Activation of Endothelial Ras Signaling Bypasses Senescence and Causes Abnormal Vascular Morphogenesis

Angiogenesis occurs during tissue remodeling, wounding, and inflammation, as well as in tumors and their activated stroma. The quiescent vascular endothelium undergoes phenotypic changes, including increased proliferation, migration, and ultimately tubulogenesis. These changes are highly coordinated, requiring input from a number of signaling molecules (1). The loss of proper regulation can have devastating consequences.

During development, abnormalities in angiogenesis are lethal to the embryo. In the postnatal state, abnormal control of vascular growth and remodeling results in vascular malformations, hemangiomas, and angiosarcoma. Abnormalities in vascular morphogenesis and repair may be more discreetly involved in vascular diseases such as stroke, aneurysms, and atherosclerosis (2). Emerging data suggest that the abnormal vasculature found in tumors can directly affect metastasis (3). Thus, an understanding of the signaling that controls proper vascular morphogenesis has the potential to affect a wide range of conditions, including tumor progression.

The small GTPase Ras is a potentially critical regulator of proper endothelial cell growth and differentiation. In humans, loss-of-function mutations in RASA1, which normally downregulates Ras, have been associated with vascular malformations (4). Patients affected with the genetic disease NF1, resulting in the loss of neurofibromin, another negative regulator of Ras, suffer from a range of cardiovascular abnormalities, in addition to their enhanced predisposition to a range of tumors (5). Existing studies in animals bolster the importance of Ras regulation in the proper establishment and maintenance of a vasculature. Mice engineered without NF1 or RASA1 suffer from embryonic lethality (6, 7) primarily from their loss of function in the vasculature, as the lethal phenotype is recapitulated by endothelial-specific knockout of these proteins (8). Mice haploinsufficient for NF1 show enhanced angiogenesis in a corneal and retinal neovascularization assay (9). The importance of normal Ras regulation in the growth control of endothelial cells is underscored by the finding that devastating tumors of endothelial cell origin, angiosarcomas, both sporadic and following exposure to vinyl chlorides, are associated with Ras mutations (10, 11). Collectively, these data suggest that unregulated Ras can compromise normal vascular growth control, morphogenesis, and function.

Despite evidence at the organism level for Ras as a critical regulator of endothelial function, limited data are available on the consequences of Ras activation at the cellular level. In other cell types, unopposed signaling from Ras is often accompanied by a cell cycle arrest or senescent phenotype, which safeguards against uncontrolled growth. This “oncogene-induced senescence” has been shown both in vivo and in vitro in an array of cell types, including melanocytes (12), fibroblasts (13), lymphocytes (14), and mammary epithelium (15). To understand the effects of sustained Ras signaling in primary endothelial cells, we used sustained and inducible expression of activated H-Ras. Surprisingly, our data suggest that human endothelial cells bypass Ras-induced senescence, with resulting autonomous growth, enhanced survival, as well as a markedly compromised ability to organize properly into vascular structures.

Human umbilical vein endothelial cells (HUVEC) were from Cascade Biologics or VEC Technologies. Cells were cultured as previously described (16). Cells were made quiescent in serum-free MCDB-131 supplemented with 1% penicillin/streptomycin and 2 mmol/L l-glutamine (serum-free), where noted. Stimulation was performed with complete growth media. Primary fibroblasts were isolated from human foreskins and grown in DMEM containing 10% fetal bovine serum and 1% penicillin/streptomycin. Normal human epidermal melanocytes were generously provided by Dr. Andrew Aplin (Thomas Jefferson University, Philadelphia, PA) and cultured in MGM-4 media purchased from Lonza. The endothelial and fibroblast coculture assay was conducted in EGM-2 media from Lonza. Human dermal microvascular endothelial cells (HDMVEC) were isolated from human foreskins following trypsinization, disassociation, and CD31 affinity purification with magnetic beads (Dynal). Isolated cells were cultured in EGM-2 MV from Lonza. Except as noted, all experiments were done with HUVECs.

Western blotting used the following antibodies: mouse anti–phospho-extracellular signal-regulated kinase (ERK), rabbit anti-ERK2, mouse anti-p16, and mouse anti-p53 (Santa Cruz Biotechnology); rabbit anti–phospho-Akt (Ser⁴⁷³), rabbit anti-Akt, mouse anti–cyclin D1, mouse anti-p21, rabbit anti–cleaved caspase-3, and rabbit anti–total caspase-3 (Cell Signaling Technology); mouse anti-hemagglutinin (HA; Covance) and mouse anti–pan-Ras (Oncogene Research, Calbiochem); and mouse anti-FLAG (Sigma). All antibodies were used at a dilution of 1:1,000 overnight at 4°C. Other conditions were the same as in ref. 16, except exposures were captured on a Kodak 4000 MM imager. All figures are representative of at least three independent experiments.

Cells were infected with either tetracycline-inducible lentivirus or retrovirus as indicated. Details are provided in Supplementary Materials and Methods.

Endothelial cells were serum starved for 24 hours, after which complete growth media were added as a mitogenic stimulus for 16 hours. Measurements of bromodeoxyuridine (BrdUrd) incorporation were performed as previously described (17). Growth assays were conducted as described previously (13). Population doublings were calculated using the following formula: population doublings = log(final cell number/initial cell number)/log2. Cumulative population doublings represent the sum of population doublings from all previous passages.

This assay was performed as previously described (18) with the following modification. Endothelial cells were mixed with fibroblasts (1:20) and seeded onto a gelatin-coated 12-well plate in EGM-2 media for 12 to 28 days (media replaced after every 2–3 days). Cells were stained live with a FITC-tagged UEA-1 lectin (Sigma-Aldrich) or fixed in 3.7% formaldehyde and visualized with either UEA-1 lectin or anti-CD31 (clone JC70A, 1:50; Dako).

To generate endothelial cells expressing activated Ras, we cloned a HA-tagged H-Ras^V12 mutant into two different viral vectors. One construct, a retroviral vector, permitted expression levels of activated Ras at levels close to the endogenous Ras (Supplementary Fig. S1B), and Ras activation was readily detected in a pull-down assay under quiescent conditions (Supplementary Fig. S1C). Recent results suggest that oncogene-induced senescence may be a function of oncogene expression levels (15). To regulate expression (and expand cells in the event Ras induced senescence), we also engineered an inducible lentivirus where expression of Ras was regulated by the stronger Tet-Op7-CMV promoter (Fig. 1A). This permits overexpression at levels at least 10-fold higher than endogenous (Fig. 1D). Importantly, this provided excellent regulation of the transgene, with no detectable expression of the HA epitope in the absence of induction (Fig. 1C). Induction with doxycycline ranging from 0.05 to 1 μg/mL allowed robust Ras expression in a dose-dependent fashion (Fig. 1C). We used a dose of 50 ng/mL doxycycline in subsequent experiments.

We next determined the signaling downstream of Ras in these cells. Both ERK and Akt are known to regulate cell responses critical to angiogenesis, including growth and survival (16, 19, 20), and are often activated in a Ras-dependent manner (16). However, chronic Ras signaling has been described to downregulate these signals in some cell types (21). We found that induction of activated Ras was sufficient to promote ERK and Akt activation that was maintained over at least a 3-day period (Fig. 1D). We observed enhanced Ras-induced signaling events even after periods of prolonged cell culture, including cells infected with low-copy number retroviruses (data not shown; Supplementary Fig. S1B). These data suggest that negative feedback of Ras signaling, implicated in inducing cellular senescence in some cell types (21), does not occur in endothelial cells. Reports also suggest that Ras activation in some primary cells activates p38 mitogen-activated protein kinase (MAPK), resulting in senescence (22–24). However, no sustained Ras-induced p38 activation in primary endothelial cells was observed (data not shown). Thus, several signals responsible for senescence in other cell types do not seem to occur in endothelial cells.

As expression of active Ras showed no downregulation of growth signaling, we reasoned that a senescence checkpoint might not be activated in primary endothelial cells following Ras activation. To directly test this, we infected primary melanocytes (a cell type well described to undergo Ras-induced senescence; ref. 12) with an inducible Ras-expressing lentivirus and compared them to Ras-expressing primary endothelial cells. We monitored population doubling and found that induction of active Ras expression resulted in slowed growth of the melanocytes within 7 days of culture and near-complete senescence by day 21 (Fig. 2A), as reported (12). This was accompanied by increases in the cyclin-dependent kinase inhibitor p16 (Fig. 2B). In contrast, melanocytes left uninduced continued to grow with similar population doubling to endothelial cells and minimal induction of p16 (Fig. 2A and B). Induction of activated Ras in human endothelial cells resulted in no appreciable change in the population doubling and no accumulation of p16. In addition, no significant increase in p21^CIP/WAF or p27^kip, proteins often associated with Ras/Raf-mediated cell cycle arrest, was detected (data not shown). In addition, we have also investigated chronic exposure to lower Ras levels using retrovirus infections (see Supplementary Fig. S2). Lastly, similar results were observed in HDMVECs (Supplementary Fig. S6). These data argue that sustained Ras signaling does not induce proliferative senescence in human endothelial cells. Importantly, Ras expression was accompanied by increased β-galactosidase staining and elevated levels of plasminogen activator inhibitor-1 (Supplementary Fig. S3; data not shown), two commonly used cell senescence markers (13, 25). These results suggest that these markers, although perhaps markers of cellular stress or metabolic status, are not reliable markers of replicative senescence.

Given the linkage between Ras mutations and endothelial cell tumors, we sought to determine if Ras was sufficient to induce the proliferation of primary endothelial cells. Endothelial cell proliferation was measured at the G₁-S transition using a BrdUrd incorporation assay. We found that in the absence of added mitogens Ras activation was sufficient to drive cells into S phase (Fig. 2C; Supplementary Fig. S2A). This was accompanied by increased cyclin D1 levels (Fig. 2D). These data suggest that activation of Ras might make endothelial cells growth factor independent, a trait often associated with cell transformation.

To determine if Ras-expressing cells were actually progressing through the cell cycle, we performed growth assays where cells were plated in growth factor–free media and counted at various times after plating. As shown in Fig. 3A and B, cells induced to express active Ras increase in number. In contrast, uninduced cells quickly begin to decline. Importantly, the expression of Ras does not render these cells completely growth factor independent, as beyond 3 days even the Ras-expressing cells begin to succumb (data not shown). Furthermore, Ras-expressing endothelial cells maintained normal contact inhibition, did not grow in soft agar, nor did they form tumors in nude mice (data not shown), suggesting that other growth-regulatory checkpoints are still intact.

Activated Ras seemed to enhance survival of endothelial cells, allowing them to survive and grow autonomously in basal media, whereas control cells failed to thrive (Fig. 3A and B; Supplementary Fig. S2B). To determine if Ras activation promoted cell survival, we used culture conditions known to be proapoptotic and directly measured the cleavage of caspase-3. The uninduced cells accumulated cleaved caspase-3 when cultured in M199 medium lacking serum. In contrast, Ras-expressing cells showed little induction of cleaved caspase-3 over the same time period (Fig. 3C and D).

Previous studies have implicated the Ras-regulated signaling pathways, phosphoinositide 3-kinase (PI3K) and ERK, as important contributors to the growth and survival of human endothelial cells (16, 17, 26, 27). To better assess the roles these signals play in our observed response to Ras, we used two well-characterized chemical inhibitors: U0126 (an inhibitor of MAPK/ERK kinase) and LY294002 (an inhibitor of PI3K). As shown in Supplementary Fig. S4A, these inhibitors show inhibition of Ras-induced signaling events specific for the targeted signal. We found that proliferation induced by Ras was strongly reduced following ERK inhibition, in contrast to PI3K inhibition (Supplementary Fig. S4B), where the effect was less pronounced. The suppression of caspase activation by Ras was attenuated by both inhibitors to similar levels, suggesting that both ERK and PI3K control signals promoting cell survival (Supplementary Fig. S4C).

Recently, Serban and colleagues (28) used adenoviral delivery of Ras effector mutants with selective signaling properties to induce angiogenesis in vivo. To genetically probe the contribution of effector pathways, we engineered endothelial cells to stably express Ras^V12S35 (Raf activation) and Ras^V12C40 (PI3K activation). Interestingly, we found that neither of these mutants was sufficient to drive endothelial cell proliferation. However, we noted that the activation of downstream effectors in these cells was quite weak compared with Ras^V12 (data not shown). To insure optimal signal strength and to directly test the contributions of the Raf and PI3K pathway to the Ras-induced changes in endothelial cells, we also engineered inducible lentiviruses to express either activated Raf (ΔN-Raf1) or an activated form of Akt (myrAkt). As shown in Fig. 4A, these showed excellent expression when induced with doxycycline, with a corresponding change in the associated cellular signaling comparable with changes seen on stimulation with growth media, except for the chronic nature of the signal. As shown in Fig. 4B, activation of neither Raf nor Akt is sufficient to induce G₁ to S transition, consistent with results obtained with the effector mutants. Importantly, there was no cell cycle arrest response, as the BrdUrd incorporation induced by growth medium was normal. These data suggest that ERK and PI3K signaling likely works in combination to drive cellular proliferation. Moreover, these data suggest that an imbalance in these signals does not result in a proliferative arrest in endothelial cells, as reported in other cell types (29, 30).

We also investigated whether the isolated signals were sufficient to alter the survival responses of the endothelial cells under apoptotic conditions. We found that expression of activated Raf1 failed to prevent cleavage of caspase-3 under conditions of serum deprivation. In contrast, caspase-3 cleavage was completely prevented by the induction of an activated Akt (Fig. 4C), consistent with the established role of this protein in endothelial survival (31).

The formation of blood vessels requires integrated growth, spatial orientation, and vascular pruning (32). As Ras-expressing cells showed abnormal growth and survival characteristics, we hypothesized that their ability to properly regulate the formation of vascular-like structures might be altered. To test this, we used a coculture assay (18). This long-term assay results in vascular structures, with morphologic features nearly indistinguishable from in vivo blood vessels, including the formation of lumens (see Supplementary Fig. S7), tight junctions, and expression of basement membranes proteins such as collagen IV (33). Whereas uninduced cells formed interconnected, branching, and looping structures, Ras-expressing endothelial cells failed to form vascular structures. Rather, these cells formed extensive planar sheets with no evidence of branching, spatial reorganization, or pruning (Fig. 5A; Supplementary Fig. S2C). This process was reversible and dependent on Ras signaling, as turning off Ras expression in the same cocultures causes the sheets to regress and tubes to form. Similarly, addition of doxycycline to induce Ras expression results in formation of sheets by cells that previously had formed vascular structures (Fig. 5B). Thus, sustained Ras signaling seems to inhibit normal vascular differentiation and induction of Ras signaling in fully differentiated structures is sufficient to induce a loss of normal morphogenesis and drive the establishment of planar sheet-like structures. Recent experiments showed that adenoviral H-Ras^V12 induced vascular permeability in vivo (28), suggesting that there may be effects of Ras activation on endothelial cell junctions. The alteration of junctional proteins has been linked to abnormal vasculogenesis. To test whether stable expression of activated Ras results in changes in endothelial cell junctional integrity, we measured changes in electrical resistance by electric cell-substrate impedance sensor (Applied BioPhysics, Inc.; ref. 34). Our data suggest that there is no defect or enhancement in junctional integrity. Similarly, we found that VE-cadherin was expressed at cell-cell junctions, although an increased level of disorganization in Ras-expressing endothelial cells was evident (Supplementary Fig. S5).

To gain insight into mechanisms underlying altered morphogenesis in the coculture assay, we used inhibitors of ERK (U0126) and PI3K (LY294002) signaling. As this phenotype was reversible, we added inhibitors to the mature structures that formed following Ras induction to investigate whether they would revert to a tubular-type phenotype. Both inhibitors visibly reduced the overall number of endothelial cells. However, we found that the addition of the PI3K inhibitor induced a phenotype shift whereby vascular structures appeared (Fig. 6A). In contrast, although the ERK inhibitor reduced the number of endothelial cells, those cells remained as highly planar structures (Fig. 6A). We next tested this with endothelial cells stably expressing Ras effector mutants: H-Ras^V12S35 known to selectively activate ERK and H-Ras^V12C40 known to selectively activate PI3K signaling (28). Consistent with the inhibitor studies, expression of H-Ras^V12S35 mutant formed relatively normal branching structures, whereas cells expressing H-Ras^V12C40 formed planar sheets (Fig. 6B). These data suggest that sustained PI3K signaling may be a principal contributor to the failed morphogenesis following sustained Ras signaling. We evaluated this further using inducible expression of either activated Raf or activated Akt. As shown in Fig. 6C, we find that expression of activated Raf results in normal morphogenic responses. In contrast, induction of activated Akt results in severely compromised morphogenesis. The expression of myrAkt results in an increase in areas of planar sheet-like structures with a small number of interspersed elongated structures. Whereas a few areas seem to undergo some degree of normal morphogenesis, the normal elongation and branching seems to be severely compromised, with an appearance of many multipolar cells (Fig. 6D). Collectively, these data argue that proper regulation of PI3K and Akt signaling is an essential component of vascular morphogenesis.

Oncogene-induced senescence was first shown in primary human fibroblasts (13), and subsequently, multiple reports describing the occurrence of this phenomenon both in vitro and in vivo (12, 14, 21, 35, 36) have appeared. Contrasting evidence showing oncogene-induced transformation and increased proliferation of primary cells also exists (37, 38). These variations can be due to differences in the inherent abilities of particular cell types to respond to oncogene activation; however, few studies have compared two cell types simultaneously. An alternative is that the extent and duration of oncogene activation dictates the decision of a cell to proliferate or prematurely growth arrest. This has been observed in vitro in NIH3T3 fibroblasts (39) and in vivo in the mammary epithelium, where low levels of activated oncogene cause proliferation and high levels induce growth arrest (15). In this study, we show that stable and sustained Ras activation, independent of expression levels, leads to senescence bypass in primary endothelial cells while simultaneously inducing growth arrest in primary melanocytes. These data strongly argue that the induction of senescence by Ras is likely to be dependent on specific factors relevant to each individual cell type.

Senescence response in melanocytes has been primarily attributed to the accumulation of the tumor suppressor p16^INK4a levels (12). We found no induction of p16 in endothelial cells, whereas the senescent melanocytes readily accumulated p16. In addition to p16, Ras-induced senescence in primary fibroblasts is also accompanied by the accumulation of tumor suppressor p53 (13). However, we find that p53 is not induced on Ras activation (data not shown) nor is its target gene p21^CIP/WAF, suggesting that p53 regulation may be differentially regulated among cell types. We also have not observed several other causative mechanisms described to induce premature senescence [e.g., the attenuation of Ras signaling events (21) or activation of p38 (24)]. As vascular heterogeneity is well established, we considered that cells from different vascular beds might show heterogeneous responses. However, we found that all of the key findings were also observed in primary cultures of HDMVECs, a cell type relevant to the formation of hemangioma and angiosarcoma (Supplementary Fig. S6). These results highlight a unique and conserved response of endothelial cells to Ras activation. Importantly, although the inherent response of the endothelial cell in isolation may be the same, the varying microenvironment of the tissue may influence the predisposition of the endothelium to manifest an altered phenotype.

Cellular senescence acts as an initial barrier preventing a benign tumor from progressing to a malignant state (40). Interestingly, senescent cells are only detected in benign but not metastatic tumors (12, 41), supporting the notion that senescence functions as a mechanism of tumor suppression. Our findings suggest that this may not be an effective checkpoint in endothelial cells and could explain the prevalence of Ras mutations in endothelial tumors (42, 43). Endothelial tumors are characterized by rapidly proliferating endothelial cells (44) that have an atypical morphology and may appear as cellular sheets (45). Our findings show that sustained Ras expression in primary endothelial cells results in enhanced proliferation, autonomous growth, and enhanced endothelial cell survival. Moreover, the ability to organize into properly formed vascular structures is compromised, with cells forming sheet-like structure like those seen in some endothelial tumors (45). Our data do not support the notion that acquisition of Ras mutations would be sufficient to induce cellular transformation, as several additional growth checkpoints were intact.

Abnormal vessels are prevalent in disorders linked to increased Ras activity. Mutations in RASA1 and NF1 genes, which downregulate Ras, have been linked to familial hemangiomas, aneurysms, cerebral vascular malformations, and an increased risk of stroke (4, 5, 46). Our data suggest that pro-proliferative signaling, defective apoptosis, as well abnormal morphogenic programming could contribute to disease conditions. Moreover, Ras mutations, even without gross manifestations, might contribute to an abnormal and unstable capillary vasculature. Recent data have shown that short-term adenoviral infection with activated Ras constructs results in enhanced angiogenesis in a mouse ear model. The adenoviral infection in these experiments targets both the endothelial and the stromal compartments, and the time of expression is limited (28); this is an important distinction because we find that the morphogenic phenotype is reversible. Thus, microenvironmental cues, expression levels, and duration of signaling may all contribute to a failure to form normal vascular structures.

We find that Ras is sufficient to drive proliferation in the absence of mitogens and this requires ERK signaling. However, our data clearly show that activation of Raf/ERK signaling through either Ras effector mutants or activated Raf is not sufficient to drive proliferation or support survival signaling. However, these sustained signals do not interfere with normal vessel morphogenesis. In contrast, chronic activation of PI3K signaling seems to alter morphogenesis. This finding is consistent with results from the transgenic expression of activated Akt in mice (26). These mice have altered vascular density, morphogenesis, and permeability. Inhibition of PI3K activity is associated with a restoration of a normalized vasculature in these mice. Prosurvival signaling alone may not be the cause of the morphogenic defects following chronic Ras signaling; rather, it seems that a dynamically regulated change in cellular programming is occurring. This is based on several observations. (a) Treatment with U0126 inhibits Ras-induced survival and decreases the cell number in the morphogenesis assay with no effect on the defect in morphogenesis induced by Ras. (b) Activation of Akt completely protects cells from apoptotic stimuli; however, partial morphogenic responses are seen in these cells, albeit severely compromised. (c) The H-Ras^V12C40 mutant induces a defect similar to H-Ras^V12, although the levels of Akt activation are much lower. Thus, PI3K-related signals, in addition to Akt, may be involved.

Aberrantly formed blood vessels are known to occur in tumors and have been implicated in metastasis (3). It seems feasible that sustained and high concentrations of angiogenic growth factors could result in sustained activation of Ras signaling. The promutagenic microenvironment of the tumor bed undergoing therapeutic interventions (rapid proliferation in the presence of mutagens and often radiation) could also result in de novo acquisition of endothelial Ras mutations. Indeed, cytogenetic abnormalities in tumor endothelial have been detected (47) and suggested as a potential barrier to therapy. Thus, mitigating or minimizing Ras signaling in the tumor vasculature may be an effective and important adjuvant therapeutic approach to vascular normalization.

No potential conflicts of interest were disclosed.

We thank C. Michael DiPersio for critical reading of the manuscript and Joseph E. Mazurkiewicz for his help with acquiring confocal images.

Grant Support: USPHS grants CA81419 (K. Pumiglia) and HL077870 (P. Vincent), American Heart Association grant 0825888D (A. Bajaj), and David E. Bryant Trust.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.