Angiogenesis, lymphangiogenesis, growth pattern, and tumor emboli in inflammatory breast cancer

A review of the current knowledge


  • Peter B. Vermeulen MD, PhD,

    Corresponding author
    1. Translational Cancer Research Group and Pathology Department, Cancer Center of the Saint-Augustinus Hospital, University of Antwerp, Wilrijk, Belgium
    • Translational Cancer Research Group, University of Antwerp/University Hospital Antwerp, Oncology Center, General Hospital, St.-Augustinus, Wilrijk, Antwerp, 2610, Belgium
    Search for more papers by this author
    • Fax: (011) 3234433019

  • Kenneth L. van Golen PhD,

    1. Department of Biological Sciences, University of Delaware, Newark, Delaware
    Search for more papers by this author
  • Luc Y. Dirix MD, PhD

    1. Translational Cancer Research Group and Pathology Department, Cancer Center of the Saint-Augustinus Hospital, University of Antwerp, Wilrijk, Belgium
    Search for more papers by this author

  • The articles in this supplement were presented at the First International Inflammatory Breast Cancer Conference, Houston, Texas, December 5-7, 2008.


This objective of the current review was to provide the reader with a comprehensive summary of the literature related to 3 important and inter-related features of the biology of inflammatory breast cancer (IBC): angiogenesis, lymphangiogenesis, and the formation of tumor emboli. Information derived from animal models of IBC as well as from translational studies using tissue samples of patients with IBC are discussed. Cancer 2010;116(11 suppl):2748–54. © 2010 American Cancer Society.

Inflammatory breast cancer (IBC) is a form of locally advanced breast cancer (LABC) characterized by a specific set of symptoms. The diagnosis is a clinical one (AJCC T4d), supported by pathologic findings that will be discussed in this review. Critical to the diagnosis is the rapid onset of symptoms and accelerated local growth, usually within 3 to 6 months. Symptoms are swelling of the breast accompanied by skin edema, also described as peau d'orange, and erythema in more than half of the breast. There may also be exquisite pain. At the time of presentation, the large majority of patients have lymph node metastases, and distant metastases are often present. Considering cell-of-origin subtypes, most cases of IBC belong to the basal, the luminal-B, or the HER2-overexpressing subtype.1 Outcome is unfavorable, but long-term survivors are reported after extensive multimodality therapy for which neoadjuvant chemotherapy is an essential element.2 Approximately 2% to 5% of all invasive breast carcinomas are IBCs, and the incidence appears to be increasing.3 The true incidence is hard to determine because of the regional differences in diagnostic criteria as well as the lack of attention to the clinical symptoms, which are occasionally only deductable from anamnesis.

Angiogenesis in Animal Models of IBC

The first, and by now well-known, animal model of IBC is the MARY-X mouse model, a stable serial transplantable xenograft derived from a 45-year-old female patient with IBC.4 At least during initial transplant generations, there is exclusive intravascular growth, both in the primary subcutaneous tumor and in the lung metastases, suggesting that MARY-X is a model suitable to study both vascularization of IBC and embolus formation. Compared with the non-IBC xenografts MDA-MB-231 and MDA-MB-468, MARY-X has the largest stromal compartment, corroborating histomorphometric analyses demonstrating low tumor/stroma ratios in IBC in patients.5 Although the initial hypothesis was that stimulation of angiogenesis occurred followed by vascular homing, more recent studies demonstrated stem cell or tumor cell vasculogenesis (also called vasculogenic mimicry) encircling solid tumor cell nests.6 This passive formation of tumor emboli has also been described in a non-IBC murine breast carcinoma model, MCH66, in which blood vessels envelope around tumor cell nests.7 A prerequisite for passive embolus formation appears to be the strong homotypic cohesion of tumor cells forming large aggregates. In IBC, this is due to an overexpressed and functional E-cadherin.8-10

The second animal model of IBC is WIBC-9, again a stable serial transplantable xenograft mouse model, derived from a 50-year-old female patient with IBC.11, 12 In this model, vascularization is the result of peripheral angiogenesis with high CD31-positive vessel density and central vasculogenic mimicry. When isolating the tumor cells by laser-assisted microdissection and/or by applying species-specific antibodies for immunohistochemistry (IHC) or probes for polymerase chain reaction (PCR), these cells were shown to express vascular markers such as α-v-β-3 integrins, the vascular endothelial growth factor (VEGF) receptors flt1 and KDR, and the angiopoietin receptor tie-2. There was elevated expression in tumor tissue and serum of human and murine angiogenic (VEGF, basic fibroblast growth factor [bFGF], angiopoietin-1 [ANG1], and interleukin-8 [IL-8]), but not of lymphangiogenic factors or their related receptors.

A third animal model of IBC is the naturally occurring canine inflammatory mammary cancer.13, 14 In approximately one-third of cases, endothelial-like neoplastic cells, indicative of vasculogenic mimicry, are observed.15 These tumors overexpress both mRNA and protein of cyclooxygenase-2 (COX-2). Vascular channels formed by carcinoma cells have also been described in COX-2 overexpressing non-IBC in patients.16

Histomorphometric Studies of Angiogenesis in Human Samples of IBC

In a group of 22 patients with non-IBC (86% of whom had high histologic grade) and 45 patients with IBC from Tunisia with poussée évolutive (PEV) scores 2 to 3 (score 3 corresponding to American Joint Committee on Cancer [AJCC] classification T4d and score 2 corresponding to a clinical history of IBC but with erythema limited to less than half of the breast), microvessel density (MVD) was measured using the Weidner method, essentially a count of blood vessels in areas of elevated density spotted at low magnification, and with antibodies directed at CD31.17 Whether graded or expressed as a continuous variable, MVD was found to be significantly higher in the IBC samples: 51% of IBC samples versus only 14% of non-IBC samples belonged to the moderate-to-high MVD category (P = .02); the median MVD was 25.5 in IBC samples versus 6.5 in non-IBC samples (P = .009). In a more elaborate study, in addition to relative vascular area by the Chalkley point-overlap method (CD34), the endothelial cell proliferation fraction (CD34-proliferating cell nuclear antigen), blood vessel maturity (CD34-α smooth muscle actin), fibrin deposition (T2G1), and hypoxia (carbonic anhydrase [CA] IX) were measured after (double)immunostaining in a group of 35 consecutive patients with IBC (AJCC T4d disease) and 104 patients with T1 to T2, N0, non-IBC.18 The choice of a nonstage-matched control group is supported by gene expression profiling results and principal component analysis demonstrating a greater resemblance of IBC to T1 to T2 breast carcinomas than to locally advanced non-IBC19 (Figure 1). The strongest argument for increased angiogenesis in IBC is the elevated fraction of proliferating endothelial cells noted in this study (15% vs 9% in non-IBC; P = .05).18 The Chalkley relative vascular area was also higher in IBC. Approximately 90% of blood vessels were immature (ie, lacking pericyte coverage) in both IBC and non-IBC cases. Approximately half of the tumors in both groups expressed CA IX. Abundant stromal fibrin deposition was present in a higher fraction of IBC cases compared with non-IBC cases (26% vs 8%, respectively; P = .02).

Figure 1.

Principal component analysis using a list of 18,182 informative genes was performed to obtain global views of the variation in gene expression among the different breast cancer samples. Inflammatory breast cancer (IBC) samples were color-coded red; non-IBC (nIBC), T1, or T2 tumors were color-coded blue; and non-IBC, T3, or T4 tumors were color-coded green. The first principal component is represented by the x axis, whereas the y axis represents the second principal component. IBC and non-IBC samples are separated along the first principal component. Unsupervised hierarchical complete linkage clustering was performed using 250 informative genes having the greatest standard deviation.

Molecular Studies of Angiogenesis in Human Samples of IBC

mRNA expression of angiogenic factors and their receptors has been quantified by real-time reverse transcriptase (RT) -PCR in 16 IBC cases and 20 non-IBC cases (stage-matched and nonstage-matched).5 Relative gene expression levels of flt1, KDR, ANG1, TIE1, TIE2, COX-2, and bFGF were significantly higher in IBC samples. There was no difference in expression levels of VEGF and ANG2. None of the factors in this study was elevated in the non-IBC group. Higher relative gene expression of COX-2 and bFGF in IBC was confirmed by IHC on a tissue microarray (TMA). There was a positive correlation of gene expression levels of markers of vascular endothelium (flt1, KDR, ANG1, TIE1, TIE2; P <.01). The relative gene expression of VEGF was found to be positively associated with the tumor/stroma ratio (r = 0.4; P = .04), whereas the relative gene expression of bFGF was negatively associated with this ratio (r = −0.5; P = .006), suggesting a difference in cell types expressing these factors in IBC.

Genome-wide expression profiling studies contain information regarding angiogenesis in IBC. When comparing 37 patients with IBC (defined as T4d disease and/or dermal lymphatic invasion) with 44 patients with non-IBC (T1-2 in 30 patients and T3-4c in 14 patients) by applying homemade 8K cDNA microarrays, unsupervised hierarchical clustering resulted in a group enriched in IBC samples.20 In this group, there was overexpression of genes belonging to a vascular cluster (ie, genes expressed by human umbilical vein endothelial cells). Supervised analysis resulted in an IBC-related gene expression signature with 85% accuracy containing overexpressed genes upstream of the MAPK pathway, related to RhoC-induced angiogenesis, and overexpression of ARNT (β-subunit of hypoxia-inducible factor 1 [HIF1]). In another study, real-time quantitative RT-PCR of >500 genes known to be associated with angiogenesis and inflammation was performed in samples of 36 patients with IBC (AJCC T4d disease) and 22 patients with non-IBC LABC.21 This resulted in 27 significantly dysregulated genes, all up-regulated in IBC. Approximately one-third of these genes were related to angiogenesis (VEGF-A, TBXA2R, PTGS2/COX-2, THBD/thrombomodulin, ANGPT2/angiopoietin 2, CCL3, CCL5, CCR5, and IL-6 [HIF1A]; P = .06). There was no difference noted with regard to expression of VEGF2, VEGF3, VEGF4, VEGFR1, VEGFR2, VEGFR3, and IL-8. In contrast to earlier reports, no difference in expression of WISP3, RhoC, and E-cadherin was found.

Anti-VEGF Treatment in Patients With IBC

Given the elevated angiogenic activity in IBC, clinical trials based on a strategy to inhibit the VEGF signal have been developed. In a first study, 18 patients were treated with a combination of a small-molecule tyrosine kinase inhibitor of VEGFR-2 and conventional chemotherapy (doxorubicin).22 The clinical response rate was as high as 90%. Plasma VEGF levels increased with treatment. Core needle biopsies were taken before treatment, after 2 cycles, and at the time of mastectomy; the MVD (Weidner method, Factor VIII-related antigen as endothelial cell marker) was lower after treatment, with the percentage of reduction significantly related to overall survival. In a second study, 20 patients with IBC and 1 patient with LABC were treated with bevacizumab, the humanized monoclonal antibody directed at all isoforms of VEGF-A, alone during Cycle 1 and combined with doxorubicin and docetaxel during Cycles 2 to 7.23, 24 Fourteen patients had a clinical partial response, 5 patients had stable disease, and 2 patients developed progressive disease. Tumor biopsies after the first cycle demonstrated effects of bevacizumab on both tumor cells and endothelial cells. Automated quantification of the expression levels of phosphorylated VEGFR-2, with 2 different phosphorylation sites assessed by IHC, demonstrated a 67% reduction in tumor cells (P = .004), combined with a highly significant increase in tumor cell apoptosis and no change in tumor cell proliferation. There was a trend toward a decrease in the endothelial cell proliferation fraction, measured after double immunostaining with antibodies directed at CD31 and Ki-67 (P = .06) and a significant decrease of the level of CD31 expression in endothelial cells (41%; P = .007). No changes were observed in MVD, VEGF-A expression, and total VEGFR-2 expression. This study also aimed to determine predictive factors for response to bevacizumab in baseline biopsies. Protein expression levels by IHC of CD31 in the vasculature, of platelet-derived growth factor receptor (PDGFR)-β in the vasculature and in tumor cells, and of VEGF-A in the tumor cells were higher in responders (respective P values of .0004, .01, and .04). These immunohistochemical data were confirmed by gene expression analysis with gene ontology (GO) categories containing genes related to VEGFR activity (eg, PDGFR-α and PDGFR-β), and genes related to cell motility, locomotion, and localization, eg, CD31.

Lymphangiogenesis in Animal Models of IBC

To the best of our knowledge, lymphangiogenesis has not been explicitly investigated in the animal models of IBC. In studies of the MARY-X model, Factor VIII-related antigen is used as a marker of the vasculature.4 This is 1 of the most blood vessel endothelium-specific markers, with a mutually exclusive vascular expression pattern when compared with the lymph vessel endothelium marker D2-40.25 In studies using the WIBC-9 xenograft, CD31 is used as a vascular marker. CD31 is also expressed on lymph vessels, although its expression is weaker than on blood vessels.11 Therefore, part of the vessels described in this model might have been lymphatic ones. In this model, mRNA levels of genes coding for lymphangiogenic factors and receptors (human VEGF-C and VEGF-D, and murine VEGFR-3) were lower compared with non-IBC xenografts.12

Histomorphometric Studies of Lymphangiogenesis in Human Samples of IBC

When comparing 29 samples from patients with IBC with 56 samples from patients with non-IBC by double immunostaining of tumor tissue sections with antibodies directed at D2-40 and Ki-67 and computer-aided histomorphometric analysis, the peritumoral fraction of proliferating lymphatic endothelial cells was found to be 3-fold higher in IBC versus non-IBC (P = .005).25 This strongly suggests ongoing lymphangiogenesis, at a higher level in IBC. Intratumoral lymph vessel area and lymph vessel perimeter were also larger in IBC samples (respective P values of .01 and .07). The non-IBC control group was comprised of patients with T1 to T4a breast carcinomas.

Molecular Studies of Lymphangiogenesis in Human Samples of IBC

Levels of mRNA of the genes coding for the lymphangiogenic factors VEGF-C VEGF-D, their receptor VEGFR-3, and the lymphatic endothelial markers PROX-1 and LYVE-1 were found to be significantly (all P values <.02) elevated in 16 IBC samples versus 20 non-IBC samples.5 In contrast with these results, comparing 36 IBC samples with 22 non-IBC LABC samples by real-time RT-PCR did not demonstrate overexpression of VEGF-C, VEGF-D, or VEGFR-3.21 In this study, however, RhoC, WISP-3, and E-cadherin were not found to be differentially expressed either, although the opposite has been repeatedly shown (for E-cadherin,8 for RhoC,26 and WISP-3).

Tumor Emboli and Growth Pattern in IBC

Although not a prerequisite for diagnosis when adopting the AJCC T4d criteria, numerous emboli, both in the tumor parenchyma and in the dermis, are often described in the pathology report. They tend to be larger than in non-IBC and are believed (to the best of our knowledge, no clear proof is available) to be responsible for the skin edema. In non-IBC, emboli are mainly located in lymph vessels; to our knowledge this has not been studied thoroughly in IBC.

Lessons From Animal Models of IBC and Confrontation With Human IBC Samples

In the MARY-X model, intravascular growth of IBC coincides with strong circumferential expression of E-cadherin in tumor cells, and an intact E-cadherin-catenin axis.9 This conserved E-cadherin expression predicts the IBC phenotype with an odds ratio of 5.6 reported in a study containing 83 IBC samples. A tissue microarray study with 34 IBC samples and 41 non-IBC samples has confirmed E-cadherin overexpression in IBC.27 In a whole-section immunohistochemical study focusing on the tumor emboli in IBC, E-cadherin expression was found to be equally strong in the emboli compared with the invasive parenchymal component in 27 of 35 (77%) samples, and stronger in 6 of 35 (17%) samples.18 Expression of E-cadherin was never found to be weaker in the intravascular emboli. Two IBC cases with lobular carcinoma were negative for E-cadherin. Even in non-IBC, an increase in E-cadherin expression has been observed in intravascular tumor emboli,28 suggesting a biological function of E-cadherin related to embolus formation. Strong homotypical interactions might be necessary for stroma-independent intravascular growth.

The heterotypical interaction with endothelial cells is weak in the emboli in the MARY-X model due to the lack of functional, although overexpressed, MUC1.10 The absence of sialyl groups (Lewis-X and Lewis-A) is responsible for dysfunction.9 In human IBC samples, overexpression of glycosylated cytoplasmic MUC1, but not of sialated MUC1, has been described,29 confirming the findings of the animal model study.

To the best of our knowledge, passive embolus formation in the animal models, as described earlier in this review, has not been solidly confirmed in human samples of IBC or non-IBC.

In the MARY-X model, tumor cells express markers that can be related to stem cell behavior.30 In addition, the IBC cell line MDA-IBC-1, described by the team from the University of Texas M. D. Anderson Cancer Center, has stem cell characteristics.31 A small study comparing 25 human IBC samples with 25 non-IBC samples suggests that, in emboli, an enrichment of tumor cells with stem cell characteristics takes place, creating a stem cell niche.30 More IBC samples than non-IBC samples had, both in the parenchymal compartment and in the embolus compartment, expression of CD133, notch 3 receptor intracellular domain, and ALDH1 (70 to 90% vs 0 to 40%, respectively). A study by Van Laere et al discusses evidence for stem cell behavior of IBC tumor cells gathered through genome-wide gene expression profiling.32

Histomorphometric Analysis of Tumor Emboli and Growth Pattern in IBC

Histologic features of the local growth of IBC in the breast gland and of the tumor emboli are described in a study comparing 35 consecutive patients with IBC with 104 patients with T1 to T2 N0 breast carcinomas.18 Necrosis was not a prominent characteristic of the parenchymal component of IBC, with only 7 cases with small necrotic foci. None of the IBC tumors contained a fibrotic focus, a central scar-like area that has replaced or is replacing necrosis, in contrast to 53% of the non-IBC samples. Hypoxia, detected by immunohistochemical demonstration of CA IX expression, was present in a comparable fraction of IBC samples and non-IBC samples (46% and 55%, respectively). This all is accordant with the smaller tumor/stroma ratio noted in IBC versus non-IBC5 within approximately 50% of IBC areas with a tumor/stroma ratio of <10% (<> 16% of non-IBCs; P<.05). Clinical examination reveals diffuse involvement of the breast, often without a palpable nodus (although by radiology a main tumor mass can be demonstrated in nearly all [98%] cases).33 Histologically, IBC has large invasive carcinoma-free areas that alternate with invasive carcinoma growing in a diffuse rather than in a nodular manner.

Nearly all IBC tumors contained lymphovascular emboli (34 of 35 cases), and in 26% of IBC cases with emboli, the tumor cells in the emboli expressed CA IX. In 23% of the IBC cases with emboli, necrosis was present in the emboli. Although not investigated in this study, necrosis in lymphovascular emboli is a very rare observation in non-IBC cases, and most likely relates to the larger size of the emboli in IBC. A peculiar hallmark of the growth pattern of IBC is the colocalization of tumor emboli and small islands of invasive carcinoma at a distance of the main tumor mass. This suggests that 1 way for IBC of fast local growth might be using blood and lymph vessels as a route of less resistance.

Tumor emboli in IBC have been compared with blastocysts because of the high content of cells with stem cell features in both conditions.30 Another condition that resembles the intravascular emboli of IBC is ductal carcinoma in situ (DCIS). Hypoxia and necrosis are indeed present in IBC emboli and centrally in high-grade DCIS (comedo-necrosis). This creates a stem cell-friendly environment: in DCIS, tumor-initiating cells with stem cell features have been shown to be present in large numbers.34 Intravascular growth of emboli, as clearly present in the MARY-X model and most likely also in the human situation, mimics what is called extensive intraductal carcinoma, in which large parts of the breast are occupied by DCIS with only small and scattered invasive areas. Both IBC emboli and DCIS are examples of stroma-independent growth, limited, respectively, by an endothelial cell layer or by a basal membrane with an intact or incomplete myoepithelial cell layer. In both conditions, this might be based on the strong cohesion of the cancer cells by elevated E-cadherin expression and diminished expression of carbohydrates responsible for heterotypic interactions.35, 36


In animal models of IBC (and non-IBC) evidence favors tumor vascularization and embolus formation being the consequence of vasculogenesis with endothelial (-like) cells originating from tumor cells (vasculogenic mimicry) or bone marrow-derived stem cells. The vascular channels formed in this way connect to capillaries originating from peripheral angiogenesis. Histomorphometric and molecular studies of human IBC samples provide evidence of increased angiogenesis, and lymphangiogenesis, in IBC mainly based on endothelial cell proliferation fraction assessment. It is less clear whether vasculogenesis is really involved. Angiogenesis has been shown in 2 major clinical trials to be a valuable target for the therapy for patients with IBC, with a dual effect on both tumor cells and endothelial cells.

The growth pattern of IBC has implications for molecular analyses based on whole-tissue aggregates. The implications of the diffuse, as opposed to a nodular, growth fashion are a low tumor/stroma ratio in IBC. This for example leads to gene expression profiles reflecting the cancer cells, the host-reactive stroma, and the intravascular emboli without the possibility to recover this component-specific information. A recent study taking the distinct components into account appears to confirm this view.37

Tumor emboli are usually large, often contain necrosis, and are present in high numbers in IBC, and might be involved in accelerated local growth, in addition to epithelial-mesenchymal transition and tumor cell motility, comparable to the growth of DCIS.

Relatively small studies are presented throughout this review and this obviously leads to a loss of information that can be solved by more intense international collaboration and the installation of an international IBC register and related tissue repository, as initiated at the First International Inflammatory Breast Cancer Conference in Houston in 2008.


This supplement was sponsored by the Houston Affiliate of the Susan G. Komen for the Cure, the National Cancer Institute, and the State of Texas Rare and Aggressive Breast Cancer Research Program. The First International Inflammatory Breast Cancer Conference was supported in part by GlaxoSmithKline, Pfizer, Eli Lilly and Company, and Cardinal Health.