TAp73 suppresses tumor angiogenesis through repression of proangiogenic cytokines and HIF-1α activity

TAp73 Loss Enhances Anchorage-Independent Growth of Transformed Cells in Vitro and Tumor Growth and Angiogenesis in Vivo.

We have previously shown that loss of ΔNp73 impairs tumor development in vivo (11). To investigate the effect of TAp73 deficiency on tumorigenesis, we transformed TAp73^+/+ and TAp73^−/− primary mouse embryonic fibroblasts with E1A and H-Ras^V12 (MEF^E1A/Ras) (Fig. S1A). No difference was found in proliferation rate in vitro (Fig. S1 B and C); however, we observed an enhanced capacity of TAp73^−/−-transformed cells for anchorage-independent growth (Fig. S1 D and E). This supports previous data showing that TAp73 knockdown (KD) by siRNA confers anchorage-independent growth to BJ-TE human fibroblasts (19). To study how TAp73 controls the establishment of tumors in vivo, transformed MEFs were injected s.c. into athymic nude mice. TAp73-deficient MEFs^E1A/Ras formed larger tumors with shorter latency compared with WT (Fig. 1 A and B), indicating that loss of TAp73 confers tumor cells a growth advantage in vivo. Strikingly, we observed that the TAp73^−/− tumors had a more vascularized appearance compared with WT, and immunofluorescence staining using blood vessel marker endomucin revealed extensive angiogenesis (Fig. 1 C and E). In light of these results, we investigated the angiogenic phenotype in tumors derived previously from ΔNp73^−/− MEF^E1A/Ras cells (11) and observed a reduction in blood vessel formation, indicating that ΔNp73 positively regulates angiogenesis (Fig. 1 D and F). To validate our findings in a spontaneous tumor model, the TAp73^−/− and ΔNp73^−/− mice were crossed to the EμMyc B-cell lymphoma model (20). In agreement with the E1A/Ras model, we found more blood vessel formation in lymphomas from TAp73^−/−/EμMyc and less in ΔNp73^−/−/EμMyc compared with p73^+/+/EμMyc mice (Fig. 1 G, H, J, and K). Taken together, this suggests that p73 isoforms have opposing roles in regulating tumor angiogenesis.

p73 Isoforms Have Opposite Effect on Endothelial Cell Sprouting in Vivo.

To determine whether the effect of p73 isoforms on tumor angiogenesis is a result of factors produced by the tumor cells and not tumor size, we used a short-term xenograft model using Tg(fli:EGFP) transgenic zebrafish (21). In this model, green fluorescent protein (EGFP) is expressed throughout the vasculature, allowing for visualization of tumor cell interactions with endogenous endothelial cells in vivo. Zebrafish embryos were injected with fluorescently labeled WT, TAp73^−/−, or ΔNp73^−/− MEFs^E1A/Ras and scored for endothelial cell sprouting 20 h after injection (Fig. 1I). Consistent with our previous results, we observed an almost twofold increase in angiogenic response toward TAp73^−/− cells compared with WT cells (Fig. 1L). In contrast, ΔNp73-deficient cells elicited a twofold decrease in angiogenic sprouting (Fig. 1L), demonstrating that p73 isoforms have opposing function in regulating tumor angiogenesis by directly affecting endothelial cell sprouting.

TAp73-Deficient Tumor Cells Increase Blood Vessel Permeability Through Disruption of Endothelial Vascular Endothelial-Cadherin Junctions.

Normally, a mature blood vessel endothelium functions as a barrier that limits vascular leakage. However, tumor cells stimulate endothelial cells to start to proliferate, migrate, and become more permeable (22). Considering the enhanced effect on endothelial cell sprouting, we next investigated the effect of TAp73 loss on blood vessel permeability. Nude mice bearing TAp73^+/+ or TAp73^−/− MEF^E1A/Ras tumors of similar size were injected with FITC-labeled high-molecular dextran. We observed more FITC-dextran leakage into the extravascular space in TAp73^−/− tumors compared with in WT tumors, demonstrating that blood vessels in TAp73^−/−-derived tumors have increased vascular permeability (Fig. 2 A and B). To determine whether this increase is caused by factors directly produced by TAp73^−/− tumor cells, we cultured TAp73^−/− and TAp73^+/+ MEF^E1A/Ras cells in hypoxia to mimic the tumor microenvironment and collected conditioned media (CM). The CM was added to confluent human dermal microvascular endothelial cells (HuDMEC) to analyze the effect on endothelial cell–cell contact (Fig. 2C). Using scanning electron microscopy (SEM), we found that HuDMECs exposed to TAp73^+/+ MEF^E1A/Ras CM still retain interendothelial bonds. In contrast, cells treated with TAp73^−/− MEF^E1A/Ras CM show a dramatic loss in interconnecting network, resulting in more gaps between the endothelial cells. This indicates that in hypoxic conditions, TAp73^−/− tumor cells are producing factors that decrease the cell–cell contact between endothelial cells, thus increasing blood vessel permeability (Fig. 2D). Endothelial cell–cell junctions are regulated by a variety of extracellular stimuli, which often lead to reorganization of interendothelial contact proteins, such as vascular endothelial (VE)-cadherin (23). Using immunofluorescence, we observed a robust reduction in VE-cadherin signal on the cell surface as well as disruption of cell contact between HuDMECs treated with CM from TAp73^−/− MEF^E1A/Ras cells exposed to hypoxia (Fig. 2 E and F), suggesting that TAp73-deficient tumor cells produce and secrete factors that disrupt endothelial cell–cell contact through reducing the cell surface localization of VE-cadherin.

Enhanced Angiogenic Response in TAp73-Deficient Tumors Is a Result of Increased Expression of Both Proangiogenic and Proinflammatory Genes.

To identify target genes involved in angiogenesis and vascular homeostasis regulated by TAp73, we profiled TAp73^−/− and TAp73^+/+ tumors, using a commercial angiogenesis RT² profiler PCR array. Among the 84 genes analyzed, 13 showed a significantly up- or down-regulation (≥twofold difference; P value ≤ 0.05) in TAp73-deficient tumors (Fig. 3 A and B; full list, Table S1), and nine of the potential target genes were validated to be deregulated in TAp73-deficient tumors (Fig. 3C). Their expression was mostly reversed in ΔNp73^−/− tumors (Fig. 3D). To investigate whether the enhanced expression of proangiogenic genes originated from the tumor cells themselves or from the tumor stroma, we analyzed the expression of our genes of interest in TAp73^−/−, ΔNp73^−/−, and WT MEFs^E1A/Ras grown both in normoxic and hypoxic conditions in vitro (Fig. 3 E and F). Among genes up-regulated in TAp73^−/− cells, we found proinflammatory chemokines known to promote both angiogenesis and leukocyte migration (Ccl2, Cxcl1, Cxcl2), suggesting that TAp73 loss creates a proinflammatory milieu within the tumor. In addition, the proangiogenic genes Vegfc, Ereg, and Tnfaip2 were shown to be up-regulated in TAp73^−/− cells. VEGF-C and epiregulin (Ereg) are both known to induce angiogenesis by promoting endothelial cell proliferation (24, 25). Furthermore, VEGF-C also induces vessel permeability (26). Interestingly, their expression was even further potentiated in hypoxia (Fig. 3E).

The proangiogenic expression was reversed on ectopic expression of TAp73β in TAp73^−/− MEF^E1A/Ras cells, as well as in human MCF7 breast adenocarcinoma cells, showing that TAp73 indeed is suppressing proangiogenic genes (Fig. S2 A–D). Intriguingly, ΔNp73^−/− MEF^E1A/Ras cells only have reduced expression of Ccl2, Cxcl1, and Cxcl2 in hypoxia (Fig. 3F), suggesting that ΔNp73 enhances expression of these chemokines specifically in low-oxygen conditions. Reintroduction of ΔNp73, and especially ΔNp73α, into ΔNp73^−/− cells restored the expression of most cytokines (Fig. S2 E and F). Next, we examined the effect of TAp73 KD in human H1299 nonsmall cell lung carcinoma cells with stable shRNA KD of TAp73, as well as MDA-MB-231 breast adenocarcinoma cells treated with siRNA against TAp73. Again, we observed an up-regulation of several proangiogenic factors, especially CCL2, CXCL1, CXCL2, and VEGF-C, on TAp73 KD (Fig. 3 G and F). We did not observe any significant effect on ΔNp73 levels on TAp73 KD or loss in the cell lines we used, showing the isoform-specificity of the KD (Fig. S3). Taken together, our findings show that TAp73 suppress the production of proangiogenic factors and that loss of TAp73 or ΔNp73 overexpression releases this repression, thus increasing tumor angiogenesis.

Importantly, the only gene significantly down-regulated in TAp73^−/− and, conversely, up-regulated in ΔNp73^−/− tumors and cells compared with WT was brain-specific angiogenesis inhibitor-1 (Bai1) (Fig. 3 C–F). The Bai1 gene encodes a large transmembrane protein that contains several thrombospondin type 1 repeats, and its extracellular part can be cleaved into antiangiogenic peptides called vasculostatins (27). Reintroduction of TAp73α or TAp73β into TAp73^−/− cells up-regulates Bai1, whereas ΔNp73α and ΔNp73β expression suppress Bai1 expression (Fig. S4 A and B). Furthermore, using H1299 cells with stable shRNA KD of TAp73 results in BAI1 down-regulation (Fig. S4 C and D). In contrast, ectopic overexpression of TAp73 isoforms, and TAp73β in particular, increases BAI1 expression in H1299 (Fig. S4E), suggesting it is a direct transcriptional target of TAp73. The BAI1 gene has been reported to contain a putative p53-response element (p53RE) (28). Analysis of publically available p53 chromatin immunoprecipitation (ChIP)-sequencing data confirmed the presence of a p53RE in intron 10 of BAI1, and we also observed a putative p53RE in intron 10 of the mouse Bai1 gene (Fig. S4F). Using ChIP, we could detect direct binding of endogenous TAp73 to intron 10 in Bai1 (Fig. S4G). Furthermore, to determine whether TAp73 is directly activating Bai1 expression, BAI1 intron 10 was cloned into a luciferase reporter and coexpressed with different p73 isoforms, as well as with p53. Ectopic expression of either TAp73β or p53 activated the Bai1 luciferase reporter. In contrast, ΔNp73 isoforms suppressed reporter activity (Fig. S4 H and I), indicating that BAI1 is a direct target of TAp73β and p53 and that high levels of ΔNp73 inhibit its expression.

Shifting the Balance Between p73 Isoforms Correlates with Angiogenesis in Patients with Breast Cancer and Affects Accumulation of Hif-1α During Hypoxia.

Next, we investigated whether the enhanced angiogenic phenotype we observed could be found in human cancer. TAp73 is rarely mutated or lost in human solid tumors; instead, ΔNp73 is often up-regulated, thus shifting the balance between the isoforms (1). Up-regulation of ΔNp73 has been associated with poor prognosis in breast cancer (18), and we therefore analyzed mRNA expression in 1,048 breast cancers from The Cancer Genome Atlas (TCGA) (29). Patient samples were separated by high or low ΔNp73 expression (99 vs. 449 patients), and through gene set enrichment analysis (GSEA), both angiogenesis and hypoxia signatures were associated with high ΔNp73 levels (Fig. 4A and Datasets S1 and S2), suggesting that shifting the balance between p73 isoforms toward ΔNp73 increases angiogenesis in human tumors.

The correlation between up-regulation of hypoxia-induced genes and high ΔNp73 levels was of particular interest, considering our findings that expression of proangiogenic target genes was further enhanced in TAp73^−/− and repressed in ΔNp73^−/− MEFs on hypoxia (Fig. 3 E and F). Transcriptional response to hypoxia occurs mainly through activation of the hypoxia-inducible factor (HIF) family of transcription factors. HIF-1α is normally rapidly degraded during normoxic conditions but is stabilized on hypoxia, dimerizes with HIF-1β, and translocates to the nucleus, where it binds to hypoxia-response elements (HRE) in target promoters (30). We thus analyzed the promoter region of the proangiogenic genes up-regulated in TAp73^−/− cells and found that all of them contained HRE, and that this regulation was conserved between mouse and human (Fig. S5), suggesting that expression of HIF-target genes is enhanced on TAp73 loss. To confirm that they indeed are direct Hif-1α target genes, we performed ChIP, using anti-Hif-1α and normal IgG on chromatin derived from TAp73^−/− cells grown in hypoxic conditions. As expected, endogenous Hif-1α was found bound to all target gene promoters, with the exception of Cxcl2 (Fig. 4B), demonstrating that they are Hif-1α target genes. To further validate the dependence on Hif-1α, we used siRNA to KD Hif-1α in TAp73^−/− cells (Fig. 4C) and assessed the expression of target genes in both normoxia and hypoxia. Importantly, all target genes were severely down-regulated in hypoxia on Hif-1α KD, showing they are regulated by Hif-1α (Fig. 4D).

Next, we investigated whether TAp73 and ΔNp73 had a direct effect on Hif-1α levels. Considering the tight regulation of HIF-1α protein, we analyzed its accumulation in response to hypoxia and found a significant increase in TAp73^−/− cells compared with WT (Fig. 4E), whereas HIF-1α protein levels decreased in ΔNp73^−/− cells compared with WT (Fig. 4F). We did not observe any differences in Hif-1α mRNA levels on deletion of either TAp73 or ΔNp73 (Fig. 4 G and H), suggesting p73 isoforms regulate Hif-1α protein stability, rather than its expression in response to hypoxia. HIF-1α protein accumulation on TAp73 loss was also confirmed in MCF7 cells treated with siRNA against TAp73 (Fig. 4I). In contrast, overexpression of TAp73 down-regulates HIF-1α levels (Fig. 4J), suggesting TAp73 normally inhibits HIF-1α protein accumulation in response to hypoxia and that on TAp73 loss, HIF-1α protein levels increases, resulting in enhanced activation of proangiogenic HIF-1α target genes (Fig. 4K).