ERK5 is required for VEGF-mediated survival and tubular morphogenesis of primary human microvascular endothelial cells

Angiogenesis defines the formation of new blood vessels from pre-existing vessels and is crucial for normal physiological vascular development, as well as for the pathological development of certain diseases, including diabetic retinopathy, atherosclerosis and cancer (Carmeliet, 2005). VEGF is a potent inducer of angiogenesis and is vital for vascular development, both during embryogenesis and in early neonatal life (Carmeliet et al., 1996; Ferrara et al., 1996; Gerber et al., 1999). Furthermore, VEGF is a key regulator of endothelial cell survival in the tumour vasculature (Benjamin and Keshet, 1997; Liu et al., 2000). It activates a number of receptor tyrosine kinases that are present on endothelial cells. VEGF receptor-1 (VEGFR-1/Flt-1) and VEGFR-2 (Flk-1/KDR) are present on vascular endothelial cells, whereas expression of VEGFR-3 (Flk-4) is confined to lymphatic endothelial cells (Cross et al., 2003; Lohela et al., 2009; Olsson et al., 2006). Numerous studies have revealed that activation of VEGFR-2 by the VEGF-A ligand is responsible for mediating the majority of physiological effects of VEGF in endothelial cells, including proliferation, migration, survival and permeability (Holmes et al., 2007). Identification of intracellular signalling pathways involved in VEGF–VEGFR-2 signalling has been subject to intense research in an attempt to decipher the crucial components downstream of VEGFR-2 (Holmes et al., 2007).

Mitogen-activated protein kinases (MAPKs) regulate the transduction of extracellular growth factor or stress-induced stimuli from the cell membrane to nuclear or cytoplasmic targets, and are crucial for the control of numerous cellular processes, including proliferation, migration and survival in response to these extracellular cues (Qi and Elion, 2005). Four major MAPK signalling cascades have been identified in mammalian cells: extracellular signal-regulated kinases 1 and 2 (ERK1/2), Jun-N-terminal kinases 1, 2 and 3 (JNK1–JNK3), p38 MAPKs (p38α, -β, -γ and -δ), and the most recently discovered MAPK, ERK5 (Chang and Karin, 2001; Roberts et al., 2009; Wang and Tournier, 2006). Each of these MAPKs is able to phosphorylate and activate a range of downstream substrates, resulting in specific intracellular responses (Lewis et al., 1998).

The ERK5 signalling module consists of MEKK2/3 (MAPK/ERK kinase kinase 2 and 3), which phosphorylates MEK5 (MAPK/ERK kinase 5), which in turn activates ERK5 by dual phosphorylation of Thr218 and Tyr220 within a T-E-Y activation motif. The physiological importance of the ERK5 signalling cascade was revealed in studies of targeted deletion of either Erk5 (Regan et al., 2002; Sohn et al., 2002; Yan et al., 2003), Mek5 (Wang et al., 2005), or Mekk3 (Yang et al., 2000) in mice. Deletion of any of these genes resulted in death between embryonic day 9.5 (E9.5) and E11.5 as a result of hemorrhaging caused by severe abnormalities in cardiovascular development and compromised vascular integrity. Importantly, targeted deletion of Erk5 in murine endothelial cells, but not in other cell types such as cardiomyocytes and hepatocytes, was also found to result in cardiovascular defects and disrupted vascular integrity, a phenotype that is identical to that of the global Erk5 knockout mouse (Hayashi et al., 2004; Hayashi and Lee, 2004). These data suggest that although ERK5 is expressed in a variety of tissues, expression of ERK5 within the endothelium is crucial to preserve vascular integrity, and to maintain normal endothelial cell function (Roberts et al., 2009).

ERK5 is activated by a range of extracellular stimuli, including the pro-angiogenic growth factors VEGF and FGF-2 (fibroblast growth factor-2) in endothelial cells (Hayashi et al., 2004). The finding that ERK5 is required for tumour-associated angiogenesis (Hayashi et al., 2005) raises the possibility that ERK5 is a crucial component of the VEGF-signalling cascade in endothelial cells. However, the physiological importance of VEGF-stimulated ERK5 activation, and the downstream targets of VEGF-induced ERK5 activity have yet to be determined.

In this study, we addressed this issue by profiling VEGF-stimulated ERK5 activation in primary human dermal microvascular endothelial cells (HDMECs), and characterised the functional effects of modulation of ERK5 activation in models of distinct stages of the angiogenic process. Activation of ERK5 was observed under conditions of both VEGF-induced HDMEC proliferation and VEGF-induced tubular morphogenesis of HDMECs in a three-dimensional collagen gel. However, only VEGF-stimulated tubular morphogenesis was affected by ERK5 downregulation. We show that ERK5 activity is required for VEGF-mediated activation of AKT and the resulting phosphorylation of BAD, leading to the prevention of endothelial cell apoptosis. Together, our data reveal for the first time, that ERK5 is a crucial mediator of VEGF-induced endothelial cell survival during conditions of tubular morphogenesis. Furthermore, our results suggest that the ERK5 pathway represents a viable target for the development of specific inhibitors aimed at preventing tumour-associated angiogenesis.

We used primary HDMECs to study the role of VEGF-mediated ERK5 activity. In response to VEGF stimulation, HDMECs proliferate on a gelatin matrix, but undergo tubular morphogenesis to form capillary vessels in a 3D collagen gel (Fig. 1A). Cell cycle analysis following stimulation with VEGF for 24 hours revealed that cells within collagen gel structures were cell cycle arrested (Fig. 1B). By contrast, VEGF-treated HDMECs plated on a gelatin matrix underwent proliferation, demonstrated by a significant number of cells in the S and G2–M phases of the cell cycle (Fig. 1B). The proliferation and differentiation of endothelial cells represent distinct aspects of the angiogenic response to VEGF in vivo, where endothelial cells must first digest the basement membrane, migrate and proliferate to form a sprout, before eventually differentiating into a functional lumen-containing vessel (Koike et al., 2004; Montesano et al., 1983). Immunoblot analysis of cell lysates using a phospho-ERK5 antibody, which is directed against the phosphorylated Thr218 and Tyr220 residues within the T-E-Y motif in the ERK5 activation loop, revealed that VEGF stimulation induced a sustained activation of ERK5 in HDMECs on both gelatin and collagen matrices (Fig. 1C,D). VEGF did not induce any changes in the total level of the ERK5 protein (Fig. 1C) in HDMECs on either matrix. In contrast to the activation of ERK5, VEGF induced a relatively transient activation of ERK1/2, which returned to basal levels by 3 hours on a gelatin matrix and by 1 hour in a collagen gel (Fig. 1C,D).

It is well established that VEGF is a potent inducer of endothelial cell proliferation in vitro (Ferrara and Henzel, 1989; Kroll and Waltenberger, 1997; Leung et al., 1989). To explore the potential involvement of ERK5 in HDMEC proliferation and tubular morphogenesis we used two separate siRNA duplexes targeting ERK5. Both of these duplexes downregulated ERK5 protein expression levels by 97%, as determined by western blot analysis (Fig. 2A). HDMECs plated on a gelatin matrix underwent a doubling in cell number following stimulation with VEGF for 3 days (Fig. 2B). ERK5 silencing had no significant effect on VEGF-induced cellular proliferation (Fig. 2B). By contrast, siRNA-mediated ERK5 silencing had a profound effect on VEGF-induced HDMEC tubular morphogenesis, because endothelial cells failed to form a network of vessels in collagen gel in response to VEGF (Fig. 2C,D). Cells appeared rounded, with pyknotic hypercondensed nuclei, indicative of cell death. Assessment of early apoptosis of HDMECs by annexin V staining of phosphatidylserine levels on the outer layer of the plasma membrane revealed that within collagen gels, cells treated with ERK5 siRNA displayed a fourfold increase in apoptosis compared with cells transfected with non-silencing (N.S.) control siRNA (Fig. 2E,F). By contrast, ERK5 siRNA-treated cells plated on a gelatin matrix showed no significant difference in the number of annexin V-positive cells, compared with that of the N.S. siRNA-treated cells (Fig. 2E,F). This data suggests that in HDMECs, ERK5 is required for VEGF-stimulated survival during tubular morphogenesis, but is not required during VEGF-induced cellular proliferation. In contrast to ERK5, siRNA-mediated silencing of ERK1 and ERK2 expression inhibited VEGF-mediated proliferation of HDMECs (supplementary material Fig. S1B). Silencing of ERK1 had a small inhibitory effect on tubular morphogenesis, with tube length reduced by 15%, whereas silencing of ERK2 had no effect on tubular morphogenesis (supplementary material Fig. S1C,D). These data reveal that, in HDMECs, ERK5 and ERK1/2 appear to have distinct roles; ERK5 regulating tubular morphogenesis and ERK1/2 regulating cellular proliferation.

The profound effect of silencing ERK5 on HDMEC tubular morphogenesis with a concomitant increase in annexin V staining (Fig. 2E), led us to analyse the activation of pathways regulating apoptosis in HDMECs. VEGF stimulated phosphorylation of the VEGFR-2 on both gelatin and collagen matrices; an effect that was not altered by silencing ERK5 expression (Fig. 3A). VEGF-induced activation of the serine/threonine kinase AKT/protein kinase B (PKB) is crucial for protecting endothelial cells against apoptosis (Fujio and Walsh, 1999; Gerber et al., 1998b). We analysed the phosphorylation status of AKT on Ser473 and Thr308 in HDMECs in response to acute VEGF stimulation when plated on a collagen matrix and a gelatin matrix. Treatment of HDMECs with siRNA duplexes against ERK5 resulted in decreased phosphorylation of AKT by approximately 50% in response to VEGF on a gelatin matrix compared with untransfected cells and cells transfected with N.S. siRNA. By contrast, on a collagen matrix the inhibition was more profound, with AKT phosphorylation reduced by at least 94% (Fig. 3A). Importantly, the total level of AKT protein was not affected on either matrix, thus indicating that the inhibition of ERK5 activity was affecting AKT phosphorylation (Fig. 3A). AKT is known to regulate the activity of a number of signalling proteins, including the pro-apoptotic protein BAD (Datta et al., 1997) and the anti-apoptotic protein BCL2 (Pugazhenthi et al., 2000). Phosphorylation of BAD on Ser136 by AKT prevents BAD from interacting with BCL2 by sequestering BAD to the 14-3-3 protein (Datta et al., 1997; Zha et al., 1996). Acute VEGF-stimulation of HDMECs resulted in increased BAD phosphorylation on Ser136; an affect that was more pronounced on the collagen matrix (Fig. 3A). Treatment of HDMECs with siRNA duplexes against ERK5 resulted in a decrease in VEGF-induced phosphorylation of BAD (Fig. 3A). Furthermore, silencing of ERK5 also inhibited the ability of VEGF to stimulate increased BCL2 mRNA expression in HDMECs: an effect that was more pronounced within a collagen gel (Fig. 3B).

Activation of the protease family of caspases represents one of the terminal stages of signal transduction pathways that lead to apoptosis in endothelial cells (Pober et al., 2009). We analysed the effect of silencing ERK5 on the activity of caspase-3/7 in HDMECs plated on a gelatin or collagen matrix 1 hour, 6 hours and 24 hours after stimulation with VEGF, using a fluorescent-based assay. Comparison of the two matrices revealed that after 24 hours under basal conditions, caspase-3/7 activity was induced on the collagen gel; an effect that was suppressed by addition of VEGF (Fig. 4F). Addition of staurosporine (STS), a known inducer of apoptosis in endothelial cells (Kabir et al., 2002), was used as a positive control and induced maximal caspase-3/7 activity on both matrices (Fig. 4A–F). Silencing of ERK5 only induced caspase-3/7 activity in HDMECs plated in a collagen matrix; an effect that was evident 1 hour after cells had been plated in a collagen gel (Fig. 4B). Loss of ERK5 induced maximal apoptosis after 6 hours, and was still evident after 24 hours; an effect that was confined to the collagen matrix (Fig. 4C–F). The presence of VEGF was not sufficient to suppress ERK5-induced cleaved caspase-3 expression at any time point (Fig. 4A–F). To confirm the role of active caspase-3 in HDMEC apoptosis caused by loss of ERK5, we analysed the levels of the 17 kDa and 19 kDa cleaved caspase-3 proteolytic fragments by western blotting. Loss of ERK5, following siRNA-mediated downregulation, resulted in increased levels of cleaved caspase-3 only in HDMECs plated in a collagen gel, an effect that could not be reversed by addition of VEGF for 6 hours (Fig. 4G,H). Addition of STS was sufficient to induce the appearance of the 17 kDa and 19 kDa cleaved caspase-3 fragments on both gelatin and collagen matrices (Fig. 4G,H). Taken together, these data show that under conditions of tubular morphogenesis, ERK5 is crucial for mediating VEGF-stimulated AKT activation, and the resultant suppression of caspase-3 activity, facilitating endothelial cell survival. Conversely, on a gelatin matrix, where proliferation occurs, ERK5 is partly required for VEGF-mediated AKT activation, but is not required for suppression of caspase-3 activity or endothelial cell survival.

To confirm the role of ERK5 in the regulation of AKT activation and suppression of cleaved caspase-3 levels, we manipulated ERK5 activity by transiently transfecting HDMECs with plasmids expressing either Flag-tagged wild-type ERK5 [ERK5(wt)], or Flag-tagged dominant-negative ERK5 [DN-ERK5(AEF)], which cannot be phosphorylated by MEK5, together with hemagglutinin (HA)-tagged constitutively active MEK5 [CA-MEK5(D)]. In agreement with previous studies (Barros and Marshall, 2005; Schweppe et al., 2006), co-transfection of HDMECs with ERK5(wt) and CA-MEK5(D) induced an increase in ERK5 phosphorylation (Fig. 5). This increase in ERK5 phosphorylation was augmented by stimulation with VEGF, and blocked by co-transfection of DN-ERK5(AEF) and CA-MEK5(D) (Fig. 5). Co-transfection of ERK5(wt) and CA-MEK5(D) resulted in enhanced phosphorylation of AKT on both Ser473 and Thr308 in both the absence and the presence of VEGF (Fig. 5). Furthermore, enhanced phosphorylation of BAD on Ser136 and increased expression of BCL2 was also observed (Fig. 5). Analysis of cleaved caspase-3 levels revealed that expression of ERK5(wt) and CA-MEK5(D) resulted in suppression of cleaved caspase-3 levels even in the absence of VEGF (Fig. 5). Co-expression of DN-ERK5(AEF) and CA-MEK5(D) did not stimulate ERK5 phosphorylation, AKT phosphorylation, BAD phosphorylation or BCL2 expression under basal conditions. Furthermore, DN-ERK5(AEF) and CA-MEK5(D) co-expression blocked VEGF-stimulated ERK5 phosphorylation, AKT phosphorylation and BAD phosphorylation. VEGF-mediated increase in BCL2 expression and suppression of cleaved caspase-3 was also blocked by co-expression of DN-ERK5(AEF) and CA-MEK5(D) (Fig. 5). Taken together, these results confirm our previous biochemical analysis with siRNA-mediated silencing of ERK5, in that ERK5 regulates activation of AKT and the downstream activation of BAD, resulting in suppression of caspase-3 activity, which ultimately leads to HDMEC survival on a collagen matrix.

In accordance with the lack of effect of siRNA-mediated downregulation of ERK5 upon HDMEC proliferation, co-transfection of HDMECs with either ERK5(wt) and CA-MEK5(D), or co-transfection of DN-ERK5(AEF) and CA-MEK5(D) had no effect upon VEGF-stimulated cellular proliferation (Fig. 6). Co-transfection of ERK5(wt) and CA-MEK5(D) stimulated HDMEC tubular morphogenesis in the absence of VEGF over 24 hours (Fig. 7A,B) and sustained VEGF-stimulated tubular morphogenesis for 5 days (Fig. 7C,D). Normal VEGF-stimulated tubular morphogenesis in this assay is maximal after 24 hours, and tubes rapidly regress after this time period; therefore the effect of sustaining VEGF-induced tube formation for 5 days was significant. By contrast, co-transfection of HDMECs with DN-ERK5(AEF) and CA-MEK5(D), did not stimulate tubular morphogenesis under basal conditions, and blocked VEGF-stimulated morphogenesis over 24 hours and 5 days (Fig. 7A–C). Together, these data demonstrate that ERK5 activity is essential for VEGF-stimulated tube formation, but is dispensable for VEGF-stimulated proliferation of HDMECs.

The ability of ERK5 to regulate tubular morphogenesis in HDMECs in a collagen gel suggested that ERK5 has the potential to regulate VEGF-stimulated angiogenesis. We used an endothelial and fibroblast co-culture assay to model events in a more representative assay of in vivo angiogenesis (Donovan et al., 2001; Sorrell et al., 2007). HDMECs were plated on a monolayer of normal human dermal fibroblasts (NHDFs) and either left unstimulated or stimulated with VEGF over a period of 5 days. The cells were then stained for expression of the endothelial-specific marker CD31 and visualised by immunohistochemistry. Quantification of the tubular networks showed that VEGF-stimulation of co-cultures resulted in a threefold increase in the endothelial tubular network after 5 days (Fig. 8A,B). siRNA-mediated silencing of either ERK5 or MEK5, using two individual duplexes directed against each target, resulted in a profound inhibition of both basal, and VEGF-stimulated tubule formation (Fig. 8A,B). The expression, and efficient siRNA-mediated knockdown of ERK5, in both NHDFs and HDMECs after 5 days was confirmed by western blotting (Fig. 8C). Analysis of cleaved caspase-3 levels in HDMEC and NHDF co-cultures by immunofluorescence revealed that siRNA-mediated silencing of ERK5 resulted in an increase in cleaved caspase-3, which co-localised with the CD31-positive HDMECs within the co-culture (Fig. 8D). Taken together, these results not only confirm our previous observation in collagen gels that ERK5 is required for VEGF-stimulated tubular morphogenesis, but also suggest that endothelial cells have a specific requirement for ERK5 to suppress caspase-3 activation and ultimately facilitate VEGF-mediated survival and tubular morphogenesis in endothelial cells.

Angiogenesis is a complex, multi-faceted process that requires the coordinated migration, proliferation and ultimate differentiation of endothelial cells to form a lumen-containing tubular vessel capable of allowing blood flow. We have used a model system to analyse the role of VEGF-mediated ERK5 activity in HDMECs during proliferation and differentiation. Our results are the first to reveal that ERK5 is specifically required for VEGF-mediated tubular morphogenesis, but is dispensable for cellular proliferation in primary HDMECs. We further define a specific role for ERK5 in the regulation of VEGF-mediated AKT phosphorylation and the suppression of caspase-3 activity during tubular morphogenesis, revealing that ERK5 regulates a defined component of VEGF-stimulated angiogenesis in endothelial cells.

VEGF stimulates a number of intracellular signalling pathways that regulate cellular proliferation, migration and survival (Holmes et al., 2007). VEGF has been reported to stimulate ERK5 activity in human umbilical vein endothelial cells (HUVECs) and mouse lung capillary endothelial cells (MLCECs) (Hayashi et al., 2004). However, the potential role of ERK5 in the mediation of VEGF-induced physiological responses had not been previously explored. In contrast to ERK1/2, VEGF-mediated activation of ERK5 was sustained under conditions of both proliferation and differentiation, with a very similar kinetic profile of activation (Fig. 1C,D). Reduction of ERK5 levels by siRNA-mediated gene silencing, revealed a specific role for ERK5 in the regulation of VEGF-mediated HDMEC tubular morphogenesis (Fig. 2C), but not proliferation (Fig. 2B). siRNA-mediated silencing of MEK5 also prevented tubular morphogenesis (supplementary material Fig. S2). The importance of ERK5 in the specific regulation of tubular morphogenesis was highlighted by the finding that co-expression of constitutively active MEK5, together with ERK5, stimulated tubular morphogenesis in the absence of VEGF, and significantly sustained VEGF-stimulated tubular morphogenesis (Fig. 7A–D). Activation of ERK5 has been previously shown to regulate the differentiation of myoblasts (Dinev et al., 2001), by regulating the pro-myogenic actions of IGF-2 (insulin-like growth factor 2) (Carter et al., 2009), suggesting that ERK5 regulates the differentiation of a number of cell types.

The lack of any effect upon manipulation of ERK5 activity in HDMECs, either by siRNA-mediated gene silencing (Fig. 2B) or overexpression (Fig. 6), after VEGF-mediated proliferation, suggests that in these cells, ERK5 activity is dispensable for this response. By contrast, knockdown of ERK1/2 clearly affected VEGF-mediated proliferation (supplementary material Fig. S1D), in agreement with the established role of ERK1/2 in the regulation of cell cycle entry (Brunet et al., 1999). The role of ERK5 in regulating cell proliferation remains unclear. Although epidermal growth factor (EGF)-mediated proliferation of fibroblasts (Kato et al., 1998) and GM-CSF (granulocyte-macrophage colony-stimulating factor)-mediated proliferation of haematopoietic cells (Dong et al., 2001) are reported to be ERK5 dependent, other research using cells from Erk5^–/– and Mek5^–/– mice has revealed that ERK5 and MEK5 are not required for cell cycle progression (Hayashi et al., 2004; Wang et al., 2005). The sensitivity of cells on a collagen matrix to ERK5 downregulation is reflected in the increased loss of VEGF-mediated activation of both AKT and BAD on this matrix, following ERK5 silencing, compared with a gelatin matrix (Fig. 3A). Activation of AKT requires its binding to phosphatidylinositol (3,4,5)-trisphosphate [PtdIns(3,4,5)P₃] via the pleckstrin homology domain and phosphorylation at Thr308 in the activation loop and phosphorylation at Ser473 in the C-terminus (Alessi et al., 1996; Stokoe et al., 1997). The mammalian target of rapamycin complex 2 (mTORC2) phosphorylates AKT at Ser473 (Jacinto et al., 2006; Sarbassov et al., 2005), whereas 3′-phosphoinositide-dependent kinase-1 (PDK1) phosphorylates AKT at Thr308 (Alessi et al., 1996). VEGF-stimulated AKT activation is regulated by phosphatidylinositol 3-kinase (PI3K) (Gerber et al., 1998b; Thakker et al., 1999). However, VEGF-mediated ERK5 activation was not inhibited by either the PI3K inhibitor wortmannin, or the mTOR inhibitor rapamycin (supplementary material Fig. S3), suggesting that ERK5 is not downstream of either PDK1 or mTORC2. Furthermore, the ability of both wortmannin and chronic exposure to rapamycin (Sarbassov et al., 2006) to inhibit VEGF-mediated phosphorylation of AKT and BAD, without inhibiting ERK5 activity, confirmed that the classical PI3K–PDK1 and mTORC2 pathways are able to regulate AKT phosphorylation and that AKT phosphorylation is required for the subsequent phosphorylation of BAD (supplementary material Fig. S3). The precise mechanism of ERK5-mediated regulation of AKT activity remains obscure. It is possible that ERK5 regulates the activity of a phosphatase, such as PP2A, which has been shown to interact with β1 integrin and dephosphorylate AKT at both Ser473 and Thr308 (Pankov et al., 2003). The recent finding that ERK5 immunoprecipitates with integrins β1, αvβ3 and αvβ5 in breast cancer cells (Sawhney et al., 2009), raises the possibility that ERK5 binds to specific integrins to coordinate signals from both the cell matrix and VEGFR-2 to regulate AKT phosphorylation and facilitate tubular morphogenesis in endothelial cells.

A role for Erk5 in the regulation of Akt phosphorylation at Ser473 and Thr308 has previously been reported in mouse embryonic fibroblasts under conditions of osmotic stress (Wang et al., 2006). In mouse neuronal cells, nerve growth factor (NGF)-mediated phosphorylation of Akt on Ser473, but not Thr308, is dependent upon ERK5 activity (Finegan et al., 2009). Recently, it has been shown that Erk5 regulates PDGF-stimulated Akt phosphorylation at Ser473 in porcine aortic endothelial (PAE) cells (Lennartsson et al., 2010). This suggests that ERK5 regulates AKT activity in response to various agonists in different cells.

Activation of AKT is important for the regulation of apoptosis because it has been shown to phosphorylate the pro-apoptotic protein BAD on Ser136, leading to its cytosolic sequestration by 14-3-3 protein (Datta et al., 1997). Furthermore, AKT also regulates expression of the anti-apoptotic protein BCL2 (Pugazhenthi et al., 2000). BCL2 and BAD heterodimerise and neutralise each others' function. BCL2 regulates apoptosis by preventing the release of cytochrome c from the mitochondria and the sequential activation of the initiator caspase-9 and terminal executioner caspase-3 pathway, ultimately resulting in endonuclease-mediated degradation of DNA, cytoplasmic and chromatin condensation and cell death (Cory et al., 2003). Consistent with the role of AKT in regulating apoptosis, we found that knockdown of ERK5 prevented VEGF-mediated phosphorylation of BAD (Fig. 3A) and also prevented expression of BCL2 mRNA in HDMECs plated in a collagen gel (Fig. 3B). ERK5 knockdown also induced caspase-3 activation on this matrix (Fig. 4A–H). Conversely, constitutive activation of ERK5 resulted in increased phosphorylation of BAD, increased expression of BCL2 and suppressed activation of caspase-3 (Fig. 5). VEGF has previously been reported to increase BCL2 levels in HUVECs (Gerber et al., 1998a) and to enhance HDMEC survival in collagen gels (Nor et al., 1999), whereas increased BAD phosphorylation has been shown to prevent apoptosis in rat sinusoidal endothelial cells (Ohi et al., 2006). Our data show that ERK5 is crucial for regulation of the apoptotic balance of these signalling molecules.

Gene ablation of Erk5 in mice results in embryonic lethality at around E10.5 as a result of abnormal cardiovascular development (Regan et al., 2002; Sohn et al., 2002; Yan et al., 2003). Histological analysis of the Erk5-null embryos has revealed that the vascular endothelial cells are irregularly aligned, and undergo apoptosis, resulting in defective angiogenesis. Previous data has also shown that ERK5 is required to protect endothelial cells against fluid shear stress (Pi et al., 2004). The requirement for ERK5 activity to protect endothelial cells from apoptosis has been ascribed to the need for ERK5-mediated phosphorylation of the myocyte enhancer factor 2C (MEF2C) transcription factor (Hayashi et al., 2004). This is based on the fact that the phenotype of Mef2c^–/– mice is similar to that of Erk5^–/–mice, with embryonic lethality resulting from cardiac and vascular malformations (Hayashi and Lee, 2004; Roberts et al., 2009). However, infection of Erk5^–/– embryos with an adenovirus encoding a constitutively active Mef2c was only able to partially protect endothelial cells from apoptosis (Hayashi et al., 2004), suggesting the existence of additional effectors downstream of ERK5 that regulate apoptosis (Olson, 2004).

Our data reveal that ERK5 is able to regulate VEGF signalling in human endothelial cells, providing a mechanism for the adverse effects on angiogenesis seen in murine Erk5^–/– embryos. Conditional knockouts of Erk5 in specific cell types including cardiomyocytes, hepatocytes and neuronal cells develop normally (Hayashi and Lee, 2004), suggesting that endothelial cells have a particular requirement for ERK5. Current anti-angiogenic drug therapy is centred around the development of inhibitors that target the VEGF ligand or VEGFR-2 (Holmes et al., 2007), which inhibit all the actions of VEGF on endothelial cells. Our results, showing that ERK5 specifically regulates tubular morphogenesis rather than proliferation, suggest that drugs targeting the ERK5 pathway in vivo may uniquely allow selective interference of a specific component of angiogenesis.

HDMECs and NHDFs were purchased from Promocell (Heidelberg, Germany). All cells were routinely grown on gelatin-coated cell culture dishes [1% w/v gelatin from porcine skin (Sigma, Poole, UK)]. HDMECs at passages (p) 2–8 were grown in EBM MV2 growth medium (Promocell) containing 5% v/v foetal calf serum (FCS), EGF (5 ng/ml), hydrocortisone (0.2 μg/ml), VEGF (0.5 ng/ml), FGF-2 (10 ng/ml), insulin-like growth factor-1 (20 ng/ml) and ascorbic acid (1 μg/ml). NHDFs (p 4–12) were grown in fibroblast growth medium (Promocell) supplemented with FGF-2 (1 ng/ml) and insulin (5 μg/ml). All cells were incubated at 37°C in humidified air containing 5% v/v CO₂.

Before stimulation with VEGF, HDMECs were serum-starved for 20 hours in EBM MV2 basal medium containing 1% (v/v) FCS. Cells were stimulated with 50 ng/ml VEGF-A₁₆₅ (Peprotech, Rocky Hill, NJ), washed in ice-cold Dulbecco's phosphate-buffered saline (DPBS) (Lonza, Basel, Switzerland) and lysed on ice in modified RIPA buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 2.5 mM EDTA, 10% (v/v) glycerol, 1% (v/v) Triton X-100, 0.1% (w/v) SDS, 0.5% (w/v) sodium deoxycholate, 1 mM sodium orthovanadate, 10 μg/ml aprotinin, 10 μg/ml leupeptin and 1 mM phenylmethylsulfonylfluoride). Lysates were clarified by centrifugation (23,000 g for 20 minutes at 4°C). For detection of cleaved caspase-3, cells were lysed in LDS sample buffer (Invitrogen, Paisley, UK) containing 5% (v/v) β-mercaptoethanol (Sigma). Protein extracts were sonicated for 10 seconds using a Soniprep 150 sonicator (Sanyo-Gallenkamp, Leicester, UK) before heat denaturation at 90°C for 5 minutes.

Proteins were resolved by SDS-PAGE on NuPAGE® 10% or 4–12% pre-cast gradient gels (Invitrogen) and transferred onto Hybond-C Extra nitrocellulose membranes (GE Healthcare, Amersham, UK). Membranes were blocked using 5% (w/v) bovine serum albumin (BSA) in Tris-buffered saline containing 0.01% (v/v) Tween 20 (TBST). Membranes were probed overnight with primary antibodies, diluted in TBST containing 2% (w/v) BSA, or 5% (w/v) non-fat milk, directed against: ERK5; AKT; Flag-Tag (DYKDDDDK Tag), HA-Tag (6E2), phospho-VEGFR-2 (Tyr1175); phospho-ERK5 (Thr218/Tyr220); phospho-p44/42 MAP kinase (Thr202/Tyr204); phospho-AKT (Thr308); phospho-AKT (Ser473); phospho-p70 S6 kinase (Thr389)(108D2); phospho-BAD (Ser136); Bcl-2 (Human Specific); and cleaved caspase-3 (Asp175) (New England Biolabs, Hitchin, UK). A primary antibody directed against MEK5 (H-94) was purchased from Santa Cruz Biotechnology (Santa Cruz, CA). An antibody directed against Actin (C-11) or Lamin B1 (H-90) (Santa Cruz Biotechnology) was used to assess protein loading in all western blot experiments. Proteins were detected using rabbit-specific secondary antibodies and ECL (enhanced chemiluminescence) western blotting detection reagents (GE Healthcare).

Total RNA was extracted using the RNeasy kit (Qiagen, Crawley, UK) according to the manufacturer's instructions. Harvested RNA was treated with DNaseI (Qiagen) before first-strand cDNA synthesis carried out by reverse transcription of 1 μg total RNA using oligo(dT)₁₈ and M-MLV Reverse Transcriptase (Invitrogen). Reaction mixtures for qRT-PCR contained 1.5 μl cDNA template, 4.5 μl distilled H₂O, 10 μl 2× Power SYBR® Green mastermix (Applied Biosystems, Warrington, UK) and 0.25 μM forward and reverse primers (Invitrogen) in each case. Primers used were: ERK5, Hs_MAPK7_1_SG QuantiTect Primer Assay (Qiagen); BCL2 forward, 5′-GTC TGG GAA TCG ATC TGG AA-3′, and BCL2 reverse, 5′-GCA ACG ATC CCA TCA ATC TT-3′; β-actin forward, 5′-ATG GAT GAT GAT ATC GCC GC-3′ and β-actin reverse, 5′-AAG CCG GCC TTG CAC AT-3′. qRT-PCR reactions were performed on a 7900 HT Fast Real-Time PCR System (Applied Biosystems, Foster City, CA) with the following parameters: 50°C for 2 minutes and 95°C for 10 minutes followed by 40 cycles at 95°C for 15 seconds and 60°C for 1 minute. Cycle threshold (C_T) values for the mRNA of interest and for the β-actin control were recorded, and changes in gene expression were determined by the Comparative C_T (2^-ΔΔC ) method (Livak and Schmittgen, 2001).

ERK5-specific short interfering RNA (siRNA) duplexes directed against the target sequences 5′-CAG ACC CAC CTT TCA GCC TTA-3′ for Hs_MAPK7_9_HP (designated ERK5 siRNA1) and 5′-CAG GAT GGC CAG GCA GAT TCA-3′ for Hs_MAPK7_5_HP (designated ERK5 siRNA2), and MEK5-specific siRNA directed against the target sequences 5′-AAG ACG TAT GTT GGA ACA AAT-3′ for Hs_MAP2K5_11_HP (designated MEK5 siRNA1) and 5′-CAA GAC GTA TGT TGG AAC AAA-3′ for Hs_MAP2K5_12_HP (designated MEK5 siRNA2), ERK1- and ERK2-specific siRNA duplexes, directed against the target sequences 5′-CTC CCT GAC CCG TCT AAT ATA-3′ for Hs_MAPK3_7 HP (designated ERK1 siRNA) and 5′-AAG TTC GAG TAG CTA TCA AGA-3′ for HS_MAPK1_10 HP (designated ERK2 siRNA) and a non-silencing (N.S.) control siRNA, directed against the sequence 5′-AAT TCT CCG AAC GTG TCA CGT-3′, were purchased from Qiagen. All siRNA duplexes were validated by qRT-PCR and western blotting prior to use (Holmes et al., 2010). HDMECs were seeded at 1×10⁵ cells per well, on gelatin-coated six-well dishes in 2 ml EBM MV2 growth medium, and incubated for 24 hours. Transfection was carried out in a total volume of 2.5 ml per well of OptiMEM medium (Gibco, Paisley, UK) containing 10 nM siRNA and 0.1% (v/v) Lipofectamine™ RNAiMAX (Invitrogen). Cells were incubated with the transfection mix for 4 hours before washing in DPBS containing Ca²⁺ and Mg²⁺ and the addition of EBM MV2 growth medium. Cells were incubated for a further 24 hours before lysis and immunoblotting. Alternatively, after washing, HDMECs were serum-starved overnight in EBM MV2 basal medium containing 1% (v/v) FCS for use in cell-based assays..

Wild-type Flag-tagged ERK5 (pcDNA3.1BMK1); dominant-negative Flag-tagged ERK5 [pcDNA3.1BMK1(AEF)] and constitutively active HA-tagged, MEK5 [pCMV5MEK5(D)] plasmids were kindly donated by Prof. Jiing-Dwan Lee (Scripps Research Institute, La Jolla, CA) (Kato et al., 1997). Constitutively active HA-tagged MEK5 was digested from pCMV5MEK5(D) with XbaI and KpnI (New England Biolabs) and subcloned into pcDNA3.1 (Invitrogen).

HDMECs were seeded on six-well gelatin-coated cell culture plates at 9.3×10⁴ and incubated for 24 hours. Growth medium was replaced with 2 ml fresh EBM MV2 growth medium immediately before transfection. Plasmid DNA (2.5 μg) was mixed with serum-free OptiMEM medium and transfected using TransPass™ HUVEC Transfection Reagent (New England Biolabs) according to the manufacturer's instructions. Cells were incubated with transfection mix for 4 hours, washed twice in DPBS followed by the addition of EBM MV2 growth medium. Cells were incubated for 24 hours, and serum starved in EBM MV2 medium containing 1% (v/v) FCS before use in assays.

PureCol™ type I collagen was purchased from Inamed Biomaterials (Leimuiden, The Netherlands). Collagen gels were prepared as described previously (Bohman et al., 2005). Serum-starved HDMECs were seeded at 1.25×10⁵ cells per well of a 24-well plate before the addition of the top layer of collagen, and stimulation with VEGF (50 ng/ml) for 20 hours. Length of tubular structures was quantified using NIH ImageJ software as described previously (Matsumoto et al., 2002).

NHDFs were seeded in fibroblast growth medium at 1.5×10⁴ cells per well on gelatin-coated 24-well plates and incubated for 3 days. On day 4, HDMECs treated with siRNA, or not, were seeded at 4.5×10³ cells per well, on top of the NHDF monolayer in EBM MV2 growth medium. On day 5, co-cultures were transfected with siRNA (as described above) for 4 hours, followed by washing in DPBS, and replacement of EBM MV2 growth medium. On day 6, cells were placed in EBM MV2 basal medium containing 1% (v/v) FCS with, or without, 50 ng/ml VEGF. On days 7 and 9, co-cultures were treated with siRNA, as described above. On day 8, cells were placed in fresh EBM MV2 medium containing 1% (v/v) FCS with, or without, 50 ng/ml VEGF. On day 10, co-cultures were fixed in ice-cold 70% (v/v) ethanol for 30 minutes before addition of blocking buffer (1% (w/v) BSA in DPBS) for 30 minutes. Monoclonal mouse anti-human CD31 antibody, (Dako Cytomation, Glostrup, Denmark) (1:750 dilution in blocking buffer) was added for 1 hour at 37°C, followed by three washes in blocking buffer. Anti-mouse IgG (whole molecule) alkaline-phosphatase-conjugated antibody (Sigma) was diluted 1:200 in blocking buffer and incubated for 1 hour at 37°C. Tubes formed from endothelial cells were visualised by the addition of the substrate SigmaFast™ BCIP/NBT (5-Bromo-4-chloro-3-indolyl phosphate/Nitro blue tetrazolium) solution (Sigma).

Tube formation in co-culture plates was analysed by photographing three randomly selected fields of view, in triplicate wells per condition, at low magnification (×4 objective) using a Nikon Eclipse TS 100 microscope attached to a Nikon DS-Fi1-L2 digital camera (Nikon, Kingston upon Thames, UK). Total tubule length was quantified using AngioQuant software (http://www.cs.tut.fi/sgn/csb/angioquant/) (Niemisto et al., 2005).

HDMECs undergoing tubular morphogenesis in 3D collagen gels were fixed with paraformaldehyde and stained with Phalloidin as described previously (Mellberg et al., 2009). When detecting cleaved caspase-3 within co-cultures, fixed co-cultures were blocked in blocking buffer [5% (v/v) normal goat serum (Vector Laboratories, Burlingame, CA) in DPBS containing 0.3% (v/v) Triton X-100] at 37°C for 30 minutes. Monoclonal mouse anti-human CD31 antibody (Dako Cytomation) was diluted 1:750, together with a 1:10 dilution of cleaved caspase-3 (Asp¹⁷⁵) antibody Alexa Fluor® 488 conjugate (New England Biolabs) in 1% (w/v) BSA in DPBS containing 0.2% (v/v) Triton X-100. Cells were incubated with the antibody solution overnight, in the dark, at 4°C with gentle rocking. Cells were washed, and incubated with Alexa Fluor® 568 goat anti-mouse IgG (Molecular Probes) diluted 1:200 in 1% (w/v) BSA in DPBS, in the dark, at room temperature for 2 hours. Finally, cells were washed in ddH₂O, and analysed using a Nikon TE2000 inverted fluorescence microscope.

HDMECs were seeded at 1.5×10⁴ cells per well in 24-well plates in EBM MV2 growth medium. After 24 hours, cells were washed in DPBS and serum-starved in 1 ml EBM MV2 medium containing 1% (v/v) FCS for 20 hours. Cells were stimulated, or not, with VEGF (50 ng/ml) and incubated for a further 72 hours. Cell numbers were quantified using a Cell Titer-Glo™ luminescent cell viability assay reagent (Promega, Southampton, UK), as previously described (Crouch et al., 1993). Luminescence was measured on a Wallac Victor² 1420 Multilabel Counter (Perkin Elmer Life Sciences, Fremont, CA).

Determination of the cell cycle distribution was carried out by FACS analysis as described previously (Mellberg et al., 2009).

Early apoptosis of HDMECs was detected as described previously (Dimberg et al., 2008).

HDMECs on gelatin or collagen matrices in 24-well plates were washed briefly in DPBS before complete cell lysis in Caspase-Glo® 3/7 reagent (Promega). Aliquots of the resultant lysates were transferred into separate wells of black-walled 96-well plates, and luminescence was measured using a Wallac Victor² 1420 Multilabel Counter.

Rapamycin and wortmannin were purchased from Calbiochem (Merck Chemicals Ltd, Nottingham, UK) and reconstituted in DMSO (Sigma). Cells were pre-incubated with indicated amounts of signal transduction inhibitors for the times shown in EBM MV2 basal medium containing 1% (v/v) FCS at 37°C, before stimulation with VEGF.

Results are expressed as mean ± s.d. Statistical analyses using an unpaired Student's t-test were performed using SPSS (version 16.0), with P<0.05 considered to be significant.

This study was supported by a CASE studentship from the Biotechnology and Biological Sciences Research Council (BBSRC) and AstraZeneca (O.Ll.R.); the North West Cancer Research Fund (NWCRF) (K.H., J.M. and M.J.C.).