In this study, we have identified several phosphorylation sites within Plk2. The sites were uncovered by unbiased mass spectrometric analysis using active Plk2 compared to inactive Plk2 in mammalian cells. There was no obvious similarity among the autophosphorylation sites, and for the most part these sites did not conform to proposed Plk2 consensus sequences (Johnson et al., 2007
). However, such consensus sequences are not stringently defined, and consist merely of flanking acidic residues. Thus, the new sites shown in this work may reveal additional sequence utilization. It is also formally possible that phosphorylation of the sites identified here could have been promoted indirectly by a Plk2-dependent signaling cascade in heterologous cells. However, unequivocal demonstration that Plk2 does autophosphorylate these sites directly requires in vitro
kinase assays using purified kinase and a modified Plk2 substrate that lacks all other potential phosphorylation sites. Such assays are difficult due to the large number of sites, and the potential requirement of priming of one site by phosphorylation at a previous site (Seeburg et al., 2008
). Regardless of direct phosphorylation, these sites are novel and important for regulation of Plk2 function.
In general, our analysis showed that the majority of phosphosites did not have major effects on Plk2 conformation as only one of the mutations (S588E) resulted in a shift to the higher migrating band similar to CA-Plk2 (, lanes 16 and 22). The mutations did, however, have significant effects on apparent protein stability in COS7 cells. Because all constructs were expressed from the same promoter and untranslated regions, changes in expression were likely due to alterations in protein half-life. This conclusion is supported by data showing that half-life determination with cycloheximide produces stability changes that correspond with steady-state levels of selected mutants compared to WT Plk2, although we cannot exclude the possibility that coding region mutations could alter translational efficiency as well. There are several mechanisms that could account for changes in protein stabilization. Plk2 levels in the cell are thought to be regulated by ubiquitin-mediated degradation by the proteasome. Phosphorylation induced ubiquitination has been demonstrated for many proteins (Hunter, 2007
) and all of the mutant sites we identified surrounding the PBD follow a similar pattern of non-phosphorylatable mutants increasing stability and phosphomimetic mutants resulting in protein levels similar to WT, suggesting these phosphorylation sites may be regulating the activity of a phosphodegron. In addition to phosphorylation up-regulating ubiquitination, it can also result in the opposite, decreasing ubiquitination and increasing stability (Dan et al., 2004
). This is consistent with the S299A mutation decreasing protein stability in COS7 cells while the S299E mutation increases its stability relative to wild-type.
Two other mechanisms may account for the 5 phosphosites that resulted in increased stability whether they were mutated to alanine or glutamic acid. One of these sites, T413 is in a PEST motif (Zimmerman and Erikson, 2007
). PEST motifs are found in many unstable proteins and have been shown to serve as targets for calcium-activated proteases and calpains as well as by the ubiquitin-proteasome pathway (Dice, 1987
, Rechsteiner and Rogers, 1996
, Tompa et al., 2004
). Indeed, the PEST motif around T413 also contains a calpain cleavage site at Q418 as predicted from the CaMPDB calpain cleavage algorithm. A mutation of any kind in this motif may result in disruption of the degradation mechanism and increased stability of the protein. Finally, it is also possible that sites for recognition by E3 ligases may be dependant on conformational changes induced by phosphorylation rather than the negative charge introduced alone. In such a case, mutation to either alanine or glutamic acid may result in an inability to form the correct conformation to expose the E3 ligase site.
These data suggest that Plk2 autophosphorylation is an important component of its protein stability in COS7 cells. The fact that one of the hyperactive mutants exhibits similar stability to WT-Plk2 (S588E) while other hyperactive mutants like S299E and CA-Plk2 are more stable than WT suggests that phosphorylation is not exclusively a signal for degradation. There may be a combinatorial effect of multiple phosphorylations with some of the sites masking or overriding the destabilizing effects of other phosphorylated sites. Notably, when expressed in neurons, the significant differences in protein expression between WT-Plk2, KD, CA, S299A, and S588A were no longer present, suggesting that the mechanisms regulating Plk2 stability in neurons may be different than in COS7 cells, possibly reflecting the widely differing roles for Plk2 in dividing and non-dividing cells.
In addition to alterations in protein stability, several mutants strongly stimulated Plk2 kinase activity. These sites were found in two distinct loci and appeared to operate by three different mechanisms, based on their gel migration patterns, which we have used as an index of conformational state. The kinase appeared to exist in either a “closed” and detergent-resistant inactive state, or in an “open” and active state. One site, S588, was found to reside in the C-terminal region between the conserved polo boxes. Because the polo boxes are thought to form an inhibitory intramolecular loop (Jang et al., 2002a
, Lowery et al., 2005
), S588 could be involved in helping mediate association between PBD and kinase domain. The S588E mutation could then destabilize this association, resulting in an active and open conformation analogous to the CA-Plk2 mutation T236E (). Why then does S588A also activate Plk2? It is possible that the serine residue itself is important in helping to mediate PBD-kinase domain interaction, and that replacement of this amino acid to alanine partially destabilizes the intramolecular complex to form a “loose” complex (). Although this mutant did not exhibit gel mobility shift, suggesting the lack of a large conformational change, the intramolecular association may be loosened enough to allow more rapid interconversion to the fully active, open configuration in the presence of a substrate. Interestingly, this site was one of two phosphosites also detected in Plk3 (; S552 in Plk3 numbering), suggesting a conserved mechanism.
Hypothetical models for Plk2 autophosphorylation regulation of kinase activity
The second site that exhibited strong activation of Plk2 was S299, which resides within the kinase domain itself. Although S299A had no observable effect, behaving much like WT-Plk2 in our assays, S299E hyperactivated Plk2 without changing its gel migration, suggesting lack of major conformational changes. It is possible that S299E also partially destabilizes the kinase-PBD intramolecular interaction, similar to S588A, leading to a “loose” complex that is more easily activated by substrate (, left). Alternatively, as this mutation is in the kinase domain itself, it could exert a direct effect on the kinase active site that may enlarge the cavity and allow substrate entry despite intact PBD-kinase binding (, right).
It should be noted that the functional significance of the identified sites could also be important in other contexts not assayed here, such as kinase activity against different classes of substrates, or targeting to specific subcellular locales when expressed at endogenous protein levels. Of the mutations that did have a strong effect on Plk2 kinase activity, we did not find any that rendered the kinase inactive, suggesting that Plk2 does not autoregulate itself in a negative feedback manner. Rather, we speculate that Plk2 may be downregulated primarily at the level of degradative clearance by the ubiquitin-proteasome system (Pak and Sheng, 2003
). However, when expressed in hippocampal neurons, the identified hyperactive mutants behaved as predicted, causing profound loss of SPAR protein and dendritic spines, similar to CA-Plk2. These results strongly indicate that these mutations are indeed functionally equivalent to a constitutively active kinase.