Biochemical Society Transactions


Protein kinases, from B to C

A.J. Cameron, M. De Rycker, V. Calleja, D. Alcor, S. Kjaer, B. Kostelecky, A. Saurin, A. Faisal, M. Laguerre, B.A. Hemmings, N. McDonald, B. Larijani, P.J. Parker


The PKB (protein kinase B) and PKC (protein kinase C) families display highly related catalytic domains that require a largely conserved series of phosphorylations for the expression of their optimum activities. However, in cells, the dynamics of these modifications are quite distinct. Based on experimental evidence, it is argued that the underlying mechanisms determining these divergent behaviours relate to the very different manner in which their variant regulatory domains interact with their respective catalytic domains. It is concluded that the distinct behaviours of PKB and PKC proteins are defined by the typical ground states of these proteins.

  • protein kinase B (PKB)
  • protein kinase C (PKC)
  • phosphoinositide-dependent kinase 1 (PDK1)
  • phosphorylation
  • pleckstrin homology domain (PH domain)
  • protein serine/threonine kinase


In mammals, cellular regulatory networks have evolved to embrace over 500 protein kinases [1]. It is evident from sequence comparisons that this proliferation of kinases has derived from multiple gene duplication events, leading to particular families and subfamilies defining branches within the now familiar kinase family tree. Within this class of regulators is the AGC branch of protein serine/threonine kinases [AGC-type kinases (protein kinase A/protein kinase G/protein kinase C-family kinases)]. This branch of the family comprises a closely related set of catalytic domains that retain a number of conserved features, with distinctions between AGC-type kinases relating to their regulation, typically conferred by their idiosyncratic regulatory domains/subunits (for reviews, see [24]). However, it is emerging that certain properties of these kinases arise from the distinctive interactions between their regulatory and catalytic domains. Here, we compare the PKB (protein kinase B; also called Akt) and cPKC (conventional PKC)/nPKC (novel PKC) subfamilies to provide insight into these distinctive inter-domain properties.

Mechanism of PKB activation

Since the original identification of PKB as an insulin- stimulated GSK3 (glycogen synthase kinase 3) kinase [5], the mechanism of activation of this group of kinases (PKBα, β and γ; Akt 1, 2 and 3) has been studied intensively and indeed reviewed comprehensively elsewhere (see [6,7]). That PKB was identified as an insulin-activated kinase reflects the fact that activation is consequent to phosphorylation of PKB on two key sites (PKBα: Thr308 activation loop site; Ser473 hydrophobic motif site) conferring extractable, stable enhancement of the protein kinase activity [8]. These phosphorylations occur in conjunction with what appears to be constitutive phosphorylation of the turn motif immediately N-terminal to the hydrophobic phosphorylation site. Based on detailed structural analysis there is now a comprehensive molecular understanding of how these phosphorylations influence the catalytic domain and together serve to stabilize an active conformer for this domain, with appropriately aligned ATP and protein/peptide substrate-binding pockets [9].

Matching the elegant structural work on PKB, biochemical studies have led to the elucidation of the upstream kinases that are responsible for phosphorylation of the activation loop site [PDK1 (phosphoinositide-dependent kinase 1)] [10,11] and the hydrophobic motif site {DNA-PK (DNA-dependent protein kinase) and TORC2 [mTOR (mammalian target of rapamycin)/rictor (rapamycin-insensitive companion of mTOR) complex 2]} [12,13]. The biochemistry, pharmacology and genetics that evidence these connections have been described in detail elsewhere [1214]. Suffice it to say that the key upstream trigger for PKB activation is an increase in PtdIns(3,4,5)P3, which is responsible for the recruitment of PKB to the plasma membrane alongside PDK1. It is in this compartment that PKB becomes phosphorylated. Recruitment is conferred by the regulatory PH domain (pleckstrin homology domain) [15], an established PtdIns(3,4,5)P3-binding domain [16]; the co-recruitment with PDK1 is also conferred by its regulatory PH domain [17]. This pattern of behaviour has generally coloured the prevailing view that the membrane co-recruitment/co-localization of PKB and its upstream kinase(s) is what drives the equilibrium of phosphorylation/dephosphorylation towards a higher stoichiometry of phosphorylation, i.e. activation.

It transpires that the above co-localization model of PKB activation falls some way short of the mark. Recent in vivo studies of PKB activation have revealed a more precisely controlled activation mechanism determined by the behaviour of the PH domain–kinase domain interface [18]. In a sense, the question is why PKB is not phosphorylated in the basal state, since part of this in vivo evidence demonstrates that PKB–PDK1 are significantly pre-complexed in the cytosol, i.e. co-localization is not a sufficient criterion for activation. This does not appear to be just a simple issue of phosphatase action, since under unstimulated conditions inhibition of PP1 (protein phosphatase 1)/PP2A with okadaic acid is sufficient to drive rapid Ser473 phosphorylation but not Thr308 phosphorylation, which is only slowly phosphorylated. Notably, this is not the case for the PH domain-deleted PKB (ΔPH-PKB), which is readily phosphorylated at both sites. The implication is that the PH domain sterically blocks Thr308 phosphorylation and that at the membrane this steric hindrance is removed [18]. Direct evidence for a PKB conformational change on membrane binding has come from the use of a genetically encoded conformational reporter for PKB [GFP (green fluorescent protein)–PKB–RFP (red fluorescent protein)]. This has also been employed to show that the plasma membrane-associated conformer once phosphorylated can dissociate from the plasma membrane and move through the cytosol in the ‘active’, phosphorylated conformation (see Figure 1).

Figure 1 PKB activation scheme

The PKB regulatory domain (triangle) and catalytic domain (bi-lobal kinase domain) interact in the ground state with the latter dephosphorylated at the two key activation sites (see text) and hence in an inactive conformation. On recruitment to membranes through the regulatory domain, there is a conformational change, which enables upstream kinases to phosphorylate the kinase domain. It is consequently stabilized in an active conformation. The phosphorylated kinase can dissociate from the membrane and remain active since the regulatory domain no longer binds this phosphorylated conformer. On dephosphorylation of the two key sites, the kinase domain takes on an inactive conformation again and can rebind the regulatory domain, returning to the ground state.

A key element of the model for PKB activation summarized here is that the regulatory domain interacts selectively with the inactive, unphosphorylated (activation loop and hydrophobic sites) kinase domain conformer. The trigger to PKB activation is in part the shift towards a loss of PH domain–kinase domain interaction. Once phosphorylated the kinase domain does not interact with the regulatory domain and this active conformer can dissociate from the membrane and distribute to various cellular sites until the protein is dephosphorylated, at which point the kinase conformation alters, the PH domain re-engages and the protein is ‘lockeD' in this inactive mode until it is re-triggered.

PKC activation

Since the first identification of PKC as a proteolytically activated kinase and subsequently as an allosterically controlled kinase [19,20], it has been evident that this kinase subfamily as isolated from cells and tissues has significant catalytic potential that can be revealed in vitro by manipulation of conditions and does not require the acute input of other proteins (e.g. upstream kinases). In fact, it is now increasingly clear that, at least for the cPKC (α, β and γ), nPKC (δ, ϵ, η and θ) and probably also the aPKC (atypical PKC) (ζ and ι) isoforms, their latent capacity is dependent, to varying absolute degrees, on phosphorylation of the same activation loop and hydrophobic motif sites as PKB (note that PKCζ and ι naturally have glutamate, ‘phosphomimetic’ residues in their hydrophobic motif sites) as well as phosphorylation of the also conserved C-terminal turn motifs (reviewed in [21,22]). While generally required for optimum activity, the allosteric input that reveals this catalytic potential means that these specific kinase domain phosphorylations serve a priming role. There has been some controversy regarding the upstream kinases involved in these priming phosphorylations, specifically whether the hydrophobic site phosphorylations of cPKC and nPKC isoforms occur via autophosphorylation [23] or through trans-phosphorylation by an upstream kinase [24]. Indeed, the same debate still exists for PKB [25] since it is not established that the TORC2 complex acts catalytically as opposed to functioning as a scaffold enabling autophosphorylation. Suffice it to say here that the ‘prejudice’ of the authors is that for both PKC and PKB these are events operating in trans and are not autophosphorylation processes.

Relevant to our understanding of the formation of the conformationally active kinase domains of PKCs, it is thought that there is co-operation between the three kinase-domain phosphorylation sites and this is based on mutagenesis. Specifically, for PKCα, particular mutations at these sites lead to metastable kinase domain conformers that retain some activity but are thermolabile, sensitive to oxidation and sensitive to dephosphorylation [26,27]. This last property has been monitored in intact cells and indeed rapid activation-induced dephosphorylation characterizes these mutants, implying greater accessibility of the remaining unmutated sites. The rationalization of this derives from the now solved kinase domain structures, in particular that of PKCθ [28], where it is evident that all three phosphate moieties make key contacts serving to retain the substrate-competent conformer. This speaks for both the co-operative action of these sites (all contribute to the thermodynamic stability of the active conformer) and to the relative inaccessibility of the phosphates when locked in by occupation of their respective binding sites.

Irrespective of the precise mechanisms involved in phosphorylation of the PKC kinase domain priming sites, the acute allosteric mechanisms of cPKC and nPKC activation involve membrane interaction through the regulatory C1 and C2 domains [4]. Consequent to or associated with this membrane recruitment, the inhibitor region within the regulatory domain, the pseudosubstrate site, is displaced from the substrate-binding pocket of the catalytic domain. The basic sequences surrounding the alanine residue that comprise the pseudosubstrate site may themselves become bound to acidic phospholipids [29]. In the absence of a structure for the holoenzyme, evidence for this interaction comes from the original observation that point mutants or deletion of pseudosubstrate sites confer a constitutive function on the mutant proteins [30] and, furthermore, that peptides based on these inhibitory sequences (where the predicted ‘target’ alanine residue is converted into a serine) are in fact very good synthetic substrates for the PKC family (see e.g. [31]). So, for the PKCs in the basal state, cells typically express a (fully) functional latent form wherein the catalytic site is occupied by an intramolecular inhibitory sequence (Figure 2). It is implicit in this model that for the formation of this latent form the catalytic domain must be in an appropriately folded conformer to act as the recipient for the inhibitory, substrate-binding pocket interaction. Circumstantial evidence for this comes from the observation that kinase-inactive mutants, where the kinase domain is predicted to be in a non/weak-substrate (ATP) binding state, are constitutively membrane-associated, consistent with an open conformation of the holoenzyme lacking pseudosubstrate–kinase domain interaction, with consequent constitutive exposure of the lipid-binding surfaces in the regulatory domain. Consistent with this interpretation is the membrane association of truncated regulatory domain constructs [32]. Further evidence on the closed nature of the holoenzyme is provided by its susceptibility to dephosphorylation in vitro and in vivo. In the absence of lipids, the latent purified PKC is a poor phosphatase substrate in vitro; in the presence of lipids, it is a much better substrate [33]. In cells, PKCs become dephosphorylated on chronic activation, indicative of an opening of the holoenzyme structure and dynamic ‘breathing’ of the exposed kinase domain enabling phosphatase access, presumably through transient exposure of these sites.

Figure 2 PKC activation scheme

The PKC regulatory domain (triangle) and kinase domain (bi-lobal domain) do not interact when the kinase domain is unphosphorylated and in an inactive conformation. On phosphorylation by upstream kinases (probably in a membrane compartment), the kinase domain takes on an active conformation. The protein remains active while membrane-associated since the inhibitory region in the regulatory domain is sequestered (see text). On release from membranes, the inhibitory pseudosubstrate site within the regulatory domain (heavy line) recognizes the active kinase domain conformer, binds and represses activity. This interaction also inhibits dephosphorylation such that even in the ground state PKCs can be highly phosphorylated.

It is evident that post-activation PKB can dissociate from membranes to target distal substrates without further regulatory inputs (see above). By contrast, in the absence of additional events, on dissociation from membranes PKCs take on their phosphorylated but autoinhibited conformation. However, there are additional layers of regulation operating on PKC proteins that can confer diffusible function. These act to lock the active conformer and include protein–protein interactions exemplified by RACKS (receptors for activated C-kinase) [34], as well as further post-translational modifications as observed for oxidation-induced activation (reviewed in [35]). The extent to which these additional layers of control operate independently of membranes is unclear; however, proteolytic activation as reported for PKCδ clearly confers a soluble function that is permissive for nuclear uptake and action [36].

The distinct PKB and PKC ground states

A critical difference in the behaviour of these two subfamilies of AGC-type kinases is the typical ground state in which they exist in cells. As described above, prior to activation of PKB, the regulatory domain interaction with the catalytic domain requires the latter to be in an inactive conformation. Once activated by phosphorylation, the regulatory domain cannot re-engage until the kinase domain is dephosphorylated and back in this inactive conformation. For PKC, the regulatory domain only interacts with the active conformer and activation requires effector binding to sequester (physically or sterically) the inhibitory pseudosubstrate site and maintain an open substrate-accessible structure.

A consequence of the distinct ground states for PKB and PKC is the requirement for acute activity of the upstream kinase(s). For PKB, these are essential and must be acutely active and available for agonist-induced PKB-dependent responses. For PKC, the pathway(s) leading to its phosphorylation may need only to have been active hours prior to agonist stimulation with lipid/effector production subsequently sufficient to trigger PKC-dependent events. It is also the case that in the context of this model of PKC behaviour, the steady state of PKC phosphorylation may be decreased under conditions where there is a degree of chronic, low-level activation. Consequently, monitoring these priming site phosphorylations may be misleading in terms of the activation state of the protein. For PKCs, it may be more informative to monitor non-priming site, post-allosteric activation, autophosphorylation, as shown for PKCα [37].


The evolution of these distinct mechanisms contrasts yet again with the next closest member of the AGC-type kinase family, PKA (protein kinase A), where the regulatory and catalytic moieties reside in distinct genes. Here, the regulatory and kinase subunits interact in a manner related to PKC, the regulatory domains have inhibitory (pseudo)substrate sequences that bind to the substrate-binding pocket of the active kinase [38]. However, unlike either PKB or PKC, where these regulatory interactions are intramolecular, activation of PKA through the binding of cAMP leads to dissociation of the catalytic subunit from the regulatory subunit, providing the opportunity for diffusion of the active catalytic subunit away from its primary regulator, defining a further distinct set of dynamics for the signal elicited.

What are the further implications of the distinguishing PKB and PKC properties? One clear and important consequence of the kinase domain conformers comprising these different PKB/PKC states is that mutant forms designed to interrogate pathway behaviour and potential therapeutic benefits may be poor surrogates of active, drug-inhibited conformers. Most but not all PKB and PKC inhibitors target the active phosphorylated forms of these proteins acting to compete with ATP for the nucleotide-binding pocket. Hence these agents target open conformation, active kinases. Inactivating mutants of PKC, wherein the wholly conserved lysine residue in the nucleotide-binding pocket is mutated, typically do not become phosphorylated in cells under normal circumstances and exist in an open membrane-associated form (A.J. Cameron and P.J. Parker, unpublished work). This by itself might not confound comparison of actions/inactions relative to drug exposure; however, the distinct open conformation can in fact undermine consistent behaviour of the active inhibited versus the inactive states.

The evolved behaviours for the PKB and PKC proteins represent distinct strategies for agonist-dependent signalling. The dynamic consequences noted here perhaps provide a rationale for why they are distinct, but clear experimental evidence on the exploitation of these dynamics (as opposed to the fact of their existence) is lacking. Based on the accumulating structural insights into the properties of these kinases, it should prove possible to modify their dynamic behaviour and determine how this influences the requirements for their signal outputs.


We are grateful to the Parker laboratory for commentary on this review.


  • Signalling: A Focus Topic at Life Sciences 2007, held at SECC Glasgow, U.K., 9–12 July 2007. Edited by B. Austen (St. George's, University of London, U.K.), C. Dart (Liverpool, U.K.), R. Empson (Royal Holloway, University of London, U.K.), P. Haris (De Montfort, Leicester, U.K.), M. Houslay (Glasgow, U.K.), P. Parker (Cancer Research UK), A. Poole (Bristol, U.K.), G. Roberts (Leicester, U.K.) and L. Roderick (Babraham Institute, Cambridge, U.K.).

Abbreviations: AGC-type kinases, protein kinase A/protein kinase G/protein kinase C-family kinases; PH domain, pleckstrin homology domain; PDK1, phosphoinositide-dependent kinase 1; PKB, protein kinase B; PKC, protein kinase C; cPKC, nPKC and aPKC, conventional, novel and atypical PKC respectively; TORC2, mTOR (mammalian target of rapamycin)/rictor (rapamycin-insensitive companion of mTOR) complex 2


View Abstract