Protein kinase D displays intrinsic Tyr autophosphorylation activity: insights into mechanism and regulation

The protein kinase D (PKD) family is regulated through multi‐site phosphorylation, including autophosphorylation. For example, PKD displays in vivo autophosphorylation on Ser‐742 (and Ser‐738 in vitro) in the activation loop and Ser‐910 in the C‐tail (hPKD1 numbering). In this paper, we describe the surprising observation that PKD also displays in vitro autocatalytic activity towards a Tyr residue in the P + 1 loop of the activation segment. We define the molecular determinants for this unusual activity and identify a Cys residue (C705 in PKD1) in the catalytic loop as of utmost importance. In cells, PKD Tyr autophosphorylation is suppressed through the association of an inhibitory factor. Our findings provide important novel insights into PKD (auto)regulation.

Protein kinases are essential for cellular life. They play a role in almost all physiological processes and their dysfunction is associated with a plethora of diseases [1]. The activity and function of many kinases is regulated through phosphorylation, in particular through phosphorylation of the activation segment [2]. This is especially the case for so-called 'RD' kinases, which contain an arginine in the highly conserved HRD motif in the catalytic loop. The Arg residue in this motif creates a basic pocket together with a Lys residue located C-terminally of the DFG motif that accommodates the phosphate group of the phospho-Ser/Thr residue in the activation loop [3]. In many cases, phosphorylation of activation segment residue(s) is dependent on the activity of an upstream kinase, but many kinases also possess an inherent autophosphorylation activity, which is physiologically relevant in many cases [4].
One of the most obvious examples of an activating autophosphorylation is found within the Receptor Tyrosine Kinase (RTK) group, where ligand-induced dimerization results in trans-autophosphorylation of the activation loop residues, and in generation of docking motifs to initiate the signaling cascade [5]. However, kinase autophosphorylation extends beyond RTKs and is actually much more common than hitherto believed. In a recent comparative literature search of all RDkinases, it was found that about 63% have autophosphorylation capabilities and 45% autophosphorylate their regulatory activation loop residue(s) [4].
Kinases can be divided by their specificity toward their targeted residues. The majority of kinases are those that create phospho-ester bonds on hydroxyl groups, and can be subdivided in Ser/Thr and Tyr kinases [6]. For some kinases, however, such a distinction cannot be made so clearly. For example, several predominantly Ser/Thr kinases have been shown to also (auto)phosphorylate Tyr residues [7][8][9][10][11][12][13][14][15][16][17][18][19]. The inverse, a Tyr kinase that also (auto)phosphorylates on Ser/Thr residues, is less common, but occurring nonetheless [20,21]. The role of this 'dual specificity' is well-described for some kinases [9,10,18,19]. Other (auto)phosphorylation events are less well understood in terms of function, and sometimes dual specificity is only seen under specific circumstances [19]. While it is known that some kinases have a dual specificity, the molecular determinants in kinases that confer to this property have not been examined in close detail.
The protein kinase D (PKD) family consists of three closely related isoforms in humans, and is part of the CAMK group [22]. They play a versatile role in (disease) biology, ranging from migration to secretion, proliferation and invasion (for a recent review see [23]). They consist of a large N-terminal domain, followed by the kinase domain. The N-terminal regulatory region encompasses an alanine/proline rich region (AP region), a tandem C1 domain binding diacylglycerol (DAG) and phorbol esters, an oligomerization domain, and a pleckstrin homology (PH) domain [24][25][26][27][28]. Deletion mutagenesis approaches have indicated that the C1 and PH domains have an autoinhibitory effect on kinase activity, which is alleviated by activation loop Ser-738/742 phosphorylation (hPKD1 numbering) [24][25][26]. These phosphorylations are exerted by several PKCs, but PKDs can also autophosphorylate on Ser-738 and Ser-742 in vitro, and on Ser-742 in cells [29,30]. Furthermore, PKD1/2 autophosphorylates in vivo at a C-terminal Ser residue (Ser-910 in PKD1). This phosphorylation is not contributing to PKD activation or maximal activity, but rather regulates the duration of PKD signaling in a cellular context [31].
Protein kinase D are also regulated through Tyr phosphorylation, for example in oxidative stress conditions. Upon exposure to H 2 O 2 , Tyr-95 and Tyr-463 in PKD1 are phosphorylated by Src and Abl respectively [32][33][34][35]. In PKD2 an additional Tyr residue is phosphorylated by Abl in oxidative stress at the P + 1 loop of the activation segment, causing an increase in substrate turnover [36].
In this paper, we describe that the PKD family members also display autophosphorylation of this Tyr residue in the P + 1 loop, and we identify the molecular determinants for this unusual activity.
Endogenous PKD1 was precipitated from HEK293 cells using a home-made antibody that recognizes the C-terminal epitope in PKD1 (EEREMKALSERVSIL). Cell lysate was incubated with antibody for 2 h and protein A beads were subsequently added. Beads were washed two times with NENT750 (50 mM Tris, pH 7.4, 1 mM EDTA, 750 mM NaCl, 0.1% NP40, 25% glycerol) and once with TBS (50 mM Tris, pH 7.4, 150 mM NaCl) prior to on-bead kinase activity assays.
Exogenous expression and purification of 6xHis-SUMO-PKD1-CAT in Sf21 insect cells was done according to standard procedures. Briefly, 6xHis-SUMO-PKD1-CAT was expressed by addition of recombinant baculovirus to a Sf21 cell culture at a cell density of 1 million cells per milliliter and a multiplicity of infection of around 1. Cell culture pellets were harvested 72 h after infection by centrifugation at 10 000 g. For protein purification, the pellet was resuspended in 50 mM Tris pH8, 300 mM NaCl, 10% glycerol supplemented with phosphatase inhibitors (Phosphostop, Roche), and protease inhibitors (cOmplete, Roche). Cells were subsequently lysed by 20 strokes in a Dounce homogenizer. The supernatant was collected by centrifugation at 15 000 g for 30 min and loaded on a Ni 2+ -NTA agarose column. After flowthrough of the soluble fraction, the column was washed with 10 column volumes of 50 mM Tris pH8, 750 mM NaCl, 20 mM Imidazole, 10% glycerol and 5 column volumes of 50 mM Tris pH8, 50 mM NaCl, 20 mM Imidazole, 10% glycerol. Proteins were eluted in 50 mM Tris pH8, 50 mM NaCl, 160 mM Imidazole, 10% glycerol. Abundance and purity of proteins in the elution fractions was verified by SDS/PAGE and peak fractions were pooled and dialyzed against 50 mM Tris pH8, 50 mM NaCl, 25% glycerol.
Exogenous expression of MBP-tagged mouse PKD1 catalytic domain in BL21-Rosetta cells was done according to standard procedures. Briefly, BL21-Rosetta cells were transformed with pMAL-C2X-mPKD1CAT and an overnight preculture was inoculated in 1L LB containing 100 lgÁmL À1 ampicillin and 25 lgÁmL À1 chloramphenicol. At OD600 of 0.6 expression of MBP-mPKD1-CAT was induced with 0.5 mM IPTG. Four hours post induction the cell pellet was collected by centrifugation at 4000 g for 20 min. MBP-mPKD1-CAT was purified according to the pMAL Protein Fusion and Purification System manual (NEB, Ipswich, MA, USA). Peak elution fractions were pooled and dialyzed against 50 mM Tris pH8, 50 mM NaCl, 25% glycerol.

Activity assays and kinetics
For in vitro autophosphorylation of PKDs, purified protein was incubated with 100 lM ATP in 50 mM Tris, pH7.4, 10 mM MgCl 2 and allowed to incubate for 30 min or the indicated time points at 30°C. Reactions were stopped by addition of SDS sample buffer and boiling at 95°C for 5 min prior to loading of an SDS/PAGE. Small unilamellar vesicles (SUVs) containing PS/PDB were added to the reaction mixture as indicated. Vesicles were prepared as follows: PDB and PS were mixed to final concentrations of 100 lgÁmL À1 PS and 250 nM PDB and dried using a Savant Speedvac concentrator (Thermo Fisher Scientific). Dried lipids were resuspended in 50 mM Tris pH7.4 and sonicated 3 times for 10 min, vesicle fragmentation was verified by the appearance of a clear solution.
For kinase assays where Syn-2 phosphorylation was assessed, the following reaction mixture was prepared: 50 mM Tris, pH7.4, 10 mM MgCl 2 , 50 ng PKD and 1.5 mgÁmL À1 Syn-2 peptide. Reactions were started with 100 lM ATP complemented with 2 lCi [c-32 P]ATP (Perkin-Elmer, Waltham, MA, USA). After 10 0 (in the linear range) the reaction was stopped by spotting 30 lL on a Whatman P81 filter paper. The filter papers were washed 3 times in 0.5% phosphoric acid, followed by one wash in 100% acetone. Subsequently the papers were air-dried and counted using the Tri-Carb 2810 TR scintillation counter (Perkin-Elmer). Data were analyzed using GRAPHPAD (PRISM, La Jolla, CA, USA).

PKD autophosphorylates on Tyr in the P + 1 loop in vitro
When studying Tyr phosphorylation of PKD by tyrosine kinases in vitro, we observed that when PKD was incubated with Mg 2+ .ATP alone, high levels of Tyr phosphorylation were detected without addition of an upstream kinase. As this phenomenon could be due to the co-precipitation of a Tyr kinase during PKD purification, we purified PKD using high ionic-strength washing steps (2M NaCl) obtaining a highly pure PKD preparation (100% on Coomassie Brilliant Blue stained SDS/PAGE, Fig. S1), devoid of any tyrosine kinases as determined via Mass-spectrometry (Table S1). Yet, this PKD preparation was still phosphorylated on Tyr residues upon incubation with Mg 2+ .ATP (Fig. 1A). Therefore, we hypothesized that PKD might autophosphorylate on Tyr residues in vitro.
To test this hypothesis, we followed in vitro Tyr phosphorylation of PKD1 in presence of the PKD inhibitors CRT0066101 and CID755673. Interestingly, we observed a dose-dependent decrease of Tyr phosphorylation upon incubation with these PKD inhibitors, while the same Tyr phosphorylation in vitro was insensitive to inhibition by the tyrosine kinase inhibitors PP2 and STI-751 (Fig. 1B). The observed phenomenon was not an artefact of PKD overexpression in human cells, since both endogenously expressed PKD1 retrieved from cells as well as heterologously expressed PKD1 catalytic domain in Spodoptera frugiperda Sf21 insect cells displayed autophosphorylation activity towards Tyr residues (Fig. S2). Furthermore, all three PKD isoforms displayed in vitro autophosphorylation on Tyr residues, albeit more pronounced for PKD2 (Fig. S3). PKD Tyr kinase activity was restricted to autophosphorylation, since we could not detect any activity towards a modified Syntide-2 peptide containing a Tyr in lieu of a Ser as the phospho-acceptor (Fig. 1C).
Next, we wondered which Tyr sites were autophosphorylated by PKD during the in vitro reaction. To identify these sites, we performed an MS analysis of full-length PKD1 subjected to a terminal autophosphorylation reaction. Interestingly, only one Tyr phosphorylation site was identified, namely Tyr-749 in the P + 1 loop of the kinase (Table 1). This was confirmed using a mutant in which this residue was substituted with Phe (PKD1 Y749F mutant). Although the activity of a P + 1 loop Y-F mutant is strongly impaired, it still displays some trans-phosphorylation activity ( Fig. 1D and [36]). However the mutant failed to autophosphorylate on Tyr residues even after 80 min in the reaction (Fig. 1E).

Molecular determinants of Tyr autophosphorylation
Next, we wondered under which conditions PKDs are able to autophosphorylate on Tyr residue(s). Autophosphorylation can occur in cis or in trans. To determine which mechanism applies to PKD Tyr-autophosphorylation, we incubated GST-tagged wild-type PKD1 with a FLAG-tagged kinase-dead (KD) PKD1 mutant, in order to be able to distinguish them by molecular weight. As shown in Fig. 2A, GST-PKD autophosphorylated efficiently on Tyr residues, but no Tyr phosphorylation could be detected for the FLAG-tagged KD construct, indicating that the autophosphorylation reaction on Tyr occurs in cis. Autophosphorylation of the activation loop Ser-738/742 residues also occurred predominantly in cis in this reaction ( Fig. 2A). Stimulation with small unilamellar vesicles (SUVs) consisting of phosphatidylserine/ Phorbol-12,13-dibutyrate (PS/PDB) resulted in an expected increase in activation loop Ser-738/742 autophosphorylation ( Fig. 2A), as previously described [29]. Surprisingly however, Tyr autophosphorylation activity was reduced after lipid stimulation ( Fig. 2A), indicating that lipid-induced conformational changes might structure PKD towards Ser autophosphorylation, while a 'lipid-devoid' conformation would promote Tyr autophosphorylation. This was further confirmed by following the kinetics of PKD purified from PDB-stimulated cells, indicating that indeed, the lipid-activated conformation of PKD disfavors Tyr autophosphorylation (Fig. 2B). The intriguing property of Tyr autophosphorylation, which is only reported for a small subset of Ser/Thr kinases, raised the question which structural features in PKD might contribute to its Tyr kinase activity.
Interestingly, in 1992, Tony Hunter already predicted linear motifs that could contribute to dual specificity of protein kinases [38]. One motif in which he observed divergence from 'normal' Ser/Thr kinases was the catalytic loop, where many dual-specificity kinases have substitutions for the conserved Arg in the HRD motif. Intriguingly, alignment of currently reported Ser/Thr kinases that exhibit dual specificity indeed confirmed that many of them have substitutions for the conserved Arg residue (Fig. 3A). This includes PKDs, which contain a Cys residue at this Arg position. Mutation of Cys to Arg in PKD1 (PKD1.C705R) resulted in an interesting enzymatic profile. In resting conditions, the activity of this mutant was decreased compared to WT PKD1: it showed less Tyr and Ser autophosphorylation compared to WT (Fig. 3B), but also the rate of Syn-2 phosphorylation was decreased compared to WT in the resting state (Fig. 3C). Upon stimulation with PS/PDB containing vesicles, activation loop Ser autophosphorylation activity potently increased in the C705R mutant, reaching even higher levels than the WT kinase, while still showing low levels of Tyr autophosphorylation (Fig. 3B). Activities towards Syn-2 were equal for both the WT and mutant after prior in vitro autophosphorylation, despite the higher activation loop Ser phosphorylation in the mutant (Fig. 3D). These data indicate that the residue at the HRD Arg position may not only stabilize the inactive conformation, but also the direct the specificity of the autophosphorylation reaction.

Tyr autophosphorylation in cells is inhibited by a factor associating with PKD
Since PKDs display this unusual activity in vitro, we wondered whether autophosphorylation was also occurring in intact cells under conditions where PKDs are known to be tyrosine phosphorylated. Therefore, we treated cells with the Tyr kinase activator/Tyr phosphatase inhibitor H 2 O 2 (Fig. 4A) or the Tyr phosphatase inhibitor pervanadate (Fig. 4B) and probed for the sensitivity of PKD1 Tyr phosphorylation to the same set of inhibitors as in Fig. 1C. Under these conditions, it was clear that Tyr phosphorylation of PKD1 was mediated exclusively by upstream tyrosine kinases, which were responsive to PP2 and STI-571 inhibition, and not via autophosphorylation, since Tyr phosphorylation was insensitive to PKD inhibition. We then wondered why this intracellular PKD activity was so divergent from its in vitro activity. Since it was clear that a lack of Tyr autophosphorylation in cells is not due to high phosphatase activity, we hypothesized that an inhibitory factor associating with PKD might block its autocatalytic activity in cells. In order to test this hypothesis, we produced different PKD1 preparations, each subjected to washing steps of different ionic strengths. If a cofactor dependent on ionic interactions bound to PKD, it could progressively dissociate with increasing strength of the washing buffer, which would subsequently increase PKD Tyr autophosphorylation. Tyr autophosphorylation by these differentially purified PKD preparations was followed in the presence of PP2 and STI-571 to exclude any phosphorylation exerted by a co-purifying Tyr kinase. The PKD inhibitor CRT0066101 was included as a control to verify that the observed Tyr phosphorylation was indeed dependent on PKD activity. As shown in Fig. 4C, Tyr autophosphorylation increased with increased ionic strength in washing steps, indicative of an inhibitory factor associating with PKD in cells via ionic interactions. Additionally, when we purified the catalytic domain of PKD1 from bacterial cells to assess in vitro Tyr autophosphorylation activity, we found that the protein was recovered as already being Tyr phosphorylated, while still showing a noticeable (albeit modest) increase in Tyr autophosphorylation in vitro (Fig. 4D). Since bacteria contain only a few 'BY' Tyr kinases (that do not share the conserved kinase fold) we wondered whether bacterial Tyr phosphorylation of PKD1 was the result of autophosphorylation activity. Therefore we induced MBP-PKD1-CAT expression in presence or absence of the PKD inhibitor CRT0066101. PKD inhibition during induction resulted in a complete loss of Tyr phosphorylation, indicating that PKD1 can autophosphorylate on Tyr residues in bacterial cells (Fig. 4E). These results are supportive of either the absence of an auto-inhibitory factor or decreased Tyr phosphatase activity in bacteria; although it should be noted that loss of phosphatase activity by itself does not increase Tyr autophosphorylation in eukaryotic cells (Fig 4A and 4B).

Discussion
Protein kinases often critically depend on phosphorylation for their regulation. In a cellular context, these regulatory phosphorylations are frequently exerted by upstream kinases. Many kinases however also possess the ability to autophosphorylate, and a recent study suggests that this might be more frequent than hitherto believed [4]. Protein kinase D is also regulated via phosphorylation and can autophosphorylate both in vitro and in cells. Known in vitro autophosphorylation sites for PKD include the activation loop Ser-738/742 residues and the C-terminal Ser-910 residue [29,31,39,40]. In cells, PKD is phosphorylated on Ser-738 mainly by upstream PKCs, while Ser-742 phosphorylation can occur as a PKC-mediated event, but under certain circumstances also as an autocatalytic event [41].
In this study, we describe the surprising finding that PKDs can also autophosphorylate at a Tyr residue in the P + 1 loop. This residue lies at an interesting position, just before the APE motif and is highly conserved in many Ser/Thr kinases. In a previous study, we found that transphosphorylation of this residue by Abl results in higher turnover of substrate phosphorylation [36]. Interestingly, the enzymatic requirements for this Tyr autophosphorylation significantly differ from those for activation loop Ser autophosphorylation. Indeed, we observed that while PKD Ser autophosphorylation is stimulated in the presence of PS/PDB containing SUVs, or when previously activated by PDB in cellulo, Tyr autophosphorylation activity is inhibited in these conditions ( Fig. 2A and schematic representation in Fig. 5). Furthermore, it seems that Tyr autophosphorylation displays a high sensitivity to CID755673-mediated inhibition, while activation loop Ser autophosphorylation is largely unaffected by this PKD inhibitor (Fig. 1C). This corroborates cellular data showing that, with this allosteric compound, activation loop phosphorylation in cells is not diminished but rather increased [42].
Interestingly, in a cellular context, the phosphorylation at Tyr-749 does not seem to be catalyzed in auto, but by bona-fide upstream Tyr kinases. In oxidative stress conditions, for example, we could show that Abl is an upstream kinase for the P + 1 Tyr residue, and that it specifically phosphorylates the PKD2 isoform due to an isoform-specific motif preceding the Tyr residue [36]. The mechanism that prohibits Tyr autophosphorylation in cells is hitherto unclear, but it likely involves the association of an inhibitory factor (Fig. 5). It is indeed not uncommon that (Tyr) autophosphorylation reactions are regulated by binding of cofactors. For example, GSK3b can autophosphorylate on Tyr residues, an activity that is dependent on Hsp90 association [18]. Furthermore, p38a autophosphorylation in cells is dependent on association with tumor growth factor-b-activated kinase 1/MAP3K7-binding protein 1 (TAB 1) [43]. In PKD, autophosphorylation on Ser-742 in cells seems to be dependent on the association with certain Ga isoforms [44].
In search for a structural explanation for its intrinsic dual-specificity, we found that PKD does not contain a conserved Arg residue in the catalytic loop HRD motif; instead, it contains a Cys. Interestingly, HRD Arg substitutions are shared by many kinases with reported dual specificity (Fig. 3A). Substitution of Cys with Arg in PKD resulted in an overall loss of activity in the autoinhibited state, a loss of Tyr autophosphorylation in both lipid-stimulated as well as unstimulated conditions and increased Ser autophosphorylation in PDB stimulated conditions (Figs 3B and 5). The Arg occurring in most kinases is adjacent to the catalytic Asp and stabilizes the activation loop in the active state by forming a basic pocket (together with a Lys residue C-terminal of the DFG motif) to bind the primary phospho-acceptor in the activation loop [45]. PKD is a non-RD kinase, but is nonetheless dependent on activation loop phosphorylation for its activity [34]. Interestingly, the Arg residue in the HRD motif is sometimes seen to pair with a conserved Glu in the aC helix in inactive kinases [2,46]. This highly conserved Glu forms the characteristic salt bridge with Lys in b strand 3 of the N-lobe in active kinases. A substitution of Cys with Arg in PKD could potentially result in the formation of such an inhibitory Arg-Glu salt bridge and stabilize the inactive conformation.
In conclusion, we report an unusual Tyr autophosphorylation activity by the PKD family of Ser/Thr kinases, an activity that has catalytic requirements distinct from its Ser autophosphorylation activities, and is dependent on the presence of a Cys residue in the HxD catalytic loop motif. Since this activity is only observed in vitro and suppressed in a cellular context, future research efforts will need to be designated to determine under which conditions Tyr autophosphorylation may be allowed to occur in a cellular context.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. Purity of PKD preparations assessed on a Coomassie brilliant blue stained polyacrylamide gel.   .  Table S1. List of proteins identified in a highly pure PKD preparation (washed with a buffer containing 2M NaCl) and their mascot scores.