Using structure-based in silico screening in combination with cell-based assays, we report here the identification and characterization of doxazosin as a novel EphA2 agonist that is independent of its α1-adrenoreceptor antagonist functions. Doxazosin directly bound to the recombinant EphA2 LBD with µM affinity and induced phosphorylation of EphA2 at similar doses in breast and prostate cancer cells, as well as glioma and hepatoma cells. Similar to native ligand ephrin-A1, doxazosin stimulation resulted in EphA2 internalization and degradation. Doxazosin treatment also inhibited both Akt and ERK1/2 kinase activities downstream of EphA2 activation. Consistent with the well-established ligand-dependent roles of Eph in negatively regulating cell motility, doxazosin retarded tumor cell migration in vitro. Moreover, in a newly established orthotopic prostate cancer metastasis model, doxazosin significantly reduced distant metastases of prostate cancer cells. We propose that, as a FDA-approved drug (Cardura®), doxazosin represents an attractive compound that can be now re-purposed for treatment of aggressive prostate cancer and potentially other malignant diseases.
Most current drug discovery efforts targeting kinases are focused on identifying small molecules that inhibit enzyme function 
. In the case of receptor tyrosine kinases, virtually all current drug development endeavors are devoted to inhibitors targeting the ATP binding pocket. Such inhibitor-focused approaches are certainly justified in lieu of the pro-oncogenic role for most RTKs in tumors. However, the unique ligand-dependent tumor suppressor functions may make development of EphA2 agonists, rather than antagonists, a fruitful strategy for targeted therapy of a variety of solid tumors. Likewise, other kinases with intrinsic tumor suppressor functions, such as LKB1 
and LATS2 
, can also be suitable targets for agonist development. The general lack of kinase agonists is often attributed to the prevailing belief that gain-of-function agonists are more difficult to develop than the loss-of-function antagonists/inhibitors. In this study, we utilized virtual screening coupled with cell-based assays to identify doxazosin as a bona fide agonist for EphA2 targeting the ligand-binding domain (LBD). Both approaches are widely utilized in contemporary drug discovery and can either be readily adapted, or developed, suggesting the feasibility to find agonists for other RTKs, including other members of the Eph subfamily.
The anti-metastatic effects of doxazosin are in keeping with the anti-migratory and anti-invasive properties of ligand activated EphA2 
. We believe that the effects are likely to be more pronounced in tumors where the ligand-independent, pro-oncogenic functions of EphA2 predominate as a result of Akt-mediated phosphorylation of EphA2 on serine 897 
. The latter scenarios can take place when ligand expression is lost or reduced relative to the often overexpressed EphA2 
. Alternatively, in some cellular contexts, EphA2 can be catalytically silenced by ephrin-As on the same cells through inhibitory cis
. In both situations, provision of exogenous agonists acting in trans
could still activate EphA2, unleashing its intrinsic tumor suppressor functions to inhibit tumor cell migration and invasion. It is important to note that, due to the complexity of in vivo
settings, we cannot completely rule out the possibility that doxazosin exerts its anti-metastatic functions by affecting targets other than EphA2 or α1-adrenoreceptor.
Previous studies analyzing the crystal structure of Eph/ephrin interactions have primarily involved EphB kinases 
. A recent study determined the crystal structure of the ligand-binding domain of EphA2 and its interaction with ephrin-A1 
. The structure indicates that only minor conformational changes occur in EphA2 upon ephrin-A1 binding, suggesting that the ephrin-binding pocket on EphA2 is formed prior to ligand binding, consistent with the “lock and key” model. This is in contrast to EphB kinases, in which significant conformational changes in the ephrin-binding pocket occur following ligand binding, as a result of “induced fit” 
. This difference in binding modes contributes to the ability of ephrin-A ligands to more readily bind to their EphA receptors with much higher affinities. In addition, while the high affinity interactions with the ligand-binding channel of EphA2 are primarily involved in ephrin-A1 binding, a second low affinity binding site is involved in EphB-ephrin-B multimerization and resulting signaling 
. In keeping with this notion, the dimeric form of ephrin-A1, and even monomeric ephrin-A1, is capable of activating EphA2 receptors, whereas multimerization of ephrin-B1 is necessary for EphB2 activation 
. Due to this enhanced binding and activation, it is suggested that EphA receptors represent better targets for small molecules than EphB receptors. Indeed, no small molecules targeting EphB receptors have yet been discovered.
Although doxazosin was discovered by virtual screening based on the EphA2 structure, it can activate EphA4 kinase at similar doses, but none of the other Eph receptors tested. This EphA2/EphA4 dual specificity has been seen previously with small molecule antagonists, which inhibited ephrin-A1 binding to both EphA2 and EphA4 kinases at high µM concentrations 
. This shared specificity suggests structural similarities between EphA2 and EphA4 in the residues necessary for interaction with doxazosin. Given the very promiscuous nature of EphA4, which cross reacts with both ephrin-A and ephrin-B, this result is not completely surprising. While the solubility problems prevented us from directly determining the EphA2/doxazosin NMR structure, we were able to model it based on the structure of EphA4/doxazosin. The NMR structures revealed that doxazosin recapitulates both hydrophobic and electrostatic interactions of ephrin-A1 with EphA2 in the crystal structure, which could contribute to its agonistic activities.
Interestingly, both doxazosin and C1 have similar binding affinities and triggered no significant secondary structure changes of the EphA4 LBD. However, doxazosin and C1 have opposite functional effects. As revealed by the present NMR studies, in addition to contacting with the EphA4 D–E and J–K loops, doxazosin has extensive contacts with residues on the convex β-strands of the ephrin binding pocket. These interactions are observed in all complexes of Eph receptors bound to their natural ephrin ligands, but are totally lacking in the EphA4-C1 complex. This strongly suggests that designed molecules need to establish interactions with residues on the J–K, D–E loops and the convex β-strands to achieve agonistic activity. Further, the dynamic stabilization on both ps-ns and µs-ms time scales upon doxazosin binding might also play a crucial role in its agonistic activity. Although challenging, further assessment of this phenomenon may be performed by combining NMR spectroscopy and molecular dynamic (MD) simulations as we recently conducted on another system 
In summary, using structure-based virtual screening and cell-based assays, we have identified and characterized doxazosin as a novel small molecule agonist for EphA2. In addition, given both its ability to inhibit prostate cancer growth and its ability to inhibit prostate cancer metastasis, doxazosin may represent a cancer therapeutic agent, particularly for aggressive prostate cancer. Future optimization of the structure of doxazosin through chemical derivatization may lead to the discovery of new EphA2 agonists with enhanced affinity, specificity, and potency for use as more effective cancer therapeutics.