The genesis and unique properties of the lymphovascular tumor embolus are poorly understood largely because of the absence of an experimental model that specifically reflects this important step of tumor progression. The lymphovascular tumor embolus is a blastocyst-like structure resistant to chemotherapy, efficient at metastasis and overexpressing E-cadherin (E-cad). Conventional dogma has regarded E-cad as a metastasis-suppressor gene involved in epithelial-mesenchymal transition. However, within the lymphovascular embolus, E-cad and its proteolytic processing by calpain and other proteases have a dominant oncogenic rather than suppressive role in metastasis formation and tumor cell survival. Studies using a human xenograft model of inflammatory breast cancer, MARY-X, demonstrated the equivalence of xenograft-generated spheroids with lymphovascular emboli in vivo with both structures demonstrating E-cad overexpression and specific proteolytic processing. Western blot revealed full-length (FL) E-cad (120 kDa) and four fragments: E-cad/NTF1 (100 kDa), E-cad/NTF2 (95 kDa), E-cad/NTF3 (85 kDa) and E-cad/NTF4 (80 kDa). Compared with MARY-X, only E-cad/NTF1 was present in the spheroids. E-cad/NTF1 was produced by calpain, E-cad/NTF2 by γ-secretase and E-cad/NTF3 by a matrix metalloproteinase (MMP). Spheroidgenesis and lymphovascular emboli formation are the direct result of calpain-mediated cleavage of E-cad and the generation of E-cad/NTF1 from membrane-associated E-cad rather than the de novo presence of either E-cad/NTF1 or E-cad/CTF1. E-cad/NTF1 retained the p120ctn-binding site but lost both the β-catenin and α-binding sites, facilitating its disassembly from traditional cadherin-based adherens junctions and its 360° distribution around the embolus. This calpain-mediated proteolysis of E-cad generates the formation of the lymphovascular embolus and is responsible for its unique properties of increased homotypic adhesion, apoptosis resistance and budding.

The lymphovascular embolus is an enigmatic entity adept at metastatic dissemination and chemotherapy resistance. Using MARY-X, a human breast cancer xenograft that exhibits florid lymphovascular emboli in mice and spheroids in vitro, we established a model where the in vitro transition stages from minced tumoral aggregates to well-formed spheroids served as a surrogate for in vivo emboli formation. MARY-X well-formed spheroids and emboli exhibited strong similarity of expression. The aggregate-to-spheroid transition stages were characterized by increased ExoC5, decreased Hgs and Rab7, increased calpains, increased full-length E-cadherin (E-cad/FL), and the transient appearance of E-cad/NTF2, a 95 kDa E-cadherin fragment and increased Notch3icd (N3icd), the latter two fragments produced by increased g-secretase. Both transient and permanent knockdowns of Rab7 in MCF-7 cells increased protein but not transcription of E-cad/FL and resulted in the de novo appearance of E-cad/NTF2, the presence of nuclear E-cad/CTF2, and increased Notch1icd (N1icd). Overexpression of Rab7 conversely decreased E-cad/FL, g-secretase (PS1/NTF), and E-cad/NTF2. Overexpression of calpains did not alter PS1/NTF but decreased E-cad/FL and E-cad/NTF2 and increased N1icd. Well-formed spheroids showed increased Rab7, absent E-cad/NTF2, decreased PS1/NTF, increased E-cad/NTF1, and increased N3icd, the latter two fragments being the direct and indirect consequences, respectively, of increased calpains (calpain 1 and calpain 2). Inhibition of calpains decreased E-cad/NTF1 but increased E-cad/NTF2 showing that calpains compete with g-secretase (PS1) for closely located cleavage/binding sites on E-cadherin and that increased calpains can shuttle even decreased