Biochemical Journal

Review article

The substrates and binding partners of protein kinase Cε

Philip M. Newton , Robert O. Messing


The ε isoform of protein kinase C (PKCε) has important roles in the function of the cardiac, immune and nervous systems. As a result of its diverse actions, PKCε is the target of active drug-discovery programmes. A major research focus is to identify signalling cascades that include PKCε and the substrates that PKCε regulates. In the present review, we identify and discuss those proteins that have been conclusively shown to be direct substrates of PKCε by the best currently available means. We will also describe binding partners that anchor PKCε near its substrates. We review the consequences of substrate phosphorylation and discuss cellular mechanisms by which target specificity is achieved. We begin with a brief overview of the biology of PKCε and methods for substrate identification, and proceed with a discussion of substrate categories to identify common themes that emerge and how these may be used to guide future studies.

  • anchoring protein
  • drug discovery
  • phosphorylation
  • protein kinase Cε (PKCε)
  • signalling


The human kinome contains over 500 protein kinases [1]. PKC (protein kinase C) (EC belongs to the AGC group of serine/threonine kinases and transduces signals derived from lipid second messengers. PKC is a family of kinases encoded by nine genes with distinct biological functions. cPKCs (conventional PKCs: α, β and γ) are activated by Ca2+ and diacylglycerol, nPKCs (novel PKCs: δ, ε, η and θ) are activated by diacylglycerol, but not by Ca2+, and aPKCs (atypical PKCs: ζ and ι/λ) are insensitive to Ca2+ and diacylglycerol, but are activated by other lipids and by phosphorylation. The ε isoform (PKCε) is widely expressed throughout the body and has important roles in the function of the nervous [2,3], cardiac [4] and immune [5] systems. It is an attractive drug target for the treatment of several conditions such as inflammation [5], ischaemia [6], addiction [7,8], pain [9], anxiety [10,11] and cancer [12,13]. There is therefore interest in developing PKCε inhibitors and in discovering additional drug targets through the identification of PKCε substrates and the signalling pathways in which they participate.

In the present review, we discuss substrates of PKCε and anchoring proteins that localize PKCε near its substrates. We have focused on substrates that have been identified by multiple methods, particularly those for which the phosphorylated residues have been determined. It is currently difficult to prove that any kinase directly phosphorylates a specific substrate in vivo and thus many surrogate measures are used to infer phosphorylation. These include in vitro phosphorylation assays with purified enzyme and substrate, radioactive labelling with 32P or use of phospho-specific antibodies to detect phosphorylation in intact tissues, and mutation of candidate phosphorylation sites to alter substrate function in vivo. It is only through a combination of these approaches that one can conclude with reasonable confidence that a particular substrate is phosphorylated by a specific kinase. Therefore, in the present review, we have limited our detailed attention to PKCε substrates for which direct in vitro phosphorylation has been demonstrated in conjunction with one or more indirect approaches using intact tissues.


PKCε shares many structural features (Figure 1) with other members of the PKC family, including a C1 domain containing two cysteine-rich motifs that bind diacylglycerols, a C2-like phospholipid-binding domain, a pseudosubstrate domain, C3 and C4 catalytic domains that contain a purine-binding site for ATP, an activation loop and a substrate recognition site [8]. A unique feature of PKCε is a six-amino-acid actin-binding motif between the C1a and C1b subdomains [14] (see also below). Like other PKC isoenzymes, PKCε is regulated by four key mechanisms: co-factor binding, phosphorylation, protein–protein interaction and regulated degradation. A major endogenous activator is DAG (sn-1,2-diacylglycerol) generated by PLC (phospholipase C)-mediated hydrolysis of phosphoinositides. PKC activation by DAG can be mimicked by the addition of other compounds that bind the C1 domain, such as tumour-promoting phorbol esters. PKCε may also be activated by other lipids, such as arachidonic acid, or PtdInsP2 [15,16].

Figure 1 Schematic diagram of PKCε primary structure

Shown are common (C) regions conserved between members of the PKC family, including the pseudosubstrate (PS) domain that binds to the substrate-recognition site when the enzyme is inactive, the neighbouring C1a and C1b domains that bind diacylglycerols, phosphatidylserine and phorbol esters, and the C2-like domain which contains a site (E14AVSLKPT21) for binding to RACK2 and a ‘pseudoRACK’ sequence (yR; H85DAPIGYD92) homologous with the corresponding binding site in RACK2. Also shown is the kinase domain, which is highly conserved between family members, and contains sites for binding of ATP and protein substrates. Regions unique to PKCε include an actin-binding domain that lies between the two C1 subdomains and the V3 hinge region which contains an autophosphorylation site (Ser368) necessary for binding of 14-3-3 proteins. Like other kinases, PKCε must be primed to exhibit full activity. Priming phosphorylation sites at the activation loop, turn motif and hydrophobic motif are shown in blue in the kinase and V5 domains. Two other sites that can be autophosphorylated in vitro are shown in green.

Like other PKC isoenzymes, PKCε must be primed through phosphorylation to display full enzymatic activity and respond to allosteric regulators. In common with other AGC kinase superfamily members, PKC isoenzymes are phosphorylated at three conserved priming sites in the catalytic domain: the activation loop (Thr566), the Thr-Pro turn motif (Thr710), and the hydrophobic Phe-Ser-Tyr motif (Ser729). PDK1 (phospholipid-dependent kinase 1) phosphorylates the activation loop, whereas the turn and hydrophobic motifs are autophosphorylated in conventional PKCs, and possibly in PKCε [17], although recent evidence indicates that mTOR (mammalian target of rapamycin) complex 2 may transphosphorylate these sites [18,19]. In cPKC isoenzymes, these phosphorylation events appear to be constitutive, whereas they are regulated in PKCε. Thus there exists a pool of unphosphorylated PKCε that can be recruited into active signalling through stimulation of receptors coupled to PI3Ks (phosphoinositide 3-kinases), which in turn activate PDK1 [17,20]. Immunoprecipitation and co-localization studies have demonstrated that unphosphorylated PKCε is associated with the anchoring protein CG-NAP [centrosome- and Golgi-localized PKN (protein kinase N)-associated protein]. This association occurs through the catalytic domain of PKCε and localizes immature PKCε to the Golgi apparatus [21].

Three additional sites in PKCε have been identified that can be autophosphorylated in vitro: Ser234, Ser316 and Ser368 [22]. However, in intact cells, it appears that these sites are transphosphorylated by cPKC isoenzymes. Phosphorylation at Ser368 is required for binding of PKCε to 14-3-3 proteins (see below). The functional consequences of phosphorylation at Ser234 and Ser316 are not yet known.


RACKs (receptors for activated C-kinase)

Following activation, PKCε translocates to specific subcellular compartments. This process involves interactions with anchoring proteins such as the RACKs, which serve to anchor PKC isoenzymes in close proximity to their substrates. Two RACKS have been cloned: RACK1, which preferentially binds activated PKCβII [23], and RACK2 (β′-COP), a coatamer protein that binds activated PKCε [24]. Disrupting the interaction between RACK2 and PKCε is a goal of some drug-discovery programmes aimed at reducing PKCε signalling [25]. Although the V1 region of PKCε and the C-terminus of β′-COP mediate a significant portion of this interaction, full-length PKCε binds also to the N-terminus of β′-COP, indicating that interactions are formed through other regions of these molecules. The PKCε–RACK2 interaction is responsible for localizing active PKCε to the Golgi apparatus [24], a localization that requires phosphorylation of PKCε at Ser729 in the hydrophobic motif [26]. In cardiac myocytes, RACK2 also localizes PKCε to myofilaments [27]. RACK1 can also interact with activated PKCε. For example, in human glioma cells, PKCε interacts with RACK1 to regulate integrin-mediated cell adhesion and motility [28].

PDLIM5 (PDZ and LIM domain protein 5)

PDLIM5 is a 63 kDa cytoplasmic protein also known as ENH1 (enigma homologue protein-1). It is a PKCε-binding protein that has been reported to bind N-type voltage-gated Ca2+ channels [29]. The function of N-type Ca2+ channels is increased by application of phorbol esters [30,31], and in transfected Xenopus oocytes, PDLIM5 facilitates this potentiation through formation of a PKCε–PDLIM5–N-type Ca2+ channel complex [29]. PKC phosphorylates the pore-forming α subunit, CaV2.2α, of N-type Ca2+ channels, although it is not clear whether PKCε is the PKC isoenzyme responsible [30,31]. The zinc finger LIM domain of PDLIM5 is essential for the formation of the PKCε–PDLIM5–N-type Ca2+ channel complex [32]. However, a recent study was unable to find an interaction between PDLIM5 and CaV2.2α, and questioned the role of PDLIM5 as an N-type channel adaptor protein [33]. Interestingly, PDLIM5 may also perform a scaffolding role for PKD (protein kinase D) and L-type Ca2+ channels in rat neonatal cardiomyocytes, to facilitate positive modulation of L-type Ca2+ channel function by PKD [34]. Most notably, this positive modulation can be inhibited by treatment with the general PKC inhibitors chelerythrine and GF109203X. There is in vitro evidence that PKCε can phosphorylate L-type Ca2+ channels [35]. Given that PKCε also binds PDLIM5 and can phosphorylate the activation loop of PKD [36], it seems reasonable to hypothesize that PKCε is involved in the PKD-dependent regulation of L-type Ca2+ channels and that PKD might participate in PKCε regulation of N-type Ca2+ channels.

TRAM {TRIF [TIR (Toll/interleukin-1 receptor) domain-containing adaptor protein inducing IFN (interferon) β]-related adaptor molecule}

Unlike RACK1, RACK2 or ENH1, some proteins that act as PKCε scaffolds or chaperones are also PKCε substrates. One such protein is TRAM. Mice lacking PKCε exhibit impaired immune function, most notably in the clearance of bacterial infections [37]. Further work on this phenotype has revealed that the response to LPS (lipopolysaccharide) is impaired in these mice. LPS is produced by Gram-negative bacteria and promotes a robust immune response via TLR4 (Toll-like receptor 4). TRAM is an adaptor protein that is activated downstream of TLR4 and activates the transcription factor IRF3 (IFN regulatory factor 3). This activation leads to the production of cytokines, including IFNβ and RANTES (regulated upon activation, normal T-cell expressed and secreted). McGettrick et al. [38] showed that PKCε phosphorylation of TRAM on Ser16 is key for signalling downstream of TLR4. This phosphorylation was demonstrated in vitro and was absent from lysates of LPS-treated fibroblasts from PKCε-knockout mice, but was restored by addition of PKCε. Antibody depletion of PKCε from lysates of LPS-treated THP1 cells also substantially reduced TRAM phosphorylation. Dephosphorylated TRAM is targeted to the cell membrane via myristoylation at its N-terminus. Following phosphorylation by PKCε, TRAM is depleted from the membrane, although it is not clear where it goes. Mutating Ser16 to alanine abolishes this effect as well as LPS-stimulated cytokine production. Recent studies have demonstrated a physical interaction between TLR4 and PKCε that is mediated by another adaptor protein, MyD88 (myeloid differentiation factor 88) [39].

14-3-3 proteins

The 14-3-3 family of proteins are encoded by seven different mammalian genes (α/β, ε, η, γ, τ/θ, ζ/δ and σ) and bind to sequence-specific phosphopeptide-binding motifs on several proteins, thereby regulating their function, stability or localization [40]. Two optimal binding motifs have been identified, mode-I (RSXpS/TXP) and mode-II (RXXXpS/TXP). The V3 hinge region of PKCε (Figure 1) contains a mode-I sequence surrounding Ser346 and a weak mode II site surrounding Ser368 [41]. GSK3β (glycogen synthase kinase 3β) phosphorylates PKCε at Ser346, whereas Ser368 is autophosphorylated [22,41]. When bound to 14-3-3 protein, PKCε is active in the absence of lipids.

During the late stages of mitosis, 14-3-3 proteins associate with the V3 region of PKCε, and the formation of this 14-3-3–PKCε complex is required for the separation of two daughter cells (cytokinesis) during abscission, the final stage of the cell cycle [41]. Mutation of either V3 phosphorylation site prevents this association and impairs completion of cytokinesis. Likewise, knockout, depletion or inhibition of PKCε impairs abscission. The mechanism by which PKCε facilitates abscission is not known, but it may involve RhoA. PKCε inhibition is associated with prolonged activation of RhoA and its persistent localization at the actomyosin ring that contracts to create a furrow before abscission. Presumably, inhibition and removal of RhoA is required for cytokinesis to be completed. The substrates that PKCε phosphorylates to regulate RhoA are not yet known, but likely candidates are RhoA GEFs (guanine-nucleotide-exchange factors) or GAPs (GTPase-activating proteins).

Histone H1

Histones are important components of chromatin that are essential for the chromosomal organization of DNA. Histone H1 was identified as a PKCε-binding protein using an overlay assay [42]. Subsequent in vitro phosphorylation analysis revealed that PKCε can phosphorylate histone H1, but not histone H2A. The individual phosphorylated residues and the functional consequences are as yet unknown, but current evidence suggests that histone H1 anchors activated PKCε in the nucleus where it can play a role in regulation of gene expression.


A common theme in cell signalling is the existence of kinase cascades that culminate in gene transcription. The activity of these cascades is under tight control and failures in regulation, resulting in aberrant gene transcription, may result in pathological conditions such as cancer. Roles for PKCε have been demonstrated in many different systems, regulating diverse functions such as cell death and nociceptor sensitization [2,3,1214]. However, within these intracellular signalling cascades, only three proteins have been positively identified as PKCε substrates: the kinases Akt/PKB (protein kinase B) and PKD, and the transcription factor STAT3 (signal transducer and activator of transcription 3).


Akt is a serine/threonine kinase important for cell survival and proliferation, angiogenesis, metabolism and protein translation [4345]. Activation of Akt requires phosphorylation at both Thr308 and Ser473. Phosphorylation of Thr308 is mediated by PDK1 [46,47]. Several other kinases phosphorylate Akt at Ser473, including ILK (integrin-linked kinase), MAPKAPK-2 [MAPK (mitogen-activated protein kinase)-activated kinase-2], DNA-PK (DNA-dependent protein kinase) and the rictor (rapamycin-insensitive companion of mTOR)–mTOR complex (reviewed in [12]). Cell culture studies have also implicated PKCε in the regulation of Akt [32,48,49]. In the heart, activated PKCε forms a complex with Akt and eNOS (endothelial nitric oxide synthase). Active Akt binds to PKCε via its PH (pleckstrin homology) domain, and formation of this three-member complex is enhanced by activation of PKCε. In vitro, PKCε phosphorylates Akt at Ser473 and increases its activity. PKCε also increases the phosphorylation of eNOS at Ser1171 and increases eNOS activity. These PKCε-mediated events appear to contribute to the cardioprotective effects of moderate ethanol consumption [50].


PKD1, formerly known as PKCμ, is a protein kinase that, like most PKC isoenzymes, is activated by phorbol esters, but contains a kinase domain that more closely resembles that of CaMKs (Ca2+/calmodulin-dependent protein kinases). PKD1 is activated in response to a variety of regulatory peptides and growth factors [51]. Like other kinases, PKD1 must be phosphorylated in its activation loop to be active. Two residues have been identified, Ser744 and Ser748, and PKCε phosphorylates both in vitro [36].


The STAT3 transcription factor is constitutively active in a variety of human cancers. STAT3 requires phosphorylation at two sites for activation: Tyr705 is phosphorylated by a Janus kinase, usually JAK1 [52], and Ser727, which resides in the transactivation domain, can be phosphorylated by several kinases, including PKCε. In prostate adenocarcinoma cells, PKCε forms a complex with STAT3 and phosphorylates STAT3 at Ser727 [53]. Knockdown of PKCε with RNA interference inhibits STAT3 Ser727 phosphorylation and reduces the activity of STAT3, measured by DNA binding and activation of transcription. These results identify a potential mechanism by which PKCε may promote cancer. They also provide, to our knowledge, the only known example of a transcription factor that is a substrate of PKCε.


GABAA (type A γ-aminobutyric acid) receptors

GABAA receptors mediate most inhibitory neurotransmission in the nervous system, and our studies with PKCε-knockout mice indicate that PKCε phosphorylates and regulates the function of these receptors. PKCε-knockout mice show increased behavioural responses to ataxic and hypnotic doses of ethanol and benzodiazepines [54]. The positive allosteric effects of these drugs on GABAA receptor function are also increased in crude synaptoneurosomes (microsacs) prepared from the frontal cortex or cerebellum of PKCε-knockout mice. In vitro phosphorylation studies and electrophysiological studies of receptors with mutations at known PKC phosphorylation sites revealed that PKCε phosphorylates Ser327 on γ2 subunits, and that phosphorylation at this site reduces the response of GABAA receptors to ethanol and benzodiazepines. A striking feature of ethanol intoxication is the development of acute tolerance to ethanol during a single drinking session, despite the persistence of elevated blood ethanol levels [55]. PKCε phosphorylation of γ2 Ser327 appears to participate in the development of this acute tolerance. Treatment in vivo with ethanol increases phosphorylation of PKCε at Ser729 in the cerebellum measured 1 h later, suggesting that ethanol increases the activity of cerebellar PKCε [56]. In vivo exposure to ethanol also increases phosphorylation of γ2 Ser327 and reduces the sensitivity of cerebellar GABAA receptors to ethanol in vitro at this time [56,57]. These findings suggest an important role for PKCε in mediating acute tolerance to the intoxicating effects of ethanol.

TRPV1 (transient receptor potential vanilloid 1)

TRPV1 is a ligand-gated, non-selective, cation channel activated by capsaicin, heat, protons, leukotrienes and anandamide [58]. It is expressed by small-diameter nociceptive neurons and integrates thermal and chemical stimuli as a polymodal receptor [5961]. TRPV1-knockout mice show reduced sensitivity to noxious thermal stimuli and reduced inflammatory thermal hyperalgesia [58,59]. PKC modulates TRPV1 function. Brief (<3 min) treatment with phorbol esters reduces the thermal threshold for activation and enhances activation by capsaicin, anandamide, ATP and bradykinin; this effect is inhibited by PKC inhibitors [6264].

In heterologous expression systems, phorbol esters also activate TRPV1 currents directly by a PKC-dependent mechanism [63,65,66], but this is evidently not commonly observed in rat DRG (dorsal root ganglion) neurons [64]. Two different groups have identified Ser502 and Ser800 as being important for phorbol ester enhancement of agonist and heat responses [67,68]. In vitro, PKCε phosphorylates Ser502 and Ser800, and in HEK (human embryonic kidney)-293 cells, intracellular application of the εV1-2 peptide inhibitor of PKCε, which inhibits PKCε binding to RACK2, reduces bradykinin-mediated enhancement of TRPV1 function [62]. Studies using the εV1-2 peptide also demonstrate that activation of neurokinin-1 receptors [57] or protease-activated-receptor 2 [69] enhances TRPV1 function via PKCε. More recently, PKD1 has been found to associate with and enhance the activity of TRPV1 through phosphorylation of TRPV1 at Ser116 [70]. This could provide a mechanism for indirect activation of TRPV1 by PKCε, since, as discussed above, PKCε phosphorylates the activation loop of PKD1, thereby stimulating PKD1 activity [36,71,72].


The cytoskeleton is composed of three structural protein elements: microtubules, intermediate filaments and actin microfilaments. PKCε regulates the function of all three of these structural elements (reviewed in [73]) by interacting with, and, in some cases, phosphorylating, proteins that form these structures.


Peripherin is an intermediate filament protein that is exclusively neuronal, being expressed at particularly high levels in the peripheral nervous system [74]. The formation of peripherin-containing protein aggregates in motor neurons is a feature of ALS (amyotrophic lateral sclerosis). Aggregates of endogenous peripherin can also form in SK-N-BE(2)C neuroblastoma cells when treated with phorbol esters. Peripherin does not appear to be directly phosphorylated by PKCε, but immunoprecipitation studies show that peripherin binds to PKCε in these cells [75]. Interestingly, the regulatory C1b domain of PKCε mediates the formation of this complex; the C2 domain does not participate, nor does peripherin bind the C1b domain of another PKC isoenzyme, PKCα. Complex formation is not regulated by PKCε activity. However, overexpression of peripherin in these cells causes peripherin aggregation, and this aggregate formation is reduced by knockdown or inhibition of PKCε and enhanced by phorbol esters or overexpression of PKCε. Thus these findings indicate an important role for PKCε in peripherin aggregation [75].


Vimentin is an intermediate filament protein found predominantly in cells of mesenchymal origin [76]. Like peripherin, vimentin is not phosphorylated directly by PKCε, but a number of PKCε-dependent phosphorylation events regulate the function of both vimentin and β1 integrins. Integrins are transmembrane proteins that mediate cell attachment to the extracellular matrix and thus are important for cell motility [77]. In motile cells, β1 integrin-containing adhesive structures are continually cycled from the trailing to the leading edge of cells in a PKCε-dependent manner [78]. The cycling of β1 integrins requires phosphorylation of vimentin; this occurs in a PKCε-dependent manner, but PKCε does not appear to be the kinase that is directly responsible [79]. In primary cultures of cardiac fibroblasts, treatment with angiotensin II promotes adhesion of cells to a collagen-1 matrix. Angiotensin II also promotes co-immunopreciptation of β1 integrin with PKCε and stimulates phosphorylation of PKCε at Ser729 in these cells. Angiotensin II does not stimulate adhesion of cells that lack PKCε, suggesting that activation of PKCε and association of PKCε with β1 integrin are required for angiotensin II-stimulated adhesion of cardiac fibroblasts to collagen [80].


Keratins are intermediate filament proteins that are expressed primarily in epithelial cells. There are numerous forms of keratin. In rat pituitary GH4C1 cells, TRH (thyrotropin-releasing hormone) stimulates PKCε phosphorylation of keratin 8. Treatment with TRH also causes PKCε and keratin 8 to co-localize. Keratin 8 is hyperphosphorylated in PKCε-overexpressing cells and is phosphorylated directly by PKCε in vitro at Ser8 and Ser23. The functional consequences of this phosphorylation are unclear, but may involve regulation of cell–cell contacts in epithelial cells [73].


Actin exists in two conformational states: globular (G-actin) and filamentous (F-actin). PKCε is unique among PKC isoenzymes in containing an F-actin-binding region. This six-amino-acid sequence (LKKQET) is located at amino acids 223–228 of PKCε. When PKCε is in the inactive state, this region is not exposed and thus actin is not bound. However, following treatment with PKC activators such as phorbol esters, diacylglycerol or arachidonic acid, PKCε undergoes a conformational change and the actin-binding region is exposed. Interestingly, binding of PKCε to F-actin stabilizes the active conformation of the enzyme [14,81]. The PKCε–F-actin interaction has profound consequences for the actin cytoskeleton, most notably promoting the formation of F-actin through several mechanisms, including inhibiting the depolymerization of F-actin, increasing the rate of actin filament elongation and overturning the inhibition of actin nucleation by thymosin β4 [82]. Functionally, the formation of a PKCε–F-actin complex appears to be important for glutamate exocytosis from dentate gyrus granule cells of the hippocampus [14].

IQGAP1 (IQ motif-containing GAP1)

The Rho-family GTPases (Rho, Rac and Cdc42) are key regulators of cytoskeletal function. These GTPases cycle between an active state (GTP-bound) and an inactive state (GDP-bound). The switch between active and inactive is mediated by their intrinsic GTPase activity. IQGAP1 is a Ca2+-sensitive GTPase scaffolding protein that binds to Rac1 and Cdc42 and promotes their active state [8385]. IQGAP1 and PKCε can be co-immunoprecipitated and IQGAP1 contains a consensus PKC phosphorylation site (Ser1443) at its C-terminus [86]. In vitro phosphorylation assays show that PKCε can phosphorylate IQGAP1 at this residue. The consequences of Ser1443 phosphorylation have not been conclusively demonstrated, but transfection of NIE-115 neuroblastoma cells with an IQGAP1 mutant that contains S1441E/S1443D mutations to mimic phosphorylation at these residues increases neurite outgrowth when cells are plated on laminin. A smaller increase in neurite outgrowth is seen when cells are transfected with native IQGAP1, whereas no increase is observed with an S1441A/S1443A mutant. Ser1441 is not phosphorylated by PKCε [87]. Interestingly, an emerging literature suggests that IQGAP1 exists in a complex containing β1 integrins [8890], whose cycling is controlled in a PKCε–vimentin-dependent manner as described above.

cMyBPC (cardiac myosin-binding protein C)

cMyBPC is a myosin-binding protein that has an essential role in cardiac contractility [91] and is a substrate for PKCε [92,93]. Transgenic mice that overexpress PKCε in the heart develop a dilated cardiomyopathy between the ages of 9 and 12 months. This pathology is associated with hyperphosphorylation of cMyBPC. In silico analysis of the cMyBPC protein sequence suggests Ser302 as a potential PKCε phosphorylation site and peptides containing this site and its surrounding sequence can be phosphorylated by PKCε in vitro. These data suggest that increased PKCε phosphorylation of cMyBPC at Ser302 contributes to aberrant cardiac function in this model of cardiomyopathy.

Cytoskeletal interaction with other PKCε substrates

PKCε interactions with the cytoskeleton may be important for the regulation of other PKCε substrates. For example, PKCε-dependent cytoskeletal regulation appears to be important for PKCε regulation of pain. Injection of adrenaline (epinephrine) into the hindpaw of rodents produces a hyperalgesia that is dependent on PKCε and can be mimicked by direct injection of PKCε activators [9]. This hyperalgesia can be blocked by agents that disrupt each of the three major cytoskeletal elements: microtubules, intermediate filaments and actin microfilaments. Disrupting the cytoskeleton does not affect baseline analgesia, suggesting that hyperalgesia mediated by PKCε activation requires an intact cytoskeleton. In a separate study, it was observed that hyperalgesia induced by a PKCε activator could be attenuated by blocking antibodies against β1 integrin [94], suggesting that PKCε-dependent β1 integrin cycling and regulation of β1 integrin by the PKCε substrate IQGAP1 are functionally important for PKCε-mediated hyperalgesia.

PKCε interactions with the cytoskeleton may promote oncogenesis. Cell culture studies have demonstrated that overexpression of PKCε is associated with anchorage-independent cell growth and increased tumorigenicity, and some cancer cell lines display increased PKCε activity (reviewed in [12]). These findings have significance in vivo, since overexpression of PKCε in the epidermis enhances development of squamous cell carcinoma after topical treatment with tumour promoters [95]. Several of the PKCε substrates and binding partners described here may contribute to the pro-oncogenic properties of PKCε. For example, the actin-binding domain of PKCε is required for the metastatic potential of cells overexpressing PKCε; injection of NIH 3T3 cell lines that overexpress PKCε into nude mice causes tumours that metastasize. Injection of NIH 3T3 cell lines that overexpress a deletion mutant of PKCε lacking the actin-binding site still cause tumours, but those tumours do not metastasize [96]. In vitro, these deletion mutants do not show the invasive morphology and phenotype of cells that overexpress native PKCε. It has been proposed that Rho GTPases and β1 integrins act downstream of PKCε in these systems to mediate some of the pro-oncogenic effects of PKCε [28,88,97].


Using a combination of in vitro phosphorylation, regulation of kinase activity in intact cells and mutation of putative phosphorylation sites in candidate substrates, it is possible to identify kinase substrates with reasonable certainty. Substrates of PKCε identified by these approaches are summarized in Table 1. However, there are many examples where not all criteria have been met and therefore a candidate substrate has not been shown to be phosphorylated directly by PKCε. For example, PKCε plays a role in regulating gap junctions and has been postulated to phosphorylate connexin43, but there are no direct in vitro phosphorylation data yet to support that conclusion [98,99]. Similarly, a role for PKCε has been demonstrated in intracellular signalling cascades, such as the MAPK pathway, without direct substrates being identified [100]. A large volume of literature points to a critical role for PKCε in ischaemic pre-conditioning [4], the phenomenon whereby brief periods of ischaemia and reperfusion confer protection from subsequent ischaemic episodes. Again, the PKCε substrates involved in this process are not yet known. One candidate is the mitochondrial permeability transition pore, which is a large-conductance multiprotein complex involved in ischaemic cell death. PKCε is a component of the complex along with the VDAC1 (voltage-dependent anion channel 1), ANT1 (adenine nucleotide translocase), hexokinase II and cyclophilin D [101]. Functional studies demonstrate that activation of PKCε attenuates pore opening, consistent with its role in ischaemic pre-conditioning. In vitro, PKCε can phosphorylate VDAC1, although the kinetics of this reaction and the phosphorylated residues are not known. Several other potential substrates have been proposed, including the mitochondrial ATPase-sensitive K+ channel and cytochrome c oxidase subunit IV (reviewed in [6]). However, direct in vitro phosphorylation of these targets has not, to our knowledge, been demonstrated.

View this table:
Table 1 Summary of the currently identified substrates and binding partners of PKCε

Even if in vitro studies demonstrate that a kinase can phosphorylate a protein directly, there is still uncertainty about whether the protein is an immediate substrate of that kinase in vivo. Developments in chemical genetics have led to a new approach that capitalizes on the fact that all kinases share a highly conserved purine-binding pocket. Shokat and colleagues have developed a strategy that utilizes the conserved features of the purine-binding domain to engineer mutants specific for ATP analogues and modified inhibitors that bind competitively with ATP at this site [102]. The mutation is made to replace a large ‘gatekeeper’ residue with a small amino acid such as glycine or alanine, and is designed to be functionally silent with respect to kinase activity and substrate specificity. This generates an AS (analogue-selective) mutant that can utilize non-natural ATP analogues with bulky residues at the N6 position. Only the AS kinase, but not wild-type kinases, can use N6-substituted analogues as phosphate donors. Therefore only unique substrates of the AS kinase are labelled by the ATP analogues. This approach has been used successfully to identify direct substrates of several kinases including JNK (c-Jun N-terminal kinase), v-Src and Cdk1 (cyclin-dependent kinase 1) [103]. We have used an AS kinase mutant of PKCε to study autophosphorylation of PKCε [22].

Several modifications to the use of AS kinases have facilitated identification of direct substrates. Most AS kinases can use ATP[S] (adenosine 5′-[γ-thio]triphosphate) analogues to label substrates, thereby producing thiophosphorylated instead of phosphorylated residues. Alkylation of the thiophosphate group with p-nitrobenzylmesylate yields a unique epitope tag that can be used to immunopurify the direct substrate and analyse the phosphorylation site by MS [104]. Alternatively, thiophosphate residues can react with iodoacetamido-biotin at low pH (<4.0) allowing the substrate to be affinity-purified using strepavidin beads [105]. In addition, some kinases can use ATP biotinylated at the γ position instead of ATP to transfer biotin directly to a substrate for affinity purification [106]. Finally, as an alternative to the AS kinase method, it is also possible to identify direct substrates using tandem affinity purification tagging [TAP-tag kinase and cross-linking of substrate and kinase using reagents such as 3,3′-dithiobis(sulfosuccinimidyl) propionate] [107].

Once a direct substrate has been identified, specific sites of phosphorylation can be determined by MS. For large substrate proteins or proteins with several sites of phosphorylation, one can use bioinformatics tools to predict phosphorylation sites, which can be validated by in vitro phosphorylation assays with substrate mutants that contain alanine substitutions at these sites. Tools that only scan for simple phosphorylation motifs suffer from a high rate of false positive hits. Other programs use a variety of techniques to reduce the false positive rate. For example, GPS ( classifies protein kinases into hierarchies and calculates the theoretically maximal false positive rate for each cluster of protein kinases. Linding et al. [108] have developed a different computational tool, NetworKIN (, that matches known phosphorylation motifs with information from several literature sources and databases to improve accuracy. Future developments in bioinformatics tools should greatly aid in the search for phosphorylation sites on substrates of all kinases, including substrates of PKCε.


This work was supported by the National Institutes of Health [grant numbers AA013588, NS053709 and AA017072 (to R.O.M.) and DA027948 (to P.M.N.)], by the U.S. Department of the Army [contract PT075682 (to P.M.N.)], and by funds provided by the State of California for medical research on alcohol and substance abuse through the University of California at San Francisco.

Abbreviations: AS, analogue-selective; cMyBPC, cardiac myosin-binding protein C; DAG, sn-1,2-diacylglycerol; ENH1, enigma homologue protein-1; eNOS, endothelial nitric oxide synthase; F-actin, filamentous actin; GABAA, type A γ-aminobutyric acid; GAP, GTPase-activating protein; IFN, interferon; IQGAP1, IQ motif-containing GAP1; LPS, lipopolysaccharide; MAPK, mitogen-activated protein kinase; mTOR, mammalian target of rapamycin; PDK1, phospholipid-dependent kinase 1; PDLIM5, PDZ and LIM domain protein 5; PKC, protein kinase C; aPKC, atypical PKC; cPKC, conventional PKC; nPKC, novel PKC; PKD, protein kinase D; RACK, receptor for activated C-kinase; STAT3, signal transducer and activator of transcription 3; TLR4, Toll-like receptor 4; TRAM, TRIF [TIR (Toll/interleukin-1 receptor) domain-containing adaptor protein inducing IFNβ]-related adaptor molecule; TRH, thyrotropin-releasing hormone; TRPV1, transient receptor potential vanilloid 1; VDAC1, voltage-dependent anion channel 1


View Abstract