Gremlin is a novel agonist of the major proangiogenic receptor VEGFR2

The bone morphogenic protein (BMP) antagonist gremlin¹  induces angiogenesis in a BMP-independent manner by binding to as-yet-unidentified endothelial cell (EC) membrane receptors and activating multiple tyrosine kinase–dependent intracellular signaling pathways in ECs.^2,3  Gremlin is produced by human tumors^4,5  and is expressed by fibroblast growth factor-2 (FGF2)–activated ECs and tumor endothelium.²  Thus, gremlin may play paracrine/autocrine roles in tumor neovascularization. The identification of the EC receptors activated by gremlin has so far been unsuccessful.

Vascular endothelial growth factor receptor-2 (VEGFR2) is the major proangiogenic tyrosine kinase receptor expressed by ECs and is activated by different members of the vascular endothelial growth factor (VEGF) family.⁶  Both gremlin and VEGFs belong to the cystine-knot protein superfamily,⁷  suggesting possible structural and/or functional similarities among these proangiogenic factors. On this basis, we investigated the capacity of gremlin to interact with and activate VEGFR2. The results demonstrate that gremlin binds and activates VEGFR2, leading to VEGFR2-dependent angiogenic responses in vitro and in vivo.

Interaction of VEGF-A and gremlin (R&D Systems) with the immobilized extracellular domain of VEGFR2 (sVEGFR2; Calbiochem) was analyzed by surface plasmon resonance (BIAcore Inc) and by competitive enzyme-linked immunosorbent assay (ELISA). VEGFR2 interaction on the EC surface was characterized by cross-linking experiments, whereas VEGFR2 dimerization was assessed by fluorescence resonance energy transfer analysis.

Motility and 3-dimensional gel invasion assays were performed on human, murine, and bovine ECs.³  When indicated, ECs were stably transfected with a pcDNA3.1 expression vector harboring the mouse VEGFR2 complementary DNA.

Alginate beads (5 μL) containing gremlin (100 ng per embryo) were grafted on the chorioallantoic membrane (CAM) of fertilized chicken eggs at day 11. After 72 hours, new blood vessels converging toward the implant were counted.

Details of experimental procedures and analysis are provided in supplemental Methods (available on the Blood Web site; see the Supplemental Materials link at the top of the online article).

Surface plasmon resonance analysis was used to assess the ability of gremlin to bind the extracellular domain of VEGFR2. As shown in Figure 1A, gremlin binds sVEGFR2 immobilized on a BIAcore sensor chip (Kd = 47 ± 15 nM). The interaction is inhibited by an excess of soluble sVEGFR2-Fc but not by sVEGFR1-Fc, sVEGFR3-Fc, or sFGFR1-Fc (Figure 1B). Similar to VEGF-E, gremlin inhibits the binding of VEGF-A to immobilized sVEGFR2-Fc in competitive ELISA but, unlike the VEGFR1-specific agonist placenta growth factor, it does not affect VEGF-A interaction with sVEGFR1-Fc (Figure 1C). To assess whether gremlin/VEGFR2 interaction occurs also under physiologic conditions, we incubated human umbilical vein ECs (HUVECs) with gremlin conjugated with a bifunctional photoactivable biotin-label transfer cross-linker (sulfo-gremlin). After photoactivation of EC-bound sulfo-gremlin, VEGFR2 immunocomplexes were separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis under reducing conditions to transfer the biotin moiety to the interacting receptor, followed by detection with streptavidin–horseradish peroxidase. As shown in Figure 1D, sulfo-gremlin forms a 250 kDa biotin–VEGFR2 complex when cross-linked to the HUVEC surface, VEGFR2 biotinylation being inhibited by an excess of unlabeled gremlin or VEGF-A. Taken together, these data demonstrate the capacity of gremlin to bind VEGFR2 and to compete with VEGF-A for receptor interaction.

We next investigated whether gremlin induces VEGFR2 autophosphorylation in ECs.⁶  In HUVECs, gremlin induced a dose-dependent phosphorylation of tyrosine phosphorylation sites Y1175 and Y951 of VEGFR2 (supplemental Figure 1a) with kinetics similar to VEGF-A (Figure 2A-B). VEGFR2 activation is also abrogated by the VEGFR2 tyrosine kinase inhibitor SU5416⁸  both in gremlin and VEGF-A–activated ECs (Figure 2C). Accordingly, VEGFR2 phosphorylation, abrogated by SU5416, was observed in fetal bovine aortic VEGFR2-GM7373 ECs stimulated by gremlin or VEGF-A after immunostaining with anti–phospho-VEGFR2 (pY1175) antibodies (supplemental Figure 2). VEGFR2 engagement leads to a rapid internalization of the activated receptor.⁹  Indeed, confocal analysis shows that both gremlin and VEGF-A induce VEGFR2 phosphorylation and internalization of the activated receptor in HUVECs (Figure 2D). In addition, fluorescence resonance energy transfer analysis of bovine aortic ECs cotransfected with enhanced yellow fluorescent protein–tagged and enhanced cyan fluorescent protein–tagged VEGFR2 highlighted the internalization of VEGFR2 dimers in the early endosomal compartment⁹  after stimulation with gremlin or VEGF-A but not with FGF2 (supplemental Figure 3). Thus, in keeping with its ability to induce focal adhesion kinase, mitogen-activated protein kinase extracellular signal-regulated kinase_1/2, and transcription factor nuclear factor κB activation,^2,3  gremlin activates VEGFR2 similar to VEGF-A. Tyrosine phosphorylation of cellular proteins (supplemental Figure 1a), chemotactic migratory response and small GTPase Rac activation (supplemental Figure 4), as well as the formation of 3-dimensional EC sprouts (Figure 2E) are similar in HUVECs stimulated by either gremlin or VEGF-A. These activities are blocked by SU5416 or by receptor-binding competitors such as neutralizing anti-VEGFR2 antibody¹⁰  or the cyclic peptide cyclo-VEGI (Calbiochem).¹¹ 

To further assess the role of VEGFR2 in mediating the proangiogenic activity of gremlin, we transfected murine aortic ECs (MAECs) expressing low levels of VEGFR2 with a murine VEGFR2 complementary DNA giving rise to VEGFR2-MAECs. The motogenic activity exerted by gremlin or VEGF-A is dramatically up-regulated in these cells compared with parental MAECs. Again, the activity of both motogens is inhibited by SU5416 and cyclo-VEGI (supplemental Figure 5).

In keeping with these in vitro observations, the angiogenic activity exerted by gremlin and VEGF-A ex vivo in the human umbilical artery ring EC sprouting assay and in vivo in the chick embryo CAM assay^2,3  is significantly inhibited by SU5416, the VEGFR2 inhibitor I (VEGFR2-KI),¹²  or by competition with an excess of free sVEGFR2 (Figure 2F-G; supplemental Figures 6-7). In all the assays, VEGFR2 inhibitors do not affect the activity of FGF2 (data not shown).

Here, we demonstrate that the angiogenic activity of gremlin is mediated by VEGFR2. Most importantly, BMP2 does not prevent gremlin from binding and activating VEGFR2 (supplemental Figure 8). Similarly, BMP4 does not affect the angiogenic activity of gremlin and its interaction with ECs.²  We therefore propose that distinct domains of gremlin may mediate the interaction with VEGFR2 or BMPs, respectively.

The capacity of gremlin to bind BMPs and to inhibit their interaction with the cognate transforming growth factor-β family receptors is thought to play a role in embryonic development¹³  and in human diseases.¹⁴  However, BMP-independent mechanism(s) may also be involved in gremlin signaling.^2,15  Our data reveal the previously unrecognized capacity of gremlin to specifically bind to and promote activation of the major proangiogenic receptor VEGFR2 in a BMP-independent manner. Thus, gremlin may exert both BMP-dependent and BMP-independent functions in different physio-pathologic conditions by inhibiting BMP-mediated transforming growth factor-β receptor activation or by a direct activation of VEGF receptor signaling, respectively. These findings extend the number of proangiogenic VEGFR2 ligands to a member of the cystine-knot BMP antagonists distinct from VEGFs.¹⁶  Gremlin is expressed by parenchymal and stromal cells in human tumors.^2,4,5  Our observations may provide novel insights for the understanding of the biology of vascular development and of angiogenesis-dependent diseases, including cancer.

This work was supported by grants from Istituto Superiore di Sanità (Oncotechnological Program), Ministero dell'Istruzione, Università e Ricerca (Centro di Eccellenza per l'Innovazione Diagnostica e Terapeutica; Cofin projects), Associazione Italiana per la Ricerca sul Cancro (AIRC), Fondazione Berlucchi, and Fondazione Cariplo (grant 2008-2264 and NOBEL Project) to M.P., from AIRC (MFAG 9161 grant) to S.M., and from Swiss National Science Foundation (grant 3100A-116507) and Oncosuisse (grant OC2 01200-08-2007) to K.B.-H.

Contribution: S.M., C.R., E.M., L.Z., K.B.-H., and M.P. designed research; S.M., C.R., E.M., V.S., D.L., and L.Z. performed research; S.M., C.R., E.M., K.B.-H., and M.P. analyzed data; and S.M., C.R., K.B.-H., and M.P. wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Marco Presta, General Pathology, Department of Biomedical Sciences and Biotechnology, University of Brescia, Viale Europa 11, 25123 Brescia, Italy; e-mail: presta@med.unibs.it.