PKCζ mediates disturbed flow-induced endothelial apoptosis via p53 SUMOylation

A hallmark of atherogenesis is focalized endothelial dysfunction, which includes altered vasoregulation, activation of inflammatory processes, and compromised barrier function caused by endothelial cell (EC) apoptosis (Hu et al., 1999; Song et al., 2008). Manifestations of dysfunctional ECs are readily observed in certain areas of the arterial tree, where disturbed flow (d-flow), hence reduced time-averaged shear stress, develops (Traub and Berk, 1998; Won et al., 2007). Steady laminar flow (s-flow) promotes release of factors from ECs that inhibit coagulation, leukocyte diapedesis, and smooth muscle cell proliferation while simultaneously promoting EC survival (Garin et al., 2007; Reinhart-King et al., 2008). Conversely, d-flow alters the profile of secreted factors and EC surface molecule expression that favors the opposite effects, thereby contributing to the development of atherosclerosis (Traub and Berk, 1998). We have previously reported the critical role of PKCζ activation in EC apoptosis (Garin et al., 2007). Importantly, s-flow interferes with PKCζ signaling, thereby down-regulating the proapoptotic effect of the kinase. In contrast, unique atheroprone signals elicited by d-flow remain unclear.

Acting as a sensor for DNA damage, the transcription factor p53 is a key regulator of the life or death of a cell, depending on whether or not the cell can cope with the damage and repair it. Although the most-studied function of p53 is its role as a transcription factor that increases the expression of proapoptotic genes (Murray-Zmijewski et al., 2008), recent studies have revealed its nontranscriptional proapoptotic activities. Cytosolic p53 directly interacts with the B cell lymphoma/leukemia-2 (Bcl-2) family member proteins Bcl-xL and Bcl-2 and antagonizes their antiapoptotic function by stabilizing the outer mitochondrial membrane (Mihara et al., 2003). Interestingly, antiapoptotic effects of p53 have also been reported (Mercer et al., 2005; Garner and Raj, 2008). It is particularly interesting that p53 inhibits apoptosis of vascular smooth muscle cells and protects against atherosclerosis formation (van Vlijmen et al., 2001; Mercer et al., 2005). However, it remains unclear how these p53 functions, especially its apoptotic effect in ECs, are regulated by flow.

SUMOylation is a posttranslational modification consisting of covalent conjugation of ubiquitin-like proteins called SUMO (small ubiquitin-like modifier) to target proteins (Hilgarth et al., 2004). It is a reversible modification that affects target protein functions, such as subcellular localization, protein partnering, DNA binding, and regulation of transcription factors (Hilgarth et al., 2004). Carter et al. (2007) proposed an interesting model for p53 nuclear export and stabilization. Masking of the C-terminal nuclear export signal (NES) results in nuclear localization of unmodified p53. A low level of ubiquitination by MDM2 exposes the NES, promoting p53 to interact with protein inhibitor of activated STATy (PIASy) and further modification by SUMOylation causes p53 nuclear export (Carter et al., 2007). These data suggest an important role of SUMOylation in p53 nuclear export. In this study, we investigate the role of PKCζ and PIASy on d-flow–mediated endothelial p53 nuclear export and apoptosis, which may contribute to EC dysfunction and subsequent atherosclerotic plaque formation. Furthermore, we show that the cytoplasmic expression of p53 and EC apoptosis are increased in ECs in the area of d-flow in vivo.

We first verified the potential role of shear stress in PKCζ activation in cultured ECs using a cone and plate type of flow apparatuses as described previously (Reinhart-King et al., 2008). To generate d-flow, we used cones with radial grooves that were 1-mm deep. Fig. 1 (A and B) shows tracks of fluorescent beads suspended in culture media when grooved and nongrooved cones were rotated at the same speed. Although the nongrooved cone created straight unidirectional tracks indicating s-flow, tracks made by the grooved cone were short and not oriented in the same direction, indicating nonlaminar movement of the media in the dish. Using these cones, we determined the effect of s- and d-flow on PKCζ activation in human umbilical vein ECs (HUVECs). D-flow increased PKCζ phosphorylation at both Thr410 and Thr560 after 10 min of stimulation (Fig. 1 C), whereas s-flow failed to activate PKCζ, although ERK5 was activated by s-flow as we reported previously (not depicted; Woo et al., 2008). PKCζ phosphorylation at Thr560 was sustained longer than phosphorylation at Thr410. These data obtained from our in vitro system are consistent with in vivo data, which showed increased PKCζ activity in d-flow area compared with s-flow area in porcine arteries (Magid and Davies, 2005), supporting the physiological relevance of our in vitro system.

Reactive oxygen species has been shown to react with NO and generate ONOO⁻, reducing NO availability and accelerating endothelial dysfunction and development of atherosclerosis (Ponnuswamy et al., 2009). Furthermore, it has been reported that ONOO⁻ was increased by d-flow both in vitro and in vivo (Hsiai et al., 2007). To test whether some effects of d-flow could be mediated by ONOO⁻, we treated ECs with ONOO⁻ and examined PKCζ activation. Indeed, ONOO⁻ also increased PKCζ phosphorylation at both Thr410 and Thr560 within 5 min of stimulation, reaching its maximum by 10 min (Fig. 1 D), suggesting that ONOO⁻-mediated PKCζ kinase activation is similar to d-flow stimulation.

We examined whether PKCζ activation was also required for d-flow and ONOO⁻-mediated EC apoptosis using two different approaches. PKCζ activity was down-regulated by inhibiting its expression by siRNA and also by expressing an adenoviral dominant-negative (DN) kinase-dead form of PKCζ (Ad-DN-PKCζ). Although d-flow and ONOO⁻ treatment increased apoptosis in control cells, a significant reduction in the number of TUNEL-positive cells was noted in PKCζ-depleted HUVECs (Fig. 2, A and B, left). When HUVECs were infected with Ad-DN-PKCζ (mutating Lys281to Met), the ONOO⁻-induced increase in TUNEL-positive cells and the expression of cleaved caspase 3 fragments (17/19 kD) were reduced compared with control cells (Fig. 2, B [center] and C). These data suggest the critical role of PKCζ kinase activity in d-flow and ONOO⁻-mediated EC apoptosis.

p53 promotes transcription of several proapoptotic genes (Mihara et al., 2003; Mercer et al., 2005; Bischof et al., 2006; Garner et al., 2007). To test whether it plays a role in the ONOO⁻-induced EC apoptosis, we first examined the effect of ONOO⁻ on p53 transcriptional activity in HUVECs that were overexpressing p53 with or without coexpression of a constitutively active PKCζ (catalytic domain of PKCζ [CATζ]; Garin et al., 2007). To our surprise, ONOO⁻ inhibited p53 transcriptional activity in control cells and also in those overexpressing p53 (Fig. 3 A). Consistent with these results, cells transfected with CATζ also showed a decreased p53 transcriptional activity (Fig. 3 B). These results strongly suggest that ONOO⁻-mediated EC apoptosis is not caused by increased p53 transcriptional activation of proapoptotic genes.

It has recently been reported that cytosolic p53 directly interacts with Bcl-2 and antagonizes its antiapoptotic stabilization of the outer mitochondrial membrane (Mihara et al., 2003; Bischof et al., 2006). Because this interaction takes place in the cytoplasm, p53 may be exported from the nucleus when PKCζ is activated by ONOO⁻. To verify this possibility, HUVECs were stimulated with ONOO⁻ for 4 h, and p53 localization was analyzed by immunostaining (Fig. 3 C). Interestingly, the anti-p53 staining contrast between the nucleus and the cytoplasm in ONOO⁻-treated cells decreased considerably, and some punctuate staining appeared in the cytoplasm (Fig. 3 C). No fluorescent signal was detected when only the secondary antibodies were used (unpublished data). ONOO⁻-induced anti-p53 staining in the cytoplasm was reduced in Ad-DN-PKCζ–transduced cells compared with adenovirus of LacZ (Ad-LacZ)–transduced cells, suggesting that ONOO⁻ induced p53 export from the nucleus in a PKCζ-dependent manner.

Next, to examine whether PKCζ is involved in ONOO⁻-mediated p53–Bcl-2 association, we expressed Ad-DN-PKCζ in HUVECs, treated them with ONOO⁻, and performed a coimmunoprecipitation assay using anti-p53 followed by immunoblotting with anti–Bcl-2. Bcl-2 coimmunoprecipitated with p53 when control cells (transduced with Ad-LacZ) were treated with ONOO⁻ for 10 min or longer, but this coimmunoprecipitation was strongly inhibited by Ad-DN-PKCζ expression (Fig. 3 D and Fig. S1, A and C). Nonimmune control IgG failed to bring down p53 and Bcl-2 (Fig. S1 A). Collectively, these results suggest a role of PKCζ in ONOO⁻-mediated p53 nuclear export and subsequent p53–Bcl-2 interaction.

Because p53 nuclear export is positively regulated by SUMOylation (Bischof et al., 2006) and also because PKCζ regulates p53 nuclear export (present study), p53 SUMOylation may be controlled by PKCζ. To test this, we cotransfected HeLa cells with Flag-tagged p53 (Flag-p53), HA-tagged SUMO3 (HA-SUMO), and constitutively active PKCζ (CATζ) and determined p53 SUMOylation by Western blotting using anti-SUMO2/3. Several bands presumably representing mono- and poly-SUMOylated p53 with apparent molecular masses of 68, 74, 82, 130, 185, and 200 kD were noted (Fig. 4 A). These bands were also labeled by anti-Flag, indicating that they are SUMOylated Flag-p53 (Fig. 4 A, middle). p53 SUMOylation was increased in CATζ-transfected cells, suggesting the role of PKCζ activity on p53 SUMOylation. In addition, these sumoylated bands were diminished in the cells transfected with the p53-K386R SUMOylation mutant (Kwek et al., 2001), further supporting that these high mass bands were SUMOylated p53.

An important role of PIASy as a SUMO E3 ligase in p53 SUMOylation has been reported (Bischof et al., 2006). To confirm that the 4–5 bands are SUMOylated p53, HUVECs with or without CATζ expression were transfected with PIASy siRNA. This should inhibit p53 SUMOylation, and indeed, PIASy depletion eliminated the appearance of these bands (Fig. 4 B). Coimmunoprecipitation using nonimmune IgG yielded no SUMOylated bands (Fig. S1 A). To see whether endogenous p53 is SUMOylated, we specifically inhibited p53 expression using siRNA and found significantly reduced SUMOylation levels in the 68-, 74-, 82-, 130-, 160-, and 185-kD bands (Fig. 4 C). Endogenous p53 SUMOylation bands were less discrete, likely caused by the combined effects of poly-SUMOylation and ubiquitination as previously described (Carter et al., 2007). SENP2 (sentrin-specific protease 2) is a de-SUMOylation enzyme that is important for both processing new SUMO proteins for conjugation as well as deconjugating SUMO from SUMOylated proteins (Cheng et al., 2004; Yeh, 2009; Witty et al., 2010). To verify the identity of the endogenous p53 SUMOylation band, we also transduced ECs with Ad-LacZ control or adenoviral SENP2 (Ad-SENP2), stimulated them with d-flow, and evaluated changes in these multiple bands (Fig. 4 C). Transduction of Ad-SENP2 completely reduced d-flow–induced p53 SUMOylation. These results together support that the 68-, 74-, 82-, 130-, 160-, 185-, and 200-kD bands are SUMOylated p53.

When HUVECs were treated with various concentrations of ONOO⁻ for 10 min, p53 SUMOylation increased in a dose-dependent manner (Fig. S1, A and B). We also found that d-flow significantly increased p53 SUMOylation after 2 h of stimulation (Fig. 4 D and Fig. S1 D). Next, we investigated whether PKCζ played a role in p53 SUMOylation by ONOO⁻ and d-flow. This SUMOylation was inhibited in cells expressing Ad-DN-PKCζ (Fig. 4 D and Fig. S1 C), suggesting a role of PKCζ in p53 SUMOylation in ONOO⁻ and d-flow–stimulated cells. In addition, Bcl-2 coimmunoprecipitated with p53 when control cells (transduced with Ad-LacZ) were treated with d-flow, but this coimmunoprecipitation was strongly inhibited by Ad-DN-PKCζ transduction (Fig. 4 D), which was also observed in ONOO⁻-treated cells (Fig. 3 D).

To test whether PKCζ plays a unique role on p53 SUMOylation and binding between p53 and Bcl-2, we transfected ECs with PKCζ siRNA and studied its effect on d-flow–mediated p53 SUMOylation and p53–Bcl-2 binding. As shown in Fig. 4 E, we found that PKCζ siRNA specifically inhibited PKCζ expression but not other PKC isoforms, including PKCι/λ, PKCα, PKCβII, and PKCδ (Fig. 4 E). Under this condition, d-flow effects on p53 SUMOylation and p53–Bcl-2 binding were significantly reduced, indicating the unique and critical role of PKCζ on d-flow–mediated p53 SUMOylation and its subsequent events.

We investigated the contribution of ONOO⁻ production on the d-flow–mediated PKCζ activation and p53 SUMOylation and found that an ONOO⁻ scavenger, ebselen, a nonselective inhibitor of NO synthesis, N-nitro-l-arginine methyl ester (L-NAME), superoxide dismutase mimetic, and an ONOO⁻ scavenger, Mn (III)tetrakis(4-benzoic acid)porphyrin chloride (Mn-TBAP), significantly inhibited d-flow–mediated PKCζ activation as well as p53 SUMOylation (Fig. 5, A–C). Moreover, these compounds strongly inhibited d-flow–induced EC apoptosis (Fig. 5, D and E), strongly suggesting a critical role of ONOO⁻ production in the d-flow–mediated signaling and subsequent apoptosis.

Because PIASy (SUMO E3 ligase) is involved in p53 SUMOylation in fibroblasts (Bischof et al., 2006), we wondered whether it was involved in the ONOO⁻- and d-flow–elicited p53 SUMOylation in HUVECs. When HUVECs were transfected with PIASy siRNA, the ONOO⁻ and d-flow–dependent p53 SUMOylation was inhibited (Fig. 6 A and Fig. 7 A). Using the PIASy-depleted cells, we studied the role of PIASy in the ONOO⁻-elicited p53–Bcl-2 association and found that the PIASy depletion inhibited this interaction (Fig. 6 B), suggesting that PIASy, which SUMOylates p53, is involved in ONOO⁻-induced p53 nuclear export and subsequent p53–Bcl-2 binding in ECs. To see whether PIASy has a role in the ONOO⁻ and d-flow–induced EC apoptosis, ECs were transfected with PIASy siRNA, challenged by ONOO⁻ or d-flow, and assayed for apoptosis. The number of TUNEL-positive cells and expression of cleaved caspase 3 induced by ONOO⁻ or d-flow were down-regulated by PIASy siRNA (Fig. 6 C and Fig. 7, B and C).

To establish the role of p53 SUMOylation on d-flow–mediated EC apoptosis, we transduced ECs with an adenovirus containing wild-type p53 (Ad-WT-p53) or the p53 SUMOylation site mutant (p53-K386R) and examined d-flow–mediated apoptosis compared with wild-type p53 in ECs. First, we confirmed that the adenoviral p53-K386R (Ad-p53-K386R) mutant significantly decreased p53 SUMOylation (Fig. 4 A). Next, we found that d-flow increased apoptosis in ECs transduced by Ad-WT-p53, but transduction of the Ad-p53-K386R mutant significantly inhibited it (Fig. 7, D and E). Because nuclear export of p53 is important for p53–Bcl-2 binding, we also used an NES mutation (L348,350A; ΔNES; O’Keefe et al., 2003) and studied d-flow–mediated EC apoptosis. As shown in Fig. S2 (A and B), in contrast to wild-type p53, we found that both p53-K386R and the ΔNES mutant stayed in the nucleus after d-flow stimulation, and the ΔNES mutant significantly inhibited d-flow–mediated apoptosis compared with wild type (Fig. 7, D and E), suggesting the critical role of p53 SUMOylation and nuclear export of p53 on d-flow–mediated EC apoptosis. Collectively, these results suggest the critical role of PIASy-mediated p53 SUMOylation and subsequent nuclear export of p53 in the ONOO⁻ and d-flow–induced EC apoptosis (Fig. S2, A and B).

As PIASy played a critical role in CATζ-mediated p53 SUMOylation (Fig. 4 B), PKCζ might directly phosphorylate PIASy and activate its E3 SUMO ligase activity. To test this possibility, we incubated recombinant PKCζ with GST-tagged PIASy fragments and found that none of the fragments was phosphorylated by PKCζ, whereas autophosphorylation of PKCζ was detected (Fig. S3). Although it does not phosphorylate PIASy, PKCζ may still interact with PIASy. We cotransfected HeLa cells with HA-tagged PKCζ and myc-tagged PIASy and performed a coimmunoprecipitation assay in which PKCζ and PIASy were coimmunoprecipitated (unpublished data). To confirm this interaction between endogenous PKCζ and PIASy, we stimulated HUVECs with ONOO⁻ for the indicated times and performed coimmunoprecipitation using anti-PKCζ. PIASy was indeed coimmunoprecipitated by anti-PKCζ, and ONOO⁻ stimulation increased the PKCζ–PIASy interaction (Fig. 8 A). Next, to determine the PIASy-binding regions of PKCζ, we generated four PKCζ-truncated mutants and evaluated their association with PIASy using a mammalian two-hybrid assay (Fig. 8, B and C). Plasmids encoding the GAL4–DNA-binding domain and PKCζ (full length or one of the truncated forms) were constructed using the pBIND vector. A plasmid containing VP16-PIASy was constructed using the pACT vector. As expected, wild-type PKCζ bound to PIASy, and ONOO⁻ increased this association (Fig. 8 B). More importantly, however, we found that the C-terminal kinase domain (aa 401–587) was required for the PKCζ–PIASy association (Fig. 8 C). Next, to determine the PKCζ binding site of PIASy, we coexpressed GST-fused PKCζ and Flag-tagged PIASy fragments in HeLa cells and performed coimmunoprecipitation using anti-Flag. We found that PIASy fragment 3 (Fr 3; aa 301–410), which contains the Siz/PIAS RING domain, interacted with PKCζ (Fig. 8 D). In a separate experiment, we examined this fragment could interfere with the PKCζ–PIASy association. HeLa cells were cotransfected with wild-type PKCζ and PIASy together with PIASy Fr 3. Consistent with the coimmunoprecipitation results, PIASy Fr 3 significantly inhibited the PKCζ–PIASy interaction (Fig. 8 E). To demonstrate the importance of the PKCζ–PIASy interaction for p53 SUMOylation, ECs expressing PIASy fragments were stimulated by d-flow. As shown in Fig. 8 F, only PIASy Fr 3 inhibited d-flow–mediated p53 SUMOylation, suggesting that p53 SUMOylation depends on PKCζ–PIASy binding.

Morphological evidence for the PKCζ–PIASy and p53–Bcl-2 association was provided by confocal microscopy of coimmunostained ECs with or without flow stimulation (Fig. 9 A). Cells cultured without flow expressed PKCζ mainly in the cytosol and PIASy in the nucleus as reported previously (Sachdev et al., 2001; Li et al., 2004). However, after d-flow stimulation, PKCζ and PIASy were colocalized in the nucleus. As for the p53 and Bcl-2 colocalization, p53 was localized in the nucleus of unstimulated cells, whereas Bcl-2 was mainly outside the nucleus as previously reported (Zhong et al., 1993; Ghosh et al., 2004). D-flow stimulation caused significant nuclear export of p53 and colocalization with Bcl-2. These data support d-flow–induced association of PKCζ–PIASy and p53–Bcl-2 in ECs.

PKCζ activation in the ECs of the lesser curvature of the aortic arch in porcine aorta was recently reported (Magid and Davies, 2005). Because activation of this kinase is proatherogenic, we investigated the expression of PKCζ, phosphorylated PKCζ, a nitrotyrosine-containing protein, p53, and apoptotic ECs using en face aorta preparations and confocal microscopy. Aortas from male wild-type C57BL/6 mice (6–8 wk old) fed with normal chow were isolated after perfusion fixation and en face preparations were made. We focused on areas designated as high probability (HP) regions (lesser curvature of aortic arch) and low probability (LP) regions (greater curvature of aortic arch) for atherogenesis (Fig. 10 A) as described previously (Iiyama et al., 1999) and in the Materials and methods section. When the endothelium was double stained with anti-PKCζ or antiphospho-PKCζ together with anti–vascular endothelial (VE)-cadherin as an EC marker, we found that the expression of total PKCζ increased in the HP area (Fig. 10 B), and especially phospho-PKCζ was significantly higher in the HP area compared with the LP area (Fig. 10 C). Quantification of these data obtained from five mice supports this conclusion (Fig. 10, B and C, bar graph). These results confirm our in vitro results that show d-flow–dependent activation of PKCζ (Fig. 1).

Next, aortas were immunostained with anti-p53 and then with TO-PRO3 for nucleus staining, and confocal microscopy was used to acquire a z series of fluorescence images. No significant fluorescent signal was observed in aorta samples treated with nonimmune rabbit IgG (not depicted) or with antigen-preabsorbed anti-p53 (Fig. S5 A). In LP areas, p53 was localized mainly in the nucleus (Fig. 10, D and F). In contrast, in HP areas, significant levels of anti-p53 staining were detected outside the nucleus, although anti-p53 staining was still associated with the nucleus. This cytoplasmic staining was localized primarily to the area underneath the nucleus (Fig. 10, E and F). These results appear to suggest that d-flow causes p53, which is present mainly in the nucleus in ECs exposed to s-flow (i.e., LP area), to translocate into the cytoplasm especially to the area directly below the nucleus.

We also examined whether ECs in the HP area underwent more apoptosis than ECs in the LP area. Aortas of 7-wk-old C57BL/6 mice were immunostained first for VE-cadherin (Fig. S4 A, red) followed by the TUNEL staining (Fig. S4 A) or were coimmunostained for annexin V (Fig. S4 B, red) and VE-cadherin (Fig. S4 B, green). Although, in the LP area, TUNEL- and annexin V–positive cells were rarely detected, such cells were frequently detected in the HP area (Fig. S4, A and B). Next, we studied the role of p53 in EC apoptosis in the 7-wk-old p53-deficient C57BL/6 mice (p53^−/−; Fig. 10, G and H). p53 was absent in lung ECs isolated from p53^−/− mice (Fig. 10 H, right). Interestingly, annexin V–positive cells in the HP area were significantly decreased in the p53^−/− mice (Fig. 10, G and H), supporting the critical role of p53 on d-flow–mediated apoptosis in vivo. Because we found the critical role of ONOO⁻ in d-flow–mediated PKCζ activation, p53 SUMOylation, and apoptosis, we investigated the expression of nitrotyrosine-containing proteins by double staining aortas with anti–VE-cadherin and antinitrotyrosine (Fig. S5, B–D). Antinitrotyrosine staining was significantly higher in HP areas than in LP areas. Our data collectively suggest that d-flow induces ONOO⁻ production, PKCζ activation, and nuclear export of p53, which increases EC apoptosis, and may prime ECs in the d-flow (HP) area to become susceptible to atherogenesis under the influence of various systemic risk factors.

Several atheroprotective signals activated by s-flow have been identified, but atheroprone signaling activated by d-flow is not well understood. The main finding of this study is that d-flow–mediated ONOO⁻ production activates PKCζ and induces p53 SUMOylation, p53 nuclear export, p53–Bcl-2 binding, and subsequent apoptosis. p53 appears to have both yin and yang effects on blood vessel physiology. Recent studies show its protective effect against cell death (Tian et al., 2000; Garner and Raj, 2008). Because p53 positively regulates the expression of p21, an antiapoptotic protein (Garner and Raj, 2008), p53 may exert its antiapoptotic effect by increasing p21 expression. Previously, Lin et al. (2000) have reported that laminar shear stress increased p53 expression and JNK-mediated p53 phosphorylation, which leads to endothelial growth arrest via increasing GADD45 and p21^cip1 expression. Therefore, it is important to emphasize here that p53 can induce growth arrest by inhibiting apoptosis via regulating p21. The interplay between cell cycle arrest and apoptosis is possibly critical in maintaining endothelial function because p53 may be able to reduce cells with DNA damages from apoptosis by preventing entry into the S phase (Garner and Raj, 2008). However, it remains unclear how p53 determines which of these two activities, pro- or antiapoptotic, to implement. Of note, most of the p53 antiapoptotic effects have been explained by its nuclear localization, as nuclear p53 protects cells from apoptosis, especially under low stress conditions (Tian et al., 2000; Garner and Raj, 2008). Indeed, our current study has revealed the antiapoptotic localization of p53 in ECs in the LP area exposed to steady laminar stress, which we consider to be a low stress condition. In contrast, ECs exposed to d-flow and ONOO⁻ had increased cytoplasmic p53 localization (i.e., nuclear export of p53), which enhanced EC apoptosis via increased p53–Bcl-2 binding.

Little information is available regarding the pathological significance of p53 subcellular localization, especially in ECs. Our study suggests that PKCζ activation by d-flow and ONOO⁻ and subsequent p53 nuclear export promote EC apoptosis. This notion is consistent with an earlier study showing that human cytomegalovirus infection of ECs induces p53 accumulation in the cytoplasm and apoptosis via a p53-dependent pathway (Shen et al., 2004). Together, these results strongly suggest the involvement of cytoplasmic p53 in EC apoptosis (Utama et al., 2006). Our en face confocal data are consistent with this idea because increased cytoplasmic anti-p53 staining and increased TUNEL and annexin V stainings are found in ECs located in the area exposed to d-flow (i.e., HP area).

In our in vitro study, we found the importance of ONOO⁻ on d-flow–mediated PKCζ activation, p53 SUMOylation, and EC apoptosis using ebselen, L-NAME, and Mn-TBAP (Fig. 5). In addition, we found that nitrotyrosine staining was significantly increased in d-flow (HP) areas in vivo (Fig. S5, B–D), where PKCζ activation, nuclear exports of p53, and EC apoptosis were increased. We used 10–100 µM ONOO⁻ as the final concentration, which is in agreement with the concentration of ONOO⁻ used in previous studies by other investigators (Alvarez et al., 2004; Levrand et al., 2005; Szabó et al., 2007). Because ONOO⁻ is a transient intermediate in free radical chemistry and is highly reactive, it is difficult to measure actual concentrations of ONOO⁻ in vivo. However, 10–100 µM ONOO⁻ are thought to be physiological because it has been suggested that the rates of ONOO⁻ production in vivo in specific compartments have been estimated to be as high as 50–100 µM/min (Alvarez et al., 2004; Levrand et al., 2005; Szabó et al., 2007). In addition, it has also been estimated that the rate of ONOO⁻ generation may reach up to 1 mM/min in an inflamed organ (lung) in vivo (Ischiropoulos et al., 1992). ONOO⁻ has several possible ways to activate PKCζ. One of the major effects of ONOO⁻ is protein tyrosine nitration, which can regulate a variety of kinases (Liaudet et al., 2009). First, ONOO⁻ can activate various receptor tyrosine kinases (Klotz et al., 2000; Zhang et al., 2000). Second, ONOO⁻ can activate Src by displacing Tyr527 from its binding site to the SH2 domain (Roskoski, 2005). Finally, ONOO⁻ can inhibit phosphatases via oxidation of cysteine-bound thiols (Takakura et al., 1999). Although the direct effect of ONOO⁻ on PKCζ remains unclear, these molecules may be involved in signaling events upstream of PKCζ. Therefore, we hypothesize that activation of a variety of kinases may be involved in ONOO⁻-mediated PKCζ activation. Further studies are necessary to determine the major signaling events.

In addition to regulating cell physiology through kinase activities, active kinases can also regulate cell physiology by altering their interaction with other molecules (Akaike et al., 2004; Boggon and Eck, 2004). PKCζ contains the pseudosubstrate autoinhibitory sequence (aa 116–122), and the release of the kinase domain (aa 268–587) from this autoinhibitory domain leads to PKCζ activation (Newton, 2001; Smith et al., 2003). In this study, we were unable to show PKCζ-mediated phosphorylation of PIASy in our in vitro kinase assay (Fig. S3), but we found that activation of PKCζ increased the PKCζ–PIASy interaction (Fig. 8). Using a mammalian two-hybrid assay, we then determined the C-terminal kinase domain of PKCζ (aa 401–587) was a PIASy binding site (Fig. 8 C). Of note, the deletion of the N-terminal autoinhibitory domain of PKCζ (aa 1–200) increased PKCζ–PIASy association, suggesting that this domain is a negative regulator not only of its kinase activity but also its binding ability with PIASy. Therefore, releasing the PKCζ N-terminal inhibitory effect (i.e., activation of PKCζ) is critical for both kinase activation and PKCζ–PIASy association, and the regulation of PIASy activity most likely depends on PKCζ-mediated increase in PKCζ–PIASy association but not phosphorylation. We also found that PKCζ binds to the RING domain of PIASy, which contains its catalytic site and may alter the structure and enzymatic activity of PIASy. Our findings suggest that d-flow–mediated PKCζ–PIASy association is critical for p53 SUMOylation to induce the endothelial p53 nuclear export and apoptosis.

C57BL/6 and p53^−/− mice with a C57BL/6 background (P53N4-M) were purchased from Taconic. All mice were maintained under pathogen-free conditions at the Aab Cardiovascular Research Institute at the University of Rochester. All animal procedures were performed with the approval of the University Committee on Animal Resources at the University of Rochester.

The plasmid encoding human HA-SUMO3 was a gift from R.T. Hay (University of Manchester, Manchester, England, UK; Tatham et al., 2001). The pcDNA3-Myc-PIASy and pcDNA-HA-PKCζ constructs were gifts from M. Takahashi (Nagoya University, Chikusa-ku, Nagoya, Japan; Matsuura et al., 2005) and J.-W. Soh (Columbia University, New York, NY; Soh et al., 1999), respectively. The p53-luciferase (p53-luc) reporter (Addgene plasmid 16442; B. Vogelstein, Johns Hopkins University School of Medicine, Baltimore, MD; el-Deiry et al., 1993) and pcDNA-Flag-p53 (Addgene plasmid 10838; T. Roberts, Harvard Medical School, Boston, MA; el-Deiry et al., 1993; Gjoerup et al., 2001) were obtained from the nonprofit Addgene plasmid repository. The Gal4 wild-type and truncated forms of PKCζ were created by inserting a KpnI–XbaI fragment generated by PCR into the pBIND vector (Promega). V16-PIASy was created by inserting human PIASy isolated from pcDNA3-PIASy into BamH1 and EcoRV1 sites of the pACT vector. Truncated forms of PIASy were constructed by inserting fragments EcoRI–XhoI or BamHI–XhoI generated by PCR into the pGEX-KG or pCMV-Taq2B vector, respectively. All constructs were verified by DNA sequencing.

Rabbit and mouse anti-PKCζ (C-20 [SC-216] and A-3 [SC-17781]), rabbit and mouse anti-p53 (FL-393 [SC-6243] and DO-1 [SC-126]), rabbit and mouse anti–Bcl-2 (N-19 [SC-492] and C-2 [SC-7382]), rabbit and mouse anti-HA (Y-11 [SC-805] and F-7 [SC-7392]), and anti-myc (A-14; SC-789) were purchased from Santa Cruz Biotechnology, Inc. The phospho-specific antibodies for p-PKCζ were purchased from Cell Signaling Technology (Thr 410; 9378) and Abcam (Thr 560; ab62372). The rabbit and mouse anti-SUMO2/3 were purchased from Abgent (AP1224a) and MBL International (M114-3), respectively. Anti-PIASy was purchased from Abcam (ab58416) and Sigma-Aldrich (P0104). Rat anti–VE-cadherin was purchased from BD (555289). ON-TARGETplus SMARTpool for human PKCζ siRNA was purchased from Thermo Fisher Scientific (L-003526-00), and a nonspecific control siRNA obtained from Invitrogen (12935-122) was used as a negative control. The Ad-DN-PKCζ was purchased from Cell Biolabs. The Ad-DN-PKCζ was generated by mutating the ATP binding site (K281M) from human PKCζ (NCBI Protein database accession no. NP_002735; Diaz-Meco et al., 1993). GST-fused active recombinant human PKCζ was purchased from Stressgen (PPK-468). Peroxynitrite was purchased form EMD. Alexa Fluor 488–conjugated mouse antinitrotyrosine was purchased from Millipore (16–226).

HUVECs were obtained from collagenase-digested umbilical cord veins (Takahashi and Berk, 1996) and collected in M200 medium supplemented with low serum growth supplement (Cascade Biological) and 5% fetal calf serum (Invitrogen). HUVECs were cultured on 0.2% gelatin-precoated dishes. For transient expression experiments, transfection of the PKCζ SMARTpool siRNA or PIASy siRNA was performed using Lipofectamine 2000 (Invitrogen) following protocols provided by the manufacturer. The Stealth RNAi Negative Control Med GC (Invitrogen), which has no homology to the vertebrate transcriptome, was used as a negative control. The cells were harvested 48 h after siRNA transfection, and protein expressions were monitored by immunoblotting with antibodies against PKCζ, PIASy, or tubulin. The target and control sequences for PIASy siRNAs were 5′-CAAGACAGGUGGAGUUGAU-3′ and 5′-UACCGUCUCCACUUGAUCG-3′.

To expose a large number of cells to flow, we used a cone and plate flow apparatus identical to the one used previously (Reinhart-King et al., 2008). Confluent HUVECs cultured in 100-mm dishes were exposed to s-flow in a flow apparatus placed in a cell culture incubator with 5% CO₂ and at 37°C for long-term experiments (shear stress = 20 dyn/cm²). To expose cells to d-flow, we used a cone with radial grooves. Shear stress induced by d-flow cannot be calculated, but we rotated the cone with 50 rpm that gives 5 dyn/cm² of laminar shear stress. We optimized and chose this speed because ECs maintained nonelongated cell shapes with this speed.

In accordance with recently published guidelines, apoptosis was quantified by multiple methodologically unrelated criteria (Galluzzi et al., 2009). To induce apoptosis, HUVECs were treated with 100 µM ONOO⁻ or d-flow. Phosphatidylserine externalization was quantified by staining cells with annexin V–Alexa Fluor 568 (Roche), and late-stage apoptosis was quantified morphologically by counting cells showing evidence of positive staining for TUNEL (Ding et al., 2000). TUNEL analyses were performed according to the manufacturer’s instructions, and the cells were counterstained with DAPI (Sigma-Aldrich) to identify nuclei. For each experiment in vitro, a total of 150–200 cells were counted from randomly selected fields of view, and TUNEL-positive cells were expressed as a percentage of total cells counted. Samples were examined using a 40× lens under a fluorescence microscope (BX51; Olympus) capable of imaging two distinct channels without adjustment of the microscope stage position. 10 random fields per sample were examined. Apoptosis was also quantified by Western blot assessment of the cleavage of the caspase 3 using the cleaved caspase 3 antibody. All measurements were performed blinded, and at least three independent experiments were performed.

Cells were plated in 12-well plates at 5 × 10⁴ cells/well. The mammalian two-hybrid assay was performed as described previously (Woo et al., 2006). In brief, cells were transfected in Opti-MEM (Invitrogen) with Lipofectamine mixture containing the pG5-luc vector and various pBIND and pACT plasmids (Promega) for 4 h. Cells were washed, and fresh DME supplemented with 10% fetal bovine serum was added. The pBIND vector contains the yeast GAL4–DNA-binding domain upstream of a multiple cloning region, and the pACT vector contains the herpes simplex virus VP16 activation domain upstream of a multiple cloning region. Various PKCζ mutants and PIASy were cloned into the pBIND and pACT vector, respectively. Because pBIND also contains the Renilla luciferase gene, the expression and transfection efficiencies were normalized with the Renilla luciferase activity. Cells were collected 36 h after transfection unless indicated otherwise, and the luciferase activity was assayed with the Dual-Luciferase kit (Promega) using a luminometer (TD-20/20; Turner Designs). Transfections were performed in triplicate, and each experiment was repeated at least three times.

HeLa cells (10⁶ cells) were plated in 100-mm dishes and transfected in Opti-MEM with Lipofectamine mixture containing the Flag-tagged PIASy-truncated mutants for 4 h. Cells were washed and cultured for 24 h in fresh complete growth medium. Cell lysates were made by incubating cells in 0.5 ml radioimmunoprecipitation assay lysis buffer for 30 min at 4°C. After centrifugation for 20 min at 12,000 rpm, the whole-cell lysates were incubated with 100 ng of recombinant GST-fused PKCζ, anti-FLAG, and IgG Sepharose beads with constant mixing for 2–5 h at 4°C. The beads were then washed three times with lysis buffer, resuspended in SDS-PAGE sample buffer, and boiled for 8 min. Samples were separated on a 15% SDS-PAGE and analyzed by Western blotting using anti-PKCζ.

To determine whether the p53 localization is regulated by PKCζ activation, HUVECs grown in 6-well plates were transduced with Ad-LacZ or Ad-DN-PKCζ for 24 h. After treatment with 100 µM ONOO⁻ for 4 h, the cells were quickly washed two times with cold PBS, fixed with 4% paraformaldehyde in PBS for 15 min, and then permeabilized with 0.2% Triton X-100 in PBS for 10 min. Cells were incubated with blocking buffer (5% goat serum and 0.1% NP-40 in PBS) for 60 min to block nonspecific binding and incubated with anti-p53 (1:200 dilution in 2% goat serum and 0.1% NP-40 in PBS) overnight at 4°C. The cells were washed three times with PBS and incubated with Alexa Fluor 546–labeled anti–rabbit IgG or Alexa Fluor 488–labeled anti–rabbit IgG₁ (1:2,000 dilution; Invitrogen) for 1 h at room temperature. The cells were counterstained with DAPI to identify nuclei. All of the images were collected using an epifluorescence microscope (BX51) equipped with a charge-coupled device camera (Spot; Diagnostic Instruments, Inc.) and an Acroplan water 60× W lens.

To determine the colocalization of p53 with Bcl-2 or PKCζ with PIASy, HUVECs were stimulated with d-flow for 3 h. After fixation, permeabilization, and 10% goat serum in PBS with 0.1% NP-40 for 1 h, the cells were incubated with mouse anti-p53 and rabbit anti–Bcl-2 or mouse anti-PKCζ and rabbit anti-PIASy (1:200) antibodies in 2% goat serum with PBS overnight at 4°C. The cells were washed three times with PBS and incubated with Alexa Fluor 546–labeled anti–mouse IgG and Alexa Fluor 488–labeled anti–rabbit IgG₁ (1:2,000 dilution) for 1 h at room temperature. The samples were analyzed using a laser-scanning confocal microscope (FV1000; Olympus) equipped with a Plapon 60× 1.42 NA oil lens objective. Quantification of the overlay image was performed using the Photoshop (CS; Adobe) program.

Cells were collected in PBS containing 10 mM N-ethylmaleimide, and cell extracts were prepared in modified radioimmunoprecipitation assay buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% NP-40, 0.1% SDS, 1 mM dithiothreitol, 1:200-diluted protease inhibitor cocktail [Sigma-Aldrich], 1 mM PMSF, 10 mM N-ethylmaleimide, and 0.1 mM iodoacetamide). Immunoprecipitation with a mouse monoclonal anti-Flag, HA, or p53 was performed as described previously (Woo et al., 2006). Bound proteins were released in 2× SDS sample buffer, resolved by SDS-PAGE, transferred onto an enhanced chemiluminescence nitrocellulose membrane (Hybond), and visualized by using the enhanced chemiluminescence detection reagents (GE Healthcare) according to the manufacturer’s instructions. p53 SUMOylation was detected by immunoprecipitation with anti-p53 followed by Western blotting with anti-SUMO2/3. Results were normalized to the lowest phosphorylation level within each set of experiments, and statistical significance was determined by comparing the mean level of the control group to each of the experimental data points.

The s-flow and d-flow areas within the aorta were identified based on the published and generally accepted anatomical locations where such flow patterns are known to occur (Iiyama et al., 1999; Hajra et al., 2000; Jongstra-Bilen et al., 2006). For example, a typical s-flow area is located in the greater curvature area and is marked as an LP region for lesion formation (Hajra et al., 2000), which is also known as a high wall shear stress area. A d-flow area is the lesser curvature area (HP region; Iiyama et al., 1999; Hajra et al., 2000; Jongstra-Bilen et al., 2006), which is also indicated as a low wall shear stress area. In the Fig. 10 legend, we indicated these two areas in the aortic arch. EC shape outlined by anti–VE-cadherin staining was also used to identify s-flow areas (elongated cell shape) and d-flow areas (irregular cell shape). It has been reported that PECAM-1 (CD31) staining, another marker of ECs, was localized to endothelial junctions in the HP region, whereas in the LP region, it is more diffuse. Therefore, the PECAM-1 staining at EC borders was much stronger in the HP region than in the LP region (Hajra et al., 2000). We found the similar tendency of VE-cadherin staining in the mouse aorta. Because these data were very consistent in every mouse, we do not think that this is caused by the damage that specifically occurred in the LP region. C57BL/6 wild-type mice (Taconic) were fed standard chow. Animals of 6–8 wk of age were euthanized by CO₂ inhalation. The arterial tree was perfused via the left ventricle with saline containing 40 USPU/ml heparin followed by 4% paraformaldehyde in PBS for 10 min. After adipose tissues were removed, aortas were cut open longitudinally and permeabilized with PBS containing 0.1% Triton X-100 and blocked by TBS containing 10% goat serum and 2.5% Tween 20 for 30 min. Aortas were incubated with 10 µg/ml rabbit anti-p53 (FL-393; Santa Cruz Biotechnology, Inc.; rabbit IgG was used as a control) and 7 µg/ml rat anti–VE-cadherin (an EC marker; BD) in the blocking solution overnight. Specificity of anti-p53 staining was tested by staining with anti-p53 preabsorbed with recombinant p53 (Santa Cruz Biotechnology, Inc.) as shown in Fig. S4. After a PBS rinse, anti–rabbit IgG and anti–rat IgG (1:1,000 dilution; Alexa Fluor 546 and 488, respectively; Invitrogen) were applied for 1 h at room temperature. Nuclei were stained using TO-PRO3 (Invitrogen). For the detection of apoptosis, we used two different methods. To perform an in situ TUNEL assay, aortas immunostained with anti–VE-cadherin (with Alexa Fluor 546 anti–rabbit IgG) were incubated in the TUNEL reaction mixture (Roche) for 1 h at 37°C according to the manufacturer’s instructions (mixture without the enzyme TdT as a control). For annexin V labeling, we injected annexin V–Alexa Fluor 568 into aortas via the left ventricle after perfusion with saline containing 40 USPU/ml heparin followed by 4% paraformaldehyde in PBS for 10 min. Then, these aortas were immunostained with anti–VE-cadherin as described in the previous paragraph. Images were acquired using a confocal laser-scanning microscope (Fluoview 300; Olympus) equipped with krypton/argon/HeNe laser lines and 20× 0.70 NA, 40× 1.0 NA, and 60× 1.4 NA objectives. For quantification of phospho- and total PKCζ expression level, 10–15 optical sections were collected at 0.3–0.5-µm increments so that z stacks of ∼4-µm-thick cell blocks from the luminal surface were obtained. Images were collected using the same confocal settings. For quantification of EC apoptosis, aortas of 7-wk-old wild-type mice (n = 7 each) were prepared. Images were acquired as described in the previous paragraph. To analyze p53 subcellular localization, we took z-stack clip images of 50 optical sections collected at 0.1-µm increments from the luminal surface. We radially extended the edge of the nucleus by the length of the nuclear radius at each point along the edge of the nucleus and delineated an area for each cell. We defined the signal level in this area as total because practically all anti-p53 staining signals (>80–90%) were included in this area. Staining signal levels in the cytoplasm and the nucleus were determined. Confocal images were obtained from s-flow and d-flow areas as shown in Fig. 10.

Data are reported as means ± SD. Statistical analysis was performed with the Prism program version 2.00 (GraphPad Software). Differences were analyzed with a one-way or a two-way repeated-measure analysis of variance as appropriate followed by Dunnett’s correction for multiple comparisons. P < 0.05 is indicated by an asterisk, and P < 0.01 is indicated by a double asterisk.

Fig. S1 shows that ONOO⁻ increases p53 SUMOylation and p53–Bcl-2 binding. Fig. S2 shows d-flow–induced p53 nuclear export analyzed by immunostaining. Fig. S3 shows whether PKCζ phosphorylates PIASy in vitro by in vitro kinase assay. Fig. S4 shows apoptosis in d-flow and s-flow area of the mouse aortic arch endothelia analyzed by TUNEL and annexin V staining. Fig. S5 shows specificity of anti-p53 staining and nitrotyrosine staining in mouse aortic arch endothelia.

This work is supported by grants from the America Heart Association to Dr. Woo (Postdoctoral Fellowship 0625957T and Scientist Development Grant 0930360N) and to Dr. Le (Postdoctoral Fellowship 4360007) and from the National Institutes of Health to Drs. Abe (HL-064839, HL-077789, and HL-102746) and Berk (HL-064839 and HL-077789). Dr. Abe is a recipient of Established Investigator Awards of the American Heart Association (0740013N).