Substrates of Mitotic Kinases

See allHide authors and affiliations

Science Signaling  28 Jun 2011:
Vol. 4, Issue 179, pp. pe31
DOI: 10.1126/scisignal.2002234


Most signaling pathways in cells involve numerous phosphorylation reactions. Some of the rules for kinase-substrate specificity are known, but a complete description of all substrates is missing. Research published in Science Signaling addresses the process of mitosis and asks how the relevant kinases recognize substrate sequence motifs and, in the cellular context, what substrates are phosphorylated and where. The results increase our molecular understanding of how individual events are coordinated during the process of cell division and show the importance of both sequence epitopes for kinase specificity and the notion of a sense of place through localization in subcellular compartments.

The human genome encodes 538 kinases (, the third most-populated family of proteins. Structural studies [to date with ~136 protein kinases (1)] have shown that the kinases in their active conformation have a similar core structure (2). They exhibit common adenosine triphosphate (ATP)–binding sites but show substrate specificity for local sequences of the protein substrate, which is for the most part engineered by a region known as the activation segment that lies between the conserved sequence motifs DFG and APE (3). This region, in many but not all kinases, contains a residue whose phosphorylation is essential for activity. Can these features generate enough specificity to ensure that the correct substrates get phosphorylated at the right time and the right place? Two recent papers address this problem (4, 5).

Identification of genuine substrates in vivo is not trivial. There are problems in achieving complete stoichiometric phosphorylation and in working with proteins in low abundance. Advances in phosphopeptide isolation, mass spectrometry, and proteomics have resulted in a conservative description of 42,000 phosphorylation sites in 11,000 different protein sequences mediated by 310 kinases [ (6)]. Although assays performed with libraries of peptide sequences have aided in the identification of optimal sequence motifs that may be phosphorylated (7), motifs with sequences that are less than optimal are often phosphorylated in cells. For example, the cyclin-dependent kinase 1 and cyclin B complex (Cdk1/cyclin B), which is the key kinase that drives cell division, has the preferred consensus substrate motif pS/T-P-X-K/R (3) where p denotes the phosphorylation site, X is any amino acid, and “/” indicates “or.” Blethrow et al. (8) used a modified kinase engineered to accept an ATP analog that no other kinase in the cell could use to uniquely label its substrates with a phosphate analog tag and identified 72 Ser-Pro phosphorylation sites in 68 proteins that were phosphorylated by Cdk1/cyclin B in HeLa cells. Whereas 68% of sites had the full consensus sequence, 32% had only the minimum requirement of a Pro in the +1 position.

Mitosis is controlled by the kinases: Cdk1/cyclin B, the Aurora family members Aurora A and Aurora B (73% identical in sequence in kinase domain), Polo-like kinase 1 (Plk1), and Nek2. Cdk1/cyclin B is the master regulator with roles in triggering nuclear envelope breakdown, microtubule dynamics, spindle formation, and the separation of condensed sister chromatids. Its activation marks progress from G2 to M phase, and its inactivation by the anaphase-promoting complex marks the transition to anaphase. Aurora A and Nek2 are required for the complex processes of centrosome maturation and centrosome separation. Aurora B is part of the chromosome passenger complex, is involved in sister chromatid condensation and segregation, and participates in the mitotic checkpoint that senses tension to ensure proper chromosome attachment to the spindle. Plk1 is localized successively to centrosomes in prophase, kinetochores in metaphase, and spindle mid-zone in telophase. It plays multiple roles, including phosphorylation of cyclin B, centrosome maturation, kinetochore function, spindle-formation, and cytokinesis. There is cross talk between the kinases: Aurora A activates Plk1 by phosphorylation of a Thr in the kinase activation segment and, in turn, Plk1 phosphorylates cyclin B to promote nuclear retention.

The papers by Alexander et al. (4) and Kettenbach et al. (5) show how kinase specificity and localization match biological events. Kettenbach et al. (5) assign the cellular activities of the mitotic kinases Aurora A, Aurora B, and Plk1 to respective substrate motifs by combining quantitative phosphoproteomics and efficient small-molecule kinase inhibition in synchronized HeLa cells. Cells arrested in metaphase with the microtubule-stabilizing drug Taxol (Bristol-Myers Squibb) were scanned for kinase substrates by determining differences in the phosphorylation status of the proteins when treated with the respective inhibitors for the individual kinases. With an additional criterion that the phosphorylation site must adhere to one of the motifs generated for that kinase by bioinformatics analysis, totals of 127, 165, and 486 sites were identified for Aurora A, Aurora B, and Plk kinases, respectively. Among the 563 proteins phosphorylated, proteins were identified in cell cycle regulation, RNA processing and splicing, centrosome assembly, nuclear transport, and DNA damage repair. The substrates listed by kinase and by location show Aurora A and Plk substrates at centrosomes; Plk1 and Aurora B substrates at the centromere and kinetochores; and Aurora A, Aurora B, and Plk1 substrates at the spindle. Among examples of new information, they show that the protein NuMA, which is a 2115–amino acid protein that binds the dynein and dynactin motor complex, is transported along microtubules to the spindle poles, and is phosphorylated on 97 sites not only by Cdk1, as observed previously, but also by Aurora A and Plk1.

Alexander et al. (4) used positional scanning oriented peptide library screening to explore potential mechanisms for substrate discrimination with the five mitotic kinases Cdk1/cyclin B, Aurora A, Aurora B, Plk1, and Nek2. The method identifies the kinase motif for optimal Vmax/Km ratio (where Vmax is the maximum velocity and Km is the Michaelis constant) for substrate turnover. The library scanning procedure led to results that emphasized the notion of both positively selected motifs and “antimotifs,” sites that had negatively selected motifs (Fig. 1). For Plk1, Aurora A, and Aurora B, optimal substrate sequences had a strong antimotif that discriminates against Pro in the +1 position. The identification of Asn in the –2 position for Plk1 substrates led to identification of two proteins involved in checkpoint control (Bub1 and p31Comet) and another in sister chromatid cohesion (Scc1) as Plk1 substrates.

Fig. 1

Selectivity in mitotic kinases. The kinase domain is in yellow, with the activation segment in magenta and activating phosphorylated Thr indicated. Shown are human phosphorylated Cdk1/cyclin B {modeled on [Protein Data Bank (PDB) code 2JGZ] Cdk2/cyclin B} with a substrate peptide from Cdk2/cyclin A (PDB 1QMZ); zebrafish phosphorylated Plk1 (PDB 3D5W); phosphorylated Aurora A in complex with a fragment of TPX2 (cyan) (PDB 1OL5); phosphorylated Aurora B in complex with a fragment of INCENP (cyan) (PDB 2BFX); and Nek2 in the nonphosphorylated state with the activation segment in an inactive conformation (PDB 2W5B). A summary (3) of the kinases’ specificity is taken from (4, 5); see original papers for more details. Φ indicates a nonpolar residue.

The critical S/T-P specificity of Cdk1 and the antimotif for Pro at +1 from all the other mitotic kinases provides distinct discrimination between Cdk1 phosphorylation sites and those for Aurora A, Aurora B, Plk1, and Nek2. The preference for Pro at +1 in Cdk1/cyclin B comes from an unusual conformation of the activation segment in this region that creates a hydrophobic site with no hydrogen bonding potential (9), which is a feature shared by kinases of the mitogen-activated protein kinase (MAPK) family (10). Structural information regarding substrate recognition for the other kinases is not available, but through modeling studies Alexander et al. show that the optimal peptide recognition may be recognized by suitable favorable contacts. Experimental verification from kinase structures complexed with substrate peptides is needed, but such crystallization studies are often difficult.

Knowledge of the optimal phosphorylation motifs and localization suggests that the kinase specificity is a combination of both positively and negatively selected sequence motifs and of nonoverlapping localizations. Alexander et al. (4) propose that kinases exist in two “spaces”: “motif space” and “localization space.” After entry into mitosis, Cdk1 is distributed throughout the cell and overlaps with all of the other kinases in localization space but not in motif space. Plk1 and the Aurora kinases do not phosphorylate the same sites despite overlapping localizations, because their motifs are mutually exclusive. Despite partially overlapping motifs, Nek2, Plk1, and the Aurora kinases do not phosphorylate the same sites because of nonoverlapping localizations: Nek2 localizes to the proximal centrioles within the core of the centrosome, whereas Aurora A and Plk1 are localized at the periphery of the pericentriolar material. Thus, the combination of positive and negative amino acid motif selection and spatial exclusivity appears to underlie the cooperative nature of mitotic kinase signaling (Fig. 1).

There are differences and little overlap in the substrates identified in the two papers. These differences may partly arise because those substrates that are phosphorylated before mitosis would not have been identified in the cell experiments reported by Kettenbach et al. The analysis by Alexander et al. identified substrates that contained the optimal motifs for the relevant kinases, several of which were well-recognized substrates that function in early mitosis that would not be actively phosphorylated in cells arrested in mid- or late mitosis. However, this may not be a complete explanation. Not every optimal motif may be phosphorylated in protein substrates. Kettenbach et al. noted that ~90% of the phosphorylation sites that they identified were found in loops and unordered regions, as has been noted for other phosphorylation sites (11, 12). In an example from glycogen metabolism, that of glycogen phosphorylase, where the substrate structures in the nonphosphorylated and phosphorylated states and the relevant kinase, phosphorylase kinase, are known, there is only one Ser residue, Ser14, that is phosphorylated. The Ser14 site is in a less-well-ordered N-terminal region, and this region can adapt without too much conformational energy to match the extended conformation required at the phosphorylase kinase catalytic site (13). The need to fit the protein substrate to the catalytic site for the epitope region recognized is a requirement for all kinases, and this can best be met by flexibility in the protein substrate.

Kinase substrate activity is engineered by many factors in addition to substrate recognition of a local epitope. Most kinases require activation by phosphorylation on the activation segment. This can be autophosphorylation, but it is usually accomplished by an upstream kinase. Often, this phosphorylation is the key event that creates the substrate recognition site for the protein substrate. Second, kinases frequently use additional subunits or domains as docking sites for substrate recognition that may include docking sites for parts of the substrate that are remote from the phosphorylation site. Third, kinases may need additional subunits. Centrosomal Aurora A needs a spindle assembly protein called TPX2 for activity, and the activation of Plk1 by Aurora A is enhanced by the binding of another protein, Bora. Aurora B associates with the chromosomal passenger proteins INCENP (inner centromere protein), borealin, and survivin that target it to centromeres. Plk1 is distinguished by a C-terminal polo box domain (PBD) that recognizes phosphorylation sites with the motif S-pT/S-P (3), which are likely to be phosphorylated by Cdk1/cyclin B. Binding of the PBD at the phosphorylation site assists localization of the kinase on the substrate, such as, for example, in regulation of the phosphatase Cdc25, which is a target for Plk1. Fourth, posttranslational modification may localize the kinase to a particular subcellular fraction. For example, Cdk1/cyclin B is cytoplasmic until Plk1 phosphorylation of cyclin B enables it be retained in the nucleus. Lastly, cells achieve specificity in their signaling networks by organizing discrete subsets of proteins on scaffolds, which involve proteins that can sequester and organize several sequential signaling proteins (14). Scaffold proteins tether and orient substrates, and they can mediate pathway branching and may also promote allosteric regulation. Colocalization can be a potent force for regulation of networks (15).

The tools for substrate identification described in the two papers have provided a reference set of substrates that are targets for particular mitotic kinases. For some of the substrates, the downstream consequences are well understood, but for others the question of what happens next is still unknown. Does the phosphorylation of the target lead to activation or inhibition or provide a docking site to recruit other proteins, and, if so, what is the molecular basis of this action? The advances and the logic and execution of the present experiments are wonderfully informative, but there is still much to know.

References and Notes

View Abstract

Navigate This Article