The step of intravasation (lymphovascular invasion), a rate-limiting step in metastasis, is greatly exaggerated in inflammatory breast carcinoma (IBC). Because nearly all human breast carcinoma cell lines grow as solitary nodules in nude/severe combined immunodeficient mice without manifesting lymphovascular invasion, this step has been difficult to study. We captured the essence of the IBC phenotype by establishing a unique human transplantable IBC xenograft, MARY-X, which manifests florid lymphovascular emboli in severe combined immunodeficient/nude mice. Comparing MARY-X with common non-IBC cell lines/xenografts, we discovered an overexpressed and overfunctioning E-cadherin/α,β-catenin axis. In MARY-X, the E-cadherin and catenins were part of a structurally and functionally intact adhesion axis involving the actin cytoskeleton. In vitro, MARY-X grew as round compact spheroids with a cell density 5–10-fold higher than that of other lines. The spheroids of MARY-X completely disadhered when placed in media containing absent Ca²⁺ or anti-E-cadherin antibodies or when retrovirally transfected with a dominant-negative E-cadherin mutant (H-2K^d-E-cad). Anti-E-cadherin antibodies injected i.v. immunolocalized to the pulmonary lymphovascular emboli of MARY-X and caused their dissolution. H-2K^d-E-cad-transfected MARY-X spheroids were only weakly tumorigenic and did not form lymphovascular emboli. A total of 90% of human IBCs showed increased membrane E-cadherin/α,β-catenin immunoreactivity. These findings indicate that it is the gain and not the loss of the E-cadherin axis that contributes to the IBC phenotype.