Our results strongly support the hypothesis that Ang2 can be either an agonist or an antagonist of the Tie2 receptor, depending upon the context. Specifically, we found that cultured ECs respond to exogenous Ang2 alone by activating the Tie2 receptor. We observed that ECs secrete Ang2 in the conditioned medium and that blockade of this autocrine loop lowers the basal activation of Tie2 and its downstream pathway, PI3K/Akt. Phenotypically, inhibition of this loop results in less resistance to apoptosis and a reduced migratory response. We asked how the agonistic effect of Ang2 compares to that of the more widely accepted naturally occurring Tie2 agonist, Ang1. We found that Ang2 is a less potent activator of Tie2 and that it binds Tie2 with lower affinity than Ang1. Either Ang1 or Ang2 can render ECs resistant to serum deprivation-induced apoptosis, but as expected, Ang1's effect is more potent. Finally, we explored what happens when Ang1 and Ang2 are applied together to ECs. Rather than observing an additive stimulation of Tie2, we found that Ang2 inhibited Ang1-induced Tie2 activation and reduced Ang1-mediated protection against EC apoptosis in a dose-dependent fashion. We conclude, therefore, that cultured ECs rely on self-produced Ang2 for trophic effects but that in the presence of the stronger Tie2 activator Ang1, Ang2 can actually dampen Tie2 signaling. Thus, we propose that Ang2 is a partial agonist/antagonist of Tie2.
The original report describing Ang2 found that it was a naturally occurring antagonist of Tie2 in ECs (28
). In contrast, more-recent reports have shown that excess Ang2 can induce Tie2 activation (8
). Kim et al. first reported that a high dose of Ang2 (800 ng/ml) could induce Tie2 phosphorylation in HUVECs in vitro and protect HUVECs from serum deprivation-induced apoptosis, probably via the PI3K/Akt pathway. However, no obvious effect was noted when less than 400 ng/ml of Ang2 was applied (22
). Ang2-induced Tie2 phosphorylation was then confirmed in HUVECs (>300 ng/ml of Ang2) and in murine brain capillary cells (31
). In those studies, Ang2 was reported to promote capillary-like endothelial tube formation in three-dimensional culture systems. Ang2 was found to have less potency than Ang1 in HUVECs (41
). Again, the PI3K/Akt pathway was reported to be involved. Our results confirm that Ang2 can activate Tie2 and extend our knowledge about this activity by showing that such activation can occur with lower doses of Ang2 (200 ng/ml) than previously reported and that the activation of Tie2 by Ang2 is not significantly dose responsive above 800 ng/ml.
While several groups have reported that exogenous Ang2 can activate Tie2, we also found that endogenous production of Ang2 is vital to the health of ECs. ECs are known to synthesize, store, and secrete Ang2 but not Ang1 (12
). We found that disruption of this autocrine loop by “extracellular” (addition of sTie2 or Ang-2-specific neutralizing antibody to culture medium) or “intracellular” (two different siRNAs against Ang2 transcript) methods reduced Tie2 phosphorylation by 60 to 80%. In addition, the levels of pTie2 in those cells show a close correlation with Ang2 levels in either culture medium or cell lysates (data not shown). ECs unable to engage in this tonic autoactivation developed reduced signaling along the critical PI3K/Akt pathway, were less resistant to apoptosis, and were unable to migrate as well. Furthermore, such cells were markedly impaired in their ability to form tubes (see Fig. S3 in the supplemental material). A similar situation arises in vivo in lymphatic vessels, where Ang1 is lacking due to the absence of periendothelial cells. Lymphatic vessel formation is highly dependent on Ang2/Tie2 signaling, as illustrated by gross lymphatic patterning defects in Ang2
null mice (15
Two other groups have reported autocrine actions of Ang2 on ECs expressing Tie2 (8
). Sharpfenecker et al. (38
) found that extracellular inhibition of Ang2 with sTie2 was ineffective at blunting autocrine Ang-2 signaling, leading them to propose an “intracrine” effect of Ang2. However, the results are not directly comparable to ours for two reasons: (a) they never knocked down endogenous Ang2 expression or used Ang2 null ECs, and (b) they were using a coculture system of ECs grown atop vascular smooth muscle cells, a point to which we will return below. On the other hand, Daly et al. (8
) found that Ang-2 blocking antibody and Ang2 siRNA were equally capable of disrupting autocrine Ang2-induced Tie2 activation. However, they observed this autocrine Ang2/Tie2 loop only in ECs that were “stressed” either by chemical inhibition of PI3K or by virally mediated overexpression of the transcription factor FOXO1. Our results clearly demonstrate that this loop also exists in and is vitally important to “unstressed” ECs under more typical culture conditions.
In comparing Ang1 and Ang2, we found that both ligands can activate Tie2 but that Ang2's signaling effect is less potent (e.g., 11.5-fold Tie2 activation with 200 ng/ml of Ang1 versus 7.5-fold with 400 ng/ml Ang2). We also observed that Ang2 binds Tie2 with ~20-fold-lower affinity than does Ang1. Though perhaps not the sole explanation, the lower binding affinity of Ang2 for Tie2 probably does contribute to its diminished potency at the receptor. Mixing experiments did not show an additive effect on Tie2 activity but rather showed that Ang2 dose-dependently inhibited Ang1 signaling. The reduction in Tie2 phosphorylation was biologically relevant, as evidenced by the progressive loss of Ang1-mediated protection from apoptosis as the Ang2 dose was increased. The mixing study is consistent with the results reported for ECs from the original Ang2 article (28
). Finally, we found that addition of either Ang1 or Ang2 inhibits the binding of the other ligand to Tie2, suggesting competition between the angiopoietins for the same binding pocket on Tie2. When excess Ang-2 is partially unseated by Ang1, Tie2 phosphorylation rises because Ang1 is the more potent agonist; when Ang1 is unseated by Ang2, Tie2 phosphorylation falls for the same reason. Figure S4 in the supplemental material shows that depletion of endogenous Ang2 by two different siRNAs results in increased Ang1-dependent Tie2 phosphorylation. This finding further supports our chief conclusion that Ang2 acts as a Tie2 antagonist when Ang1 is present.
To our knowledge, this is the first description of an endogenous partial agonist/antagonist for a receptor tyrosine kinase. Small-molecule partial agonists have been described for opioid, adrenergic, among others (11
). One non-G-protein-coupled-receptor example of partial agonist compounds is the selective estrogen receptor modulator tamoxifen (2
). However, we are aware of no other endogenous multiligand-single-receptor system that provides a precedent for the actions mediated by Ang2. As such, the Ang/Tie2 system offers investigators a novel opportunity to study the endogenous mechanisms that modulate signal transduction through receptor tyrosine kinases.
Ang2 may be a “dummy” ligand that weakly activates Tie2, but in high enough concentrations, it can “unseat” local Ang1 from its receptor. Others have repeatedly shown that Ang1 is highly multimerized in its native form whereas Ang2 is less aggregated (4
). The multimerization of Ang1 may account for its ability to strongly activate Tie2 (25
). Ang1 aggregates would theoretically have reduced entropy for binding Tie2 monomers, since the binding of any one Ang1 molecule to any one Tie2 favors the binding of additional noncovalently tethered Ang1 molecules to nearby Tie2 monomers. In this way, Ang1 aggregates would favor the clustering and cross-phosphorylation of cell surface Tie2. An excess of Ang2 in the presence of Ang1 could disrupt Ang1-Tie2 clusters and favor the formation of lower-order multimers—e.g., dimers and trimers—of Ang2-Tie2, resulting in some degree of receptor activation but significantly less than that achieved by large clusters of Tie2 induced by comparable concentrations of Ang1. In this fashion, Ang2 could antagonize Ang1 signaling. However, in a system devoid of Ang1, Ang2 would weakly activate Tie2.
Close inspection of our results also suggests that differences between Ang1/Tie2 and Ang2/Tie2 signaling may exist not only at the level of the receptor but also in the strength of the downstream signal. Whereas 400 ng/ml of Ang2 induces Tie2 phosphorylation to ~65% of 200 ng/ml of Ang1 (Fig. ), the effect of the same doses on Akt phosphorylation is significantly less (Fig. ). This suggests not only that Ang2 is a lower-affinity binder of Tie2 with a less-potent phosphorylation effect at the receptor but that postreceptor signaling is also attenuated compared to Ang1. While the reason for this is unclear and merits further investigation, several possibilities exist. First, our results do not exclude the possibility that Ang2 and Ang1 promote phosphorylation of different sets of tyrosine residues in the cytoplasmic tail of Tie2. For example, tyrosine 1100 is required for the recruitment of the SH2-domain-containing p85 subunit of PI3K (21
). Interestingly, treatment with PI3K inhibitors only partially abrogates the Ang1-signaled migratory response (21
), suggesting the involvement of other tyrosines in this response, such as tyrosine 1106, which may recruit Dok-R and activate intracellular GTPases that favor migration (20
). A second possibility is that the enhanced clustering of Tie2 monomers by Ang1 versus Ang2 may increase the cell surface localization of PI3K (10
). The enzymatic components of this cascade will tend to amplify small upstream differences as signals are transduced downstream, which could explain how a small difference in Tie2 activation leads to progressively larger differences in PI3K and Akt activation. Last, Ang1 is known to have at least one non-Tie2 receptor (6
). We cannot rule out that coreceptors on HUVECs are partly responsible for differences in signal transduction efficiency.
Ang1 is made primarily by periendothelial cells (i.e., pericytes and vascular smooth muscle cells), whereas Ang2 is synthesized by ECs. Given that Ang1 is more matrix bound than Ang2 and that circulating Ang2 concentrations are generally low (<2 ng/ml) in unperturbed individuals (17
), one could expect a high degree of basal Tie2 activation. This is indeed the case in organs throughout the body (44
). In the Ang1-rich in vivo setting of blood vessels, local increases in Ang2 could precisely attenuate the tonic activity of Tie2. The ability to carefully decrease Tie2 activation would be beneficial in certain settings—i.e., to destabilize formed vessels, to promote local inflammation, or to permit a local increase in vascular permeability.
The finely tuned balance of endogenous Tie2 ligands may be upset in diseases like cancer and sepsis. Solid tumors can strongly upregulate the production of Ang2, in part through the induction of hypoxia-inducible factors in the tumor cells (32
). Extra Ang2 may contribute to local downregulation of Tie2 activity. In turn, this could enable the necessary destabilization of blood vessels that initiates tumor angiogenesis. In fact, a compound related to the Ang2 neutralizing antibody used in this report also possesses potent antiangiogenic and antitumor effects in murine xenograft models (34
In sepsis—a syndrome of disseminated infection that results in several vascular changes, including increased permeability and endothelial inflammation—we and others have found high levels (20 to 200 ng/ml) of circulating Ang2 (17
) and have observed relative Tie2 inhibition in animal models of sepsis (unpublished results), implying that the rise in Ang2 is antagonizing local Ang1/Tie2 signaling. The observation that heavily supraphysiological circulating levels of Ang2 in vivo (~10 μg/ml) can also activate Tie2 (8
) highlights the importance of regulating Ang1 and Ang2 within a narrow range. One implication of the present work is that reduction of circulating Ang2 (e.g., by a neutralizing antibody) in a disease such as sepsis would not be expected to have as positive an impact on Tie2 activity as administration of excess Ang1.
This study has several limitations. Our data suggesting an agonist role for Ang2 are derived from in vitro studies in which Ang1 expression was absent or barely detectable. Future studies using coculture of ECs with vascular smooth muscle cells or an endothelial cell-specific knockout of Ang2 in mice should further clarify the role of Ang2 in the maintenance of vascular homeostasis. Second, the biochemistry underlying the ability of a ligand to act as both an agonist and an antagonist merits further investigation. Third, additional cell biological responses in Ang1-Ang2 mixing experiments should be measured, such as changes in monolayer permeability or surface expression of inflammatory adhesion molecules. Related to this, Ang1 and Ang2 may activate distinct postreceptor signaling pathways within ECs. Finally, coreceptors or other angiopoietin receptors, such as Tie1, may be involved in the modulation of signals through Tie2 (48
In conclusion, using one model system, we have found that Ang2 can act as either a weak Tie2 agonist or a dose-dependent Tie2 antagonist. To our knowledge, this is the first description of an endogenous partial agonist/antagonist system targeted to one receptor. Ang2 is an agonist only in the absence of Ang1, and Ang1-deprived ECs rely on autocrine Ang2 stimulation for cell survival and an appropriate migratory response. Ang2 binds Tie2 with less affinity than Ang1, and the latter is a more potent activator of Tie2. When they are combined, Ang2 inhibits Ang1 signaling and downstream effects, thus becoming an antagonist. These findings help reconcile divergent reports on the biological actions of Ang2 and have significant implications for the design of therapies targeted to the angiopoietin-Tie2 pathway.