SCF^β-TRCP suppresses angiogenesis and thyroid cancer cell migration by promoting ubiquitination and destruction of VEGF receptor 2

Multi-subunit Cullin–Ring complexes comprise the largest known class of E3 ubiquitin ligases (Petroski and Deshaies, 2005). Cullins directly interact with Roc1, a Ring finger protein, and the Cullin–Roc1 complex comprises the core module of a series of E3 ubiquitin ligases. Thus, we started our investigation by examining whether a specific Cullin–Ring complex interacts with VEGFR2. We found that Cullin 1, but not other members of the Cullin family that we examined, specifically interacts with VEGFR2 (Fig. 1, A and B). This suggests that the SCF complex, which contains Cullin 1, might be involved in the regulation of VEGFR2 stability. In keeping with this notion, depletion of endogenous Cullin 1, but not Cullin 4A, resulted in the up-regulation of VEGFR2 expression in human microvascular endothelial cells (HMVECs; Fig. 1 C). Next, we sought to explore which F-box protein, when complexed with Cullin 1, is responsible for VEGFR2 degradation. β-TRCP, one of the well characterized F-box proteins, binds to its substrates by recognizing a specific DSG(XX)S phosphodegron motif, within which the two serine residues are phosphorylated (Frescas and Pagano, 2008). We noticed that the cytoplasmic tail of VEGFR2 contains three DSG(XX)S motifs that could potentially be recognized by β-TRCP and are conserved among different species (see Fig. 3 A). This prompted us to examine whether β-TRCP interacts with VEGFR2 in vitro. Using coimmunoprecipitation, we found that both β-TRCP1 and β-TRCP2 interact with VEGFR2 (Fig. 1 D). The interaction between VEGFR2 and β-TRCP1 was abolished when the C-terminal WD40 repeat motif of β-TRCP1, which has been shown to mediate the interaction with most of its substrates (Wu et al., 2003), was mutated (Fig.1 D). Furthermore, we detected an interaction between endogenous VEGFR2 and endogenous β-TRCP1 in HMVECs (Fig. 1 E) and further demonstrated that phosphatase treatment abolished the interaction between VEGFR2 and β-TRCP1 (Fig. 1 F). In support of an important role for β-TRCP in regulating VEGFR2 abundance, depletion of either β-TRCP1 or β-TRCP2 led to up-regulation of VEGFR2 levels (Fig. 1 G). More importantly, depletion of β-TRCP did not affect significantly VEGFR2 mRNA levels (see Fig. 4 D), suggesting that the observed increase in VEGFR2 abundance was mainly through posttranscriptional mechanisms. In support of this notion, using cycloheximide, we demonstrated that the half-life of VEGFR2 was markedly extended after depletion of β-TRCP1 (Fig. 1, H and I).

It has been demonstrated that β-TRCP substrates need to be phosphorylated within their degron sequences by specific kinases before their recognition and destruction by the SCF^β-TRCP complex (Jin et al., 2003; Cardozo and Pagano, 2004). We therefore sought to identify the upstream kinase that phosphorylates VEGFR2 and triggers its destruction. To this end, we transfected 293T cells with HA-VEGFR2 and Flag–β-TRCP1, along with different kinases (CKIδ [Inuzuka et al., 2010] and GSK3β [Liu et al., 2002]) that have been previously identified to play a critical role in facilitating β-TRCP–mediated proteolysis of the target protein. CKIδ, but not GSK3β, promoted the destruction of ectopically expressed VEGFR2 (Fig. 2 A). Furthermore, this degradation of VEGFR2 was completely blocked when treated with the proteasome inhibitor MG132, suggesting the involvement of 26S proteasome (Fig. 2 A). Importantly, the R474A mutant form of β-TRCP1 that is impaired to interact with VEGFR2 (Fig. 1 D) failed to promote VEGFR2 destruction by CKIδ (Fig. 2 B), suggesting a critical role of β-TRCP1 in CKIδ-mediated destruction of VEGFR2.

In further support of a direct role of CKI in VEGFR2 protein stability, VEGFR2 levels were markedly up-regulated in a concentration-dependent manner when HMVECs were treated with D4476, a specific CKI inhibitor (Fig. 2, C and D). Furthermore, coimmunoprecipitation analysis revealed that CKIδ, but not CKIα or CKIε isoforms, specifically interacts with VEGFR2 (Fig. 2 E). Consistent with this finding, depletion of endogenous CKIδ, but not other CKI isoforms (CKIα and CKIε), led to a significant increase in endogenous VEGFR2 abundance (Fig. 2 F). Consistent with a negative role of CKIδ in VEGFR2 stability control, depletion of CKIδ significantly extended endogenous VEGFR2 half-life in HMVECs (Fig. 2, G and H). In a reciprocal set of experiments, we found that overexpression of CKIδ markedly decreased the half-life of VEGFR2 (Fig. 2 I).

Most β-TRCP substrates contain one or several DSG(XX)S degrons that are recognized by β-TRCP when they are properly phosphorylated by one or more kinases (Frescas and Pagano, 2008). We noticed that there are three DSG(XX)S degrons present in the C-terminal tail of VEGFR2 that are conserved among different species (Fig. 3 A; and Fig. S1, A–C). Interestingly, most of the Ser/Thr sites within the putative degrons are predicted to be CKI sites (Fig. S1, E and F). To test the significance of each putative phosphodegron, we created individual point mutations in the DSG motifs by replacing both serine residues with alanine to examine whether these mutants resist CKI-mediated VEGFR2 degradation (Fig. 3 A). We found that individual mutation of each DSG motif (labeled DSG1, DSG2, and DSG3) had no significant effect on β-TRCP–mediated destruction of VEGFR2 in the presence of CKIδ (Fig. 3 B), suggesting that the phosphorylation of all three DSG degrons might be involved in the degradation of VEGFR2. To test this hypothesis, we generated a DSG triple mutant that contains serine to alanine substitutions in all three DSG degrons (labeled as DSG-T). However, VEGFR2–DSG-T could still be degraded by β-TRCP1/CKIδ with similar efficiency (Fig. 3 E). Furthermore, the half-life of DSG-T was not significantly increased in response to VEGF-A (unpublished data), indicating that there are other potential degrons present in VEGFR2 mediating its destruction. Using the ScanSite program, we identified another potential CKI phosphorylation site (DDTD) in the C-terminal tail of VEGFR2 (Fig. 3 A and Fig. S1 F). Thus, we generated additional VEGFR2 mutants (Fig. 3 A) to evaluate the contribution of the DDTD site in β-TRCP–mediated VEGFR2 destruction. Using in vitro kinase assays, we showed that the three identified DSG sites, as well as the newly identified DDTD site, are the major CKI phosphorylation sites. Combinational inactivation of the three DSG degrons (DSG-T), or DDTD alone, significantly reduced CKI-mediated phosphorylation of VEGFR2 (Fig. 3 C). Moreover, combinational mutation of all four putative degrons (9A) further decreased CKI-mediated in vitro phosphorylation of GST-VEGFR2 (Fig. 3 C). Consistent with this result, CKI treatment triggered the interaction between GST-VEGFR2 and β-TRCP1 in vitro (Fig. 3 D). In contrast, mutation of the three putative DSG degrons (DSG-T) led to a reduced interaction between GST-VEGFR2 and β-TRCP1. Interestingly, mutation of all four degrons (9A) led to further reduction in detectable interaction between GST-VEGFR2 and β-TRCP1 (Fig. 3 D), which reflects their deficiency in CKI-mediated phosphorylation (Fig. 3 C). Moreover, consistent with their decreased interactions with β-TRCP1, these mutants are also deficient in β-TRCP1/CKI–mediated proteolysis (Fig. 3 E). Importantly, compared with WT-VEGFR2, 9A-VEGFR2 could similarly activate the downstream ERK signaling pathway (Fig. 3 F) and promote cell migration in a VEGF-A–dependent manner (unpublished data), indicating that 9A-VEGFR2 is functional in vivo. We further showed that β-TRCP1 promotes in vitro VEGFR2 ubiquitination in a CKI-dependent manner (Fig. 3 G) and that the inefficient destruction of 9A-VEGFR2 observed in Fig. 3 E is likely a result of deficient ubiquitination of VEGFR2 (Fig. 3 G).

Because VEGFR2 is predominantly expressed in endothelial cells and is involved in the regulation of angiogenesis, we next investigated whether depletion of endogenous β-TRCP affects VEGFR2 protein abundance in HMVECs under physiological conditions. Depletion of either β-TRCP1 or β-TRCP1+2 in HMVECs (Fig. 4, B and C) led to up-regulation of VEGFR2 protein levels (Fig. 4 A). In contrast, depletion of endogenous β-TRCP did not affect the expression of other VEGF receptors (VEGFR1 and VEGFR3) or CCM3, a protein recently reported to modulate VEGFR2 function (Fig. 4 A; He et al., 2010). Furthermore, β-TRCP depletion did not affect cell viability compared with shGFP control cells (unpublished data). These results suggest that β-TRCP might regulate angiogenesis through specifically targeting VEGFR2 for ubiquitination and destruction. Furthermore, β-TRCP depletion did not cause a marked change in VEGFR2 mRNA levels (Fig. 4 D), suggesting that β-TRCP regulates VEGFR2 expression mainly through a posttranscriptional mechanism. Consistently, the half-life of endogenous VEGFR2 was increased in β-TRCP–depleted HMVECs compared with shGFP-treated cells (Fig. 4 E). More importantly, we found that depletion of β-TRCP results in increased activity of VEGFR2 downstream signaling pathways as indicated by increased p-VEGFR2 (pTyr1175), p-AKT (pSer473), and p-ERK1/2 levels (Fig. 4 F; Shibuya and Claesson-Welsh, 2006).

Consistent with the role of β-TRCP in regulating VEGFR2 abundance and the activities of downstream signaling pathways, compared with shGFP-treated HMVECs, β-TRCP–depleted HMVECs exhibited increased migration (Fig. 5, A and B). Furthermore, using in vitro angiogenesis (matrigel tube formation) assays, we found that depletion of endogenous β-TRCP in HMVECs leads to a significant increase in their capacity to form tubes as compared with control cells infected with shGFP (Fig. 5, C and D). VEGF-A addition significantly increased the number of tubes in shGFP cells; however, VEGF-A addition did not increase the number of tubes either by shβ-TRCP1 or shβ-TRCP1+2 cells (Fig. 5, C and D), suggesting that because these cells already have increased ability to form tubes, further addition of VEGF-A did not increase tube formation in shβ-TRCP cells. Furthermore, additional depletion of VEGFR2 blocked the elevated tube formation in β-TRCP–depleted HMVECs, indicating that VEGFR2 may be the major route through which β-TRCP regulates the tube formation ability of HMVECs (Fig. 5, E and F), while reintroducing β-TRCP into the β-TRCP–depleted HMVECs retarded their in vitro tube formation (Fig. 5, G and H). These results therefore indicate a causal relationship between loss of β-TRCP and elevated in vitro tube formation in HMVECs.

Because both enhanced cellular migration and elevated in vitro tube formation are positive markers for increased angiogenesis, these results support a potential role for β-TRCP in regulating the angiogenesis process through modulating VEGFR2 stability. However, we recognize that the in vitro cell-based assays may not accurately model the angiogenesis process under physiological conditions. To obtain direct evidence to support the potential role of β-TRCP in regulation of angiogenesis, we used a well characterized in vivo zebrafish model (Nasevicius et al., 2000; Habeck et al., 2002) to evaluate how β-TRCP governs VEGFR2 stability to influence angiogenesis.

It has been previously demonstrated that both VEGF-A and VEGFR2 are required for vasculogenesis in zebrafish (Nasevicius et al., 2000; Habeck et al., 2002). Loss of VEGF-A function severely affects the formation of the entire vasculature, whereas VEGFR2 dysfunction impairs the formation of blood vessels generated by sprouting (Nasevicius et al., 2000; Habeck et al., 2002). Thus, we sought to examine whether inactivation of β-TRCP by morpholino (MO) treatment affects vasculogenesis and/or angiogenesis in zebrafish. Interestingly, zebrafish does not contain a homologue of human β-TRCP1 but contains two paralogues of β-TRCP2 (called Fbxw11a and Fbxw11b). Embryos injected with control or MOs specifically inactivating Fbxw11a or Fbxw11b alone or in combination were monitored for alterations in blood vessel formation over 72 h after fertilization. Although a small dose-dependent decrease of embryo viability was observed with either the Fbxw11a or Fbxw11b MO alone, at the highest concentration (100 µM) injected, >90% of the zebrafish embryos were still viable (Fig. S2). In contrast, combinatorial MO injection significantly reduced embryo viability, such that at a concentration of 100 µM, 100% lethality was observed. Combinatorial low-dose MO injection (20 µM) resulted in reduced viability compared with controls (61.5%) but allowed the embryos to develop with overall normal gross morphology allowing analysis (Fig. S2). These data indicate that there is partial functional redundancy between the two β-TRCP2 paralogues.

To further characterize the potential role of β-TRCP2 in angiogenesis, we used the low-dose (20 µM) combinatorial MO strategy to simultaneously inactivate both Fbxw11a and Fbxw11b during zebrafish vascular development in two endothelial reporter lines: Kdrl:EGFP (Fig. 6, A–F; Cross et al., 2003) and fli1a:EGFP (unpublished data; Lawson and Weinstein, 2002). Whereas large blood vessels such as the dorsal aorta appeared to form normally, β-TRCP2–MO-injected embryos exhibited dysregulated vessel formation with increased sprouting as compared with controls (Fig. 6, A and B). Fine analysis of the head/anterior trunk (blue boxed region) and posterior trunk/tail (red boxed region; Fig. 6 C–F) demonstrates that compared with the normal blood vessel formation observed in the control embryos (Fig. 6, C and E, yellow arrows), there is a significant increase in the number of branching blood vessels initiated in β-TRCP2–MO-treated zebrafish (Fig. 6, D and F, red arrows). Significantly, unlike the symmetrical pairs of vessels formed at regular intervals along the somite boundaries in controls, in β-TRCP2–MO-injected embryos vascular sprouts were commonly disorganized, originating at several points along the length of each somite, or with multiple branching connections converging at a single point. These results were confirmed by in situ hybridization (unpublished data), as well as combinatorial splice-blocking MOs, demonstrating the specificity of the effect and significant reduction of gene expression of both Fbxw11a and Fbxw11b RT-PCR (Fig. 6 G) compared with uninjected sibling controls. This elevated blood vessel development was found to correlate with increased expression of VEGFR2 (Fig. 6, H–J) after MO-mediated knockdown of β-TRCP2 (Fig. 6 G), consistent with our in vitro data using HMVECs. However, it was recently reported by Kanarek et al. (2010) that the β-TRCP1/2 double knockout mice are nearly normal, with severe defects in spermatogenesis caused by the stabilization of SNAIL. As the authors used shβ-TRCP2 to knock down endogenous β-TRCP2 in the β-TRCP1^−/− mice background to achieve the β-TRCP1/2 double knockout phenotype, it is possible that the residual expression of β-TRCP2 allows the shβ-TRCP2/β-TRCP1^−/− mice to function relatively normally. Further studies are required to explore whether deletion of β-TRCP1 and/or β-TRCP2 in mice also results in abnormalities in angiogenesis and vascular development.

Thyroid cancer accounts for up to 95% of primary cancers involving the endocrine system. Papillary thyroid cancer (PTC) is the most common type of thyroid cancer, representing 75–85% of all thyroid cancer cases, and various signaling pathways, such as BRAF and MAPK (i.e., ERK1/2), have been identified as participants in the initiation and progression of human PTC (Khurana et al., 2005; Knauf and Fagin, 2009; Nucera et al., 2009, 2010; Keefe et al., 2010; Liu et al., 2010; Chakravarty et al., 2011). Importantly, as in many other human cancers, VEGFR2-induced angiogenesis was found to play a critical role in thyroid cancer angiogenesis (Keefe et al., 2010; Ringel, 2011). In keeping with this notion, recent clinical trials have proposed VEGFR inhibitors, such as axitinib and sorafenib, as anti-thyroid cancer therapeutics (Cohen et al., 2008; Gupta-Abramson et al., 2008), although the underlying mechanism of their therapeutic effects remains unknown. This prompted us to further examine whether β-TRCP negatively governs VEGFR2 stability in the human thyroid cancer, and furthermore, whether de-regulation of this destruction pathway contributes to the pathological progression of human thyroid cancers. In keeping with our hypothesis, we observed a robust inverse correlation between β-TRCP1 and VEGFR2 protein expression levels in the human thyroid cancer cell lines we examined (Fig. 7, A and B). In contrast to SW1736 and 8505C anaplastic thyroid cancer (ATC)–derived cell lines, or the FTC236 metastatic follicular cancer–derived cell line (Fig. 7, A and B), which display high expression of β-TRCP1, the human BCPAP cell line, a poorly differentiated PTC with squamous metaplasia-derived cell line, has very low β-TRCP1 protein expression and elevated VEGFR2 protein expression levels (Fig. 7 A). In support of this result, we also detected an inverse correlation between β-TRCP1 protein expression levels and angiogenic profile (i.e., CD31 immunoexpression) in an array of poorly or well differentiated human PTC with squamous metaplasia foci (Table S1), which were similar to the initial tumor from which the BCPAP cell line was established (Fig. 7 C). In particular, we found that ∼50% of PTC with squamous metaplasia showed high expression levels (3+) of β-TRCP1, both in the cytoplasm of tumor cells or in the endothelial cells, and low/moderate CD31 (1+/2+) or absent VEGFR2 protein expression levels (Fig. 7, C and D). Whereas PTC with low β-TRCP1 protein expression levels (1+) showed increased intratumoral angiogenesis (CD31 immunoexpression; Fig. 7, C and D), and intriguingly ∼6% of PTC cells and 40% of PTC-associated endothelial cells/high-powered fields express VEGFR2 (Fig. 7 C).

Importantly, ectopic expression of β-TRCP1 led to a significant reduction in the endogenous VEGFR2 abundance in the BCPAP cell line (Fig. 7 E) that can be reversed by treatment with the proteasome inhibitor MG132 (Fig. 7 F), arguing that β-TRCP1 is a major negative regulator of VEGFR2 abundance. Furthermore, there was no marked difference in VEGFR2 mRNA levels between empty vector (EV) and β-TRCP1–expressing BCPAP cells, indicating that the regulation of VEGFR2 abundance mainly occurs at the posttranscriptional level (Fig. 7, G and H). These results indicate that in BCPAP (Fig. 7 A) or in poorly differentiated follicular thyroid cancer (FTC)–derived cells (FTC133; Fig. 7 B), β-TRCP1 negatively regulates VEGFR2 abundance, possibly through a ubiquitination-dependent degradation pathway, as it does in human endothelial cells (Fig. 4). Importantly, we found that in human PTC clinical samples there is also a clear inverse correlation between β-TRCP1 expression and several angiogenesis markers (Fig. 7, C and D). This prompted us to further investigate whether elevated VEGFR2 expression in human PTC cells could affect the angiogenesis process of the surrounding endothelial cells.

To this end, we found that the BCPAP with elevated VEGFR2 levels secreted an increased amount of VEGF-A protein (2010 ± 580 pg/ml) compared with the ATC-derived cells SW1736 and 8505C (517 ± 129 and 0.0 pg/ml, respectively) expressing low levels of VEGFR2 (Fig. 8 A), whereas ectopic expression of β-TRCP1 significantly reduced the secreted VEGF-A (470 ± 98 pg/ml in BCPAP-HA-β-TRCP1 cells compared with 967 ± 99 pg/ml in BCPAP-EV cells; Fig. 8 B). Ectopic expression of β-TRCP1 did not affect the mRNA levels of VEGF-A (unpublished data), indicating that β-TRCP1 might regulate VEGF-A through a posttranscriptional mechanism. Whereas additional investigation is warranted to uncover the molecular mechanisms underlying the effect of β-TRCP1 on secretion of VEGF-A, we found that VEGF-A contains an evolutionally conserved DSG-like phosphodegron (Fig. S3), suggesting an interesting possibility that β-TRCP1 may regulate the stability of VEGF-A precursor to indirectly govern its secretion. Importantly, in response to VEGF-A, the downstream signaling pathways of VEGFR2, as detected by the levels of phospho-VEGFR2 (pTyr1175), phospho-Akt (pSer473), and phospho-ERK1/2, are significantly decreased in BCPAP-HA–β-TRCP1 cells compared with BCPAP-EV cells (Fig. 8 C). In a similar fashion to the endothelial cells (Fig. 5, A and B), β-TRCP1–expressing BCPAP cells exhibited significantly less migration in response to VEGF-A compared with the BCPAP-EV cells (Fig. 8, D and E). These results suggested that β-TRCP1 plays a critical role in governing cellular migratory ability, presumably in part via regulating the stability of VEGFR2, in both human endothelial and BCPAP thyroid cancer cell lines.

However, this could not explain why elevated VEGFR2 expression in a subset of human thyroid cancer cells, by itself, could lead to increased angiogenesis which is typically associated with human endothelial cells. To address this question, we used a trans-well system of co-cultured BCPAP cells and human endothelial cells grown in the absence of serum to examine whether manipulation of the β-TRCP1–VEGFR2 signaling pathway in BCPAP cells could affect cellular migration of human endothelial cells (Fig. 9 A). Interestingly, we found that VEGFR2 levels in BCPAP cells play a critical role in determining migratory ability of the co-cultured HMVECs. Very few HMVECs migrated when they were co-cultured with β-TRCP1–expressing BCPAP cells compared with BCPAP-EV cells (Fig. 9, B and C). This is likely a result of the elevated secretion of VEGF-A from the BCPAP-EV cells because the use of VEGF-A blocking antibody could efficiently abolish cellular migration of the co-cultured HMVECs (Fig. 9, B and C). Furthermore, VEGFR2 knockdown BCPAP (shVEGFR2) cells showed a significant decrease in cell migration compared with control (shGFP) BCPAP cells (Fig. 9, D and E). Overall, these results indicated that elevated VEGFR2 expression in human thyroid cancer cells could not only regulate their own migratory ability via an autocrine mechanism but also regulate the migration of the surrounding human endothelial cells, presumably through a paracrine mechanism. The latter mechanism might offer a possible molecular mechanism for increased VEGFR2 expression that is frequently observed in various human cancers and its correlation with increased angiogenesis.

As recent studies suggest that oncogene addiction could be used as a rationale for targeted anti-cancer therapy (Weinstein and Joe, 2006), we wanted to understand whether expression levels of VEGFR2 in various human thyroid cancer cells correlate with their sensitivity to the VEGFR2 inhibitor, sorafenib. Interestingly, we found that the BCPAP cell line that exhibited increased VEGFR2 protein expression showed increased sensitivity to sorafenib, compared with SW1736 and 8505C ATC-derived cell lines with much lower expression levels of VEGFR2 and corresponding higher levels of β-TRCP1 (Fig. 9 F). At 3 µM sorafenib, the cell viability was significantly decreased in BCPAP cells, whereas no significant decrease was observed in SW1736 and 8505C cells at this sorafenib concentration (Fig. 9 F). Because sorafenib has been proposed as an efficient inhibitor of VEGFR2 kinase, these results indicate that human thyroid cancer cell lines with high levels of VEGFR2, such as BCPAP, might be in part addicted to VEGFR2 activity/functions, and thereby be more sensitive to VEGFR2 inhibitors including sorafenib. Consistent with this notion, we further showed that ectopic expression of β-TRCP1, which efficiently reduced endogenous VEGFR2 levels (Fig. 9 G), could significantly reduce BCPAP’s sensitivity to sorafenib. At 3 µM sorafenib concentration, the BCPAP-EV cell viability was significantly reduced when compared with a slight reduction of cell viability in BCPAP-HA–β-TRCP1 cells (Fig. 9 G). Collectively, these results identify a critical role of β-TRCP1 in determining the cellular sensitivity and therapeutic efficacy of sorafenib treatment in human thyroid cancer cells. They further suggest that thyroid cancer patients with low β-TRCP1 expression might respond better to the sorafenib (and maybe to other TKIs) treatment.

Here, we show for the first time a possible mechanism by which CKI-dependent phosphorylation of VEGFR2 at specific sites in its C-terminal tail triggers SCF^β-TRCP-mediated VEGFR2 ubiquitination and destruction. Importantly, our study demonstrated that β-TRCP depletion leads to stabilization of VEGFR2 in human endothelial cells as well as in human thyroid cancer cells, leading to increased sensitivity to VEGF-A–induced stimuli. Correspondingly, this stimuli increased cell migration and enhanced tube formation in β-TRCP–depleted HMVECs. Consistent with the conclusions drawn from our in vitro cell-based assays, we also observed an increased number of branching vessels in zebrafish (Weinstein, 2002) when β-TRCP2 was specifically depleted by MO treatment (Fig. 6, A–F). Moreover, the increased formation of blood vessels was associated with increased VEGFR2 protein levels in β-TRCP2 knockdown zebrafish. These findings indicated a prominent role for β-TRCP in regulating angiogenesis, possibly through the timely destruction of VEGFR2.

Furthermore, we observed a strong inverse correlation between β-TRCP1 and VEGFR2 in BCPAP thyroid cancer cell line (Fabien et al., 1994) as well as in primary PTC with squamous metaplasia (Fig. 7, A and B). For poorly differentiated thyroid cancer patients who fail to respond to the canonical treatment, there is generally neck recurrence and distance metastases leading to low survival rates; thus, rational targeted therapies are needed (Baudin and Schlumberger, 2007). The pathogenesis of squamous transformation of the thyroid gland remains unclear because the gland only contains squamous epithelium under inflammatory processes or neoplasms (Kleer et al., 2000; Walvekar et al., 2006). Furthermore, of the targeted therapies that are used in clinical trials in patients with thyroid cancers, including PTC, sorafenib that targets VEGFR2 signaling has been considered a promising therapeutic intervention (Gupta-Abramson et al., 2008; Lam et al., 2010). However, recent clinical trials based on the use of sorafenib in patients with aggressive/metastatic/progressive thyroid cancers (including PTC or FTCs, etc.) have observed a partial response rate of ∼20% and a stable disease rate of ∼53.3%. Furthermore, patients that histologically had an ATC showed a progressive disease (Gupta-Abramson et al., 2008), indicating that aggressive thyroid cancers may elicit unknown resistance mechanisms to sorafenib.

To this end, we have found for the first time that overexpression of β-TRCP1 markedly depleted VEGFR2 in the BCPAP cells (Fig. 7 E). More importantly, enhanced migration of HMVECs was observed when co-cultured with BCPAP-EV (control) cells. This migration was completely blocked in the presence of antibody to VEGF-A (Fig. 8, D and E), suggesting that the VEGF-A secreted from BCPAP cells induced the HMVEC cell migration. In contrast, the migration of HMVECs was significantly reduced when β-TRCP–overexpressing BCPAP cells were used in the co-cultured/transwell system (Fig. 9, A–C). These results suggested that by altering VEGFR2 and VEGF-A levels, β-TRCP suppresses angiogenesis via a paracrine-mediated mechanism.

Overall, these results pointed to a prominent role for β-TRCP in the negative regulation of VEGFR2 levels in PTC cells. Notably, we also found an inverse correlation between the VEGFR2 expression levels and their pharmacological efficacy/sensitivity to the VEGFR2 inhibitor sorafenib (Fig. 9, F and G). Poorly differentiated human PTC cells with elevated VEGFR2 are more sensitive to sorafenib. Therefore, our results suggest that defects in the β-TRCP/CKI signaling pathway might contribute to increased VEGFR2 protein expression, as well as PTC cell migration and sensitivity to VEGFR2 inhibitor, sorafenib. Furthermore, our work also provides a possible molecular basis for targeted use of VEGFR2 inhibitors, such as sorafenib or other TKIs, to treat more aggressive PTCs refractory to conventional therapies based on β-TRCP expression levels as a novel biomarker. Our results suggest that PTC patients with low expression levels of β-TRCP may respond more favorably to sorafenib treatment, whereas PTC patients with elevated β-TRCP expression might be poor responders. As genomic loss of β-TRCP has been identified in certain types of cancers (Frescas and Pagano, 2008), our results indicate that dysfunction in either β-TRCP or CKI might lead to stabilization of VEGFR2, which could result in hyperactivation of the VEGF-A–VEGFR2 pathway that facilitates angiogenesis and cancer cell migration.

In summary, our study provides a possible molecular mechanism for the frequent elevation of VEGFR2 displayed in many types of human cancers. Therefore, drugs that activate the β-TRCP and/or CKIδ signaling pathways to induce VEGFR2 destruction could be used to treat various types of human cancers that are associated with increased angiogenesis. In contrast, the identification of both β-TRCP and CKIδ as the upstream components involved in regulating VEGFR2 stability also provides the rationale for developing β-TRCP and CKI inhibitors as potential novel drugs to induce angiogenesis. These pharmacological agents could be used to facilitate new blood vessel formation for the treatment of cardiovascular diseases, such as coronary artery disease or heart injury, that are frequently observed in the aged population.

HeLa human cervical carcinoma and 293T human embryonic kidney cell lines were cultured in DME medium containing 10% FBS and antibiotics (streptomycin and penicillin). The HMVEC line was cultured in endothelial growth medium (EGM) containing 2% serum and 0.2% bovine brain extract (Shao and Guo, 2004). Human thyroid cancer cell lines: the BCPAP human thyroid cancer cell line harboring the BRAF^V600E mutation was established from the primary tumor of a 76-yr-old woman with poorly differentiated PTC with squamous metaplasia (Fabien et al., 1994) and provided by G. Damante (University of Udine, Udine, Italy); and the 8505C human metastatic cell line (human ATC) was purchased from Deutsche Sammlung von Mikroorganismen und Zellkulturen (German collection of microorganisms and cell culture) harboring the BRAF^V600E mutation. The human metastatic ATC-derived cell line SW1736 harboring the BRAF^V600E mutation was provided by N.E. Heldin (University of Uppsala, Uppsala, Sweden). The FTC133 human FTC-derived cell line and the FTC236 human metastatic FTC-derived cell line were provided by O.H. Clark (University of California, San Francisco Medical Center, San Francisco, CA). BCPAP, FTC133, and FTC236 cells were cultured in DME medium with 10% FBS and antibiotics. SW1736 and 8505C cells were cultured in RPMI medium containing 10% FBS and antibiotics.

pIRES-VEGFR2-HA was obtained from J. Huot and F. Houle (Laval University Cancer Research Center, Quebec, Canada). VEGFR2 mutants were generated with the QuikChange XL Site-Directed Mutagenesis kit (Agilent Technologies) according to the manufacturer’s instructions. Various VEGFR2 cDNAs were subcloned using the Pfu polymerase (Agilent Technologies) into the pGEX vector to create GST-VEGFR2 in frame fusion proteins. Short hairpin RNAs (shRNA lentivirus vectors), i.e., shRNA–β-TRCP1, shRNA–β-TRCP1+2, shRNA-GFP, and various CKI constructs were described previously (Jin et al., 2003; Shirogane et al., 2005). Myc–β-TRCP1, Myc–β-TRCP2, Myc–β-TRCP1-R474A, and HA–β-TRCP1 constructs were described previously (Jin et al., 2003; Wu et al., 2003; Shirogane et al., 2005). shβ-TRCP–resistant β-TRCP1 cDNA was described previously (Inuzuka et al., 2010) and was subcloned into Blasticidin-encoding lentiviral vector obtained from Addgene. Myc-Cullin 1, Myc-Cullin 2, Myc-Cullin 3, Myc-Cullin 4A, and Myc-Cullin 5 constructs were gifts from J. DeCaprio (Dana-Farber Cancer Institute, Boston, MA). shRNA constructs against various CKI isoforms were obtained from J. Jin (The University of Texas Health Science Center, Houston, TX). Lentiviral shRNA constructs against GFP and CKIδ were obtained from W. Hahn (Dana-Farber Cancer Institute, Boston, MA). Lentiviral shRNA constructs against VEGFR2 were obtained from Thermo Fisher Scientific, and the specific clone number TRCN0000001687 was used to effectively knock down VEGFR2 in HMVEC and BCPAP cell lines. shRNA lentiviral vectors against Cullin-1 and Cullin-4A were gifts from J. Wade Harper (Harvard Medical School, Boston, MA).

5 × 10⁵ HEK 293T cells were grown in 60-mm plates and transfected using Lipofectamine (Invitrogen) in OptiMEM (Invitrogen) for 48 h according to the manufacturer’s instructions. After infection, the cells were selected with 1–2 µg/ml puromycin for 72 h to eliminate the uninfected cells before collecting the whole cell lysates (WCLs) for the subsequent biochemical assays.

Either lentiviral or retroviral infections (viral transductions) were performed as follows: 6 × 10⁵ HEK 293T cells were seeded in 60-mm dishes and cotransfected the next day with each lentivirus or retrovirus vector, along with helper plasmids (i.e., gag-pol and VSV-G were used for lentiviral infections). Media with progeny virus from transfected HEK 293T cells was collected every 24 h for 2 d, and then filtered with 0.45-µm filters (Millipore) and freshly used to transduce 293T, HeLa, HMVEC, and BCPAP cells overnight in the presence of 8 µg/ml Polybrene (Sigma-Aldrich). Finally, stable transduced cells were treated with 2.5 µg/ml blasticidin or 1–2 µg/ml puromycin (Sigma-Aldrich) for selection. Gene or protein knockdown or overexpression in the transduced cells was confirmed by real-time RT-PCR and Western blot analysis. All assays were performed in triplicate.

Anti c-Myc, polyclonal anti-HA, anti-CKIδ (H-60), anti-CKIα, anti-Cullin 1, anti–FLK-1 (A-3) AC, anti–β-TRCP, anti-CCM3, and anti–human VEGF-A antibodies were purchased from Santa Cruz Biotechnology, Inc. Anti-Cullin5 was from Bethyl Laboratories, Inc., and anti-Cullin4 (2699S) and anti–β-TRCP (D13F10) antibodies were purchased from Cell Signaling Technology. Anti-CKIε antibody was purchased from BD. Anti-tubulin, polyclonal anti-FLAG, monoclonal anti-FLAG, peroxidase-conjugated anti–mouse secondary antibody, peroxidase-conjugated anti–rabbit secondary antibody, anti-Flag agarose beads, and anti-HA agarose beads were purchased from Sigma-Aldrich. Monoclonal anti-HA antibody was purchased from Covance. Anti-GFP antibody was obtained from Invitrogen. Anti-VEGFR1, anti-VEGFR2, anti-VEGFR3, anti–phospho-VEGFR2-1175, anti–phospho-AKT (ser473), anti–total-AKT1, anti-pERK1/2, and anti–total-ERK1/2 antibodies were purchased from Cell Signaling Technology. Sorafenib was purchased from Enzo Life Sciences and VEGF-A was obtained from Dr. Lawler’s laboratory.

Cells were lysed in EBC (50 mM Tris, pH 8.0, 120 mM NaCl, and 0.5% NP-40) buffer supplemented with protease inhibitors (Complete Mini; Roche) and phosphatase inhibitors (phosphatase inhibitor cocktail set I and II; EMD). The protein concentrations of the lysates were measured using a protein assay reagent (Bio-Rad Laboratories) on a DU-800 spectrophotometer (Beckman Coulter). The lysates were then resolved by SDS-PAGE and immunoblotted with the indicated antibodies. For immunoprecipitation, 20 h after transfection, cells were treated with 10 µM MG132 overnight before harvesting for immunoprecipitation procedures. 800 µg of protein lysates were incubated with the appropriate antibody (1–2 µg) overnight at 4°C, followed by the addition of carrier beads. Immunocomplexes were washed five times with NETN buffer (20 mM Tris, pH 8.0, 100 mM NaCl, 1 mM EDTA, and 0.5% NP-40) before being resolved by SDS-PAGE and immunoblotted with indicated antibodies.

Cells were plated into 6-cm tissue culture dishes 20 h before transfection. When cells reached an appropriate confluence, they were transfected with 2.0 µg HA-VEGFR2, along with 1.0 µg Flag-β-TRCP1 and 0.1 µg of a plasmid encoding GFP as an internal control, in the presence or absence of 0.4 µg Myc-CKIδ. For half-life studies, 20 µg/ml cycloheximide (CHX; Sigma-Aldrich) was added to the medium 40 h after transfection. At various time points thereafter, cells were lysed and protein concentrations were measured. 30 µg of the indicated WCLs were separated by SDS-PAGE and protein abundances were measured by immunoblot analysis.

RNA was extracted with an RNeasy mini kit (QIAGEN), and the reverse transcription reaction was performed with the ABI TaqMan Reverse Transcription Reagents (N808-0234). After mixing the resulting template with VEGFR2 (Hs00911700_m1), β-TRCP1 (Hs001182707_m1), or β-TRCP2 (Hs00362667_m1) primers and TaqMan Fast Universal PCR Master Mix (4352042; Applied Biosystems), the real-time RT-PCR reaction was performed with the ABI-7500 Fast Real-time PCR system (Applied Biosystems). VEGF-A copies number/10⁶ 18S were assessed by absolute quantification using real-time RT-PCR (Shih and Smith, 2005) using the VEGF-A and 18S primer sequences provided by the Multigene Transcriptional Profiling Core Facility (Center for Vascular Biology Research, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA).

Binding to immobilized GST proteins was performed as described previously (Wei et al., 2004). Where indicated, the GST-VEGFR2 proteins were incubated with CKIδ in the presence of ATP for 1 h before the binding assays.

CKIδ was purchased from New England Biolabs, Inc. The CKI in vitro kinase assays were performed according to the manufacturer’s instructions (New England Biolabs, Inc.). In brief, 5 µg of indicated GST fusion proteins were incubated with purified active CKIδ in the presence of 5 µCi γ-[³²P]ATP and 200 µM cold ATP in the CKI reaction buffer for 30 min. The reaction was stopped by the addition of SDS-containing lysis buffer, resolved on SDS-PAGE, and detected by autoradiography (Inuzuka et al., 2010).

The in vitro ubiquitination assays were performed as described previously (Jin et al., 2005). To purify SCF^β-TRCP1 complex, 293T cells were transfected with vectors encoding GST–β-TRCP1, Myc–Cul-1, Myc-Skp1, and HA-Rbx1. The SCF^β-TRCP1 (E3) complexes were purified from the WCLs using GST-agarose beads. Before the in vitro ubiquitination assay, indicated GST-VEGFR2 proteins were incubated with purified, recombinant active CKIδ in the presence of ATP (with kinase reaction buffer as a negative control) at 30°C for 30 min. Afterward, the kinase reaction products were incubated with purified SCF^β-TRCP1 (E3) complexes in the presence of purified, recombinant active E1, E2 (UbcH5a and UbcH3), ATP, and ubiquitin at 30°C for 45 min. The reactions were stopped by the addition of SDS-PAGE sample buffer and the reaction products were resolved by SDS-PAGE and probed with the indicated antibodies (Inuzuka et al., 2010).

Migration assays were performed according to manufacturer’s instructions using chambers (Corning) that did not contain Matrigel. In brief, shGFP- or shβ-TRCP cells (2.5 × 10⁴ HMVECs/0.5 ml) without serum were seeded into the insert. The bottom face of the filters (with 8-µm pore) were coated with type I collagen. 5% FBS in growth medium was added into the wells as a chemoattractant. In brief, these trans-wells/chambers were used to study the effect of β-TRCP knockdown on the migration of transduced HMVECs in the presence or absence of 30–100 ng/ml VEGF-A. The migration assay was performed for 5–18 h in culture and then inserts were fixed in methanol for 2 min, stained with Giemsa for 30 min, washed in distilled water, and air dried. Additionally, shGFP or shVEGFR2 BCPAP cells were grown in 10% serum for 24 h, and then 1 × 10⁵ cells/0.5 ml were seeded into the insert (in the presence or absence of 100 ng/ml human VEGF-A, added into the insert along with the cells) of the transwell and 5% FBS in growth medium was added into the wells as chemoattractant. Migration assay was run for ∼4 h. BCPAP cells (EV or HA-β-TRCP1) were cultured (2 × 10⁴ cells/0.5 ml/insert of transwell) were grown for 24 h with low serum (2%) and then serum-free overnight before migration assay. 200 ng/ml VEGF-A was used as chemoattractant in the well. Migration assay was run for ∼18 h.

We also generated a trans-well cell co-culture system (Fig. 9 A). BCPAP cells (EV or HA-β-TRCP1) were cultured (2 × 10⁴ cells/well) in 24-well plates. HMVECs (5 × 10⁴ cells/0.5 ml) were placed in inserts and placed on top of each well in serum-free medium. Therefore, these two types of cells were separated by the insert membrane. The secreted VEGF-A from the BCPAP cells (which were grown in the absence of serum) served as chemoattractant to induce migration of HMVECs (Fig. 9 A-C). The assay was conducted in the presence or absence of 5–10 µg human VEGF-A antibody for ∼44 h, and the migrated HMVECs were photographed and quantified as described. In all assays, migrated cells were counted (number of cells/field or number of cells/well) using a 10 or 20× objective, and four fields were chosen per well with two well per each condition. Furthermore, migrated cells were also photographed using a camera attached to the microscope.

The in vitro angiogenesis assays (tube formation assays) were performed as previously described (Ponce, 2001). In brief, control shGFP and shβ-TRCP HMVECs (4 × 10⁴/0.5 ml) in EGM medium containing 2% serum were seeded on Matrigel (BD) and treated with or without 100 ng/ml VEGF-A. After 5 h of incubation, cells were photographed using the camera attached to the microscope. The number of tubes was counted using a 10 or 20× objective and four fields were chosen per well with two wells per each condition.

Zebrafish were maintained according to Institutional Animal Care and Use Committee (IACUC-BIDMC) protocols. The MO sequences were used to target Fbxw11a (5′-AGGTTTTGTCCTCCATCTCCGTCTC-3′; or the splice form, 5′-GAAATACAACACACACCTTGAACAC-3′) and Fbxw11b (5′-ATTGTTGTTCTAGCTGAATCCAACC-3′; or the splice form, 5′-AAGAGCACCCAAACACTTACCATCT-3′). Control and knockdown MOs specifically targeted to β-TRCP2 were injected into wild-type or endothelial reporter (Kdrl:EGFP and fli1a:EGFP) zebrafish at the one cell stage (three independent replicate experiments were performed). At 72 h after fertilization, viability counts were enumerated and/or embryos were imaged by fluorescence microscopy using a microscope (Discovery; Carl Zeiss) to assess alterations in vascular development in two independent zebrafish reporter lines: Kdrl:EGFP (Cross et al., 2003) and fli1a:EGFP (Lawson and Weinstein, 2002). In situ hybridization and RT-PCR were conducted using standard zebrafish protocols (www.zfin.org). The following RT-PCR primers were used: fbxw11a forward 5′-ACAAAACCCTGGAGCTGATG-3′, and reverse 5′-GTCCAGGAACGACAGGATGT-3′; and fbxw11b forward 5′-CATTTTCACTGCGAGTTTCG-3′, and reverse 5′-TCAAAGAGCTGGATGCACAC-3′. Anti-VEGFR2 antibody (Abcam) was used to detect endogenous VEGFR2 in zebrafish. This antibody was raised with synthetic nonphosphopeptide derived from human VEGFR2 around the phosphorylation site of tyrosine 1059 (P-D-YP-V-R), a region which shares a high degree of homology with the zebrafish VEGFR2 protein.

Human thyroid cancer cell lines (1 × 10⁴ cells/well) were cultured in medium containing 0.5% serum and plated in a 96-well sterile culture plate. Cells were treated with or without various concentrations of sorafenib for 48 h. Cell viability was measured using the CellTiter-Glo Luminescent Cell Viability assay kit according to the manufacturer’s instructions (Promega).

BCPAP, EV-BCPAP, HA-β-TRCP1-BCPAP, SW1736, and 8505C thyroid cancer cells were cultured in 6-cm dishes with serum-free medium overnight. The next day, the medium was collected and secreted VEGF-A protein levels were determined by ELISA kit (R&D Systems) according to the manufacturer’s instructions.

sh-GFP or sh-β-TRCP transduced HMVECs were serum starved overnight and then collected, fixed with 75% ethanol, and stained with propidium iodide for flow cytometry analysis according to Nucera et al., (2010) using FACS (BD) to assess sub-G1 cell populations.

All tissue specimens were fixed with 10% buffered formalin phosphate and embedded in paraffin blocks. Histopathology evaluation was performed by an endocrine pathologist (P.M. Sadow) on hematoxylin and eosin (H&E)–stained tissue sections of 15 human thyroid carcinomas under the approval of Massachusetts General Hospital IRB committee (P.M. Sadow, 2011P000013). These were visualized with a microscope (BX 41; Olympus) and a camera (Q COLOR 5 photo; Olympus). 4-µm-thick sections of 10 formalin-fixed human thyroid carcinomas with squamous metaplasia foci (6 classical PTC, 1 PTC tall cell variant, 1 PTC diffuse sclerosing variant, 1 squamous cell carcinoma, and 1 poorly differentiated thyroid carcinoma) classified according to the World Health Organization (for squamous metaplasia features read Kleer et al., 2000). Tumor size was >0.4 cm (range: 0.4–4.5 cm; mean: 1.6 cm). 4 out of 10 thyroid carcinomas showed extra thyroidal extension, 4 out of 10 thyroid carcinomas showed lymphovascular invasion, and 4 out of 10 thyroid carcinomas showed neck lymph nodes metastasis (Table S1). These thyroid tumor tissues were used for immunohistochemical (IHC) procedures. After baking overnight at 37°C, deparaffinization with xylene/ethanol and rehydration were performed. IHC was performed using the β-TRCP (Santa Cruz Biotechnology, Inc.), VEGFR2 (2479; Cell Signaling Technology), VEGF-A (C1; Santa Cruz Biotechnology, Inc.), D2-40 (marker of lymphatic vessels; Covance), and CD31 (marker of endothelial cells; M20; Abcam) primary antibodies. The sections, treated in a pressure cooker for antigen retrieval (Biocare Medical), were incubated at 123°C in Citrate buffer (Target Retrieval Solution; DAKO), cooled, and washed with PBS. Antigen retrieval was performed for 60 min at RT. The primary antibody was detected using a biotin-free secondary antibody (K4011; Envision system; DAKO) and incubated for 30 min. All incubations were performed in a humid chamber at room temperature. Slides were rinsed with PBS between incubations. Sections were developed using 3,3′-diaminobenzidine (DAB; Sigma-Aldrich) as a substrate and were counterstained with Mayer’s Hematoxylin. The immunohistochemical markers were assessed semi-quantitatively using the following scoring method: 0, negative; 1+, 1–10% positive cells (low expression); 2+, 11–50% positive cells (moderate expression); and 3+, >50% positive cells (high expression).

Results were expressed as mean ± SD. Statistical comparisons were made using an unpaired Student’s t test and one way ANOVA (Excel; Microsoft). Densitometry analysis was performed using ImageJ software (National Institutes of Health) and values were normalized to tubulin. Differences were considered statistically significant when P < 0.05.

Fig. S1 lists the DSG(XX)S phosphodegron sequences in VEGFR2 recognized by β-TRCP, as well as potential and predicted CKI-mediated phosphorylation sites in VEGFR2. Fig. S2 presents zebrafish embryo viability data after treatment with MOs that target β-TRCP2. Fig. S3 lists the β-TRCP–recognizable DSG-like canonical degron consensus in VEGF-A among various species. Table S1 lists the β-TRCP1 protein expression levels and angiogenesis markers in human thyroid cancers.

We thank Rong Shao (Veterinary and Animal Sciences, University of Massachusetts, Amherst, MA) for providing the HMVEC line, and Jacques Huot and Francois Houle (Laval University Cancer Research Center, Québec, Canada) for the pIRES-VEGFR2-HA construct. We thank Harold F. Dvorak, Alex Toker, and Issac Rabinovitz for constructive suggestions. We thank Mark A Duquette, Shideh Khazerounian, Mei Zheng, and Nina Hu for technical assistance.

W. Wei is a DOD Prostate Research Program New Investigator and American Cancer Society Scholar. This work was supported in part by the National Institutes of Health (NIH) grants to W. Wei (GM089763 and GM094777) and J. Lawler (CA130895). S. Shaik is supported by a NRSA T-32 training grant, H. Inuzuka was awarded a K01 grant (AG041218) from NIH, and H. Fukushima is supported by a JSPS Postdoctoral Fellowship. C. Nucera was the recipient of a Ph.D. fellowship from the Italian Ministry of Education, Universities and Research (MIUR). C. Nucera is also funded by the American Thyroid Association for a Thyroid Cancer Grant.

The authors have no conflicting financial interests.