3 binding was seen. Since PP2A C subunit, Mst3, and CCM3 are other core striatin binding partners, the results support the model that Mob3 binds directly to striatin. However, the question of whether striatin binds two molecules of Mob3 or uses two binding domains to interact with a single molecule of Mob3 still needs to be resolved. Given our other data on the organization of the striatin complex, it is tempting to speculate that binding of two striatin domains to one Mob3 molecule might function in part to bring Mst3 and PP2A into proximity in the 3D structure of striatin.
The results of this study show that striatin-associated PP2A is the phosphatase responsible for negatively regulating the phosphorylation and activation of Mst3 in striatin family complexes. Four separate striatin point mutants deficient in PP2A binding showed hyperphosphorylation of associated Mst3. Moreover, hyperphosphorylated, gel-shifted Mst3 could also be detected in lysates from cells expressing these mutants, indicating that these mutants cause the hyperphosphorylation of a significant portion of the Mst3 in the cell. Using Mst3 activation loop point mutants, autophosphorylation site T178-specific antibody, and okadaic acid to inhibit PP2A and activate Mst3 in vivo [37
], we demonstrated that gel-shifted Mst3 was indicative of Mst3 activation and autophosphorylation in vivo. This conclusion is further supported by previous studies that reported a gel shift of Mst3 upon autophosphorylation in vitro [19
], and a consequential increase in kinase activity after autophosphorylation [38
]. Our data further show that the Mst3 gel shift represents an underestimation of the amount of Mst3 autophosphorylated on threonine 178, the reported site of Mst3 autophosphorylation for Mst3b [19
], because phosphorylation at both threonines 178 and 182 on Mst3b appears to be required to obtain the observed gel shift. Thus, striatin-associated PP2A negatively regulates the phosphorylation and activation of Mst3, and likely of the other related kinases in striatin family complexes, Mst4 (which is also gel-shifted upon okadaic acid treatment) and STK25, as well. The fact that reduction in striatin-associated PP2A is sufficient to cause activation of Mst3 suggests that one way the Mst3 and the other GCKIII kinases could be regulated is by modulation of striatin-associated PP2A activity or modulation of PP2A's access to Mst3 in striatin family complexes. An important goal of future research on these complexes will be to determine if these mechanisms exist, and if so, to elucidate them.
Previous data suggested that threonine 182 was not a major autophosphorylation site on Mst3 [19
]. However, this data was largely based on peptide phosphorylation studies. We (Figure ) and others [19
] have found that incubation of immunoprecipitated Mst3 in the presence of ATP and manganese generates the hypershifted form of Mst3 in vitro. Considered together with the current result that mutation of threonine 182 to alanine completely prevents the gel-shift of Mst3 without blocking the phosphorylation of threonine 178, this finding supports the notion that threonine 182 is a major autophosphorylation site of Mst3. The fact that there is no detectable gel shift when either threonine 178 or threonine 182 is mutated to alanine is intriguing. One possible explanation is that simultaneous phosphorylation at both of these sites is required to produce the gel shift. Another possibility is that there may be a preference for sequential phosphorylation of Mst3 activation loop phosphorylation sites, with phosphorylation at threonine 178 being a prerequisite for phosphorylation at threonine 182, which in turn is required for the observed gel shift to occur. Evidence for an interdependence in activation loop phosphorylations has been noted previously for other kinases such as Chk2 [39
]. Availability of a phospho-specific antibody for threonine 182 of Mst3 would be helpful for distinguishing between these and other possibilities.
Deletion of the calmodulin-binding domain (Δ(148-166) striatin), which has no effect on striatin oligomerization or binding of Mob3 to striatin and little effect on PP2A association with striatin (Figure ), caused a 3- to 4-fold increase in both Mst3 and Mst4 association with striatin. This result is consistent with the possibility that the calmodulin-binding domain negatively regulates the binding of Mst3 and Mst4 in striatin family complexes. The effect of this deletion mutant is not simply due to the removal of amino acids in this general area of the striatin sequence because adjacent deletions on either side of the calmodulin-binding domain had either no effect or only a modest 25% increase in Mst3 binding. Given that we and others have shown that striatin family members associate with calmodulin in a calcium-dependent manner [8
], calcium may regulate Mst3 and Mst4 recruitment to striatin complexes by regulating calmodulin binding to striatin. However, it is also possible that removal of these specific sequences modifies Mst3 binding in a manner that will not be recapitulated by changes in calcium. Further experimentation will be required to distinguish between these possibilities.