Generation of striatin mutants for structure-function analysis of striatin complexes
To facilitate dissection of the molecular organization of striatin complexes, we generated a set of deletion mutants and a complementary set of point mutants of striatin for use in our experiments. Because all cells tested to date contain endogenous striatin, all striatin mutants were constructed with an N-terminal Hemagglutinin (HA)-epitope tag to allow specific immunoprecipitation and detection of the exogenously expressed wild-type and mutant striatins. Figure 1A shows the known domain organization of striatin while Figure 1B shows a schematic of the different striatin mutants used in this study on the same scale as Figure 1A.
The WD-repeats of striatin are not required for PP2A binding but contribute to Mob3 association
The A and C subunits of PP2A are known to bind striatin [8] but the region of striatin that binds to PP2A is unknown. As a first approach to identifying striatin sequences important for PP2A association, we compared the relative ability of HA-epitope tagged wild-type striatin and two HA-tagged striatin C-terminal deletion mutants, K270Stop and F457Stop (Figure 1), to bind PP2A in vivo. Both bind PP2A (Figure 2A). Quantitation of the ratio of PP2A bound to striatin shows that loss of seven of eight of striatin's WD-repeats (F457Stop) has no effect on PP2A association while loss of all eight WD-repeats plus the central region of striatin (K270Stop) has only a small effect on PP2A binding that was not statistically significant (Figure 2B). These data indicate that residues 270-780 of human striatin are largely dispensable for PP2A association and that the first 269 amino acids of striatin contain the primary PP2A binding domain.
To determine the importance of C-terminal striatin sequences for Mob3/striatin complex formation, the same immunoprecipitates were probed for Mob3. Mob3, like PP2A, also binds to both of the C-terminal deletion mutants (Figure 2A). Thus, amino acids 1-269 in striatin also contain a domain that binds Mob3, either directly or indirectly. However, quantitation of results from multiple experiments shows that Mob3 binds at reduced levels to both of the C-terminal mutants (Figure 2B). Loss of seven WD-repeats (F457Stop) or all eight WD-repeats plus the central region of striatin (K270Stop) results in a 29% or 45% reduction in Mob3 binding, respectively. These results demonstrate that the WD-repeats are important for efficient association of Mob3 and that residues between 270 and 457 may also contribute. Thus, striatin residues 1-269 contain binding sites for both Mob3 and PP2A.
Both N- and C-terminal sequences of striatin associate with Mob3 but only N-terminal striatin sequences associate stably with PP2A
The reduced binding of Mob3 to both of the striatin C-terminal deletion mutants suggested that striatin sequences beyond residue 269 might bind Mob3. To test for this possibility, immunoprecipitates of Δ(3-309) striatin, an N-terminal deletion mutant of striatin (Figure 1), were probed for the presence of Mob3. The Δ(3-309) striatin mutant bound Mob3 at approximately 40% the level of wild-type striatin (Figure 2C-D). Thus, there are at least two distinct domains within striatin (aa1-269 and aa310-780) capable of interacting directly or indirectly with Mob3. PP2A C subunit, on the other hand, did not bind to Δ(3-309) striatin (Figure 2C-D), demonstrating that PP2A associates stably only with residues located in the N-terminal region of striatin.
The coiled-coil and caveolin-binding domains of striatin, but not the calmodulin-binding domain of striatin, are necessary for oligomerization and for PP2A binding
To further localize the PP2A-binding domain within the N-terminal 269 amino acids of striatin, HA-tagged striatin mutants lacking previously identified N-terminal protein-interaction domains (caveolin-binding, coiled-coil, or calmodulin-binding) were created (Figure 1). Co-immunoprecipitation was utilized to test the abilities of these mutants to bind PP2A C subunit (Figure 3). The coiled-coil domain of striatin is required for binding PP2A, since a coiled-coil deletion mutant, Δ(70-116), is unable to specifically co-precipitate PP2A C subunit (Figure 3A-B). It has been reported that the coiled-coil domain is the oligomerization domain of the striatin family of proteins [15]. In agreement with this, Δ(70-116) striatin fails to oligomerize with SG2NA and zinedin, as demonstrated by the absence of SG2NA and zinedin in immunoprecipitates of this mutant (Figure 3A-B). Δ(70-116) also shows reduced ability to bind Mob3, binding only 40% of wild-type levels. Considering the fact that there is a Mob3 binding domain in aa1-269 of striatin that can bind approximately 55% of the wild-type level of Mob3 (Figure 2B), these results suggest that the coiled-coil domain contains binding sites for both PP2A and Mob3.
In addition to the coiled-coil domain, striatin has been reported to bind caveolin-1 through a caveolin-binding domain found within amino acid residues 53-63 [10]. To determine the role, if any, of the caveolin-binding domain in striatin association with PP2A and Mob3, a mutant of striatin lacking the caveolin-binding domain, Δ(53-66) striatin, was analyzed. Co-immunoprecipitation studies revealed that Δ(53-66) striatin binds little to no PP2A (Figure 3A-B). In contrast, the Δ(53-66) striatin mutant could co-precipitate Mob3 at near wild-type levels, confirming that loss of PP2A binding is specific and not due to protein misfolding. Interestingly, this mutant is completely defective in oligomerization as indicated by its inability to bind SG2NA and zinedin (Figure 3A-B), indicating that one or more residues within 53-66 are essential for striatin oligomerization.
Finally, we examined the importance of the calmodulin-binding domain in the N-terminus of striatin by analyzing the striatin deletion mutant, Δ(148-166), which deletes this entire domain (Figure 1). Loss of striatin's calmodulin-binding domain has only a small effect on PP2A C subunit binding, and no significant effect on Mob3 binding or oligomerization with SG2NA and zinedin (Figure 3C-D).
Residues in the coiled-coil domain of striatin are critical for PP2A C subunit association independent from their role in oligomerization
The fact that the coiled-coil domain of striatin is required for its association with PP2A suggested that PP2A might bind directly to this region. If so, the introduction of point mutants in the coiled-coil domain may disrupt PP2A binding without interfering with striatin oligomerization. To test this hypothesis, we compared the peptide sequence of the coiled-coil domain of striatin from several species and identified residues that were completely conserved. Four such residues were Arg88, Lys89, Arg100 and Arg101. Pairs of these residues were substituted with uncharged serine residues or residues of opposite charge in an attempt to disrupt striatin/PP2A association. The resulting double point mutants, R88S/K89E and R100S/R101E striatin, were then analyzed for PP2A, Mob3, and SG2NA binding. The mutants R88S/K89E and R100S/R101E are able to bind only 40% and 10% of wild-type levels of PP2A C subunit, respectively (Figure 4A-B). Binding of Mob3 was not reduced. Both mutants retained the ability to efficiently bind SG2NA at wild-type (R88S/K89E) or near wild-type (81%; R100S/R101E) levels (Figure 4A-B), indicating that effects on PP2A C subunit can be separated from effects on oligomerization. These results support the hypothesis that PP2A binds directly to the coiled-coil domain of striatin.
To further investigate PP2A binding in the coiled-coil domain, four conserved hydrophobic residues (leucines 84, 94, and 105, and isoleucine 102) were replaced with alanine. Two triple mutants (L84A/L94A/I102A and L84A/L94A/L105A) reduced PP2A binding to 32% to 17% of wild-type levels, while striatin oligomerization remained intact (Figure 4C-D). No statistically significant reduction in the binding of Mob3 was seen. Together, these results show that critical residues in the PP2A binding domain of striatin are within the central portion of the coiled-coil domain between residues 84 and 105.
To define whether the coiled-coil domain was sufficient to bind PP2A and to test the importance of the amino acids preceding the caveolin-binding domain for PP2A binding, two additional mutants (P132Stop and 46-131) were constructed (Figure 1). Both of these mutants contain the coiled-coil domain. Because the caveolin-binding domain is also required for PP2A binding (Figure 3) both mutants were designed to retain this domain as well, but mutant 46-131 lacks most of the amino acids that precede it. PP2A binds efficiently to both mutants, demonstrating that the binding site for PP2A is within striatin amino acids 46-131 and that the first forty-five amino acids of striatin are dispensable for PP2A binding (Figure 4E).
Determinants within striatin residues 191-344 are critical for binding to Mst3 and Mst4 kinases and to CCM3
We next investigated the association of striatin with Mst3 kinase, a recently discovered component of striatin family complexes [9]. Initial experiments showed that Mst3 does not bind to Δ(3-309) striatin (Figure 5A), indicating that striatin residues before 310 are critical for its association. However, Mst3 bound well to Δ(70-116) striatin (Figure 5A) and Δ(53-66) (data not shown), indicating that Mst3 association with striatin does not require the coiled-coil or caveolin-binding domains or oligomerization of striatin. This result also indicates that the loss of PP2A binding to Δ(70-116) seen in Figure 3A-B is not simply due to misfolding of this mutant.
To further delineate the striatin residues needed for Mst3 binding, the ability of Mst3 to associate with a set of deletion mutants spanning regions of unknown function between the coiled-coil domain and residue 344 of striatin (Figure 1) was measured. The results of this analysis (Figure 5B-C) indicate that deletion of the residues between the coiled-coil domain and the calmodulin domain and deletions after the calmodulin-binding domain up to residue 193 cause no reduction in Mst3 binding. However, deletion of residues 191-269 causes a ~33% reduction in Mst3 binding while loss of residues 270-344 causes an ~85% loss of Mst3 binding. Deletion of striatin amino acids 191-269 and 270-344 also reduced Mst4 binding by 33 ± 5% and 75 ± 3%, respectively (average ± range of two experiments). These results indicate that the Mst3 and Mst4 binding sites probably span across amino acid 269, with the strongest interactions being C-terminal to this residue.
Because CCM3 has been reported to bind and stabilize Mst3 and Mst4 [21], the importance of striatin residues 191-269 and 270-344 for CCM3 binding was tested. Since commercially available CCM3 antibodies did not consistently detect CCM3, we established a stable FLAG-tagged CCM3-expressing cell line as described in Methods and analyzed the ability of this CCM3 to associate with HA-tagged wild-type and mutant striatins by co-immunoprecipitation. Deletion of striatin residues 191-269 and 270-344 reduced CCM3 binding ~40% and ~90%, respectively (Figure 6A-B), suggesting that CCM3 and the Mst3 and Mst4 kinases may bind as a complex to this region of striatin.
The calmodulin-binding domain of striatin negatively regulates association with the Mst3 and Mst4 kinases
The deletion mutant, Δ(115-149) striatin, had a statistically significant ~25% increase in the amount of Mst3 bound (Figure 5C). Since this region is just before the calmodulin-binding domain of striatin, we tested the ability of Mst3 to bind to Δ(148-166) striatin, in which the calmodulin-binding domain has been deleted. At the same time, we tested an additional mutant, Δ(227-309) striatin, spanning the region that seemed to affect Mst3 binding the most. The results indicate that Δ(227-309) striatin is almost completely defective in Mst3 binding, while still binding Mob3 at 75% of wild-type levels (Figure 7A-B). In striking contrast, Mst3 bound ~3.5-fold better when the calmodulin-binding domain of striatin was deleted (Figure 7A-B). Interestingly, probing of the same immunoprecipitates for the related striatin-associated kinase, Mst4, revealed a less dramatic reduction (38%) in Mst4 binding to Δ(227-309) striatin but a similar ~4-fold increase in binding to the calmodulin-binding domain deletion mutant.
Striatin-associated PP2A negatively regulates the phosphorylation of Mst3 kinase
Previously, two-dimensional analysis of SG2NA complexes from 32P-inorganic phosphate-labeled cells revealed an ~52 kDa unknown protein whose phosphorylation increased dramatically upon treatment of cells with okadaic acid at concentrations known to inhibit PP2A [14]. The size and estimated isoelectric point of that phosphoprotein are similar to Mst3, raising the possibility that striatin-associated PP2A regulates the phosphorylation state of striatin-associated Mst3. To test this possibility, we took advantage of the fact that phosphorylation slows the mobility of many proteins. Figure 8A shows that upon incubation with ATP and manganese in vitro, a portion of Mst3 undergoes a gel shift on SDS-polyacrylamide gel electrophoresis (SDS-PAGE). The gel shift does not occur in the absence of ATP or if the kinase inhibitor staurosporine is included (Figure 8A), indicating that the shift occurs due to autophosphorylation of Mst3 or phosphorylation of Mst3 by a tightly associated kinase.
Mst3 in lysates from untreated human 293 cells exists predominantly as one major band and a minor, more slowly migrating, upper band (Figure 8B, first lane). Upon treatment with 100 nM okadaic acid, the Mst3 in the lower band shifts to the upper band over time until at 4 h the upper band becomes the predominant band (Figure 8B). It was previously shown that treatment of mammalian cells with 100 nM okadaic acid for 6 h-24 h inhibits most cellular PP2A without significantly inhibiting PP1 because of the slow rate at which okadaic acid enters cells [31]. Considered together, these results indicate that Mst3 phosphorylation may be regulated by PP2A.
Okadaic acid-induced phosphorylation of Mst3 could be due to inhibition of PP2A in the striatin complex or to inhibition of another PP2A holoenzyme responsible for regulating the phosphorylation state of Mst3. To distinguish between these two possibilities, we tested whether selectively reducing the amount of PP2A associated with striatin would increase the steady-state phosphorylation level of Mst3 in those striatin complexes. We reasoned that if PP2A in the striatin complex were responsible for dephosphorylation of striatin-associated Mst3, then mutant striatins with reduced PP2A binding would bind phosphorylated Mst3 but dephosphorylate it at a reduced rate, causing the accumulation of hyperphosphorylated Mst3 in those striatin complexes.
To test this prediction, the four PP2A-deficient striatin coiled-coil domain point mutants were used. In cells expressing these mutants, PP2A that is not complexed with striatin (~98% of PP2A in the cell [8]) will be unaffected; only the amount of PP2A associated with HA-striatin will be affected. We compared the ratio of the upper (hyperphosphorylated) and lower bands of Mst3 associated with the PP2A-deficient striatin mutants with the ratio of these Mst3 bands associated with wild-type striatin. Figure 8C (HA-IP) shows the results of a representative experiment. Mst3 associated with striatin mutants deficient in binding PP2A has a greater proportion of the upper band than Mst3 associated with wild-type striatin, indicating that there is an increased amount of hyperphosphorylated Mst3 associated with the PP2A-deficient striatin mutants. Quantitative analysis of several experiments showed that Mst3 hyperphosphorylation was significantly increased in all four of our PP2A-deficient striatin complexes (Figure 8D; black columns), with the greatest increase in the R100S/R101E mutant shown in Figure 4 to be the most defective in binding PP2A. The reduction of PP2A binding to striatin causes the hyperphosphorylation (gel shift) of a large fraction of the associated Mst3, as can be seen most clearly by visual comparison of the Mst3 bands in the wild-type and R100S/R101E lanes in the HA-IP panel of Figure 8C. More than half of Mst3 associated with the R100S/R101E mutant is in the upper band.
To determine the effect on the total Mst3 population in the cell, we also analyzed the amount of hyperphosphorylation (upper band) of Mst3 in lysates from cells expressing exogenous wild-type or PP2A-deficient mutant striatins. Figure 8C-D (lysate panel and white columns) show that expression of PP2A-deficient striatin does cause an increase in total hyperphosphorylated Mst3 in cells. The most robust increase, ~2.5-fold, was again induced by expression of the striatin mutant most deficient in binding PP2A, R100S/R101E.
To definitively prove that the gel shift in Mst3 observed with PP2A-deficient striatin mutants is due to increased phosphorylation and not some other modification, we tested whether gel-shifted Mst3 associated with R100S/R101E striatin could be eliminated by treatment with purified PP2A. The slower migrating form of Mst3 disappeared when denatured R100S/R101E striatin immunoprecipitates were incubated with purified PP2A but not when they were incubated with PP2A and okadaic acid (Figure 8E). Together, these results indicate that PP2A in the striatin complex regulates the phosphorylation state of striatin-bound Mst3.
PP2A negatively regulates the activation of striatin-associated Mst3 kinase
To gain insight into the functional significance of the PP2A-regulated phosphorylation of Mst3 in the striatin complex, we determined the sites of phosphorylation on Mst3 necessary for the observed gel shift. To do this, the three reported activation loop phosphorylation sites of Mst3, threonine residues 172, 178, and 182 in Mst3 isoform b (PhosphoSitePlus [32]), were individually mutated to alanine to prevent their phosphorylation. After transfection of constructs expressing FLAG-tagged wild-type and mutant Mst3 isoform b proteins into HEK293 cells and treatment of the cells with okadaic acid or vehicle control (DMSO), lysates were analyzed by immunoblotting with anti-FLAG antibody (Figure 9; compare top two panels). The results show that mutation of threonine 172 causes a small reduction in the ratio of shifted to unshifted Mst3 bands while mutation of either threonine 178 or 182 to alanine abolishes the okadaic-acid induced gel shift of Mst3.
Next an autophosphorylation site-specific (pT178) antibody was used to further analyze the T172A, T178A, and T182A Mst3 mutants. Because Mst3, Mst4, and STK25 have similar sequences in their activation loops, this antibody reacts with all of these kinases when activated by autophosphorylation. To specifically examine Mst3, lysates of okadaic acid-treated cells expressing the FLAG-tagged Mst3 isoform b proteins were first denatured to disrupt complexes by heating with SDS and reducing agent as described in Methods. Then anti-FLAG immunoprecipitates of wild-type and mutant Mst3 proteins were prepared and immunoblotted with anti-phospho-Mst3 (pT178) antibody and with anti-FLAG antibody (Figure 9; two middle panels). The results confirm the specificity of the anti-phospho-Mst3 (pT178) antibody since only IgG background is seen in the T178A pMst3 immunoblot lane while a strong band of the T178A mutant protein is seen in the FLAG-Mst3 (FL-Mst3) immunoblot panel. The results also reveal that the T182A Mst3 mutant is phosphorylated robustly on threonine 178 and that Mst3 phosphorylated on threonine 178 is present in both the lower and upper bands of the Mst3 doublet in the wild-type and T172A lanes. Thus, phosphorylation of threonine 178 is not sufficient to generate the upper Mst3 band, but appears to be a prerequisite for its formation. No threonine 178 phosphorylation was detected on any of the FLAG-tagged Mst3 proteins when FLAG-Mst3 immunoprecipitates were prepared from vehicle control (DMSO)-treated cells (Figure 9; bottom two panels), indicating that the Mst3 autophosphorylation on threonine 178 seen with okadaic acid-treated cells was induced by okadaic acid and not pre-existing. Together, the results in this section show that both the upper and lower Mst3 bands contain phosphorylated Mst3 species but generation of the upper, hyperphosphorylated band appears to require at a minimum phosphorylation of both threonine 178 and threonine 182 in the activation loop of Mst3. Thus, the gel shift seen with PP2A-deficient striatin mutants is indicative of, but probably an underestimate of, activation and autophosphorylation of Mst3.