Embryonic Protein Nodal Promotes Breast Cancer Vascularization

Metastatic breast cancer cells exhibit stem cell–like properties (1, 2) and the expression of hESC-associated genes is associated with poor prognosis in breast cancer patients (3, 4). Nodal, a morphogen from the TGF-β superfamily, is a hESC-associated protein that becomes reexpressed during cancer progression (1, 5–7). Nodal signals via a SMAD2/3-dependent pathway following activation of a receptor complex, including ALK4/7, ActRIIB, and the coreceptor Cripto-1 (8). Recent studies show that Nodal promotes a dedifferentiated phenotype in melanoma, glioma, prostate cancer, and pancreatic cancer and that it increases cancer cell invasion and tumorigenicity in these cancer types (5, 7, 9, 10). Nodal expression is also correlated with breast cancer progression, such that Nodal is absent in normal breast tissues, yet is aberrantly expressed in invasive breast cancer lesions (6).

The acquisition of stem cell markers has been linked to the enhancement of hallmarks of cancer, such as self-sufficiency in growth signals and cellular invasion. Angiogenesis has also been correlated with the acquisition of stem cell markers. For example, CD105⁺ human renal carcinoma cells, expressing Nestin, Nanog, and Oct3/4, promote angiogenesis by shedding microvesicles that contain mRNAs for proangiogenic growth factors (11). In addition, a mesenchymal-like subclass of high-grade gliomas has been shown to exhibit elevated expression of proangiogenic factors, including VEGF and PECAM (CD31), and a recent study revealed that glioma stem cell–like populations promote vascularization via a combinatorial mechanism involving the secretion of VEGF and stromal derived factor-1 (12, 13). Lastly, one study has reported that Nodal regulates VEGF expression and vessel density in gliomas (14). However, this latter study did not directly assess the role of Nodal in the regulation of angiogenesis or vasculogenesis.

Here, we show that Nodal protein is correlated with high microvascular density in human breast cancer tissues, and that it causes vascular recruitment in vivo. In vitro, we found that Nodal increases the ability of breast cancer cells to promote tube formation by endothelial cells, and that it regulates the expression of VEGF and PDGF in these breast cancer cells. Finally, using inducible short hairpin RNAs (shRNA), we determined that Nodal can be targeted in established tumors to decrease vascular density and induce necrosis. In adults, Nodal is largely tumor specific; hence, it may be a useful target for the treatment of breast cancer.

Please see Supplementary Data for full methods.

Tissue was obtained from the Ontario Tumor Bank. Experiments involving animals were approved by the Animal Use Subcommittee at the University of Western Ontario (Protocol no. 2008-101).

Immunohistochemistry (IHC) for Nodal or CD31 was conducted using antibodies as described (Supplementary Table S1). To verify the specificity of Nodal staining, monoclonal and polyclonal antibodies were compared (Supplementary Fig. S1). In confirmation with previous studies, these antibodies revealed comparable staining patterns (15). We proceeded with the monoclonal antibody, as previously described (5, 6, 16). Cytoplasmic staining for Nodal was seen in all cases. Stained sections were blindly and independently scored by 2 pathologists (S.J. Done and J. Moreno) to derive an Allred score (17). Values were dichotomized using a total Allred score of 7 and 8 as high expression and 6 or less as low expression. Microvascular density (MVD) was blindly and independently scored based on CD31 immunohistochemical staining, following the recommendations from the International Consensus on Evaluation of Angiogenesis in Solid Human Tumors (18). Data were dichotomized using the average number of vessels among all samples as the cutoff between high and low MVD values. All scoring values obtained by S.J. Done and J. Moreno had a Pearson correlation coefficient (rp) of 0.73 or higher. A total of 83 cases of 90 had both Nodal and CD31 MVD scores, which were used for statistical analyses. For clinical characteristics of these 83 patients, see Supplementary Table S2. Assessment of correlations was carried out using Pearson correlation and a 2-sided t test.

Human umbilical vein endothelial cells [HUVEC; American Type Culture Collection (ATCC)], and primary human adult microvascular endothelial cells (HMVECs; Invitrogen) were maintained as per instructions. All breast cancer cell lines were obtained from and validated by ATCC and were maintained as per instructions. Multiple constructs were used to knockdown or induce Nodal signaling in each of the cell lines used (Supplementary Methods).

Total RNA was isolated and reverse transcriptase PCR (RT-PCR) was carried out using human primer/probe sets (Supplementary Table S3).

Protein lysates were prepared and Western blotting was conducted as previously described (6) using appropriate antibodies (Supplementary Table S1). In accordance with previous studies (5, 6, 19, 20), up to 3 Nodal species were detected: pro-Nodal bands at approximately 39 kDa, multiple processed (glycosylated etc.) bands at approximately 50 kDa, and a cleaved Nodal band at approximately 15 kDa. These bands are proportionally expressed (20).

Tube formation, migration, viability, and proliferation were assayed. In some cases, rhNodal (100 ng/mL; R&D), VEGF (20 ng/mL; Invitrogen), and/or platelet-derived growth factor (PDGF; 20 ng/mL; Invitrogen) were added to the conditioned medium after it was collected. An amount of 20 ng/mL of VEGF and PDGF was used in accordance with previous reports (21, 22).

VEGF and PDGF were measured using Aushon SearchLight Chemiluminescent Angiogenesis Arrays (#84694; Aushon Biosystems) or by Western blot analysis.

In vivo angiogenesis assays (Trevigen) were carried out as per instructions.

One collagen-enmeshed grid containing 2 × 10⁶ cells was placed on each chorioallantoic membrane (CAM), and newly formed blood vessels were scored and quantified as previously described (23).

MDA-MB-231 cells were transfected with a doxycyclin-inducible control or Nodal-targeted shRNA. A total of 500,000 cells were injected into the flank of 6- to 8-week-old female Nude mice (Crl:NU-Foxn1^nu; Charles River). Doxycyclin was administered 2 weeks following onset of palpable tumors via diet (0.625 g doxycyline hyclate/kg; Harlan Laboratories). Immunofluorescence for CD31 and Nodal was conducted, and RNA was extracted to measure the expression of VEGF and PDGF. To assess necrosis, 3 tumor sections spaced evenly through tissue block were stained with hematoxylin and eosin (H&E). Necrosis was calculated as a percentage of the total tumor area. To assess hypoxia, animals were injected with hypoxyprobe-1 (60 mg/kg) and IHC for hypoxyprobe adducts was carried out on tissue sections.

Multiple comparisons for parametric data were carried out using one-way ANOVA and a Tukey–Kramer Comparisons Post-Hoc test. ANOVA on ranks and a Mann–Whitney rank sum test were used for nonparametric data. Student t test was used to compare 2 variables. All statistical tests were 2-sided and comparisons were considered statistically significant at P < 0.05.

Angiogenesis is a rate-limiting step in the metastatic cascade, and high MVD is correlated with a poor clinical outcome (24, 25). Given recent evidence linking tumor angiogenesis to stem cell markers, we sought to determine whether Nodal expression correlates with MVD in human breast cancers. Immunohistochemical analysis of Nodal and CD31 (to demarcate endothelial cells) in breast cancer patients was examined using serial sections (25). Breast cancer tissue from 90 patients was procured from the primary tumor site; 89 via surgical resection and 1 via excisional biopsy. The average age of the patients was 61 years (ranging from 27–90). The majority of the cases were invasive disease, diagnosed as being grade 3 (47%) or grade 2 (37.4%). Of the 90 samples, 83 could be scored for both Nodal and MVD and were used for statistical analyses. Clinical characteristics of these patients are listed in Supplementary Table S2. IHC for both Nodal and CD31 was heterogeneous, varying among individuals and within sections in extent, intensity, and localization. Nodal was generally confined to epithelial-like tumor cells (Fig. 1A; Supplementary Figs. S1 and S2). Nodal staining was not correlated to breast cancer grade (P = 0.140) or to estrogen receptor (ER) (P = 0.682), progesterone receptor (PR; P = 0.801) or HER-2 (P = 0.589) status. However, high levels of Nodal were associated with a high VD (n = 83; rp = 0.3, P = 0.0076), suggesting a clinical association between the acquisition of Nodal expression and vascular recruitment (Fig. 1A and B; Supplementary Fig. S2).

To explore whether Nodal regulates breast cancer angiogenesis, we first tested the effects of Nodal on endothelial cell tube formation and migration using HUVECs or HMVECs treated with conditioned media from stably transfected breast cancer cells. For our loss-of-function experiments, we used MDA-MB-231 cells, which express high levels of Nodal, and for our gain-of-function approach, we used T47D cells, which express relatively low levels of Nodal (Fig. 2A; ref. 6). We found that conditioned media from MDA-MB-231 cells transfected with a Nodal-targeted shRNA (231+shNodal) significantly reduced endothelial branching (HUVEC, n = 3, P < 0.001; HMVEC, n = 3, P < 0.001) and migration (HUVEC, n = 4, P < 0.001; HMVEC, n = 3, P < 0.001) compared with conditioned media from MDA-MB-231 cells expressing a control shRNA (231+shControl; Fig. 2B–E, H–J). This effect was not due to altered cell viability or proliferation (Fig. 2F, G, K). In complementary Nodal gain-of-function experiments, we treated HUVECs with conditioned media from T47D cells transfected with an empty vector (T47D+EV) or a Nodal expression construct (T47D+Nodal; Fig. 2L). We found that tube formation (n = 3, P < 0.01) and migration (n = 4, P < 0.05) were significantly increased in response to conditioned media from T47D+Nodal cells, compared with T47D+EV cells (Fig. 2M–O). This effect was not due to a change in viability (Fig. 2P).

To determine whether Nodal was acting directly on the endothelial cells or indirectly by regulating the expression of angiogenic factors in the cancer cells, we added recombinant human Nodal (rhNodal) to the collected conditioned media during the course of the tube formation assays. Although migration of HUVECs was rescued by rhNodal treatment, migration of HMVECs and tube formation for both endothelial cell types were not affected (Fig. 2C–F, H–K). This suggests that Nodal acts on some endothelial cells directly, but that Nodal also promotes angiogenesis indirectly, perhaps by upregulating the secretion of proangiogenic proteins by the cancer cells.

In support of this concept, although Nodal inhibition with shRNA did not significantly alter cellular proliferation of MDA-MB-231 when conditioned medium was collected, it did cause a significant reduction in PDGF (n = 3, P = 0.018) and VEGF (n = 3, P = 0.007) protein (Fig. 3A–D). Furthermore, addition of rhNodal (100 ng/mL) to 231+shNodal cells rescued VEGF expression in these cells (Fig. 3E). To validate these observations, we measured VEGF and PDGF protein in response to Nodal knockdown in MDA-MB-468 cells using a protein array. We found that Nodal inhibition caused decreased expression of PDGF and VEGF in conditioned media (n = 3, P < 0.03) and cell lysate (n = 3, P < 0.001; Fig. 3F–J). In complementary gain-of-function experiments, treatment of T47D cells with rhNodal did not affect cellular proliferation (Fig 3K and L) but did cause a significant increase in VEGF and PDGF mRNA expression (n = 3, P < 0.05; Fig. 3M). Furthermore, treatment of T47D or MCF-7 cells with increasing concentrations of rhNodal (0, 50, or 100 ng/mL) resulted in an increase in VEGF and PDGF protein expression (Fig. 3N). Similarly, using a protein array, we found that transfection of T47D cells with a Nodal expression construct caused a significant increase in PDGF protein expression in conditioned media (n = 3, P = 0.045) and cell lysate (n = 4, P = 0.029; Fig. 3O and P) and a significant increase in VEGF protein expression in cell lysate (n = 3, P = 0.029; Fig. 3Q–S). Lastly, we found that activation of Nodal type I receptor is required for VEGF and PDGF regulation, as treatment of T47D cells with SB431542 inhibited rhNodal-induced upregulation of these proteins (Fig. 3T).

We next investigated whether we could rescue the effects of shNodal-conditioned media on endothelial cells with recombinant VEGF and/or PDGF. We added these proteins to the conditioned media after it was collected from the cancer cells (i.e., during the incubation on endothelial cells). Unlike the addition of rhNodal, which did not rescue angiogenic phenotypes, we found that addition of VEGF and/or PDGF to 231+shNodal conditioned media during the tube formation and migration assays did restore both phenotypes in HUVEC and HMVEC cultures (Fig. 4A–C, E–G). This was not because of a significant change in viability (Fig. 4D and H).

To test the significance of our in vitro results, we carried out Directed In Vivo Angiogenesis Assays (DIVAA). Briefly, cancer cells were seeded into matrix-containing angioreactors, which were implanted subcutaneously into mice. Endothelial cells that invaded the angioreactors were quantified using a fluorescein isothiocyanate–lectin–based assay (26). We determined that 468+shControl cells (n = 7, P = 0.007) and 231+shControl cells (n = 10, P = 0.003) efficiently recruit vasculature within 10 days, and that this ability is mitigated when Nodal is knocked down (Fig. 5A–D). Moreover, this in vivo effect of Nodal knockdown on angiogenesis was rescued in 468+shNodal cells by the addition of VEGF (Fig. 5C and D). In contrast to our in vitro studies, we were able to partially recover the effects of Nodal knockdown with the addition of rhNodal (Fig. 5C and D). This may have been because the cancer cells in the implants responded to the rhNodal by reexpressing VEGF and PDGF. Alternatively, Nodal may affect other proangiogenic cell types in vivo, such as hematopoietic stem cells, not accounted for in our endothelial cell assays. In accordance with our in vitro studies, T47D+Nodal–transfected cells induced significantly higher vessel recruitment compared with controls (n = 10, P < 0.001; Fig. 5E and F).

As proof-of-principle, we used a chick CAM assay to assess the role of Nodal during vascularization in vivo over a 3-day period (23). We found that 231+shControl cells were able to induce almost 2-fold greater angiogenesis than 231+shNodal cells (n ≥ 37, P < 0.001; Fig. 5G). Furthermore, inclusion of recombinant VEGF (20 ng/mL) with 231+shNodal cells on the CAM resulted in a rescue of the angiogenic phenotype (Fig. 5G). When tumor formation was assayed over 5 days, we found that 231+shNodal cells were unable to form tumors compared with controls (Fig. 5H). Furthermore, despite promoting vascularization, VEGF did not rescue 231+shNodal tumor formation (Fig. 5H), suggesting that Nodal may regulate multiple aspects of tumorigenesis in this model.

We next evaluated the effects of targeting Nodal in established tumors. MDA-MB-231 cells transfected with a doxycycline-inducible Nodal shRNA (231+ishN) or Control shRNA (231+ishC) were injected into the flanks of mice. Control (–Dox) groups were given regular chow for the entire experiment, whereas +Dox groups were given chow-containing doxycycline during weeks 2 to 4 of the experiment (after tumor initiation; Fig. 6A and B). By administering doxycycline to animals with established tumors, we avoided significant changes in tumor size between treatments, minimizing confounding effects of tumor growth on our analysis of angiogenesis. Of note, we found no significant difference in tumor size over time between treatments (n ≥ 4, P = 0.4; Fig. 6C). Following tumor excision, immunofluorescence localization revealed that the distribution of Nodal mirrored the distribution of CD31 across treatments, and that CD31 expression in 231+ishN +Dox tumors was limited to the tumor periphery. Indeed, we found an approximately 50% reduction in CD31 staining in 231+ishN +Dox tumors compared with all other treatments (n ≥ 4, P = 0.04; Fig. 6D and E). In cases in which Nodal expression rebounded following doxycycline treatment in 231+ishN mice (a common limitation of shRNA technology), CD31 expression was similarly restored (Fig. 6F). In accordance with these findings, we found that 231+ishN +Dox exhibited a significant reduction in VEGF expression compared with all other treatment groups (n ≥ 3, P = 0.003; Fig. 6G). Significant differences in PDGF expression were not, however, detected.

As a corollary to our CD31 and VEGF measurements, we found that tumor necrosis was significantly higher and frequently extended closer to the tumor periphery in 231+ishN +Dox tumors compared with controls (n ≥ 4, P < 0.001; Fig. 7A and B). Furthermore, by injecting mice with Hypoxyprobe-1 before sacrifice and then immunostaining for this compound to delineate low oxygen levels, we determined that hypoxic regions surrounded areas of necrosis, and that 231+ishN +Dox tumors were more hypoxic than control tumors (Fig. 7C).

This study shows that the Nodal expression correlates with MVD in human breast cancer tissue, and that this embryonic protein enhances the angiogenic potential of breast cancer cells. Importantly, Nodal can be targeted within growing tumors to inhibit vascularization and promote necrosis. Because vascular density is correlated with metastasis and poor prognosis, our study strongly implicates Nodal as a prognostic indicator and therapeutic target for breast cancer.

In accordance with previous studies in glioma, Nodal promoted VEGF and PDGF expression in all of the human breast cancer cell lines tested (14). However, we found that there were differences in protein expression between lysate and conditioned media within a particular cell type, suggesting that Nodal may be involved in regulation of protein trafficking, protein stability, or secretion. Furthermore, although Nodal was consistent in promoting VEGF and PDGF, a full analysis of a proangiogenic signature was not conducted, leaving the possibility that Nodal may activate other cytokines to mediate vascularization.

When comparing rhNodal rescue treatments from in vitro versus in vivo experiments, we found that rhNodal differentially rescued the angiogenic phenotype. For instance, addition of rhNodal to shNodal-conditioned media after it was collected from the cancer cells did not rescue tube formation in vitro. However, addition of rhNodal to shNodal-containing DIVAA angioreactors did induce a partial rescue of vascularization. This suggests that rhNodal is able to induce signaling and downstream targets when in contact with cancer cells or with other cell types in the animal but is not able to propagate signaling when in contact with only endothelial cells. We speculate that this is due to differential expression of Nodal receptor components, for example, Cripto, across cell types. Although some studies indicate that Nodal may signal in a Cripto-independent manner (27), others suggest that Cripto is required for Nodal-induced phenomena. Moreover, Cripto has been linked to breast and colon tumorigenesis, and one study reported that Cripto promotes HUVEC migration in vitro and angiogenesis in vivo (28, 29). Future studies exploring patterns of Cripto expression on endothelial cells or endothelial progenitors, and the role of Cripto in Nodal-induced angiogenesis would be valuable for developing targeted therapies.

Although we showed that Nodal induces angiogenesis by regulating proangiogenic factors and promoting endothelial cell recruitment, it is possible that additional mechanisms of neovascularization also exist. One possibility is that in vivo, bone marrow–derived (BMD) progenitor cells are recruited to the tumor. Indeed, BMD endothelial progenitors can incorporate into the luminal walls of neovessels during breast cancer progression (30) and inhibition of BMD endothelial progenitor cell mobilization/function has been shown to decrease tumor vascularization (30, 31). Given the role of Nodal in the regulation of stem cell fate and morphogenesis, it is conceivable that it also regulates the recruitment of endothelial progenitors.

To recapitulate therapeutic intervention, we designed an assay wherein Nodal was inhibited after tumor initiation. We found that there was a decrease in vascularization when Nodal was inhibited in these tumors. In accordance with these observations, one study in melanoma showed a decrease in lung colonization following administration of a function-blocking antibody targeted to Nodal (32). Colonies that formed in mice treated with anti-Nodal antibody exhibited signs of apoptosis inclusive of cytoplasmic swelling and vacuole expansion; however, the antiangiogenic effects of Nodal inhibition was not directly explored. We suggest that Nodal inhibition reduced tumor growth, at least in part, by preventing angiogenesis.

It is not clear from our experiments whether vascular reduction via Nodal inhibition would yield a positive or negative response with respect to metastasis, particularly given our findings that Nodal knockdown results in tumor hypoxia. Tumor hypoxia is associated with cellular programs that yield enhanced metastasis. Moreover, hypoxia promotes tumor-initiating cell populations (33) and induces therapeutic resistance to chemotherapy and radiation (34, 35). Furthermore, overexpression of hypoxia-inducible factor α (HIF-1α) promotes metastatic phenotypes and is associated with increased breast cancer relapse and decreased patient survival (36–38). HIF proteins also upregulate of a number of stem cell–associated transcription factors that induce tumor-promoting programs, including TWIST1 and ZEB1/2 (39–43). We show that Nodal inhibition causes decreased VEGF expression and increased tumor necrosis concomitant with tumor hypoxia; however, we have previously shown that hypoxia induces Nodal expression through HIF-1α (20), and VEGF is one of the best-characterized targets of HIF proteins (44). Hence, studies that measure metastases following Nodal inhibition and that target Nodal and hypoxia in combination warrant investigation.

Collectively, we have shown that Nodal supports a proangiogenic niche. This discovery provides mechanistic insight into recent studies showing that cancer cells expressing stem cell–associated genes are better able to recruit vasculature than more differentiated cancer cells (13). Importantly, we have shown that Nodal is correlated with high MVD in human breast cancers, and that blocking Nodal expression in established tumors results in a marked reduction in vascularization concomitant with tumor necrosis. Our discovery suggests a novel role for Nodal as a target for the treatment of breast cancer.

No potential conflicts of interest were disclosed.

Conception and design: D.F. Quail, G. Zhang, D.A. Hess, L.-M. Postovit

Development of methodology: D.F. Quail, L.A. Walsh, G. Zhang, D.A. Hess, L.-M. Postovit

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D.F. Quail, L.A. Walsh, G. Zhang, S.D. Findlay, J. Moreno, L. Fung, D.A. Hess

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D.F. Quail, L.A. Walsh, G. Zhang, J. Moreno, L. Fung, A. Ablack, J. Lewis, S.J. Done, D.A. Hess, L.-M. Postovit

Writing, review, and/or revision of the manuscript: D.F. Quail, L.A. Walsh, G. Zhang, A. Ablack, J. Lewis, S.J. Done, D.A. Hess, L.-M. Postovit

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.F. Quail, D.A. Hess

Study supervision: L.-M. Postovit

This work was supported by the Canadian Institutes for Health Research (MOP 89714, MOP 119589, and PLS 95381) and the Cancer Research Society to L.-M. Postovit. D.F. Quail is supported by fellowships from the Canadian Breast Cancer Foundation and a CIHR training grant. L.A. Walsh is a recipient of a Natural Sciences and Engineering Research Council Alexander Graham Bell Canada Graduate Scholarship. J. Moreno is supported by the Terry Fox Foundation Strategic Health Research Training Program in Cancer Research at CIHR (TGT-53912). L.-M. Postovit is the recipient of the Premier New Investigator Award from the CIHR. D.A. Hess is the recipient of a new investigator award and MacDonald Scholarship for the Canadian Heart and Stroke Foundation.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.