Distinguishing fibrovascular septa from vasculogenic mimicry patterns

Identification of Vasculogenic Mimicry in Histological Samples

Vasculogenic mimicry has been identified in several malignancies and is generally associated with aggressive tumor growth and increased risk of metastasis. Patterned matrix can be identified in light microscopy of tumor sections stained with periodic acid-Schiff (PAS) without hematoxylin counterstain. In this chapter, the process is comprehensively described including tissue sources, formalin fixation and paraffin embedding, staining protocols, and the method for pattern identification in the microscope. Specific pattern types are illustrated in figures, and a number of pitfalls are detailed. The text can be used as a guideline by any researcher or clinician that wishes to evaluate histological samples for the presence of vasculogenic mimicry.

VE-cadherin fusion protein substrate enhanced the vasculogenic mimicry capability of hepatocellular carcinoma cells

To investigate the VE-cadherin-based intercellular crosstalk in tumor cells, a fusion protein consisting of a human VE-cadherin extracellular domain and an immunoglobulin G Fc region (hVE-cad-Fc) was prepared as an artificial extracellular matrix (ECM) for the culture of hepatocellular carcinoma cells (Bel7402 cells). Compared with cells cultured on TCPS and collagen coated plates, the Bel7402 cells cultured on a series concentration of hVE-cad-Fc coated plates showed elongated morphology, inhibited proliferation, and significantly enhanced migration and secretion of ECM compounds and cytokines in a concentration-dependent manner. When the concentration of hVE-cad-Fc reached 15 μg mL-1, the Bel7402 cells formed a patterned network with positive-staining of PAS. The high-density hVE-cad-Fc substrate markedly elevated the expression of endogenous VE-cadherin and EphA2, and subsequently activated the intracellular signal transduction pathways including VE-cadherin/PI3K/MMPs and VE-cadherin/EphA2/FAK/p-VE-cadherin axes as well as epithelial-mesenchymal transition (EMT). Therefore, the high-density hVE-cad-Fc substrate was able to induce the vasculogenic mimicry (VM) formation of Bel7402 cells, and exhibit the potential for the construction of an in vitro tumor model of VM. Moreover, hVE-cad-Fc appeared to be a promising candidate for the investigation of VE-cadherin mediated cell-cell interactions in tumor development.

Tumor angiogenesis: causes, consequences, challenges and opportunities

Tumor vascularization occurs through several distinct biological processes, which not only vary between tumor type and anatomic location, but also occur simultaneously within the same cancer tissue. These processes are orchestrated by a range of secreted factors and signaling pathways and can involve participation of non-endothelial cells, such as progenitors or cancer stem cells. Anti-angiogenic therapies using either antibodies or tyrosine kinase inhibitors have been approved to treat several types of cancer. However, the benefit of treatment has so far been modest, some patients not responding at all and others acquiring resistance. It is becoming increasingly clear that blocking tumors from accessing the circulation is not an easy task to accomplish. Tumor vessel functionality and gene expression often differ vastly when comparing different cancer subtypes, and vessel phenotype can be markedly heterogeneous within a single tumor. Here, we summarize the current understanding of cellular and molecular mechanisms involved in tumor angiogenesis and discuss challenges and opportunities associated with vascular targeting.

Overview of advances in vasculogenic mimicry - a potential target for tumor therapy

Vasculogenic mimicry (VM) describes the process utilized by highly aggressive cancer cells to generate vascular-like structures without the presence of endothelial cells. VM has been vividly described in various tumors and participates in cancer progression dissemination and metastasis. Diverse molecular mechanisms and signaling pathways are involved in VM formation. Furthermore, the patterning characteristics of VM, detected with molecular imaging, are being investigated for use as a tool to aid clinical practice. This review explores the most recent studies investigating the role of VM in tumor induction. Indeed, the recognition of these advances will increasingly affect the development of novel therapeutic target strategies for VM in human cancer.

Collateral Damage Intended-Cancer-Associated Fibroblasts and Vasculature Are Potential Targets in Cancer Therapy

After oncogenic transformation, tumor cells rewire their metabolism to obtain sufficient energy and biochemical building blocks for cell proliferation, even under hypoxic conditions. Glucose and glutamine become their major limiting nutritional demands. Instead of being autonomous, tumor cells change their immediate environment not only by their metabolites but also by mediators, such as juxtacrine cell contacts, chemokines and other cytokines. Thus, the tumor cells shape their microenvironment as well as induce resident cells, such as fibroblasts and endothelial cells (ECs), to support them. Fibroblasts differentiate into cancer-associated fibroblasts (CAFs), which produce a qualitatively and quantitatively different extracellular matrix (ECM). By their contractile power, they exert tensile forces onto this ECM, leading to increased intratumoral pressure. Moreover, along with enhanced cross-linkage of the ECM components, CAFs thus stiffen the ECM. Attracted by tumor cell- and CAF-secreted vascular endothelial growth factor (VEGF), ECs sprout from pre-existing blood vessels during tumor-induced angiogenesis. Tumor vessels are distinct from EC-lined vessels, because tumor cells integrate into the endothelium or even mimic and replace it in vasculogenic mimicry (VM) vessels. Not only the VM vessels but also the characteristically malformed EC-lined tumor vessels are typical for tumor tissue and may represent promising targets in cancer therapy.

Structural and functional identification of vasculogenic mimicry in vitro

Vasculogenic mimicry (VM) describes a process by which cancer cells establish an alternative perfusion pathway in an endothelial cell-free manner. Despite its strong correlation with reduced patient survival, controversy still surrounds the existence of an in vitro model of VM. Furthermore, many studies that claim to demonstrate VM fail to provide solid evidence of true hollow channels, raising concerns as to whether actual VM is actually being examined. Herein, we provide a standardized in vitro assay that recreates the formation of functional hollow channels using ovarian cancer cell lines, cancer spheres and primary cultures derived from ovarian cancer ascites. X-ray microtomography 3D-reconstruction, fluorescence confocal microscopy and dye microinjection conclusively confirm the existence of functional glycoprotein-rich lined tubular structures in vitro and demonstrate that many of structures reported in the literature may not represent VM. This assay may be useful to design and test future VM-blocking anticancer therapies.

Disseminated Tumor Cells Persist in the Bone Marrow of Breast Cancer Patients through Sustained Activation of the Unfolded Protein Response

Disseminated tumor cells (DTC), which share mesenchymal and epithelial properties, are considered to be metastasis-initiating cells in breast cancer. However, the mechanisms supporting DTC survival are poorly understood. DTC extravasation into the bone marrow may be encouraged by low oxygen concentrations that trigger metabolic and molecular alterations contributing to DTC survival. Here, we investigated how the unfolded protein response (UPR), an important cytoprotective program induced by hypoxia, affects the behavior of stressed cancer cells. DTC cell lines established from the bone marrow of patients with breast cancer (BC-M1), lung cancer, (LC-M1), and prostate cancer (PC-E1) were subjected to hypoxic and hypoglycemic conditions. BC-M1 and LC-M1 exhibiting mesenchymal and epithelial properties adapted readily to hypoxia and glucose starvation. Upregulation of UPR proteins, such as the glucose-regulated protein Grp78, induced the formation of filamentous networks, resulting in proliferative advantages and sustained survival under total glucose deprivation. High Grp78 expression correlated with mesenchymal attributes of breast and lung cancer cells and with poor differentiation in clinical samples of primary breast and lung carcinomas. In DTCs isolated from bone marrow specimens from breast cancer patients, Grp78-positive stress granules were observed, consistent with the likelihood these cells were exposed to acute cell stress. Overall, our findings provide the first evidence that the UPR is activated in DTC in the bone marrow from cancer patients, warranting further study of this cell stress pathway as a predictive biomarker for recurrent metastatic disease.

CD133-targeted niche-dependent therapy in cancer: a multipronged approach

Cancer treatment continues to be challenged by the development of therapeutic resistances and relapses in the clinical setting, which are largely attributed to tumor heterogeneity, particularly the existence of cancer stem cells (CSCs). Thus, targeting the CSC subpopulation may represent an effective therapeutic strategy. However, despite advances in identifying and characterizing CD133(+) CSCs in various human cancers, efforts to translate these experimental findings to clinical modalities have been slow in the making, especially in light of the growing awareness of CSC plasticity and the foreseeable pitfall of therapeutically targeting CSC base sorely on a surface marker. We, and others, have demonstrated that the CD133(+) CSCs reside in complex vascular niches, where reciprocal signaling between the CD133(+) CSCs and their microenvironment may govern niche morphogenesis and homeostasis. Herein, we discuss the multifaceted functional role of the CD133(+) cells in the context of their niche, and the potential of targeting CD133 as a niche-dependent approach in effective therapy.

Opportunities and challenges in tumor angiogenesis research: back and forth between bench and bed

Angiogenesis is essential for tumor growth and metastasis. Many signaling pathways are involved in regulating tumor angiogenesis, with the vascular endothelial growth factor pathway being of particular interest. The recognition of the heterogeneity in tumor vasculature has led to better predictions of prognosis through differential analyses of the vasculature. However, the clinical benefits from antiangiogenic therapy are limited, because many antiangiogenic agents cannot provide long-term survival benefits, suggesting the development of drug resistance. Activation of the hypoxia and c-Met pathways, as well as other proangiogenic factors, has been shown to be responsible for such resistance. Vessel co-option could be another important mechanism. For future development, research to improve the efficacy of antiangiogenic therapy includes (a) using tumor-derived endothelial cells for drug screening; (b) developing the drugs focusing on specific tumor types; (c) developing a better preclinical model for drug study; (d) developing more accurate biomarkers for patient selection; (e) targeting the c-Met pathway or other pathways; and (f) optimizing the dose and schedule of antiangiogenic therapy. In summary, the future of antiangiogenic therapy for cancer patients depends on our efforts to develop the right drugs, select the right patients, and optimize the treatment conditions.

Suppression of α5 gene expression is closely related to the tumorigenic properties of uveal melanoma cell lines

Cancer aggressiveness is related to the ability of cancer cells to escape the anchorage dependency toward the extracellular matrix, a process regulated by the integrin α5β1 and its ligand fibronectin. Here, we characterized the expression of the α5 gene in human uveal melanoma cell lines with distinct tumorigenic properties and investigated some of the mechanisms underlying the variations of their malignancy. Strong and weak expression of α5 was observed in cells with no (T108/T115) and high (T97/T98) tumorigenic properties, respectively. Expression and DNA binding of the transcription factors Sp1, activator protein 1 (AP-1) (both acting as activators), and nuclear factor I (NFI) (a strong repressor) to the α5 promoter were demonstrated in all cell lines. A reduced expression of AP-1 combined with a dramatic increase in NFI correlated with the suppression of α5 expression in T97 and T98 cells. Restoring α5 expression in T97 cells entirely abolished their tumorigenicity in immunodeficient mice. These uveal melanoma cell lines might therefore prove particularly useful as cellular models to investigate α5β1 function in the pathogenesis of invasive uveal melanoma.

VEGF as a biomarker for metastatic uveal melanoma in humans

PURPOSE

High levels of serum VEGF have been reported in many types of cancers, especially in the metastatic stage. The aim of this study was to examine the potential of VEGF serum level as a tumor marker for metastases in uveal melanoma (UM) patients.

MATERIALS AND METHODS

Levels of serum VEGF were analyzed by ELISA for 23 UM patients (none of whom developed metastases within 5 years from diagnosis) at the time of diagnosis, soon after treatment, and 3 years later, and compared with serum VEGF levels of 39 metastatic patients, 58 10-year disease-free (10yDF) patients, and 23 healthy subjects. VEGF ratios were calculated per patient between diagnosis and after treatment, and between diagnosis and 3 years later. Matched pairs univariate analysis was performed for 17 metastatic patients for whom sera were available from before and after the diagnosis of metastases. Patients were followed biannually with liver ultrasonography and liver function tests for the presence of metastases.

RESULTS

The inter-patient VEGF level range was large (e.g., the range for the metastatic patients was 46-1892 pg/ml). The mean ± SD levels for the control, 10yDF, and metastatic groups were: 329.65 ± 190.0, 407.66 ± 261.9 and 453.52 ± 270.2, respectively (p = 0.2456). The mean VEGF level ratio from after treatment to diagnosis was 1.08 (p = 0. 0024), and the ratio from 3 years after diagnosis to diagnosis was 1.53 (p = 0.0009). The mean ± SE post/premetastatic levels ratio was 1.35 ± 0.21 (p = 0.0595).

CONCLUSIONS

Serum VEGF increased significantly after metastases developed. However, the wide inter-patient variance precludes the use of any cut-off level to determine the metastatic status of an individual patient based on a single VEGF serum level. An increase in VEGF on serial measurements may indicate the development of metastases. Further investigation is warranted to assess VEGF's value as a predictive marker for metastatic disease.

VEGFR-1 expressed by malignant melanoma-initiating cells is required for tumor growth

Melanoma growth is driven by malignant melanoma-initiating cells (MMIC) identified by expression of the ATP-binding cassette (ABC) member ABCB5. ABCB5(+) melanoma subpopulations have been shown to overexpress the vasculogenic differentiation markers CD144 (VE-cadherin) and TIE1 and are associated with CD31(-) vasculogenic mimicry (VM), an established biomarker associated with increased patient mortality. Here we identify a critical role for VEGFR-1 signaling in ABCB5(+) MMIC-dependent VM and tumor growth. Global gene expression analyses, validated by mRNA and protein determinations, revealed preferential expression of VEGFR-1 on ABCB5(+) tumor cells purified from clinical melanomas and established melanoma lines. In vitro, VEGF induced the expression of CD144 in ABCB5(+) subpopulations that constitutively expressed VEGFR-1 but not in ABCB5(-) bulk populations that were predominantly VEGFR-1(-). In vivo, melanoma-specific shRNA-mediated knockdown of VEGFR-1 blocked the development of ABCB5(+) VM morphology and inhibited ABCB5(+) VM-associated production of the secreted melanoma mitogen laminin. Moreover, melanoma-specific VEGFR-1 knockdown markedly inhibited tumor growth (by > 90%). Our results show that VEGFR-1 function in MMIC regulates VM and associated laminin production and show that this function represents one mechanism through which MMICs promote tumor growth.

The therapeutic promise of the cancer stem cell concept

Cancer stem cells (CSCs) are a subpopulation of tumor cells that selectively possess tumor initiation and self-renewal capacity and the ability to give rise to bulk populations of nontumorigenic cancer cell progeny through differentiation. As we discuss here, they have been prospectively identified in several human malignancies, and their relative abundance in clinical cancer specimens has been correlated with malignant disease progression in human patients. Furthermore, recent findings suggest that clinical cancer progression driven by CSCs may contribute to the failure of existing therapies to consistently eradicate malignant tumors. Therefore, CSC-directed therapeutic approaches might represent translationally relevant strategies to improve clinical cancer therapy, in particular for those malignancies that are currently refractory to conventional anticancer agents directed predominantly at tumor bulk populations.

Modeling the behavior of uveal melanoma in the liver

PURPOSE

To model the behavior of uveal melanoma in the liver.

METHODS

A 15-muL suspension of metastatic MUM2B or either primary OCM1 or M619 uveal melanoma cells was injected into the liver parenchyma of 105 CB17 SCID mice through a 1-cm abdominal incision. Animals were killed at 2, 4, 6, or 8 weeks after injection. Before euthanatization, 3% FITC-BSA buffer was injected into the retro-orbital plexus of one eye of three mice. Liver tissues were examined by light and fluorescence microscopy, and were stained with human anti-laminin. Vasculogenic mimicry patterns were reconstructed from serial laser scanning confocal microscopic stacks.

RESULTS

OCM1a cells formed microscopic nodules in the mouse liver within 2 weeks after injection and metastasized to the lung 6 weeks later. By contrast, M619 and MUM2B cells formed expansile nodules in the liver within 2 weeks and gave rise to pulmonary metastases within 4 weeks after injection. Vasculogenic mimicry patterns, composed of human laminin and identical with those in human primary and metastatic uveal melanomas, were detected in the animal model. The detection of human rather than mouse laminin in the vasculogenic mimicry patterns in this model demonstrates that these patterns were of tumor cell origin and were not co-opted from the mouse liver microenvironment.

CONCLUSIONS

There are currently no effective treatments for metastatic uveal melanoma. This direct-injection model focuses on critical interactions between the tumor cell and the liver. It provides for translationally relevant approaches to the development of new modalities to detect small tumor burdens in patients, to study the biology of clinical dormancy of metastatic disease in uveal melanoma, to design and test novel treatments to prevent the emergence of clinically manifest liver metastases after dormancy, and to treat established uveal melanoma metastases.

Comparing vasculogenic mimicry with endothelial cell-lined vessels: techniques for 3D reconstruction and quantitative analysis of tissue components from archival paraffin blocks

We previously described techniques to generate 3-dimensional reconstructions of the tumor microcirculation using immunofluorescence histochemistry and laser scanning confocal microscopy on serial sections from archival formalin-fixed, paraffin-embedded tissues. By aligning sequential z-stacks in an immersive visualization environment (ImmersaDesk), the need to insert fiduciary markers into tissue was eliminated. In this study, we developed methods to stitch overlapping confocal z-series together to extend the surface area of interest well beyond that captured by the confocal microscope objective and developed methods to quantify the distribution of markers of interest in 3 dimensions. These techniques were applied to the problem of comparing the surface area of nonendothelial cell-lined, laminin-rich looping vasculogenic mimicry (VM) patterns that are known to transmit fluid, with the surface area of endothelial cell-lined vessels in metastatic uveal melanoma to the liver in 3 dimensions. After labeling sections with antibodies to CD34 and laminin, the surface area of VM patterns to vessels was calculated by segmenting out structures that labeled with laminin but not with CD34 from those structures labeling with CD34, or CD34 and laminin. In metastatic uveal melanoma tissues featuring colocalization of high microvascular density [66.4 microvessels adjusted for 0.313 mm2 area (range 56.7 to 72.7)] and VM patterning, the surface area of VM patterns was 11.6-fold greater (range 10.8 to 14.1) than the surface provided by CD34-positive vessels. These methods may be extended to visualize and quantify molecular markers in 3 dimensions in a variety of pathologic entities from archival paraffin-embedded tissues.

Epigenetic reversion of breast carcinoma phenotype is accompanied by changes in DNA sequestration as measured by AluI restriction enzyme

The importance of microenvironment and context in regulation of tissue-specific genes is well established. DNA exposure to or the sequestration from nucleases detects differences in higher order chromatin structure in intact cells without disturbing cellular or tissue architecture. To investigate the relationship between chromatin organization and tumor phenotype, we used an established three-dimensional assay in which normal and malignant human breast cells can be easily distinguished by the morphology of the structures they make (acinus-like versus tumor-like, respectively). We show that these phenotypes can be distinguished also by sensitivity to AluI digestion in which the malignant cells resist digestion relative to nonmalignant cells. Treatment of T4-2 breast cancer cells in three-dimensional culture with cAMP analogs or a phosphatidylinositol 3-kinase inhibitor not only reverted their phenotype from nonpolar to polar acinar-like structures but also enhanced chromatin sensitivity to AluI. By using different cAMP analogs, we show that cAMP-induced phenotypic reversion, polarization, and shift in DNA organization act through a cAMP-dependent protein-kinase A-coupled signaling pathway. Importantly, inhibitory antibody to fibronectin produced the same effect. These experiments underscore the concept that modifying the tumor microenvironment can alter the organization of tumor cells and demonstrate that architecture and global chromatin organization are coupled and highly plastic.

Demonstrating circulation in vasculogenic mimicry patterns of uveal melanoma by confocal indocyanine green angiography

PURPOSE

Vasculogenic mimicry patterns, formed by highly invasive melanoma cells, connect to endothelial cell-lined blood vessels and contain fluid in vitroand in vivo. This study was designed to determine if fluid leaks into vasculogenic mimicry patterns without circulation, or if fluid circulates in and clears from these patterns.

METHODS

Indocyanine green (ICG) laser scanning confocal angiography (Heidelberg Retinal Angiograph (HRA); Heidelberg Engineering, Heidelberg, Germany) was performed on nine patients with posterior choroidal melanoma in an institutional setting. Blood was drawn before the ICG injection and from the contralateral arm of the ICG injection site and 1 min after the injection. Outcome measures include time to first filling of retinal vessels and vasculogenic mimicry patterns and the time at which no fluorescence could be detected by the HRA instrument. After fluorescence was no longer detected in vessels or patterns, the tubes containing the patient's blood was imaged by the Heidelberg HRA.

RESULTS

Looping vasculogenic mimicry patterns were detected focally in five patients within 30 s after injection and were detectable up to 12 min post-injection. Blood drawn before ICG injection did not autofluoresce but ICG-containing blood pooled in the tube continued to fluoresce at 1-month post-injection.

CONCLUSIONS

Vasculogenic mimicry patterns are not part of the endothelial cell-lined vascular system and fluid enters these patterns through leakage. The rapid infusion of ICG into these patterns after injection and the disappearance of fluorescence detectable by the Heidelberg HRA suggest that fluid circulates in these patterns and does not accumulate as a stagnant pool.

The heterogeneous distribution of monosomy 3 in uveal melanomas: implications for prognostication based on fine-needle aspiration biopsies

CONTEXT

The detection of monosomy 3 in uveal melanomas has repeatedly been associated with adverse outcome. Fine-needle aspiration biopsy is being used to detect monosomy 3 in these tumors, based on the assumption that this chromosomal abnormality is distributed homogeneously throughout the tumor.

OBJECTIVE

To study the distribution of monosomy 3 in primary uveal melanoma by fluorescence in situ hybridization (FISH).

DESIGN

We studied 50 enucleated eyes with uveal melanoma. In all 50 tumors we performed cytogenetic analysis and FISH using a DNA-specific probe for the centromere region of chromosome 3 on cultured tumor cells. In addition, the percentage of tumor cells with monosomy 3 was assessed by FISH on nuclei, isolated from paraffin-embedded tissue and compared to results of FISH on regular histology sections of the paraffin-embedded tissue.

RESULTS

Combining karyotyping and FISH on cultured cells identified monosomy 3 in 19 (38%) of 50 tumors, whereas FISH on nuclei isolated from paraffin-embedded tissue showed 31 (62%) of 50 as having monosomy for chromosome 3. FISH analysis on paraffin sections showed tumor heterogeneity for copy number of chromosome 3 in at least 7 cases.

CONCLUSIONS

FISH analysis on paraffin sections shows that heterogeneity of monosomy of chromosome 3 is a frequent phenomenon in uveal melanoma. FISH on nuclei isolated from paraffin-embedded tissue identifies a higher frequency of monosomy 3 than the traditional combination of karyotyping and FISH on cultured uveal melanoma cells. The practice of assigning patients to risk categories based on fine-needle aspiration biopsy samples from primary uveal melanoma may be subject to error based on the heterogeneous distribution of monosomy 3 in these tumors.

Tumor cell plasticity in uveal melanoma: microenvironment directed dampening of the invasive and metastatic genotype and phenotype accompanies the generation of vasculogenic mimicry patterns

The histological detection of laminin-rich vasculogenic mimicry patterns in human primary uveal melanomas is associated with death from metastases. We therefore hypothesized that highly invasive uveal melanoma cells forming vasculogenic mimicry patterns after exposure to a laminin-rich three-dimensional microenvironment would differentially express genes associated with invasive and metastatic behavior. However, we discovered that genes associated with differentiation (GDF15 and ATF3) and suppression of proliferation (CDKNa1/p21) were up-regulated in highly invasive uveal melanoma cells forming vasculogenic mimicry patterns, and genes associated with promotion of invasive and metastatic behavior such as CD44, CCNE2 (cyclin E2), THBS1 (thrombospondin 1), and CSPG2 (chondroitin sulfate proteoglycan; versican) were down-regulated. After forming vasculogenic mimicry patterns, uveal melanoma cells invaded only short distances, failed to replicate, and changed morphologically from the invasive epithelioid to the indolent spindle A phenotype. In human tissue samples, uveal melanoma cells within vasculogenic mimicry patterns assumed the spindle A morphology, and the expression of Ki67 was significantly reduced in adjacent melanoma cells. Thus, the generation of vasculogenic mimicry patterns is accompanied by dampening of the invasive and metastatic uveal melanoma genotype and phenotype and underscores the plasticity of these cells in response to cues from the microenvironment.

Functional gene expression analysis uncovers phenotypic switch in aggressive uveal melanomas

Microarray gene expression profiling is a powerful tool for generating molecular cancer classifications. However, elucidating biological insights from these large data sets has been challenging. Previously, we identified a gene expression-based classification of primary uveal melanomas that accurately predicts metastatic death. Class 1 tumors have a low risk and class 2 tumors a high risk for metastatic death. Here, we used genes that discriminate these tumor classes to identify biological correlates of the aggressive class 2 signature. A search for Gene Ontology categories enriched in our class-discriminating gene list revealed a global down-regulation of neural crest and melanocyte-specific genes and an up-regulation of epithelial genes in class 2 tumors. Correspondingly, class 2 tumors exhibited epithelial features, such as polygonal cell morphology, up-regulation of the epithelial adhesion molecule E-cadherin, colocalization of E-cadherin and beta-catenin to the plasma membrane, and formation of cell-cell adhesions and acinar structures. One of our top class-discriminating genes was the helix-loop-helix inhibitor ID2, which was strongly down-regulated in class 2 tumors. The class 2 phenotype could be recapitulated by eliminating Id2 in cultured class 1 human uveal melanoma cells and in a mouse ocular melanoma model. Id2 seemed to suppress the epithelial-like class 2 phenotype by inhibiting an activator of the E-cadherin promoter. Consequently, Id2 loss triggered up-regulation of E-cadherin, which in turn promoted anchorage-independent cell growth, a likely antecedent to metastasis. These findings reveal new roles for Id2 and E-cadherin in uveal melanoma progression, and they identify potential targets for therapeutic intervention.