PP1 and PP2A use opposite phospho-dependencies to control distinct processes at the kinetochore

Background:

There are about 400 Ser/Thr kinases in the human genome but only 28 known Ser/Thr phosphatases (1, 2). How does this small group of phosphatases accurately regulate the phospho-state of hundreds of substrates?

PP1 and PP2A-B56 belong to the same PPP family of phosphatases. They share a similar 280 residue catalytic domain, the activities of which are comparable (3). Yet one does not appear to fully compensate the loss of the other in vivo (4), suggesting that each performs many unique functions. Why and how is a given phosphatase better-suited for certain roles over another? What features make it distinct?

This study addresses these questions in the context of the kinetochore. During spindle assembly, the kinetochore needs to perform two critical functions. It must (A) make stable, accurate attachments with microtubules and (B) signal the cell to maintain mitotic state until (A) has been achieved. These two functions are controlled by the phospho-states of the kinetochore protein Ndc80 and MELT motifs of the Knl1 protein respectively. Aurora B kinase phosphorylation of Ndc80 destabilises incorrect kinetochore-microtubule interactions. MELT motif phosphorylation by Mps1 kinase signals the cell to remain in mitosis. These two key phospho-modifications are also regulated by kinetochore phosphatases.

Both PP1 and PP2A-B56 are recruited to the Knl1 subcomplex on the kinetochore. PP1 directly binds the N-terminus of Knl1 to a RVSF motif. PP2A-B56 binds a LSPIIE motif of a Knl1 interactor. Interestingly, both these motifs are phosphorylated. And rather curiously, this has opposite effects on binding of the respective phosphatase. PP1 binding is reduced by phosphorylation of RVSF (5, 6), while PP2A-B56 binding is enhanced by LSPIIE phosphorylation (7, 8, 9).

Why are two similar phosphatases recruited to the same scaffold? Do they perform different functions? How are their individual activities regulated? Do their dissimilar binding patterns play a role?

 

Key findings:

The authors first establish that the PP1 and PP2A-B56 do not perform identical functions at the kinetochore. Deleting LSPIIE, but not RVSF, results in increased Ndc80 phosphorylation and chromosome alignment defects. This implies that in a wild-type kinetochore, only PP2A-B56 directly regulates phospho-Ndc80.

Removal of either LSPIIE or RVSF results in mitotic arrest and high phospho-MELT. However, both phosphatases do not directly dephosphorylate MELT. PP2A-B56 is known to promote PP1 association to the kinetochore by dephosphorylating Knl1 (10). When these Knl1 phosphosites are mutated, removal of the LSPIIE motif no longer causes mitotic arrest, while deleting RVSF still results in arrest and high phospho-MELT. Collectively, these results indicate that only PP1 directly regulates mitotic arrest/exit. While PP2A-B56 indirectly facilitates PP1 recruitment, it does not dephosphorylate MELT motifs on wild-type kinetochores (even though it is detectable there).

Two hypotheses were tested to explain these differences. First – the identity of the phosphatase could be critical to each function. However, deleting the motif recruiting one and artificially recruiting the other in its place had no measurable effects.

Second, the authors considered whether the exact binding position of the phosphatase matters to function by say, controlling substrate accessibility. But deleting the motif recruiting either phosphatase and recruiting the same phosphatase to a different site on Knl1 did not impact its function. Ndc80 dephosphorylation occurred normally when PP2A-B56 was repositioned and MELT dephosphorylation was not impacted by PP1 repositioning.

If the two enzymes can functionally substitute for each other and their positions within the Knl1 subcomplex is immaterial to signalling, how are PP1 and PP2A-B56 restricted to distinct functions at the kinetochore?

As mentioned previously, a key difference between the two enzymes is the mechanism by which they are recruited to their respective motifs. To explore this aspect, the authors focused on a critical unexplained observation that despite being present at the kinetochore, PP2A-B56 does not dephosphorylate MELT motifs in the absence of PP1. They take advantage of a motif LPTIHE which binds PP2A-B56 with an affinity similar to LSPIIE but independently of phosphorylation. Amazingly, in the absence of RVSF (and thereby PP1), if the motif recruiting PP2A-B56 is changed to LPTIHE, PP2A-B56 is now able to substitute for PP1 and induce cells to exit mitosis.

The authors next construct a mathematical model for this network. It assumes that kinase activities involved in the system are fixed and that both PP1 and PP2A-B56 dephosphorylate the following substrates with the same kinetics, namely MELT, Ndc80, RVSP and LSPIIE. The only difference is that PP2A-B56 is recruited by phosphorylation of the motif it binds to, and PP1 is not. Strikingly, such a model is able to recapitulate their experimental findings.

Finally, in a wider search for such motifs, the authors find that RVxF and LxxIxE are present in ~700 proteins and a large fraction of these contain phosphorylatable residues in key positions. Quite interestingly, structural examination of PP1 and B56 reveals how a negatively charged surface on the former and a basic groove on B56 would explain the opposite behaviour towards a negatively charged phosphate in the motifs they bind.

What I like about this preprint:

This paper tackles a complex but widely-applicable problem with a nice combination of clear experiments, mathematical modelling and a wider analysis of the implicated motifs in other contexts. The key finding of this paper that inverse phospho-dependencies in phosphatase binding can create unique functional differences between similar enzymes, is unexpected and novel.

Questions for the authors:

1. How critical are binding kinetics to the manifestation of these feedback loops? For example, if FRAP recovery of PP2A-B56 accumulation at the kinetochore was compared between LSPIIE- and LPTIHE-driven binding, would they be significantly different? And does that matter for the striking result seen in Figure 4E?
2. Among the validated proteins with RVxF/LxxIxE motifs that are phosphorylated, are the predicted binding partners more often kinases and phosphatases? That is, is such phospho-dependence more likely to impact a certain type of signalling network and do you predict that tyrosine kinases/phosphatases may be regulated this way, albeit with different motifs?

References: 

1. Almo et al., Struct. Funct. Genomics, 2007
2. Manning et al., Science, 2002
3. Ingebritsen and Cohen, Eur J Biochem, 1983
4. Kauko et al., bioRxiv, 2018
5. Kim et al., J Bio Chem, 2003
6. Nasa et al., Sci Signal, 2018
7. Wang et al., Protein Cell, 2016
8. Wang et al., Structure, 2016
9. Hertz et al., Mol Cell, 2016
10. Nijenhuis et al., Nat Cell Biol, 2014

Tags: kinetochore, phosphatases, signalling networks

doi: https://doi.org/10.1242/prelights.12195

Read preprint (3 votes)