Protein kinase D activity controls endothelial nitric oxide synthesis Free

Vascular endothelial growth factor (VEFG) is a key physiological regulator of both vascular development during embryogenesis (vasculogenesis) and blood vessel formation in the adult (angiogenesis). VEGF is mainly secreted from mesenchymal, stromal and epithelial sources to act on endothelial cells in both arteries and veins, and its receptors are transmembrane proteins that belong to the receptor tyrosine kinase family (Ferrara et al., 2003; Ho and Kuo, 2007; Olsson et al., 2006). The angiogenic effects of VEGF are primarily mediated by VEGF receptor 2 (VEGFR2) (Ho and Kuo, 2007). Upon VEGF binding to endothelial cells, VEGFR2 undergoes dimerization and tyrosine kinase autophosphorylation, which, in turn, activates various signaling cascades (Byrne et al., 2005). The best-characterized actions of VEGF on endothelial cells are its ability to promote growth, stimulation of DNA synthesis and proliferation (Parenti et al., 1998). In addition, VEGF is also a survival factor for endothelial cells in vitro, preventing apoptosis induced by serum starvation in a process mediated, at least partially, by the phosphoinositide 3-kinase (PI3K)–Akt pathway (Fujio and Walsh, 1999; Gerber et al., 1998a; Gerber et al., 1998b). Furthermore, a constitutively active Akt is sufficient to promote survival of serum-starved endothelial cells (Gerber et al., 1998b).

At present, two isoforms of the protein kinase D (PKD) family, PKD1 and PKD2 (also known as PRKD1 and PRKD2), have emerged as crucial molecular players not only in VEGF signaling and angiogenesis but also in other important signaling pathways in the cardiovascular system (Evans and Zachary, 2011; Ha and Jin, 2009). It has also been confirmed that VEGF-mediated PKD activation in endothelial cells promotes both proliferation and migration (Hao et al., 2009; Qin et al., 2006; Wong and Jin, 2005), with PKD activation involving PKC-mediated PKD phosphorylation (Zugaza et al., 1996). Although PKD activity functionally seems to regulate migration, proliferation and apoptosis in endothelial cells, the identification of PKD substrates in endothelial cells has remained elusive (Avkiran et al., 2008).

As early as 1993, Brock and co-workers reported that the vasodilatation induced by VEGF on endothelial cells was dependent on the activity of endothelial nitric oxide synthase (eNOS), given that pre-incubation with eNOS inhibitors such as L-NMMA blocked the relaxation of coronary arteries induced by VEGF (Ku et al., 1993). Subsequently other reports showed that systemic administration of eNOS inhibitors to rabbits bearing a corneal implant blocked the migration of endothelial cells induced by VEGF (Ziche et al., 1997). In addition, pharmacological doses of VEGF are known to stimulate endothelial nitric oxide (NO) formation and reduce blood pressure in animals and humans (Hood et al., 1998; Janvier et al., 2005), whereas inhibition of VEGFR2 using a specific antibody rapidly increases blood pressure in mice (Facemire et al., 2009). In humans, VEGF is also known to maintain a normal endothelial control of vasomotor tone given that injection of a monoclonal anti-VEGF antibody into the brachial artery inactivates circulating VEGF, hence, decreasing endothelium-dependent vasodilatation within 15 min (Thijs et al., 2013).

VEGF-dependent phosphorylation of eNOS Ser 1179 and the subsequent boost in NO synthesis has been for some time considered to be mediated directly by activated Akt (protein kinase B) (Fleming, 2010; Forstermann and Sessa, 2012; Michel and Vanhoutte, 2010). However, it is becoming clear that Akt is not the only activated kinase to target eNOS Ser1179 upon VEGF treatment of endothelial cells. Expression of a dominant-negative Akt mutant in bovine endothelial cells only partially inhibits phosphorylation of eNOS Ser1179 induced by VEGF (Boo et al., 2002). Furthermore, a very evident phosphorylation on eNOS Ser1179 can still be observed in lung endothelial cells derived from Akt-knockout mice injected with VEGF (Schleicher et al., 2009).

Therefore, given that VEGF-mediated actions proceed through the activation of various protein kinases, including PKD, it is feasible that PKD could be involved in the phosphorylation-dependent activation of eNOS induced by this growth factor. Supporting this idea, Hao and co-workers showed in a recent report that an antibody recognizing phosphorylated PKD substrates prominently labeled a band of ∼135 kDa in human endothelial cells treated with VEGF. Importantly, the intensity of this band decreased partially after PKD2 silencing (Hao et al., 2009). Here, we have analyzed the possible participation of PKD in eNOS phosphorylation and activation. We show that, indeed, eNOS is a PKD substrate both in vitro and in transfected cells. In addition, we show that PKD1 and PKD2 can associate with eNOS in cells and also that silencing of these PKD isoforms abrogates eNOS-mediated wound-healing process in VEGF-treated endothelial cells. Finally, we also report that mice injected in the carotid artery with a PKD inhibitor display a clearly diminished response towards VEGF-mediated vasodilatation, a process that requires eNOS activation. In summary, we have identified eNOS as the first PKD substrate specific to endothelial cells and shed light on a new regulatory mechanism of eNOS activity with importance for the understanding of various circulatory pathologies.

Given that VEGF activates PKD isoforms in endothelial cells, we wondered whether PKD could be using eNOS as a direct substrate in the endothelium and, if this was the case, what was the possible functional outcome. Given that VEGF triggers an array of signaling pathways in cultured endothelial cells, we decided to initially perform in vitro kinase assays using recombinant proteins. We incubated recombinant bovine eNOS with purified active catalytic domain of PKD1 fused to GST (PKD1-cat active) in the presence of ATP. In order to identify phosphorylated residues in eNOS, the product of the in vitro kinase reaction was digested with trypsin and subjected to high-pressure liquid chromatography (HPLC) and peptide fragmentation by matrix-assisted laser-desorption ionization–time-of-flight (MALDI TOF)/TOF mass spectrometry (Fig. 1A). Several hundred eNOS-derived peptides were obtained, but only one significant phosphopeptide was identified. De novo sequencing of an eluted tryptic peptide with a mass of 1174.502 Da revealed that it corresponded to sequence TQpSFSLQER (residues Thr1177–Arg1185 of bovine eNOS) and the phosphorylated residue was assigned to the serine residue present at the third position (pS; b₃ in Fig. 1A). Therefore, eNOS Ser1179, which is located within the C-terminus, was the serine residue phosphorylated by PKD1 in vitro. In order to corroborate the mass spectrometry data, we used a commercially available phosphospecific antibody recognizing this phosphorylated site (against eNOS-pSer1179) and performed an in vitro kinase assay as above followed by immunoblot analysis. The eNOS-pSer1179 antibody only detected phosphorylated eNOS when pre-incubated with active catalytic domain of PKD1 in the presence of ATP (Fig. 1B).

To validate our in vitro data, we further investigated whether in bovine aortic endothelial cells (BAECs) endogenous eNOS would become phosphorylated by ectopically expressed PKD. Transfection of GFP-tagged wild-type PKD1 or PKD2 into BAECs led to a robust increase in endogenous eNOS phosphorylation on Ser1179 (Fig. 2A). Detection of a high degree of autophosphorylation on the overexpressed PKD1 and PKD2 using a phosphospecific antibody (against PKD-pSer916), which reflects the activation state of both isoforms (Matthews et al., 2000), indicated that they were active when transfected into BAECs. In addition, the increased appearance of vasodilator-stimulated phosphoprotein (VASP) phosphorylated at Ser239, a read-out for NO release (Sartoretto et al., 2009), suggested that eNOS phosphorylation on Ser1179 mediated by PKD1 and PKD2 resulted in enzymatic activation and increased NO synthesis. BAECs presented a basal level of eNOS Ser1179 phosphorylation, which indicated this particular site was already phosphorylated in endothelial cells by endogenous serine kinases, including PKD. To check PKD content in BAECs, we performed immunoblot analysis and detected bands corresponding to both PKD1 and PKD2 isoforms (Fig. 2B). Because little is known about endogenous PKD expression in endothelium, we also examined the presence of PKD in frozen tissue samples of mouse carotid arteries and its possible colocalization with eNOS by confocal immunomicroscopy. Fig. 2C shows that eNOS was present in the monolayer of endothelial cells whereas the PKD signal was detected in tunica media and also in endothelium, where it colocalized with eNOS.

Next, we continued exploring eNOS phosphorylation on Ser1179 by PKD using an eNOS construct where this serine residue had been mutated. For that purpose, we used HEK293T cells co-transfected with wild-type eNOS or the non-phosphorylatable mutant eNOS-Ser1179Ala (eNOS^SA) together with GFP-tagged wild-type PKD1 or PKD2 (PKD1wt and PKD2wt) (Fig. 3A). As shown in the figure, although there is a basal phosphorylation of wild-type eNOS on Ser1179, transfection of both PKD1 and PKD2 increases this phosphorylation approximately twofold. As expected, no phosphorylation could be detected for eNOS^SA either in the presence or absence of transfected PKD1 or PKD2. In order to further analyze the contribution of PKD activity to the phosphorylation of eNOS Ser1179, HEK293T cells were co-transfected with both wild-type eNOS and eNOS^SA constructs together with GFP-tagged PKD1 wild-type (PKDwt), or constitutively active (PKD1ca) or kinase-inactive (PKD1ki) mutants (Fig. 3B). The increase in eNOS Ser1179 phosphorylation was more significant when wild-type or constitutively active PKD1 were overexpressed, and nonexistent when the kinase-inactive mutant was used (Fig. 3B), indicating that phosphorylation of this site in eNOS is indeed regulated by PKD activity. In these cells, the Ser1179 phosphorylation signal was only found after wild-type eNOS transfection, and accordingly, no signal was detected when the eNOS^SA mutant was used.

It has been previously reported that, mechanistically, eNOS phosphorylation on Ser1179 enhances the rate of electron flux from the reductase to the oxygenase domain of the protein and reduces the relative Ca²⁺ requirement for the enzyme, thus increasing NO synthesis (McCabe et al., 2000). In order to analyze the effect of PKD phosphorylation on eNOS activity, COS-7 cells were transfected with wild-type eNOS or its non-phosphorylatable Ser1179Ala mutant in the presence or absence of PKD1wt, PKD1ca or PKD2wt and eNOS activity was measured as NO release with the fluorescent probe 4,5-diaminofluorescein diacetate (DAF2-DA) (Fig. 3C) and also as [¹⁴C]L-citrulline formation when [¹⁴C]L-arginine was used as a substrate (Fig. 3D). As a positive control, we also transfected COS-7 cells with inducible nitric oxide synthase (iNOS), given that this isoform releases large amounts of NO and its activity is not dependent on the intracellular Ca²⁺ concentrations. Interestingly, the non-phosphorylatable eNOS mutant (eNOS^SA), when transfected alone or together with PKD1wt, was unable to release detectable NO levels in the absence of added Ca²⁺ ionophores. In contrast, the NO-releasing activity of transfected wild-type eNOS was low but increased significantly when co-transfected with PKD1wt. This effect was slightly but significantly further enhanced by activating PKD1 with phorbol 12, 13-dibutyrate (PDBu), and was blocked after incubating the cells with the eNOS inhibitor L-NAME (Fig. 3C). Likewise, the conversion of [¹⁴C]L-arginine into [¹⁴C]L-citrulline in lysates obtained from COS-7 cells transfected with wild-type eNOS was augmented significantly when co-transfected with PKD1wt, PKD2wt or PKD1ca. Finally, even when wild-type eNOS was co-transfected with PKD1ca the amount of [¹⁴C]L-citrulline formed was significantly diminished when the eNOS inhibitor L-NAME was present (Fig. 3D). Thus, our data indicate not only that PKD1 phosphorylation of eNOS activates NO synthesis but also that phosphorylation displaces the Ca²⁺/calmodulin-binding curve towards lower Ca²⁺ concentrations. Hence, PKD1 phosphorylation of eNOS activates the enzyme and further confirms the proposal that phosphorylation of Ser1179 mediates the Ca²⁺-independent activation of eNOS (Dimmeler et al., 1999; Fulton et al., 1999; McCabe et al., 2000).

We next analyzed whether PKD1 or PKD2 were able to associate with eNOS. For that purpose, we transfected HEK293T cells with full-length wild-type eNOS or a mutant that lacked the C-terminus (eNOSΔ27) including the phosphorylation site, together with wild-type GFP–PKD1 and GFP–PKD2 (PKD1 or PKD2). At 2 days after transfection PKD was immunoprecipitated from total lysates using anti-GFP antibody and the presence of eNOS in the immunocomplexes was assessed by detecting the signal for eNOS by immunoblotting (Fig. 4A). Our results showed that wild-type eNOS associated with both PKD isoforms, and that the eNOS phosphorylation site was not involved in the association because an eNOS mutant lacking the C-terminus still associated with PKD1 and PKD2. Next, we examined which PKD1 domains could be mediating this association and whether PKD activity could affect it using GFP–PKD1wt, the constitutively active mutant (PKD1ca), and two mutants in which the cysteine-rich domain (CRD) or the pleckstrin homology domain (PH) had been deleted (PKD1^ΔCRD and PKD1^ΔPH, respectively) (Fig. 4B). These constructs were co-transfected with wild-type eNOS in the presence or absence of PDBu treatment to activate the kinase. Then, lysates were immunoprecipitated with an anti-GFP antibody. As shown in the figure, GFP by itself could not associate with eNOS (lanes 1 and 2), whereas all the PKD constructs tested were able to associate with full-length eNOS independently of PDBu treatment. Hence, the association between the eNOS and PKD1 is not mediated by the CRD or the PH domains and does not require the activation of the kinase or the presence of the phosphorylatable eNOS Ser1179 residue.

In order to complement our studies, we examined which region of eNOS could be mediating this association. HEK293T cells were co-transfected with PKD1wt, or the empty vector, together with different eNOS constructs: full-length wild-type eNOS, a C-terminus truncated mutant (eNOS Δ27), the NADPH-binding domain (NADPH, amino acids 988–1205) or the heme-oxygenase domain (HEME, amino acids 1–521) (Fig. 4C). At 2 days after transfection and before lysis, cells were left untreated or treated with PDBu for 15 min to activate PKD1. Then, lysates were immunoprecipitated with an anti-eNOS polyclonal antibody and PKD and eNOS were detected. PKD activation by PDBu, although increased eNOS phosphorylation, did not potentiate the formation of eNOS–PKD1 complexes. PKD co-immunoprecipitation was observed with full-length eNOS and with its C-terminus mutant Δ27 but was hardly detectable with the NADPH-binding domains of eNOS and was undetectable with the isolated heme-oxygenase domain. Thus, our results indicate that eNOS associates to PKD1 through residues located between the CaM-binding sequence and the FAD- or FMN-binding subdomains.

The fact that VEGF treatment of BAECs leads to the activation of PKD (Wong and Jin, 2005) led us to explore VEGF-induced eNOS phosphorylation. We examined the activation of endothelial PKD through the addition of PDBu for 15 min (Fig. 5A) or VEGF for 30 min (Fig. 5B) and found a substantial increase in PKD Ser916 phosphorylation, of approximately twofold, with both stimuli. In identical conditions, eNOS Ser1179 phosphorylation is significantly augmented (Fig. 5A,B), hence, indicating that endogenous PKD isoforms might be responsible, at least in part, for the observed eNOS phosphorylation at this position. Furthermore, the vasodilator peptide bradykinin also rapidly induced PKD activation with the concomitant eNOS Ser1179 phosphorylation (supplementary material Fig. S1). Importantly, the timecourse of VEGF-induced eNOS phosphorylation at Ser1179 showed a fast and increasing effect up to the 30 min assayed (Fig. 5C). Thus, these results indicate a correlation between endogenous PKD activation in endothelial cells and eNOS phosphorylation on Ser1179.

It has been described that in non-stimulated endothelial cells, eNOS targets to specific intracellular domains, including the Golgi and cholesterol- and sphingolipid-rich microdomains of the plasma membrane (caveolae), and that upon stimulation its localization changes from membranous compartments to the cytosol, with this translocation being important for activation and inactivation of the enzyme (Forstermann and Sessa, 2012; Michel and Vanhoutte, 2010). By contrast, PKD is mainly cytosolic, showing some association to intracellular compartments, including the Golgi, in unstimulated cells, but it can rapidly translocate to different subcellular compartments depending on the cellular context and stimulation conditions (Rozengurt, 2011). Here, we analyzed the subcellular localization of endogenous eNOS in BAECs transfected with GFP-tagged PKD1 by confocal microscopy. Transfected BAECs were left untreated or stimulated with VEGF for 15 min, then fixed and immuno stained using an anti-eNOS antibody. As shown in Fig. 5D, in resting cells both eNOS and PKD1 were excluded from the cell nucleus and enriched in endomembranes and perinuclear areas of the cell (very likely the Golgi) where a high degree of colocalization could be observed. After VEGF stimulation, both proteins suffered a bulk translocation, with their subcellular distribution becoming more cytoplasmic and less condensed in perinuclear areas. No apparent plasma membrane localization of eNOS could be detected in agreement with the previously reported cytoplasmic immunolocalization of activated Ser1179 phosphorylated eNOS (Fulton et al., 2002). In the case of PKD1, some translocation to certain plasma membrane regions could also be observed. This subcellular redistribution is in agreement with previous studies in human umbilical vein endothelial cells (HUVECs) that reported that, whereas in the absence of VEGF stimulation PKD1 was almost completely absent from the plasma membrane, a short stimulation of endothelial cells with VEGF induced PKD1 translocation to plasma membrane protrusions, likely lamellipodia (Di Blasio et al., 2010). Similar results were obtained when GFP–PKD2 was transfected (data not shown).

We next examined the role of endogenous PKD in BAEC migration, an eNOS-dependent process that plays key roles in angiogenesis (Borniquel et al., 2010; López-Rivera et al., 2005). For that purpose, we performed a wound healing assay by stimulating cells with VEGF or PDBu, in the presence of eNOS, PKD or Akt inhibitors (Fig. 6). Cells were monitored for 24 h up to wound closure. As shown in Fig. 6A, treatment with the eNOS inhibitor L-NAME blocked migration induced by the pro-angiogenic factor VEGF and also by PDBu, two compounds that, as demonstrated in Fig. 5A,B, activate endogenous PKD1 and PKD2, and increase phosphorylation of eNOS on Ser1179. Pretreatment of BAECs with the PKD inhibitor Gö6976 and with Akt inhibitors significantly decreased the sealing of the wound in 24 h (Fig. 6B; supplementary material Fig. S2). These results indicate that PKD activity is required for the migration of endothelial cells in vitro, and also that PKD1 and PKD2, and, to a lesser extent Akt are involved in this process.

Next, in order to analyze the role of PKD in eNOS Ser1179 phosphorylation, BAECs were incubated with VEGF in the presence or absence of Gö6976 (Fig. 7A; supplementary material Fig. S3). Inhibition of endogenous PKD1 and PKD2 significantly diminished the amount of eNOS-pSer1179 detected in western blots, hence underpinning the role of this kinase in eNOS activation in endothelial cells. However, PKD1 and PKD2 might activate eNOS and stimulate angiogenic processes through short- and long-term mechanisms. Although it has been reported that PKD1 is involved in angiogenesis (Qin et al., 2006; Wong and Jin, 2005), recent results point towards a prominent role of PKD2 in proliferation and migration of endothelial cells (Hao et al., 2009). In order to determine the role of each individual isoform in eNOS activation in BAECs we performed gene knockdown using lentivirus that produce specific short hairpin RNAs (shRNAs) targeted against either PKD1 or PKD2. We previously checked that those validated shRNAs were also effective for gene silencing of bovine PKD1 and PKD2 (Fig. 7B). Importantly, the knockdown of PKD1 or PKD2 reduced the VEGF-dependent eNOS phosphorylation on Ser1179 in BAECs (supplementary material Fig. S4A). When wound healing assays were performed the PKD inhibitor Gö6976 and, to a greater extent, the specific lentivirus-mediated silencing of either PKD1 or PKD2 had a severe effect on the BAEC wound healing process (Fig. 7C). In addition, eNOS inhibition with L-NAME in PKD1- or PKD2-slienced BAECs revealed the need for NO in the wound healing process (supplementary material Fig. S4B). Hence, our results indicate both PKD isoforms are involved in the VEGF-dependent eNOS activation and migration of BAECs in a wound-healing assay. This is in contrast with previous reports that suggested that PKD1 silencing had only a minor effect on HUVEC migration, whereas PKD2 was the key player in this process (Hao et al., 2009).

It is generally accepted that the eNOS-generated NO stimulates cyclic guanosine monophosphate (cGMP) synthesis by soluble guanylate cyclase which, in turn, leads to relaxation of vascular smooth muscle (Gruetter et al., 1979). In order to study the role of PKD in vascular function, endothelium-dependent vasodilatation was analyzed. For that purpose, changes in vessel diameter in the right carotid artery of mice were measured by high-resolution ultrasound imaging. We have previously shown that PKD and eNOS colocalize in the endothelial monolayer in mouse carotid (Fig. 2C). Hence, we tested the response to VEGF, a molecule with well-characterized vasodilatatory properties, in the presence and absence of PKD inhibition. VEGF injection in the mouse tail caused a rapid vasodilatation of the carotid artery with the maximal measured diameter increase (over 1.8-fold) observed after 2 min (Fig. 8A). Conversely, PKD inhibition using Gö6976 for 10 min followed by the subsequent injection of VEGF significantly reduced vasodilatation, causing the carotid artery to have a maximum diameter of 0.18 mm at 4 min post-injection (Fig. 8B). In addition, injection of the PKD inhibitor Gö6976 per se had a clear vasoconstrictor effect on mice (Fig. 8B,C). To exclude potential intimal smooth muscle abnormalities, we evaluated endothelium-independent vasorelaxation by infusing 10⁻⁷ mol/l sodium nitroprusside, which led to normal vessel dilatation, as expected (data not shown). These results indicate that the NO signaling pathway is compromised when PKD is inhibited, which is reflected in the vasoconstrictor effect of PKD inhibition and the partial loss of the VEGF-induced vasodilatatory response. Hence, PKD-mediated phosphorylation of eNOS might be responsible, at least in part, for the vasodilatory action of VEGF.

eNOS-derived NO, one of the key players in vascular homeostasis, not only modulates blood pressure and vascular tone but also displays multiple antiatherogenic roles including antithrombotic, antiproliferative, anti-inflammatory and antioxidant effects (Forstermann and Sessa, 2010). The enzymatic levels of eNOS in the endothelium are in part regulated through phosphorylation on serine, threonine and tyrosine residues (Fleming, 2010; Forstermann and Sessa, 2012). In particular, eNOS phosphorylation on Ser1179 is without doubt the best characterized post-translational modification of the enzyme, stimulating the flux of electrons within the reductase domain and leading to enzymatic activation. Although an atomic structure of the full-length reductase domain of eNOS is not available, comparison with the neuronal NOS (nNOS) crystal structure (Garcin et al., 2004) suggests that the regulatory eNOS C-terminal tail very likely adopts a helical conformation, which probably fits within a negatively charged groove across the FAD–FMN interface, shielding the flavins from solvent. Hence, the activation observed upon eNOS phosphorylation on Ser1179 very likely reflects the displacement of this α-helix and the concomitant increased electron transfer within the reductase domain. In addition, eNOS Ser1179 phosphorylation (very frequently accompanied by coordinated dephosphorylation of Thr495) diminishes the Ca²⁺ sensitivity of the enzyme (Fulton et al., 1999; McCabe et al., 2000; Schneider et al., 2003).

The use of commercially available anti-pSer1179 antibodies for eNOS has revealed that in unstimulated cultured endothelial cells, Ser1179 is only marginally phosphorylated. Upon the application of fluid shear stress (Dimmeler et al., 1999; Gallis et al., 1999), acute mechanical stretch (Hu et al., 2013) or UV light (Park et al., 2011), eNOS becomes rapidly phosphorylated on Ser1179. Likewise, treatment of endothelial monolayers with VEGF (Fulton et al., 1999; Michell et al., 1999), insulin (Salt et al., 2003), estrogens (Haynes et al., 2000), endothelin-1 (Liu et al., 2003) or bradykinin (Fleming et al., 2001; Schneider et al., 2003) leads to eNOS Ser1179 phosphorylation. To date, it has been shown that at least six protein kinases [Akt, AMPK, PKA, cGK-I/PKG, Chk1 and Ca²⁺/calmodulin-dependent protein kinase II (CaMKII)] are involved in the direct or indirect phosphorylation of eNOS on Ser1179, although the ones involved in each particular physiological response vary with the stimuli applied. For example, whereas shear stress elicits Ser1179 phosphorylation mediated by PKA (Boo et al., 2002), treatment with insulin (Montagnani et al., 2001), estrogens (Hisamoto et al., 2001), and sphingosine-1-phosphate (Morales-Ruiz et al., 2001) mainly leads to phosphorylation of eNOS through Akt. By contrast, eNOS phosphorylation on Ser1179 due to the treatment of endothelial cells with bradykinin, Ca²⁺ ionophore or thapsigargin is considered to be mediated through CaMKII activation (Fleming et al., 2001; Schneider et al., 2003).

It has been historically accepted that VEGF-mediated eNOS Ser1179 phosphorylation and subsequent activation was selectively due to Akt activity. In fact, we and others have reported that activated Akt phosphorylates eNOS on Ser1179 (Dimmeler et al., 1999; Fulton et al., 1999; Michell et al., 1999). Moreover, VEGF treatment of endothelial cells results in an activation of Akt (Fontana et al., 2002; Fulton et al., 1999; Michell et al., 1999). However, expression of a dominant-negative Akt mutant in BAECs only partially inhibits phosphorylation of eNOS Ser1179 induced by VEGF (Boo et al., 2002). The presence of another kinase participating in this step can be inferred even more convincingly from the fact that isolated endothelial cells from Akt-knockout mice injected with VEGF through the jugular vein still show a very prominent Ser1179 phosphorylated eNOS band in immunoblots (Schleicher et al., 2009). Here, we identify PKD1 and PKD2 as new kinases that are also able to phosphorylate eNOS on Ser1179. Given that the effects of VEGF on PKD activation are well established (Hao et al., 2009; Qin et al., 2006; Wong and Jin, 2005), it is feasible that PKD could be responsible for VEGF-mediated effects on eNOS activation that are independent of Akt.

eNOS forms a cellular complex with the majority of the protein kinases known to act on eNOS Ser1179, such as Akt (Michell et al., 1999), CaMKII (Ching et al., 2011), Chk1 (Park et al., 2011) and AMPK (Hess et al., 2009). We also see a clear association between PKD1 and PKD2 with eNOS and show that the reductase domain of eNOS is involved in this process. However, eNOS phosphorylation is not required for its association with either of the PKD isoforms because eNOS mutants with a deleted C-terminus are still able to form a complex. From our studies, we cannot conclude whether eNOS and PKD interact directly or need additional proteins. In the case of Akt, activation of the kinase by VEGF stimulation and subsequent phosphorylation of eNOS on Ser1179 is required for the formation of eNOS–Hsp90 heterocomplexes (Brouet et al., 2001; Fontana et al., 2002). Although the physiological meaning of the association of eNOS to the kinases that use it as substrate is unclear, it has been suggested that, at least in the case of Akt and Chk1, that chaperone Hsp90 might perhaps function as a scaffold for the recruitment of eNOS and the kinase leading to eNOS activation (Fontana et al., 2002; Park et al., 2011). Interestingly, this chaperone is also required for VEGFR2 signaling to eNOS (Duval et al., 2007). Given the role of Hsp90 in VEGF and eNOS crosstalk, this chaperone could also be mediating the association between PKD1–PKD2 and eNOS in endothelial cells.

Our results clearly show that eNOS is a new PKD substrate. The consensus sequence of PKD protein substrates presents a hydrophobic residue such as Leu or Ile at position −5, together with a basic residue such as Lys or Arg at position −3 (Döppler et al., 2005; Hutti et al., 2004; Nishikawa et al., 1997). Comparison of this sequence with that of eNOS (supplementary material Table S1) shows that the eNOS phosphorylation site partially fails to fully meet this requirement because a basic Arg residue is present at the −5 position. Interestingly, the autophosphorylation motif located at the C-terminus of PKD1 and PKD2 does not fit within the consensus phosphorylation sequence either because an acidic Glu residue is present at the −3 position. Various amino acids present within the kinase recognition site of eNOS are also present in other reported PKD substrates, such as the Thr present at the −2 position also found in PI4KIIIb (Hausser et al., 2005) and HDAC5 (Vega et al., 2004), Gln at the −1 position as in HDAC5 (Vega et al., 2004) or Phe at +1 position as in Kidins220 (Iglesias et al., 2000).

Growing evidence indicates that the functional activities of PKD family in the circulatory system and in vascular biology are of utter importance, particularly those regulating proliferation, migration and tubulation of endothelial cells that are crucial for angiogenesis (Evans and Zachary, 2011). In this regard, VEGF-dependent PKD activation is known to be required for endothelial cell migration (Hao et al., 2009; Shin et al., 2012), proliferation (Hao et al., 2009) and tumor angiogenesis (Azoitei et al., 2010). Our study shows that VEGF treatment activates both PKD1 and PKD2 in BAECs, a result that is in accordance with studies performed in other endothelial cells (Evans et al., 2008; Evans and Zachary, 2011; Ha et al., 2008; Ha and Jin, 2009; Qin et al., 2006). We also find that pharmacological inhibition of PKD or specific silencing of PKD1 or PKD2 hampers BAEC migration in wound healing, a process known to be dependent on eNOS.

The role of endothelial PKD isoforms in vascularization of tumors is becoming more evident, given that recent data indicate that the abnormal proliferation of HUVECs observed in the presence of dimethylhydrazine is dependent on PKD1. Downregulation of this kinase using siRNA inhibited abnormal proliferation and formation of vascular neoplasms (Nam et al., 2012). PKD is known to regulate cell migration through different mechanisms, with the one associated with its role in the regulation of actin dynamics very well documented (Olayioye et al., 2013). Indeed, many PKD substrates regulate cellular migration through this mechanism, resulting in their phosphorylation and inhibition or enhancement of migration (Olayioye et al., 2013).

Here, we presented an additional mechanism by which PKD can control endothelial cell migration in response to VEGF, by having a direct action on eNOS phosphorylation and activation. Whether all these mechanisms are interconnected and how important the contribution of each of them is to endothelial function remains to be determined. In summary, our findings reinforce the pivotal role played by PKD in endothelial cell biology, adding to the list of PKD functions its ability to directly phosphorylate and activate eNOS, a crucial enzyme for endothelium physiology.

VEGF, phorbol-12,13-dibutyrate (PDBu), Gö69761, Tricibirine, Akt 1/2 kinase inhibitor, L-N^Gnitroarginine methyl ester (L-NAME), 4,5-diaminofluorescein diacetate (DAF2-DA), protein-A–Sepharose, 2′,5′-ADP-Sepharose, adenosine 2′(3′)-monophosphate mixed isomers, (6R)-5,6,7,8-tetrahydrobiopterin dihydrochloride (BH₄) and ATP were from Sigma (St Louis, MO). Ni-NTA resin was from Qiagen (Chatsworth, CA) and glutathione–Sepharose was from Amersham (Buckinghamshire, UK). L-arginine was purchased from Calbiochem (Merck Millipore, Darmstadt, Germany). [¹⁴C]L-arginine was from GE Healthcare (Uppsala, Sweden). Phosphatase inhibitors PhosSTOP were from Roche. Mouse monoclonal antibodies anti-eNOS and anti-eNOS-phospho-Ser¹¹⁷⁹ (Ser¹¹⁷⁷ in human eNOS and Ser¹¹⁷⁹ in bovine eNOS) were from BD Transduction laboratories. Mouse monoclonal antibodies recognizing total VASP and PKD2, and rabbit polyclonal antibodies anti-VASP-phospho-Ser239, anti-PKCμ (recognizing total PKD1/2) and anti-PKCμ-phospho-Ser⁹¹⁰ were from Santa Cruz Biotechnology (Santa Cruz, CA). We produced antibodies against eNOS and GFP by immunizing rabbits with purified bovine eNOS or GFP. Anti-tubulin monoclonal antibody, horseradish-peroxidase-conjugated anti-rabbit and anti-mouse IgG secondary antibodies and the fluorescent secondary antibodies labeled with Cy2 or Cy3 were from Sigma.

The molecular cloning, recombinant expression and purification of eNOS were as previously described (Rodríguez-Crespo et al., 1996; Rodríguez-Crespo et al., 1997; Rodríguez-Crespo et al., 1999).

The C-terminal region of PKD1 (Ser558–Leu918; PKD1cat) containing the full-length catalytic domain was amplified from pBS-PKD1. The PCR product was cloned in pDONR201 by a recombination reaction with BP clonase (GATEWAY system, Invitrogen Life Technologies; Carlsbad, CA), to generate the construct pENTR-PKD1cat. After automated sequencing, PKD1cat was subcloned in pDEST15 using LR clonase. This vector for prokaryotic expression generates PKD1cat fused to glutathione S-transferase (GST; GST-PKD1cat) of approximate molecular mass of 65 kDa, which was purified following standard methods and stored at −20°C. This protein is constitutively active because it lacks the regulatory autoinhibitory domain.

We followed the procedures described previously (Navarro- Lérida et al., 2006; Navarro-Lérida et al., 2007).

We followed the procedure described previously (Navarro- Lérida et al., 2004). Briefly, COS-7 cells were transfected with the desired construct and, at 48 h after transfection, cells were harvested, sonicated for 5 s and separation of [¹⁴C]L-arginine and [¹⁴C]L-citrulline was performed in a Dowex resin. We used ∼1 µCi of [¹⁴C]L-arginine per condition.

We followed the procedures already published by our group (Herranz et al., 2012). Flow dilatation was measured before and after infusion of VEGF (30 µl of a 10 µg/ml solution in water), the PKD inhibitor Gö6976 (30 µl of a 225 µg/ml solution in DMSO), or both substances. Readings were taken immediately after injection (0 min) and after 1, 2, 4 and 10 min. To test endothelium-independent vasorelaxation, sodium nitroprusside was infused at 10⁻⁷ mol/l (30 µl) and recordings taken as above. Each condition was tested in at least four animals. Image analyses were performed off-line from recorded loops using the automated system software provided by the manufacturer (Visualsonics). The study was conducted following the guidelines of the Spanish Animal Care and Use Committee, according to the guidelines for ethical care of experimental animals of the European Union (2010/63/EU). This study conforms to the Guide for the Care and Use of Laboratory Animals published by the U.S. National Institute of Health (NIH Publication no. 85–23, revised 1996).

Phosphorylation was determined performing an in vitro kinase assay. Briefly, purified full-length eNOS and GST–PKD1cat were mixed in kinase buffer (10 mM Tris-HCl pH 7.5, 5 mM MgCl₂ and 1 mM dithiothreitol), and subjected to an in vitro kinase assay for 1 h at room temperature in the presence or absence of 100 µM final concentration of ATP. Samples were analyzed by mass spectrometry or immunoblotting.

In vitro phosphorylated eNOS was digested with trypsin and analyzed by HPLC followed by MALDI TOF/TOF and peptide fragmentation and de novo sequencing in the Proteomic Studies Unit (Unidad de Proteómica; Facultad de Farmacia Parque Científico de Madrid, UCM, Madrid, Spain).

BAEC migration was determined using a scratch wound assay as previously reported (Borniquel et al., 2010; López-Rivera et al., 2005).

Validated shRNA cloned in pLKO.1 vector, against human PKD1 (two constructs: TRCN0000195251 and TRCN0000002124), named shPKD1-A and B, respectively, and against human PKD2 (TRCN0000001948 and TRCN0000001950) named shPKD2-A′ and B′ were purchased from Sigma (MISSION® shRNA lentiviral plasmids). Lentiviral production was performed in HEK293T cells following standard procedures.

We are grateful to María Luisa Hernáez Sánchez (UCM, Madrid, Spain) for proteomic studies and Iván Ventoso (CBM, Madrid, Spain) for help with the lentivirus construction and transduction.