Activation is only the beginning: mechanisms that tune kinase substrate specificity

Protein kinases are a large family of enzymes that transfer phosphate groups from ATP to other protein substrates. This phosphate transfer modulates the structure and/or activity of substrates on a rapid time scale, making kinases essential regulators of cellular signaling. To appropriately respond to changing conditions, each particular kinase must recognize its substrates within the complex environment of the cell and target only those proteins on the correct amino acid residues. Furthermore, many kinases have dozens to hundreds of substrates, but they do not act on all of them all at once; rather, they are able to target specific substrates at the right time and in the right place, in response to a distinct signal. This is what we are referring to as substrate specificity — the chemical features of the enzyme and/or the substrate that determine what is capable of being phosphorylated as a substrate, what is not capable of being a substrate, and what distinguishes substrates from each other such that they can be picked out from a kinase's substrate repertoire under the right conditions. It is important to note that different references use different terms for this property, most commonly selectivity rather than specificity [1], and that usage has changed somewhat over time. Regardless, the fascinating biochemical question remains: how are kinases able to achieve such specificity and differentiate between multiple potential substrates? Filling this knowledge gap may not only enhance our understanding of cell signaling but also facilitate development of targeted therapeutics to treat diseases such as cancer, in which specific kinase pathways may be misregulated. This review will examine emerging work on four serine/threonine kinases (Aurora B; calcium/calmodulin dependent kinase II, CaMKII; glycogen synthase kinase 3β, GSK3β; and casein kinase 1δ, CK1δ) from diverse kinase families to describe mechanisms by which kinases may change their substrate specificity, ranging from the structural level to the cellular level (Figure 1).

The structures of protein substrates are a fundamental aspect of kinase specificity. Kinases recognize distinct linear peptide motifs, which comprise the amino acids surrounding the phosphorylation site [2]. Different enzymes or families of kinases are capable of phosphorylating substrates with different motifs. Additional aspects are often layered on top of the amino acid sequence in the motif, such as the three-dimensional structure of the substrate, which must have chemical and steric characteristics that allow at least a transient interaction with the kinase active site [3]. Motifs are not static, however. A kinase's preferred motif may change due to conformational changes in the kinase that alter kinase-substrate interactions. For example, substrate motifs for GSK3β and CK1δ differ depending on whether conserved phosphorylation sites in the catalytic domains are occupied [4,5], and Aurora B favors TP motifs in mitosis but switches to SP motifs during cytokinesis [6].

Several kinases recognize motifs that include post-translational modifications, most commonly phosphorylation. Prior phosphorylation, or ‘priming,’ of a serine or threonine by one kinase generates the preferred motif for a different kinase. The substrate is only available when it is primed, allowing multiple signals to converge through the activities of the priming kinase(s) and phosphatase(s) that remove the priming site in addition to the kinase that recognizes the primed motif. Both GSK3β and CK1δ prefer motifs that include priming phosphorylation [7–11], though they act on unprimed substrates as well [12–19]. Primed substrates tend to have greater binding affinities and faster rates of phosphorylation due to the structural characteristics of the kinases’ substrate binding grooves [20,21]. As discussed in more detail below, regulatory mechanisms often affect primed and unprimed substrates differently, allowing for some separation of function. It was initially believed that primed substrates of GSK3β tended to be downstream of insulin and growth factor signaling, while unprimed substrates were in the Wnt pathway [13,22–24]. However, the discovery that CK1α primes β-catenin for GSK3β phosphorylation [25,26] goes against this trend, and for CK1δ, there is no clear correlation between whether a substrate is primed and which pathway it participates in, highlighting the need for additional mechanisms that regulate the substrate specificities of kinases with multiple functions.

The structures of kinases themselves are quite dynamic, with a variety of conformational changes and post-translational modifications that affect their substrate specificity. A classic example of this is phosphorylation of a conserved threonine in the activation loop, often called the T-loop, which sits near the substrate binding groove in the catalytic domain [27]. In many kinases, T-loop phosphorylation is required for the kinase to adopt an active conformation [27–29]. This occurs through assembly of the catalytic C spine and regulatory R spine, two chains of hydrophobic residues that span the catalytic domain and through their interaction align the kinase for efficient phosphoryl transfer from ATP to the protein or peptide substrate [28,30–33]. In addition, active kinase conformations require coordination of Mg²⁺ by the DFG motif, formation of a conserved salt bridge in the αC helix to position ATP, and docking of the activation loop against the C-lobe to form the substrate binding surface [27,28,32,33].

In Aurora B, T232 is the canonical T-loop site, and its autophosphorylation activates the kinase [34–36]. T232-phosphorylated Aurora B is activated even further upon binding to the chromosomal passenger protein INCENP, and INCENP binding can activate Aurora B even when T232 is not phosphorylated [37–43]. A conserved region at the C-terminus of INCENP called the IN-box inserts a hydrophobic motif into the N-lobe of Aurora B to promote the active conformation of the αC helix and activation loop [37]. INCENP is also a substrate of Aurora B, and phosphorylation of the TSS motif in the IN-box maximally activates the kinase toward other substrates [37,39,44–47]. A crystal structure of the closely-related Aurora C kinase bound to TSS-phosphorylated INCENP illustrates how these sites stabilize the activation loop and αC helix in concert with phosphorylated T198, which is the T-loop site homologous to Aurora B T232 (Figure 2A) [47]. Recently, molecular dynamics simulations support a model in which Aurora B is activated in a similar manner and also emphasize the importance of conformational dynamics in kinase activation, as T232 and TSS-phosphorylated Aurora B/IN-box was quickly stabilized by interactions with nearby basic residues, and the phosphorylated complex was overall less flexible compared with the unphosphorylated [39]. Binding to survivin, another component of the chromosomal passenger complex (CPC), has also been shown to activate Aurora B [48]. This multi-step activation process, in which allosteric conformational changes and dynamics play major roles [39], allows more complex regulation of Aurora B throughout the cell cycle.

In addition to T232, Aurora B is phosphorylated on a second site in the activation loop, S227, by PKCε [49,50]. Recent work has shown that phosphorylation of S227 in combination with T232 promotes phosphorylation of specific substrates at the centromere during the metaphase-anaphase transition [49] and at the midbody during abscission [50,51], while ATP binding and activity toward other substrates, such as histone H3, is unchanged [49,50]. Molecular dynamics simulations showed that when both S227 and T232 are phosphorylated, the activation loop is in a different conformation than when T232 alone is occupied [49]. The loop is less flexible, and the conformation of INCENP also differs [49]. At the metaphase-anaphase transition, S227 phosphorylation directs Aurora B activity toward S29 on TopoIIα, rather than T1460 [49,52,53]. The conformational changes induced by S227 phosphorylation allow TopoIIα S29 to fit in the substrate binding groove and come into proximity to the catalytic aspartate, while T1460 binds in an unproductive orientation [49]. This promotes resolution of sister chromatid catenation and delays anaphase entry until non-disjunction is corrected [49].

Later on during cell division, phosphorylation at S227 promotes resolution of the abscission checkpoint [50], which delays cytokinesis when genetic material remains in the cleavage furrow [54–57]. In this context, S227 phosphorylation requires PKCε binding to 14-3-3 and localization to the midbody ring [51]. Occupancy of S227 increased Aurora B-mediated phosphorylation of Borealin S165 and increased localization of CHMP4C to the midbody ring, allowing abscission to proceed [50,54–57]. In addition to Borealin S165, S227-phosphorylated Aurora B showed a 10-fold increase in phosphorylation of other substrates, such as H3F3A and desmin [50], suggesting that the conformational changes induced by S227 phosphorylation may have additional effects in other cellular processes. Together, INCENP binding and phosphorylation combined with phosphorylation of the activation loop on T232 and S227 change the substrate binding interface on Aurora B, influencing which substrates are preferentially phosphorylated and at what rate.

Other protein kinases do not require T-loop phosphorylation to adopt an active conformation [27], but changes in the activation loop often still influence substrate phosphorylation. GSK3β, for example, is phosphorylated on a tyrosine, Y216, rather than a threonine in the activation loop [58–60], and it maintains the structural hallmarks of an active kinase without any phosphorylation [20,61,62]. Similar to several other kinases [63], Y216 is autophosphorylated shortly after translation during protein folding, and this facilitates substrate phosphorylation but is not strictly required for activity [20,22,58,60,64–66]. The fact that Y216 is autophosphorylated is itself an example of substrate switching — GSK3β must act as a tyrosine kinase to target Y216 intramolecularly, even though it is a serine/threonine kinase when targeting exogenous substrates [66]. GSK3β is capable of tyrosine autophosphorylation only when in a transitional intermediate in complex with the chaperone Hsp90; fully mature GSK3β typically phosphorylates serines and threonines [66]. However, recent work has shown that a bacterial protein, SteE, can interact with host cell GSK3β and promote tyrosine phosphorylation of STAT3 [67], so this switch to tyrosine phosphorylation may be more widespread than previously appreciated.

Once Y216 is phosphorylated, it is able to interact with a positively-charged binding pocket, which stabilizes it in an orientation away from the substrate binding groove (Figure 2B) [5,20,22,62]. Unphosphorylated Y216 is also able to sample this conformation but lacks this stabilizing interaction, so the rate of substrate phosphorylation increases ∼5-fold upon phosphorylation [5,20,22,62]. Primed substrate binding promotes the same orientation, even when Y216 is not phosphorylated, because the primed phosphorylation site binds in the same pocket, demonstrating why Y216 phosphorylation is not an absolute requirement for catalytic activity [5,20,22,62]. Y216 phosphorylation promotes the greatest increase in GSK3β activity towards unprimed substrates that have a basic residue in the +1 position, likely because these substrates would clash with the positively-charged pocket [5,62]. When Y216 is not phosphorylated, a small, hydrophobic residue is favored at the +1 position, suggesting that GSK3β would preferentially target a different subset of substrates than when Y216 is phosphorylated [5].

Other kinases, such as CaMKII and CK1δ, are not phosphorylated on the T-loop at all. CK1δ constitutively displays the structural features of an active kinase in the absence of T-loop phosphorylation, leading to the initial hypothesis that CK1δ was constitutively active [27]. However, recent work is uncovering a variety of ways in which CK1δ activity is regulated. The activation loop of CK1δ plays a dynamic role in shaping the substrate binding cleft, and its position has been shown to switch the specificity of the kinase between two different target sites on the substrate PER2 [68], similar to the change in specificity of Aurora B on TopoIIα [49]. When the loop is up (Figure 2C, dark blue), the substrate binding cleft is more constrained, and CK1δ preferentially targets PER2 S478, which leads to PER2 degradation and a short circadian period [68]. When the loop is down (Figure 2C, teal), the substrate binding cleft is more open, and CK1δ preferentially targets PER2 S659 and four downstream sites, which leads to PER2 stabilization and longer circadian period [68]. It is important to note that the activation loop is plastic, and the kinase samples both of these conformations, and likely others in between; in fact, both the loop-up and loop-down conformations can be seen in the same crystals [21,68]. The loop-up conformation seems to be rare in wildtype CK1δ under standard conditions, but a short period mutant of CK1δ called tau (R178C) intrinsically favors the loop-up conformation, thus promoting PER2 degradation [68–70]. This activation loop switch regulates circadian rhythms via PER2 phosphorylation state and protein level [15,68,71], and it will be interesting to learn whether other CK1δ substrates are regulated in the same manner.

CK1δ also autophosphorylates a conserved site, T220, which is not located in the activation loop but can change its conformation and the global substrate profile of the kinase [4]. T220 lies on helix αG, which is proximal to the substrate binding groove and, in other kinases, is known to regulate kinase function through a variety of different binding interactions [28,62,72]. When T220 is occupied, helix αG is prone to unwinding, becoming more flexible and affecting the conformation of the activation loop, which changes the shape of the substrate binding groove [4] (Figure 2D). Some substrates, such as casein, LRP6, and YAP1 are phosphorylated more slowly when T220 is phosphorylated, while others such as p63 are phosphorylated more rapidly [4]. In the fission yeast homologues of CK1δ, phosphorylation at T220 regulates processes including the response to heat stress, the abundance of meiotic mRNA, and a mitotic checkpoint [4].

These examples illustrate that activation loop dynamics are complex and varied among different kinases. While the adoption of an active kinase conformation upon activation loop phosphorylation is well-documented for many, it is far from universal [27]. There could be multiple sites of phosphorylation on the activation loop, as in Aurora B [49,50], a phosphorylated tyrosine rather than threonine, as in GSK3β [58–60], or no phosphorylation, as in CaMKII and CK1δ. Other types of post-translational modifications can also affect T-loop phosphorylation, allowing potential cross-talk between different signaling systems [73–75]. The location of the activation loop immediately proximal to the substrate binding surface suggests that even subtle changes in its average conformation or in its distribution of conformational states could impact substrate binding. As we learn more about the specificity of different kinases on a variety of substrates, it seems plausible that changes in activation loop conformation could affect substrates differently, as seen with Aurora B [49] and CK1δ [68], allowing for fine-tuned changes in substrate specificity rather than, or in addition to, bulk catalytic activity.

In addition to T220, CK1δ and closely related CK1s autophosphorylate several sites on their C-terminal tails [76–82], and these sites regulate CK1δ substrate specificity via a different mechanism than the T220 site. The unphosphorylated C-terminus does not interact with the catalytic domain; however, the autophosphorylated C-terminal tail acts as a pseudosubstrate, binding in the substrate binding cleft and competitively inhibiting substrate phosphorylation [76–82]. Though the affinity between the catalytic domain and the autophosphorylated tail is weak, the interaction occurs in cis and is therefore promoted by the high local concentration of the tail [76]. One hypothesis is that the relative affinities of different substrates for the catalytic domain compared with the tail may influence how readily each substrate can be targeted [76]. Substrates with a high affinity may displace the lower-affinity tail even when it is phosphorylated [76], while phosphatases may dephosphorylate the tail, relieving autoinhibition to increase kinase activity or allow lower-affinity substrates to be targeted [76,83–86]. While there are crystal structures of two primed substrate peptides bound to the CK1δ catalytic domain (Figure 3A,B), and the interaction of the autophosphorylated tail with the substrate binding groove is supported by biochemical, NMR, and hydrogen-deuterium exchange mass spectrometry evidence [76,78,82], there is not currently an atomic-resolution model of this pseudosubstrate interaction. In the future, it would be informative to compare the conformations of tail-bound and substrate-bound CK1δ and to understand how this pseudosubstrate interaction regulates substrates involved in different cellular processes.

GSK3β also utilizes a phosphorylated pseudosubstrate domain [5,13,20]. S9 in the GSK3β N-terminus can be phosphorylated by a variety of kinases in response to cellular conditions, notably downstream of insulin and growth factor signaling [87]. Like in CK1δ, the phosphorylated pseudosubstrate domain binds the substrate binding groove with a relatively low affinity that is enhanced by its close proximity, and it competes with primed substrate binding to a positively-charged docking site [5,13,20]. Recently, a structure of GSK3β bound to a peptide from its substrate β-catenin was solved [88] (Figure 3C), and the phosphorylated N-terminus interacts with the GSK3β catalytic domain in a similar manner [5] (Figure 3D). When S9 is phosphorylated, this pseudosubstrate interaction can inhibit substrate phosphorylation to different extents depending on the substrate, thereby modulating GSK3β substrate specificity [13,89]. Phosphorylation of primed peptides from glycogen synthase and eIF2B was strongly suppressed by a peptide from the pseudosubstrate domain containing pS9, but the unprimed versions of these peptides were not affected [13]. Conversely, phosphorylation of a primed peptide from Tau was only slightly inhibited by S9 phosphorylation of GSK3β, while phosphorylation of the unprimed Tau peptide was more strongly suppressed [89].

Binding to other proteins, namely Axin in the β-catenin destruction complex, can also protect GSK3β from S9 phosphorylation, allowing for separation of GSK3β function downstream of Wnt from GSK3β function downstream of insulin and growth factors [90–93]. Recently, in vitro reconstitution and kinetic analysis confirmed that phosphorylated S9 interferes with substrate binding to GSK3β to decrease reaction rates [93]. Axin binding to GSK3β promoted an allosteric conformational change that made S9 less accessible to upstream kinases, preventing S9 phosphorylation in the destruction complex and funneling GSK3β activity toward β-catenin [90–93].

For both CK1δ and GSK3β, pseudosubstrate binding does not lead to complete inhibition of catalytic activity but rather establishes a new steady-state between the kinase and its pool of substrates [76,87]. Therefore, this mechanism has the potential to direct kinase activity toward a subset of substrates depending on cellular conditions. For GSK3β, this is exemplified by the separation between Wnt-mediated and insulin- and growth factor-mediated signaling [90,91]. For CK1δ, the biological implications remain to be investigated.

CaMKII has a regulatory domain that is similar to a pseudosubstrate; however, it is autoinhibitory until relieved by calcium/calmodulin (CaM) binding. CaMKII activity toward many, but not all, substrates is significantly stimulated by CaM binding when intracellular Ca²⁺ concentrations increase, such as during induction of long-term potentiation (LTP) in neurons [94]. CaM binds the regulatory domain of CaMKII, causing a conformational shift that opens access to the substrate catalytic domain, and then allows for CaMKII to become active [95–97] (Figure 3E). The conformation of the regulatory domain depends on CaM binding, where its N-terminal region is helical and C-terminal region is disordered when CaM is unbound, and vice versa for when CaM is bound [72]. Autophosphorylation of a site in the regulatory domain (T286 in the α isoform) allows CaMKII to maintain consistent, ‘autonomous’ activity even after CaM releases it, at ∼20% of its CaM-activated level [95,98–105]. Two other autophosphorylation sites on the regulatory domain, T305 and T306, can be targeted after CaM dissociates to prevent CaM rebinding to the regulatory domain [106,107]. All three of these sites work together to maintain CaMKII in an autonomous state until they are dephosphorylated.

Specific ‘high-autonomy’ substrates are preferentially phosphorylated by the autonomous form of the kinase, at rates much higher than the ∼20% observed for other substrates. These substrates often play a role in synaptic plasticity within neurons, specifically LTP and long-term depression (LTD). For example, CaMKII-mediated phosphorylation of S567 on GluA1, which is a subunit of AMPA-type glutamate receptors, inhibits synaptic localization of these receptors during LTD [108]. GluA1 S567 was phosphorylated at the same rate whether CaM was associated or not, and autonomous phosphorylation of GluA1 was 5-fold greater than autonomous phosphorylation of other substrates, including another site on GluA1, S831, which behaved as a canonical substrate [108].

CaMKII phosphorylates another substrate known as AKAP79/150 (A-kinase anchoring protein; AKAP79 in humans, AKAP150 in rats) more efficiently in its autonomous state [109]. In the presence of CaM, AKAP79/150 was slightly phosphorylated, but without CaM, this substrate was robustly phosphorylated to a level ∼350% greater than when CaM was present [109]. AKAP79/150 localizes protein kinase A (PKA) to synapses during LTP. During LTD, autonomous CaMKII phosphorylation removes AKAP79/150 from synapses, which delocalizes PKA [110]. CaM can also bind directly to AKAP79/150 and block CaMKII's phosphorylation sites when intracellular Ca²⁺ is high during LTP. At baseline Ca²⁺ levels, CaM releases AKAP79/150, allowing CaMKII to target AKAP79/150 and switch to promoting LTD [109]. High-autonomy substrates such as GluA1 and AKAP79/150 showcase how CaMKII's specificity is regulated by its autoinhibitory mechanism, and how this mechanism can be utilized to switch between different types of substrates in response to Ca²⁺ during LTP and LTD.

The pseudosubstrate domains in CK1δ, GSK3β, and CaMKII illustrate how intramolecular binding interactions can modulate kinase substrate specificity. Other protein-protein interactions can also have this effect through the formation of protein complexes and higher order structures [111]. For example, CaMKII forms a dodecameric holoenzyme, and this is necessary for autophosphorylation of T286, T305, and T306 in trans [112]. Though the activation loop is not phosphorylated, it forms a docking site for the hub domain of the adjacent monomer, allowing assembly of a compact multimer (Figure 4A) [112]. The conformation of the dodecamer influences the accessibility of the regulatory domain to CaM binding, as a more extended multimer facilitates activation by CaM (Figure 4B) [113]. When intracellular Ca²⁺ is high, CaM can bind to the regulatory domains on two neighboring catalytic subunits, and one of them can phosphorylate the other at T286, resulting in an individual autonomous subunit. In a state of constant intracellular Ca²⁺, this multimer will have a few of its catalytic subunits T286-autophosphorylated at a time. During frequent periods of high intracellular Ca²⁺, such as during LTP induction, T286 is autophosphorylated on many of its catalytic subunits, resulting in greater autonomous CaMKII activity [95].

CaMKII also interacts with GluN2B, an NMDA-type glutamate receptor subunit that plays a role in LTP and LTD in hippocampal neurons [114]. GluN2B binding localizes CaMKII to excitatory synapses during LTP induction, where it targets specific substrates such as GluA1. Initial GluN2B binding requires T286 autophosphorylation or CaM-binding, but subsequent activity can be maintained if CaM is released or T286 is dephosphorylated, promoting longer-term autonomous activity [95,115,116]. Therefore, interaction with GluN2B simultaneously regulates the autonomous activity and the localization of CaMKII.

GSK3β also participates in multiple types of protein-protein interactions that regulate its substrate specificity. One of the most well studied is the β-catenin destruction complex, which is scaffolded by Axin and promotes proteasomal degradation of the transcription factor β-catenin in the absence of Wnt ligand [117] (Figure 4C). When GSK3β is associated with the destruction complex, it is protected from S9 phosphorylation and specific for substrates in the Wnt pathway [90–93], and binding to Axin increases the rate at which GSK3β phosphorylates β-catenin [22–24,118]. Recent work that quantified GSK3β phosphorylation kinetics on different substrates in the presence or absence of Axin demonstrated that interaction with Axin actually lowers the affinity of GSK3β for its substrates, but then rescues this specifically for β-catenin, thereby directing Axin-bound GSK3β away from other substrates and toward substrates in the Wnt pathway [93,118]. When concentrations of competing substrates are in excess over β-catenin, they act as competitive inhibitors, and association with Axin lowers GSK3β’s affinity for these inhibitors, thereby increasing the rate of β-catenin phosphorylation [118].

GSK3β’s binding to Axin is itself in competition with other proteins. The binding site for FRAT/GBP (Frequently rearranged in advanced T-cell lymphomas in mammals [119], GSK3-binding protein in Xenopus [120]) overlaps with that of Axin [121,122] (Figure 4D). FRAT-associated GSK3β decreased phosphorylation of Axin and β-catenin, the opposite effect of Axin-associated GSK3β [123]. By inhibiting GSK3β activity toward its substrates in the Wnt pathway, FRAT promotes β-catenin-mediated transcription during development [123–126]. Phosphorylation of primed substrates outside of the Wnt pathway was not affected by the interaction of GSK3β and FRAT [123–125]. GSK3β also exists in other complexes, and this raises the model that different scaffolding proteins may craft distinct subpopulations of GSK3β that are targeted toward specific substrates [118].

In addition to allosterically modulating the activity of a kinase [111], scaffolding proteins can often regulate the subcellular localization of their binding partners, as described for CaMKII and GluN2B above. The dynamic localization of Aurora B during cell division is another prime example. Aurora B initially localizes to centromeres in prometaphase and metaphase, the central spindle in anaphase, and the midbody in telophase and cytokinesis [127–129]. At each location and each time during cell division, Aurora B phosphorylates specific substrates; this spatiotemporal separation allows Aurora B to participate in different processes as cell division progresses, from chromosome condensation to the spindle assembly checkpoint to abscission [129]. Association with INCENP and other members of the CPC is essential for Aurora B localization, and, reciprocally, Aurora B kinase activity is required for CPC localization [42,45,128]. The proteins that interact with Aurora B and the CPC have been shown to change between mitosis and cytokinesis [6], and specific binding interactions are required for some functions of Aurora B. For example, CPC binding to chromatin and microtubules simultaneously may facilitate substrate recognition during spindle assembly [130], and CPC binding to the kinesin Mklp2 is required for relocalization to the central spindle in anaphase [131]. The complex pattern of Aurora B localization establishes gradients of kinase activity within dividing cells to help regulate Aurora B substrates involved in different aspects of mitosis [132–134].

Through these layers of regulation — from Angstrom-scale changes in the substrate binding interface, to conformational changes in protein domains, to the formation of different protein complexes, to changing spatial and temporal localization within cells (Figure 1) — we are only beginning to understand how kinases are able to orchestrate cellular signaling events. These mechanisms of course are not exhaustive or mutually exclusive, and we have not been able to discuss all of the work on kinase specificity here, but we suggest that these may be a starting point to understand common themes that regulate many different types of protein kinases. Simple changes in catalytic activity often cannot fully explain physiological kinase function, so when measuring kinase activation, it's worth asking, ‘toward what?’ An increase in the rate of phosphorylation of one substrate may not necessarily reflect increased rates of phosphorylation toward all substrates, especially for kinases with many substrates that are involved in different signaling pathways. Therefore, it is imperative that we continue to identify and validate the depth of kinases’ substrates and rigorously examine how activity toward different substrates may be regulated by different mechanisms. Combining knowledge about structural conformations and dynamics, signaling pathways, and cellular outcomes will advance our understanding of kinase substrate specificity.

The authors declare that there are no competing interests associated with the manuscript.

Conceptualization: L.K.C., S.N.C. Investigation: L.K.C., S.N.C. Visualization: S.N.C. Writing — original draft: L.K.C., S.N.C. Writing — review and editing: L.K.C., S.N.C. Supervision: S.N.C. Funding acquisition: S.N.C.

Figure 1 contains images that were modified (recolored, resized, and combined) from Servier Medical Art under a CC-BY 3.0 Unported license. We thank Kathy Gould and members of the Cullati laboratory for helpful discussions.

CPC

chromosomal passenger complex

LTD

long-term depression

LTP

long-term potentiation

PKA

protein kinase A