Tumor cell vasculogenic mimicry: from controversy to therapeutic promise

Abstract

In 1999, The American Journal of Pathology published an article entitled "Vascular channel formation by human melanoma cells in vivo and in vitro: vasculogenic mimicry," by Maniotis and colleagues, which ignited a spirited debate for several years and earned distinction as a citation classic. Tumor cell vasculogenic mimicry (VM) refers to the plasticity of aggressive cancer cells forming de novo vascular networks, which thereby contribute to perfusion of rapidly growing tumors, transporting fluid from leaky vessels, and/or connecting with the constitutional endothelial-lined vasculature. The tumor cells capable of VM share a plastic, transendothelial phenotype, which may be induced by hypoxia. Since VM was introduced as a novel paradigm for melanoma tumor perfusion, many studies have contributed new findings illuminating the underlying molecular pathways supporting VM in a variety of tumors, including carcinomas, sarcomas, glioblastomas, astrocytomas, and melanomas. Facilitating the functional plasticity of tumor cell VM are key proteins associated with vascular, stem cell, and hypoxia-related signaling pathways, each deserving serious consideration as potential therapeutic targets and diagnostic indicators of the aggressive, metastatic phenotype.

Figure 1

Tumor cell vasculogenic mimicry (VM) in vitro and in vivo. A and B: Aggressive human melanoma cells begin to form VM networks (arrows) in vitro in three-dimensional collagen I gels by the end of day 1 (A) and mature into more extensive VM networks (arrows) by day 3 (B). C: By day 14, similar network structures (arrows) can be perfused with a fluorescent dye (the injection site is indicated by an asterisk). D: H&E-stained histological section of high-grade human ovarian cancer shows RBCs within tumor cell-lined channels (arrowhead). E and F: Scanning electron microscopy of ovarian cancer cells cultured on three-dimensional collagen I matrices reveals formation of tubular structures (E) that contain a hollow lumen lined externally by flattened cancer cells when fractured in preparation (F). G and H: Immunostaining for Ln5 γ2 chain protein reveals VM networks in three-dimensional collagen I cultures of aggressive melanoma cells (G), similar to the VM laminin networks seen in a patient's melanoma histological tumor section (H). I: VM networks (arrows) are also seen in a histological section of aggressive melanoma coexpressing both EPHA2 (red) and VE-cadherin (green) under immunofluorescence microscopy (I). Original magnification: ×100 (A, B, and G); ×150 (E); ×200 (C and H); ×630 (D and I); ×2980 (F).

Figure 2

Laser capture microdissection (LCM) of melanoma VM versus endothelial cell angiogenesis. A: Holes left after excision of tumor cells by LCM from non-network regions (nests) in three-dimensional cultures of aggressive human melanoma cells. B: Isolated VM networks of melanoma cells on an LCM collection cap. C: Melanoma nests isolated from the VM culture in A on an LCM collection cap. D and E: Human microvascular endothelial cells (MV) on a Matrigel three-dimensional matrix (D) and isolated vascular networks of MV cells on an LCM collection cap (E). F: Using a gene array, the ratios were determined for relative gene expression in the melanoma cells isolated from the nests versus the networks under the different conditions, and three groups were identified: group 1, angiogenesis-specific; group 2, ECM and cell adhesion-specific expression for aggressive melanoma cells in networks versus nests on collagen I matrices; and group 3, angiogenesis-specific for aggressive melanoma cells versus MV on Matrigel. Scale bar = 400 μm.

Figure 3

Schematic model of signaling pathways implicated in tumor cell VM. Only signaling molecules that have been specifically modulated using antisense oligonucleotides, small inhibitory RNAs, blocking antibodies, small molecule inhibitors, or transient transfections, with demonstrated ability to directly affect VM, are depicted. These molecules are categorized as vascular (red), embryonic/stem cell (green), tumor microenvironment (purple), and hypoxia signaling pathways (blue). Molecules shaded with two different colors demonstrate overlap between major VM signaling pathways. Involvement of Gal-3, IL-8, cAMP, and EPAC 1/Rap1 in VM has been previously reviewed by Seftor et al. Question marks indicate the potential involvement of a protein and/or downstream effector protein or proteins in modulating VM in aggressive cancer cells, for which the underlying signaling pathway or pathways are not yet clearly defined. EP3, prostaglandin E receptor EP3 subtype (encoded by PTGER3).

Figure 4

Endostatin disrupts angiogenesis but not VM. A: Bright-field micrograph of human microvascular endothelial cells-1 (HMEC-1) cultured on a three-dimensional collagen I gel for 3 days in the presence of 10 μg/mL of recombinant human (rh) endostatin. B: Bright-field micrograph of PAS-stained aggressive human melanoma cells cultured on a three-dimensional collagen I gel for 6 days in the presence of 10 μg/mL of recombinant human endostatin. C: Western blot analysis shows the relative differences among endostatin receptor α₅-integrin subunit protein expression levels in whole-cell lysates from aggressive human melanoma cells (C8161), human umbilical vein endothelial cells (HUVEC), and HMEC-1 cells. β-Actin was used as a control for equal loading, and protein amounts were determined relative to β-actin. D: Semiquantitative RT-PCR shows the relative expression levels of integrin α₅ subunit mRNA in C8161, HUVEC, and HMEC-1 cells. GAPDH was used as a control for both loading and relative levels of expression. Original magnification: ×100.