# NEOVASCULARIZATION, ANGIOGENESIS AND VASCULOGENIC MIMICRY IN CANCER

EDITED BY : César López-Camarillo, Laurence A. Marchat, Naureen Starling and Erika Ruiz-Garcia PUBLISHED IN : Frontiers in Oncology

#### Frontiers eBook Copyright Statement

The copyright in the text of individual articles in this eBook is the property of their respective authors or their respective institutions or funders. The copyright in graphics and images within each article may be subject to copyright of other parties. In both cases this is subject to a license granted to Frontiers. The compilation of articles constituting this eBook is the property of Frontiers.

Each article within this eBook, and the eBook itself, are published under the most recent version of the Creative Commons CC-BY licence. The version current at the date of publication of this eBook is CC-BY 4.0. If the CC-BY licence is updated, the licence granted by Frontiers is automatically updated to the new version.

When exercising any right under the CC-BY licence, Frontiers must be attributed as the original publisher of the article or eBook, as applicable.

Authors have the responsibility of ensuring that any graphics or other materials which are the property of others may be included in the CC-BY licence, but this should be checked before relying on the CC-BY licence to reproduce those materials. Any copyright notices relating to those materials must be complied with.

Copyright and source acknowledgement notices may not be removed and must be displayed in any copy, derivative work or partial copy which includes the elements in question.

All copyright, and all rights therein, are protected by national and international copyright laws. The above represents a summary only. For further information please read Frontiers' Conditions for Website Use and Copyright Statement, and the applicable CC-BY licence.

ISSN 1664-8714 ISBN 978-2-88963-985-4 DOI 10.3389/978-2-88963-985-4

#### About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

#### Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

#### Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

#### What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

## NEOVASCULARIZATION, ANGIOGENESIS AND VASCULOGENIC MIMICRY IN CANCER

Topic Editors:

César López-Camarillo, Universidad Autónoma de la Ciudad de México, Mexico Laurence A. Marchat, Instituto Politécnico Nacional, Mexico Naureen Starling, Royal Marsden NHS Foundation Trust, United Kingdom Erika Ruiz-Garcia, National Institute of Cancerology (INCAN), Mexico

Citation: López-Camarillo, C., Marchat, L. A., Starling, N., Ruiz-Garcia, E., eds. (2020). Neovascularization, Angiogenesis and Vasculogenic Mimicry in Cancer. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-985-4

# Table of Contents

*06 Editorial: Neovascularization, Angiogenesis and Vasculogenic Mimicry in Cancer*

César López-Camarillo, Erika Ruiz-García, Naureen Starling and Laurence A. Marchat

*09 HypoxamiRs Profiling Identify miR-765 as a Regulator of the Early Stages of Vasculogenic Mimicry in SKOV3 Ovarian Cancer Cells*

Yarely M. Salinas-Vera, Dolores Gallardo-Rincón, Raúl García-Vázquez, Olga N. Hernández-de la Cruz, Laurence A. Marchat, Juan Antonio González-Barrios, Erika Ruíz-García, Carlos Vázquez-Calzada, Estefanía Contreras-Sanzón, Martha Resendiz-Hernández, Horacio Astudillo-de la Vega, José L. Cruz-Colin, Alma D. Campos-Parra

and César López-Camarillo *23 Targeting Vascular Endothelial Growth Factor in Oesophagogastric Cancer: A Review of Progress to Date and Immunotherapy Combination Strategies*

Oliver Butters, Kate Young, David Cunningham, Ian Chau and Naureen Starling

*36 Crosstalk Between Long Non-coding RNAs, Micro-RNAs and mRNAs: Deciphering Molecular Mechanisms of Master Regulators in Cancer*

Eduardo López-Urrutia, Lilia P. Bustamante Montes, Diego Ladrón de Guevara Cervantes, Carlos Pérez-Plasencia and Alma D. Campos-Parra


Jun Lu, Qin Shi, Lele Zhang, Jun Wu, Yuqing Lou, Jie Qian, Bo Zhang, Shuyuan Wang, Huimin Wang, Xiaodong Zhao and Baohui Han


*110 IL27R*a *Deficiency Alters Endothelial Cell Function and Subverts Tumor Angiogenesis in Mammary Carcinoma*

Annika F. Fink, Giorgia Ciliberti, Rüdiger Popp, Evelyn Sirait-Fischer, Ann-Christin Frank, Ingrid Fleming, Divya Sekar, Andreas Weigert and Bernhard Brüne


Shiu-Wen Huang, Hung-Yu Yang, Wei-Jan Huang, Wei-Chuan Chen, Meng-Chieh Yu, Shih-Wei Wang, Ya-Fen Hsu and Ming-Jen Hsu

*154 Nanoparticle Delivery and Tumor Vascular Normalization: The Chicken or The Egg?*

George Mattheolabakis and Constantinos M. Mikelis


Julie-Ann Hulin, Ekaterina A. Gubareva, Natalia Jarzebska, Roman N. Rodionov, Arduino A. Mangoni and Sara Tommasi


Daniel Delgado-Bellido, Concepción Bueno-Galera, Laura López-Jiménez, Angel Garcia-Diaz and F. Javier Oliver


Erik Lizárraga-Verdugo, Melisa Avendaño-Félix, Mercedes Bermúdez, Rosalio Ramos-Payán, Carlos Pérez-Plasencia and Maribel Aguilar-Medina

# Editorial: Neovascularization, Angiogenesis and Vasculogenic Mimicry in Cancer

#### César López-Camarillo<sup>1</sup> \*, Erika Ruiz-García<sup>2</sup> , Naureen Starling<sup>3</sup> and Laurence A. Marchat <sup>4</sup>

<sup>1</sup> Posgrado en Ciencias Genómicas, Universidad Autónoma de la Ciudad de México, Mexico City, México, <sup>2</sup> Laboratorio de Medicina Translacional y Departamento de Tumores Gastro-Intestinales, Instituto Nacional de Cancerología, Mexico City, México, <sup>3</sup> Royal Marsden NHS Foundation Trust London, London, United Kingdom, <sup>4</sup> Programa en Biomedicina Molecular y Red de Biotecnología, ENMyH-Instituto Politécnico Nacional, Mexico City, México

Keywords: vasculogenic mimicry, angiogenesis, lncRNAs, microRNAs, therapy

#### **Editorial on the Research Topic**

#### **Neovascularization, Angiogenesis and Vasculogenic Mimicry in Cancer**

#### Edited by:

Emilio Hirsch, University of Turin, Italy

#### Reviewed by:

Akiko Mammoto, Medical College of Wisconsin, United States Daniel Delgado-Bellido, Consejo Superior de Investigaciones Científicas (CSIC), Spain Louise Reynolds, Queen Mary University of London, United Kingdom

\*Correspondence:

César López-Camarillo genomicas@yahoo.com.mx

#### Specialty section:

This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology

> Received: 01 April 2020 Accepted: 05 June 2020 Published: 16 July 2020

#### Citation:

López-Camarillo C, Ruiz-García E, Starling N and Marchat LA (2020) Editorial: Neovascularization, Angiogenesis and Vasculogenic Mimicry in Cancer. Front. Oncol. 10:1140. doi: 10.3389/fonc.2020.01140 Vasculogenesis refers to the development of new vessels from primordial endothelial stem cells, whereas angiogenesis denotes the formation of vessels from pre-existing capillary structures. In particular, angiogenesis is a complex cellular mechanism required for the formation of new blood vessels from the existing vasculature or from bone marrow-derived endothelial progenitors, allowing homeostasis. In contrast, pathological angiogenesis found in tumors leads to accelerated tumor growth and development at early stages of carcinogenesis (1). During angiogenesis a plethora of pro-angiogenic factors are secreted by cells present in tumor microenvironment including cancer cells, cancer associated-fibroblasts, pericytes, and immune cells to induce vascularization via activation of pre-existing host endothelium. On the other hand, vasculogenic mimicry (VM) describe the ability of tumors cells to organize themselves in patterned three-dimensional (3D) channel-like structures resembling vascular networks in order to acquire nutrients and supports the requirements for tumor growth. VM has been reported in diverse types of tumors such as breast, melanoma, lung, glioblastoma, ovarian, and prostate cancers, among others, indicating that it may be designated as a novel cancer hallmark (2). Importantly, the existence of tumoral VM in oncologic patients has been associated with low free-disease survival and worst prognosis. VM operates in an independent via, or simultaneously, with classical endothelial vessels and angiogenesis. These cellular processes are regulated by diverse cellular lineages including pericytes and multiple pro-angiogenic and anti-angiogenic factors activated by hypoxic microenvironment, which become deregulated in tumors leading to pathological angiogenesis. Pericytes secrete growth factors that stimulate endothelial cells proliferation, and migration; as well as proteases secretion that contribute to modulate the surrounding extracellular matrix (3). In addition, pericytes are recruited by VM-positive cells in order to stimulate sprouting and to provide structural support of the growing vascular-like networks (4). Also, multiple proteins, signaling networks and regulatory non-coding RNAs (ncRNAs) activate cell proliferation, extracellular matrix remodeling, invasion and metastasis mechanisms to promote neovascularization, angiogenesis and VM. However, the balance and interplay among signaling transductors, and ncRNAs is poorly understood. Therefore, there is a great interest in the development of anti-angiogenesis and anti-vasculogenic mimicry strategies that could inhibit tumor vascularization.

In this Research Topic we have organized a collection of original research and review articles that examine the more recent progress in neovascularization, angiogenesis, and VM. To incite readers to check the complete collection, here we introduce several representative contributions

**6**

made for authors. After the first study published by Maniotis et al. (5), the presence of VM and the experimental approaches used to confirm its presence in tumors and cell cultures have remained controversial. In an opinion paper on VM, Valdivia et al., describe the state of art, and the contentious topics around VM. Authors debate about the utility of the current tools used to demonstrate the presence of VM, specifically the Periodic Acid Schiff (PAS) positive stain of tumor tissues and the commonly used in vitro models. They conclude that intercellular connections occurring in cell monolayers when cultured in a Matrigel matrix at early times could not represent VM. Moreover, authors raise doubts on the validity of PAS+ staining to detect the presence of VM in patient tissues, and finally outline the requirement for new biomarkers of VM for clinical use. Remarkably, authors make an urgent call for reliable in vitro and in vivo VM models which must be approved by the scientific community in order to better explain the mechanisms governing this phenomenon.

Two reviews summarize the actual state of the art in the molecular mechanisms of VM in breast and ovarian cancers. Andonegui-Elguera et al. summarize the mechanisms of VM development in breast cancer, including the participation of signaling proteins and the functional relationships between cancer stem cells, the epithelial-mesenchymal transition and VM. Also, they discuss the clinical significance of VM in prognosis with special emphasis in the opportunities of targeting VM for triple negative breast cancer therapy. Ayala-Domínguez et al. summarize the mechanisms of VM in gynecological ovarian cancer. They reviewed the actual knowledge of angiogenesis, vasculogenesis, and vessel co-option mechanisms with a special focus in the signaling pathways and microRNAs involved in VM regulation. In addition, authors discuss the clinical implications of the potential targeting of molecules involved in VM for ovarian cancer therapy.

Tumor hypoxia is one of the most important mechanisms to activate angiogenesis and VM (6). Salinas-Vera et al. examine the role of hypoxia-regulated microRNAs (dubbed as hypoxamiRs) during the onset of VM in ovarian cancer. They identify the modulation of 11 hypoxamiRs which are predicted to participate in VM and angiogenesis with potential clinical implications. Also, authors demonstrate that miR-765 modulates the initiation of 3D capillary-like arrangements via activation of the VEGFA/AKT1/SRC axis in SKOV3 ovarian cancer cells.

The limited clinical outcome of anti-angiogenic therapy depends, at least in part, on the inefficient tumor perfusion that limits both the diffusion of chemotherapeutic agents and the antitumoral functions of immune cells. Pathological tumor angiogenesis is characterized by an immature and disorganized vasculature architecture leading to enhanced permeability and retention effect which may results in cancer cells intravasation and increased metastasis (7). Therefore, vascular normalization has emerged as a new concept and a complementary therapeutic approach aiming to normalize the tumor vasculature. Mattheolabakis and Mikelis describe an overview of the nanoparticles used for simultaneous delivery of anti-angiogenic and chemotherapeutic drugs, which may take advantage of the leaky and tortuous tumor vasculature to diffuse out of the tumor vessels, aiming to achieve vascular normalization and higher efficacy for anticancer therapies.

In a mini review paper, Fernández-Cortés et al. discuss the role of tumor microenvironment constituted by tumor associated macrophages, cancer-associated fibroblasts, cancer stem cells, stromal cells and pericytes in VM acquisition. They emphasize on the phosphorylation of the VE-cadherin frequently expressed in endothelial cells and diverse types of aggressive tumors, and its role in VM formation. Also, authors describe the current therapeutic agents targeting FAK/Y658 VE-cadherin and VE-PTP/TIE-2 which have been proposed to impair VM. Another study by Delgado-Bellido et al., showed that melanoma cells undergoing VM express the VE-cadherin phosphatase VE-PTP which complexed with VE-Cadherin and p120. VE-PTP knockdown results in degradation of complex and enhanced autophagy suggesting a pivotal role for VE-PTP in VM formation.

In the last decade, the altered expression of ncRNAs such as microRNAs and long non-coding RNAs (lncRNAs) have been reported in diverse types of tumors where they regulate the expression of oncogenes and tumor suppressor genes driving tumorigenesis. Four reviews focus on the pivotal roles of ncRNAs in cancer are presented in this Research Topic. López-Urrutia et al., review the crosstalk between lncRNAs, microRNAs and mRNAs with a special emphasis in neovascularization, VM, and angiogenesis. Authors consider that the current knowledge on the lncRNA/microRNAs/mRNA axis in these related cellular processes is still limited and deserves further scrutiny. Likewise, Hernández de la Cruz et al., discuss the role of tumor microenvironment, the epithelial-mesenchymal transition and signaling pathways in VM formation in solid tumors. Also, they describe the regulation networks of lncRNAs-microRNAs and their potential impact in personalized cancer treatments. Authors remark that microRNAs and lncRNAs can be potential biomarkers for prognosis, and predictors of therapy response. An epigenetic perspective on the regulatory roles of ncRNAs in neovascularization and angiogenesis in normal and cancer cells is provided by Hernández-Romero et al.. Authors describe several microRNAs and their epigenetic targets involved in angiogenesis and vascular diseases. Also the role of lncRNAs as scaffolds for epigenetic players in neovascularization and angiogenesis is discussed. They conclude that microRNAs and lncRNAs could influence the epigenetic mechanisms in endothelial cells and tumors, thus making ncRNAs as promising epigenetic targets for therapy. Finally, Cao et al. describe the role of ncRNAs regulating the mechanisms of lymphangiogenesis in lymphatic development and discuss their potential as therapeutic targets.

Gastric and esophageal cancers are the third and sixth leading causes of cancer related death worldwide, respectively. Butters et al. present a review about the progress in targeting VGFA and the immunotherapy combination strategies in oesophagogastric cancer. Authors summarize the phase III studies targeting VEGF and the clinical trials focused in the study of immune-checkpoint inhibitors and anti-angiogenic compounds in OG cancer therapy. On the other hand, Lizárraga-Verdugo et al., review the actual knowledge of CSCs research in gastrointestinal cancers (GIC). Cancer stem cells (CSCs) are a small subpopulation of cells present in discrete tumor niches that exhibit a stem-cell phenotype similar to progenitors such as self-renewal, differentiation and maintenance of tumor growth and heterogeneity. Such cells have been found and isolated from diverse types of tumors, and they represent attractive targets for therapy (8). Authors remark that CSCs from GIC are able to transdifferentiate into endothelial-like cells and pericytes, two important lineages for maintenance of cancer vascular niche, thus opening opportunities for therapy intervention of angiogenesis and VM.

Quintero-Fabián et al. emphasize the importance of the matrix metalloproteinases (MMPs) that participate in the degradation of basement membrane to stimulates the cancer cell growth and spreading in various types of cancer. They also analyze the roles of MMPs, cytokines, and immune system cells in the angiogenic events in cancer cells.

In conclusion, investigations of vascular diseases continue to be essential toward the development of new therapeutic strategies

#### REFERENCES


that produce more successful treatments for localized and metastatic cancers. We believe that the experimental discoveries and opinions presented in this Research Topic may have a major impact on oncology research and treatment and will inspire future research. We hope this Research Topic will fuel further interests in Scientifics, general readers and Scholars.

#### AUTHOR CONTRIBUTIONS

All authors listed have made substantial contributions to the Research Topic, and approved it for publication.

#### ACKNOWLEDGMENTS

We thank all the authors and reviewers of this Frontiers Research Topic for their excellent contribution. We also thank to the editorial team at Frontiers for their invaluable support.


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2020 López-Camarillo, Ruiz-García, Starling and Marchat. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# HypoxamiRs Profiling Identify miR-765 as a Regulator of the Early Stages of Vasculogenic Mimicry in SKOV3 Ovarian Cancer Cells

Yarely M. Salinas-Vera<sup>1</sup> , Dolores Gallardo-Rincón<sup>2</sup> , Raúl García-Vázquez <sup>3</sup> , Olga N. Hernández-de la Cruz <sup>1</sup> , Laurence A. Marchat <sup>3</sup> , Juan Antonio González-Barrios <sup>4</sup> , Erika Ruíz-García<sup>2</sup> , Carlos Vázquez-Calzada<sup>5</sup> , Estefanía Contreras-Sanzón<sup>1</sup> , Martha Resendiz-Hernández <sup>1</sup> , Horacio Astudillo-de la Vega<sup>6</sup> , José L. Cruz-Colin<sup>7</sup> , Alma D. Campos-Parra<sup>8</sup> and César López-Camarillo<sup>1</sup> \*

#### Edited by:

Monica Venere, The Ohio State University, United States

#### Reviewed by:

Gareth I. Owen, Pontificia Universidad Católica de Chile, Chile Vijay Pandey, National University of Singapore, Singapore

\*Correspondence:

César López-Camarillo genomicas@yahoo.com.mx

#### Specialty section:

This article was submitted to Cancer Molecular Targets and Therapeutics, a section of the journal Frontiers in Oncology

Received: 09 January 2019 Accepted: 23 April 2019 Published: 14 May 2019

#### Citation:

Salinas-Vera YM, Gallardo-Rincón D, García-Vázquez R, Hernández-de la Cruz ON, Marchat LA, González-Barrios JA, Ruíz-García E, Vázquez-Calzada C, Contreras-Sanzón E, Resendiz-Hernández M, Astudillo-de la Vega H, Cruz-Colin JL, Campos-Parra AD and López-Camarillo C (2019) HypoxamiRs Profiling Identify miR-765 as a Regulator of the Early Stages of Vasculogenic Mimicry in SKOV3 Ovarian Cancer Cells. Front. Oncol. 9:381. doi: 10.3389/fonc.2019.00381 <sup>1</sup> Posgrado en Ciencias Genómicas, Universidad Autónoma de la Ciudad de Mexico, Mexico City, Mexico, <sup>2</sup> Laboratorio de Medicina Translacional y Departamento de Tumores Gastro-Intestinales, Instituto Nacional de Cancerología, Mexico City, Mexico, <sup>3</sup> Programa en Biomedicina Molecular y Red de Biotecnología, Instituto Politécnico Nacional, Mexico City, Mexico, <sup>4</sup> Laboratorio de Medicina Genómica, Hospital Regional 1 de Octubre ISSSTE, Mexico City, Mexico, <sup>5</sup> Departamento de Infectómica y Patogénesis Molecular, CINVESTAV-IPN, Mexico City, Mexico, <sup>6</sup> Laboratorio de Investigación Translacional en Cáncer y Terapia Celular, Hospital de Oncología, Centro Médico Nacional Siglo XXI, Mexico City, Mexico, <sup>7</sup> Subdirección de Investigación Básica, Instituto Nacional de Medicina Genómica, Mexico City, Mexico, <sup>8</sup> Laboratorio de Genómica, Instituto Nacional de Cancerología, Mexico City, Mexico

Vasculogenic mimicry (VM) is a novel cancer hallmark in which malignant cells develop matrix-associated 3D tubular networks with a lumen under hypoxia to supply nutrients needed for tumor growth. Recent studies showed that microRNAs (miRNAs) may have a role in VM regulation. In this study, we examined the relevance of hypoxia-regulated miRNAs (hypoxamiRs) in the early stages of VM formation. Data showed that after 48 h hypoxia and 12 h incubation on matrigel SKOV3 ovarian cancer cells undergo the formation of matrix-associated intercellular connections referred hereafter as 3D channels-like structures, which arose previous to the apparition of canonical tubular structures representative of VM. Comprehensive profiling of 754 mature miRNAs at the onset of hypoxia-induced 3D channels-like structures showed that 11 hypoxamiRs were modulated (FC>1.5; p < 0.05) in SKOV3 cells (9 downregulated and 2 upregulated). Bioinformatic analysis of the set of regulated miRNAs showed that they might impact cellular pathways related with tumorigenesis. Moreover, overall survival analysis in a cohort of ovarian cancer patients (n = 485) indicated that low miR-765, miR-193b, miR-148a and high miR-138 levels were associated with worst patients outcome. In particular, miR-765 was severely downregulated after hypoxia (FC < 32.02; p < 0.05), and predicted to target a number of protein-encoding genes involved in angiogenesis and VM. Functional assays showed that ectopic restoration of miR-765 in SKOV3 cells resulted in a significant inhibition of hypoxia-induced 3D channels-like formation that was associated with a reduced number of branch points and patterned tubular-like structures. Mechanistic studies confirmed that miR-765 decreased the levels of VEGFA, AKT1 and SRC-α transducers and exerted a negative regulation of VEGFA by specific binding to its 3'UTR. Finally, overall survival analysis of a cohort of ovarian cancer

**9**

patients (n = 1435) indicates that high levels of VEGFA, AKT1 and SRC-α and low miR-765 expression were associated with worst patients outcome. In conclusion, here we reported a novel hypoxamiRs signature which constitutes a molecular guide for further clinical and functional studies on the early stages of VM. Our data also suggested that miR-765 coordinates the formation of 3D channels-like structures through modulation of VEGFA/AKT1/SRC-α axis in SKOV3 ovarian cancer cells.

Keywords: ovarian cancer, vasculogenic mimicry, hypoxia, miR-765, VEGFA

#### INTRODUCTION

Tumor vasculogenic mimicry (VM) is a novel cancer hallmark formerly described in malignant melanoma cells which involves the formation of patterned three dimensional (3D) channels networks by tumor cells (1). These tubular networks resemble embryonic vasculogenesis, and they describe the ability of certain types of aggressive cancer cells to express endotheliumassociated genes (2). Tumor VM occur de novo without or in combination with blood vessels formation changing our conventional acceptance that classical angiogenesis is the only means by which cancer cells acquire a nutrients supply to nourish tumors. Studies supporting these assumptions have demonstrated that in vivo the 3D channels contain plasm, erythrocytes and blood flow with a hemodynamics similar to those occurring in endothelial vessels (3). Evidences for VM have been found in other solid tumors and cancer cell lines such as in glioblastoma (4), breast (5, 6), prostate (7), lung (8), hepatocellular (9) and ovarian cancers (10, 11), among others. This morphologic plasticity have been associated to aggressive tumor phenotypes, increased metastasis and tumor progression of certain types of cancers. Moreover, meta-analysis studies have established a definitive association between VM with poor clinical poor prognosis in human cancer patients (12). Remarkably, tumor VM may contribute to the resistance of diverse type of tumors against anti-angiogenic therapy (13, 14). Therefore, the exploration of the multiple roles of VM in cancer hallmarks, especially in drug resistance, would broaden our knowledge and eventually ameliorate the treatment efficacy in cancer.

Cellular features underlying VM are diverse although they may summarized as follows: (i) vascular-like tubules are lined by tumor cells in combination or not with endothelial cells forming complex 3D mosaic patterns; (ii) VM cells achieve remodeling of extracellular matrix and tumor microenvironment; (iii) 3D channels assembled during VM connects with the tumor microcirculation system providing blood and supplies for tumor growth, (iv) VM provides also a perfusion route for metabolic waste; and (v) in tumor tissues VM cells showed Periodicacid Schiff (PAS) positive and CD31 negative staining which provides a new tool for potential use in clinical practice (15). Nonetheless, in vitro reports on VM are still debatable because only few studies provide solid evidence of 3D tube formation (1, 16–19) or use malignant melanoma or ovarian cancer cell lines previously confirmed to form tubular 3D structures (19–21). In an outstanding paper from Owen's lab this controversy was addressed by characterizing VM in vitro using SKOV3, HEY and other ovarian cancer cell lines, as well as spheres and primary cultures derived from ovarian cancer ascites (19). Using dye microinjection, X-ray microtomography 3Dreconstruction, and confocal microscopy studies they confirmed that glycoprotein-rich lined 3D tubular structures are present in in vitro cultures and were able of conducting fluids. This study highlights the importance of confirmatory in vitro assays for VM, and surprisingly suggested that many of 3D cellular networks reported in the literature may not represent genuine VM (19).

Diverse molecular mechanisms and signaling pathways have been described to be involved in VM formation (22–24). Moreover, it has been described that aggressive tumor cells undergoing VM showed specific gene-expression profiles that resembles that of an undifferentiated, embryonic-like cells (2). Molecular mechanisms operating in VM have been extensively studied recently with some master regulators identified (25). For instance, hypoxia inducible factor 1-α (HIF-1α) greatly promotes VM formation in response to hypoxia as it occurs in angiogenesis (26). The role of other proteins and signaling pathways that promote cell proliferation, migration, invasion and matrix remodeling during tumor VM also has been described. These include factors such as the vascular endothelialcadherin (VE-cadherin) (21, 27), epithelial cell kinase (EphA2) (18), phosphoinositide 3-kinase alpha (PI3K-α) (6), matrix metalloproteinase (MMPs), laminin 5 (Ln-5) γ2 chain, focal adhesion kinase (FAK) (23–25) and proto-oncogene tyrosineprotein kinase SRC-α (6). Although important advances in deciphering the molecular mechanism underlying VM, the finetuning modulation and the role of non-coding RNAs in the early stages of VM remains poorly understood.

During the last decades, the study of non-coding RNAs in cancer biology has exploded revealing unsuspected functions in tumorigenesis. MicroRNAs (miRNAs) are non-coding singlestranded small RNAs of 21-25 nucleotides in length that function as negative regulators of gene expression (28). MiRNAs function as guide molecules in post-transcriptional gene silencing by partially complementing with the 3′ -end of target transcripts resulting in mRNA degradation or translational repression in cytoplasmic P-bodies (29). These small non-coding RNAs may target a plethora of regulatory molecules driving tumorigenesis. Recent studies showed that some miRNAs have a pivotal role in VM in diverse types of solid tumors. For instance, miR-26b targets EphA2 a VM regulator in glioma (30). In breast cancer, miR-204 exerts a fine-tuning regulation of the synergistic transduction of PI3K/AKT1/FAK mediators critical in VM formation (6). In ovarian cancer only two studies about the role of miRNAs, specifically miR-200a and miR-27b, have been reported (31, 32), indicating that detailed miRNAs functions in VM regulation in ovarian cancer remains to be elucidated. In the present investigation, we reported a novel miRNAs signature activated during the hypoxia-induced 3D channels-like networks formation in ovarian cancer cells. Also, we provide functional data suggesting a role for miR-765 in VM through regulation of VEGFA/AKT1/SRC-α axis.

### MATERIALS AND METHODS

#### Cell Lines

Human ovarian cancer cell line SKOV-3 was obtained from the American Type Culture Collection (ATTC HTB-77), and routinely grown in Dulbecco's modification of Eagle's minimal medium (DMEM) supplemented with 10% fetal bovine serum and penicillin-streptomycin (50 unit/ml; Invitrogen, Carlsbad, CA, USA).

#### Periodic Acid Staining

3D-cultures were fixed in 4% formaldehyde in phosphate buffered solution (PBS) 1X for 30 min at room temperature. Coverslips were incubated with 0.5% periodic acid for 5 min, washed with PBS 1X for 5 min and Schiff reagent for additional 15 min. Then, cells were washed with PBS 1X for 5 min. Later they were incubated with hematoxylin for 1 min and washed in tap water for 5 min. Samples were dehydrated and mounted in coverslip using a synthetic mounting medium for microscopy.

#### Three Dimensional (3D) Cultures

Experiments were performed with 70–80% confluent cell cultures. 3D cultures were prepared for confocal microscopy analysis as follow: 18 × 18 mm glass coverslips were acetonewashed, air-dried and placed in 6-well culture plates, coated with 50 µL of Matrigel per coverslip and air-dried for 60 min at room temperature. Cell cultures were trypsinized, and 60,000 cells were resuspended in 200 µL of culture medium, which was seeded on matrigel-coated coverslips. Cells were incubated at 37◦C for 3 h to allow its adhesion to the matrix and then covered with 3 ml of culture medium.

#### Immunofluorescence Analysis

Briefly, 3D-cultures were fixed in 4% formaldehyde in PBS 1X for 30 min at room temperature. Coverslips were incubated with 0.1% Triton X-100 for 3 min. Following washing with PBS 1X, cells were blocked for 40 min at room temperature with 0.2% BSA in PBS 1X, and incubated with Phalloidin 1X (Abcam, ab235138) for 30 min at room temperature. Stained cells were then washed with PBS 1X for 15 min and mounted for confocal microscopy.

### RNA Isolation

Total RNA was extracted using 500 µl Trizol (Invitrogen, Carlsbad, CA) for 1 × 10<sup>4</sup> cells/well as described the manufacturer. RNA integrity was assessed using capillary electrophoresis system Agilent 2100 Bioanalyzer. Samples with a RNA integrity >5 were processed.

### MicroRNAs Expression Profiling

The Megaplex TaqMan Low-Density Array (TLDA) v 3.0 (Applied Biosystems, Foster City, CA) platform was used to measure the expression of 754 human specific miRNAs in parallel. Briefly, total RNA (600 ng) was retro-transcribed using stem-loop primers specific for each miRNA in order to obtain complementary DNA (cDNA) templates. Subsequently, a preamplification step of 12 cycles was included to increase the concentration of low-level miRNAs. The pre-amplified products were loaded into the TLDA and reactions were started using the 7900 FAST real-time thermal cycler (ABI). RNU44 and RNU48 expression was used as internal control. For statistical analysis miRNAs levels were measured by quantitative reverse transcription polymerase chain reaction (qRT-PCR) in TLDA using the comparative Ct (211Ct) method. All analyses were done using R (HTqPCR and gplots-bioconductor). The Ct raw data were determined using an automatic baseline and a threshold of 0.2. A fold change (FC) (log2 RQ) value >1.5 was used to define the differentially expressed miRNAs. An adjusted t-test was used to evaluate the significant differences in Ct values between groups. To identify subgroups defined by miRNA expression profiles, an unsupervised clustering analysis using Spearman correlation and average linkage was used.

#### Bioinformatics Analysis

MiRNA targets were identified using TargetScan 7.0 (http:// www.targetscan.org/vert\_71/), and PicTar (http://www.pictar. org/) softwares. Only target genes that were predicted by the two algorithms were selected for further analysis. Gene ontology and enrichment cellular pathway analyses were performed using David tool.

### Transfection of miR-765 Mimic

MiRNA-765 mimic (AM17100 ThermoFisher), and pre-miRnegative control scramble (AM17110 ThermoFisher) were transfected in SKOV3 cells using siPORT amine transfection agent. Briefly, miR-765 (80 nM) and scramble (80 nM) were individually added to wells containing 1 × 10<sup>4</sup> cells cultured in DMEM for 48 h. Then, overexpression of miR-765 was confirmed by quantitative RT-PCR at 48 h postransfection using total RNA. MiR-765-expressing cells were used for downstream analysis.

### 3D Channels-Like Networks Inhibition Assays

3D channels-like networks experiments were performed through 3D-dimensional cultures on matrigel. Firstly, SKOV3 cells (1 × 10<sup>4</sup> cells/well) were transfected with pre-miR-765 (80 nM) or scramble (30 nM) negative control as previously described. The cells were cultured in 96-well plate covered with geltrex matrix (50 µl). Afterward, cells were incubated at 37◦C in 5% CO<sup>2</sup> atmosphere in hypoxia conditions (1% O2) for 48 h. Then, the formation of 3D channels formation was induced by seeding cells on matrigel and then capillary-like structures were observed under an inverted microscope (Iroscope SI-PH) and imaged during 0, 6, and 12 h. Two observers individually counted the number of branch points and tubular structures. Data were expressed as mean ±S.D. p<0.05 was considered as statistically significant.

#### Western Blot Assays

30 µg of whole protein extracts were separated on 12% SDS-PAGE and transferred to 0.2µm nitrocellulose membrane (Bio-Rad) and then incubated with the following primary antibodies: anti-AKT1 (1:1000, C74H10 Cell signaling), anti-SRC-α (1:1000; sc-130124 Santa Cruz), anti-VEGFA (1:500, ab183100 abcam) and anti-GAPDH (1:1000, sc-365062 Santa Cruz). Densitometry analysis of immunodetected bands in Western blots assays were performed using the public domain myImage Analysis software.

#### Luciferase Gene Reporter Assays

DNA fragments of the 3'UTR of VEGFA gene containing the predicted miR-765 binding sites were cloned into p-miRreport vector (Ambion) downstream of luciferase gene. All constructs were verified through automatic sequencing. Then, recombinant pmiR-LUC-VEGFA plasmid was transfected into SKOV3 cells using lipofectamine 2000 (Invitrogen). At 24 h after transfection, pre-miR-765 (80 nM) and scramble were cotransfected with lipofectamine RNAi max (Invitrogen). Then, 24 h after transfection firefly and Renilla reniformis luciferase activities were both measured by the Dual-Glo luciferase Assay (Promega) using a Fluoroskan AscentTM Microplate Fluorometer. Firefly luciferase activity was normalized with Renilla reniformis luciferase.

### Kaplan Meier Analysis

Overall survival analysis using Kaplan Meier plotter for miR-765, VEGFA, AKT1, and SRC-α genes in ovarian cancer patients were evaluated as previously described (33, 34). Briefly, we used the Start KM plotter for ovarian cancer tool that use genome-wide for mRNA expression data and overall survival clinical information of cancer patients, which were downloaded from Gene Expression Omnibus GEO (Affymetrix HG-U133A, HG-U133A 2.0, and HG-U133 Plus 2.0 microarrays) and The


Cancer Genome Atlas TCGA, whereas for miRNAs expression we used Start miRpower for pan-cancer as implemented in the KM plotter at the URL (http://kmplot.com/analysis/index. php?p=backgroundr). To define the prognostic value of genes the samples were split into two groups according to various quantile expression of miR-765 (n = 485) and VEGFA, AKT1 and SRC-α genes in ovarian cancer patients (n = 1435). A Kaplan-Meier survival plot compared the two patient cohorts, and the hazard ratio with 95% confidence intervals and logrank P-value were calculated.

#### Statistical Analysis

Experiments were performed three times by triplicate and results were represented as mean ±S.D. One-way analysis of variance (ANOVA) followed by Tukey's test were used to compare the differences between means. A p < 0.05 was considered as statistically significant.

### RESULTS

### MicroRNAs Modulated During Hypoxia-Induced 3D Channels-Like Structures Formation in Ovarian Cancer Cells

To investigate the role of hypoxia in expression of miRNAs associated with the initial phases of vasculogenic mimicry (VM), firstly we established an in vitro model for three-dimensional (3D) channels-like structures formation representative of the early stages of VM. We have chosen the SKOV3 ovarian cancer cells which were previously unequivocally demonstrated to form vasculogenic mimicry in vitro after 4 days incubation in hypoxia (19). Here, SKOV3 cells were grown in confluent monolayers under hypoxia (1% O2) or normoxia conditions during 48 h. Then, cells were seeding on matrigel and incubated for 0, 6, and 12 h to track the formation of 3D capillary-like structures, which represent the stages previous to VM formation. Results showed that SKOV3 cells grown in normoxia hardly exhibited the formation of cellular networks after 6 and 12 h incubation on matrigel (**Figures 1A–C**). When cells were grown in hypoxia, a dramatical increase in extend of cellular networks was observed during the course of time. SKOV3 cells exhibited the typical morphologic changes indicative of 3D channels-like networks formation after 0 and 6 h incubation on matrigel (**Figures 1D,E**). Remarkably, after 12 h incubation a significant and gradual increase in networks was found (**Figure 1F**). Quantification of the number of cellular networks showed that these structures were significantly augmented from 98 ± 4 to 172 ± 7 after 6 and 12 h incubation, respectively (**Figure 1G**). Likewise, the number of branch points was significantly increased from 43 ± 2 to 71 ± 4 after 6 and 12 h, respectively (**Figure 1H**). At 12 h, positive PAS staining was found mainly along the length of the cellular networks suggesting the existence of extracellular matrix compounds (**Figures 1I,J**). To evaluate the potential presence of tubular structures with a hollow tube, SKOV3 cells were stained with rhodamine-phalloidin and analyzed by confocal microscopy (**Figures 1K,L**). Immunofluorescence images of the cellular networks showed very discrete elevated structures with tubular-like appearances as observed in bright field and red channel (**Figures 1M,N**). A confocal microscopy Z-stack reconstruction of 12 h old 3D-cultures of SKOV3 cells hardly showed the presence of proper tubular structures with hollow centers (**Figures 1N,O**). These findings indicate that after 48 h hypoxia and 12 h incubation on matrigel, no clear tubules with hollow centers were generated by SKOV3 cells. Instead of

TABLE 1 | Modulated microRNAs after 48 h hypoxia in SKOV3 ovarian cancer cells and predicted targets with functions associated to cancer.


TABLE 1 | Continued



<sup>a</sup>Uniprot database name; VM, vasculogenic mimicry.

we found cellular networks which were organized and lined in a time-dependent manner.

In order to identify the set of miRNAs regulated by hypoxia (hypoxamiRs) before VM formation, we profiled 667 mature miRNAs using Taq Man Low Density Arrays (TLDAs) after 48 h hypoxia. Our results showed that 11 unique hypoxamiRs were significantly modulated (FC>1.5; p < 0.05) in SKOV3 cells. Of these 9 miRNAs were downregulated (miR-765, miR-660, miR-218, miR-198, miR-518b, miR-148a, miR-1290, miR-193b, miR-222) and 2 upregulated (miR-486-3p, miR-138) in comparison to control cells grown without hypoxia (time 0) (**Figure 2**). Next, we were wondering if expression levels of the set of modulated miRNAs may have clinical implications in ovarian cancer. Therefore, we performed overall survival analysis using Kaplan Meier tool (Start miRpower pan-cancer) which utilize genome-wide transcriptome data and overall survival clinical information from a large cohort of ovarian cancer patients (n = 485) with a follow-up of 180 months as described in material and methods (33, 34). To define the prognostic value of genes the samples were split into two groups according to quantile expression of miRNAs. A Kaplan-Meier survival plot compared the two patient cohorts, and the hazard ratio with 95% confidence intervals and logrank P-value were calculated. Results showed that high expression of miR-138 (HR = 1.80, logrank P = 5.3e-07) and low levels of miR-765 (HR = 0.77, logrank P = 0.05), miR-193b (HR = 0.86, logrank P = 0.25), and miR-148a (HR = 0.63, logrank P = 0.0001) genes were associated to low overall survival of ovarian cancer patients (**Figure 2**).

#### HypoxamiRs Regulate Cellular Pathways Associated With Cancer

Predictive analysis of the set of regulated hypoxamiRs suggested that they might impact common cellular processes and signaling pathways related with tumorigenesis (**Figure 3A**). The signaling pathways enriched were TGF-β, WNT, mTOR, AMPK, estrogen receptor and RAP1. Computational predictions also indicated that these miRNAs may target a number of genes involved in VM and angiogenesis including HIF-1A, HIF-1AN, HIF-3A, PTGFRN, AKT1, VEGFA, VEGFB, VEGFC, PDGFR, TGF-βR2, MMP2, PTK2, SRC, SHC3, and GRB2, among others (**Table 1**). In particular, we focused in miR-765 for further functional analysis because: (i) it was severely downregulated after hypoxia (FC < 32.02; p < 0.05), (ii) it was predicted to target a number of genes involved in VM (**Figure 3B**), and (iii) there is no reports about the functions of miR-765 in ovarian cancer neither in tumor VM.

#### Hypoxia-Suppressed miR-765 Inhibits Channels-Like Networks Formation

To examine the functional role of miR-765 on 3D channelslike networks, we restored its expression in SKOV3 cells by transfection of specific RNA mimics. Then, 3D channelslike networks formation was induced by 48 h hypoxia as described before. Non-transfected and scramble-treated cells were included as controls. Interestingly, ectopic restoration of miR-765 produced a dramatic inhibition of 3D channelslike networks formation (**Figure 4A**). A significant reduction

of the number of branch points (up to 85%) and capillary tubes (up to 92%) were found in miR-765-transfected cells in comparison to control cells (**Figure 4B**). To discard pleiotropic effects of miR-765 overexpression in cell survival of ovarian cancer cells, we performed cell viability assays. Data showed no significant changes in viability of miR-765-expressing SKOV3 cancer cells at the tested concentrations which indicate that the effect of miR-765 in 3D channels-like networks impairment was specific (**Figure 4C**).

### MiR-765 Downregulates VEGFA, AKT1 and SRC-α and Directly Target VEGFA

Because the bioinformatics predictions of gene targets suggested that several signaling pathways such as VEGFA, AKT, and SRC/FAK, could be affected in SKOV3 cells transfected with miR-765, we proceed to evaluate the changes in expression of the aforementioned proteins using available antibodies in Western blot assays (**Figure 5**). Results showed that VEGFA protein was expressed at low levels in cells cultured under normoxia conditions, but its expression was significantly increased under hypoxia. Moreover, we observed a significant decrease in VEGFA levels in SKOV3 cells transfected with miR-765 mimics in comparison to non-treated and scramble transfected controls cells (**Figures 5A,B**). Likewise, a significant decrease in both SRC-α and AKT1 levels was found in cells transfected with miR-765 mimics in comparison to control cells (**Figures 5A,C,D**). No significant changes were observed in GADPH levels used as control. Computational predictions also showed that miR-765 may target a number of protein-encoding genes with known roles in VM. Of these, we focused in the study of VEGFA as it was downregulated by miR-765 and it contain a potential miR-765 binding site at 3′UTR (**Figure 5E**). To corroborate whether miR-765 can exert posttranscriptional repression of VEGFA, we performed luciferase reporter assays. A DNA fragment corresponding to 3′UTR of VEGFA was cloned downstream of the luciferase-coding region of pmiR-LUC vector (**Figure 5E**). In addition, a mutated version of the miR-765 binding site at the VEGFA 3'UTR was included as a plasmid control. Data showed that ectopic expression of miR-765 and co-transfection of recombinant VEGFA 3′UTR wild type plasmid into SKOV3 cells resulted in a significant reduction of the relative luciferase activity in comparison with

controls (**Figure 5B**). In addition, when mutated sequence was assayed no significant changes in luciferase activity were found. Altogether these data confirmed that VEGFA is a novel target of miR-765.

### Expression Levels of miR-765, VEGFA, AKT1, and SRC-α Correlate With Poor Patient's Outcome

Then we were wondering if changes in expression levels of miR-765, VEGFA, AKT1 and SRC-α have clinical implications in ovarian cancer. Thus, we performed overall survival analysis using Start Kaplan Meier plotter for ovarian cancer which use genome-wide transcriptome data and overall survival clinical information from a large cohort of ovarian cancer patients (n = 1485) with a mean follow-up of 170 months. To define the prognostic value of genes the samples were split into two groups according to various quantile expression of VEGFA, AKT1 and SRC-α genes. A Kaplan-Meier survival plot compared the two patient cohorts, and the hazard ratio with 95% confidence intervals and logrank P-value were calculated. Results showed that low levels of miR-765 (HR = 0.77, logrank P = 0.05) and high expression of its targets VEGFA (HR = 1.38, logrank P = 1.8e-05), AKT1 (HR = 1.19, logrank P = 0.0071), and SRC-α (HR = 1.39, logrank P = 0.000092) signaling genes were associated to low overall survival of ovarian cancer patients (**Figure 6**).

### DISCUSSION

Tumor VM is a highly orchestrated cellular mechanism in which highly aggressive and metastatic tumor cells form vascularlike 3D networks to provide an efficient and functional fluidconducting system for blood and oxygen supply, as an alternative to classical vasculogenesis. This morphologic plasticity is associated to high aggressiveness, increased metastasis and tumor progression of certain types of cancers. In clinical VM has been related with low overall survival and resistance to current anti-angiogenic therapies (60). Remarkably, tumor VM can be potentially targeted by novel therapeutic agents, thus currently diverse investigations in the search of novel VM regulators are undergoing. In order to contribute with the understanding of the role of small non-coding RNAs in the molecular mechanisms responsible for VM, here we have uncovered a novel set of miRNAs modulated at the early onset of hypoxia-induced 3D channels-like structures formation, previous to the proper formation of tubules indicative of VM in SKOV3 cells. We fist

cancer patients (n = 1435). Kaplan-Meier survival plots compared the two patient cohorts, and the hazard ratio with 95% confidence intervals and logrank P-value

set-up an in vitro model, and using PAS staining we confirmed that SKOV3 cells efficiently form 3D channels-like networks in agreement with other studies (16, 61–63). It's important to note that the tubular-like structures we have analyzed here, may not reflect VM properly, but they represents the early stages of VM and the morphological and transcriptional programs activated by 48 h hypoxia, previous to VM appearance. It's important to remarks the urgency of confirmatory in vitro assays for proper VM in the different types of cancer, as a recent report (19) surprisingly suggested that many of structures reported in the literature at early times of hypoxia may not represent VM, as we can confirms in the present study.

Hypoxia is an important activator of VM, thus we decided to search for the miRNAs regulated by hypoxia (hypoxamiRs) during initial stages of VM, as it remains largely unknown in ovarian cancer. Our results showed that 11 hypoxamiRs were significantly modulated. Of these 9 miRNAs were downregulated (miR-765, miR-660, miR-218, miR-198, miR-518b, miR-148a, miR-1290, miR-193b, miR-222) and 2 upregulated (miR-486-3p, miR-138) (**Figure 2**). Interestingly, high expression of miR-138 and low levels of miR-765, miR-193b, and miR-148a genes were associated to low overall survival suggesting a potential clinical value in ovarian cancer patients (**Figure 2**). However, we cannot drawn a solid connection between outcome and VM in patients, as we have collected the clinical data from KMplot databases, and unfortunately no VM presence/absence data is available for the cohort of patients analyzed here. Thus, we have limited the conclusions only to a correlation between miRNAs regulated by hypoxia and the overall survival. On the other hand, several of the regulated miRNAs have been previously associated with tumorigenesis in diverse types of cancer. For instance, miR-660 was reported as

were calculated.

downregulated in in lung cancer patients and its transient and stable overexpression using RNA mimics reduced migration, invasion, and proliferation properties and increased apoptosis in p53 wild-type lung cancer cells (64). Likewise, miR-218-5p expression was lower in cervical cancer tumors in comparison with normal tissues. MiR-218-5p suppressed the progression of cervical cancer via LYN/NF-κB signaling pathway (65). In addition, miR-138 promotes cell proliferation and invasion on colorectal cancer (66), and it contributes to resistance to therapy in multiple myeloma and non-small cell lung cancer (67, 68). Of the set of regulated hypoxamiRs, we focused in the study of functional relationships between miR-765 and 3D channels-like formation. Recently, miR-765 have been reported as upregulated or downregulated in diverse types of malignancies such as esophageal squamous cell carcinoma (69), melanoma (70), osteosarcoma (71), oral squamous cancer (72) and hepatocellular carcinoma (73). Nevertheless, miR-765 functions in ovarian cancer and tumor VM remains largely unknown. Our data showed that the ability of SKOV3 cells to develop 3D channels-like structures formation under hypoxia was significantly reduced after transfection of miR-765 mimics. This may be explained as the target predictions indicate that miR-765 may regulate genes associated to the cell proliferation, matrix remodeling, migration, and invasion, angiogenesis, and VM formation. Indeed, we demonstrated that miR-765 was able to downregulate the VEGFA, AKT1 and SRC-α signaling transducer critical in VM. Also important is the fact that expression of miR-765 and its aforementioned gene targets have a potential clinical value as its deregulation was associated with worst outcome in ovarian cancer patients (**Figure 6**). Main limitations of the present study are denoted by the use of a single cell model, which however, permit us to delineate important conclusions about the hypoxamiRs modulated in SKOV3 cells, and guide us to the analysis of miR-765 and its role in 3D channels-like structures formation.

#### REFERENCES


Nonetheless, we understand the need to extend our initial findings in additional ovarian cancer cell lines in future studies. Also, a limitation of the present study is that we specifically analyzed here the early stages of VM (after 48 h hypoxia); thus the potential role of the revealed miRNAs signature at later stages of proper VM is unknown. Taken altogether, we propose that miR-765 may regulate 3D channels-like structures formation through both direct and indirect targeting of signaling transducers. Also, we suggested that miR-765 could impair VEGFA by direct binding to VEGFA and AKT1; as well as by indirect downregulation of SRC-α which in turn may block the VEGFA/AKT1 signaling transduction. In conclusion, in the current work we provide a novel set of regulated hypoxamiRs and experimental data supporting an unexpected role for VEGFA/AKT1/SRC-α axis in 3D channels-like structures formation in SKOV3 cells. As novel therapies targeting hypoxic cancer cells are needed to improve therapy treatment of cancer, we consider that our data are relevant and deserves further in vivo validation.

#### AUTHOR CONTRIBUTIONS

YS-V, RG-V, OH, EC-S, and MR-H conducted all the experiments. JG-B performed the microRNAs profiling. CV-C performed the confocal microscopy. JC-C provide advice in cell cultures. DG-R, ER-G, HA, and AC-P contributed to experimental design, intellectual input, and interpreting data. CL-C and LM wrote the manuscript.

#### ACKNOWLEDGMENTS

The authors acknowledge the Grupo Mexicano de Investigación en Cáncer de Ovario for research funding and support. The authors also acknowledge the Universidad Autónoma de la Ciudad de Mexico for support.


migration by suppressing SHC-RAF-ERK and AKT signaling. Neuro Oncol. (2010) 12:941–55. doi: 10.1093/neuonc/noq048


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Salinas-Vera, Gallardo-Rincón, García-Vázquez, Hernández-de la Cruz, Marchat, González-Barrios, Ruíz-García, Vázquez-Calzada, Contreras-Sanzón, Resendiz-Hernández, Astudillo-de la Vega, Cruz-Colin, Campos-Parra and López-Camarillo. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Targeting Vascular Endothelial Growth Factor in Oesophagogastric Cancer: A Review of Progress to Date and Immunotherapy Combination Strategies

Oliver Butters † , Kate Young† , David Cunningham, Ian Chau and Naureen Starling\*

*Gastrointestinal Unit, Royal Marsden Hospital, London, United Kingdom*

#### Edited by:

*Giuseppe Di Lorenzo, Azienda Ospedaliera Universitaria Federico II, Italy*

#### Reviewed by:

*Shumei Zhai, Shandong University, China Ajaikumar B. Kunnumakkara, Indian Institute of Technology Guwahati, India*

#### \*Correspondence:

*Naureen Starling naureen.starling@rmh.nhs.uk*

*†These authors have contributed equally to this work and share first authorship*

#### Specialty section:

*This article was submitted to Cancer Molecular Targets and Therapeutics, a section of the journal Frontiers in Oncology*

Received: *13 December 2018* Accepted: *24 June 2019* Published: *16 July 2019*

#### Citation:

*Butters O, Young K, Cunningham D, Chau I and Starling N (2019) Targeting Vascular Endothelial Growth Factor in Oesophagogastric Cancer: A Review of Progress to Date and Immunotherapy Combination Strategies. Front. Oncol. 9:618. doi: 10.3389/fonc.2019.00618* In 2014, the survival benefits seen in REGARD and RAINBOW studies led the way for the regulatory approval of ramucirumab in the second line setting in oesophagogastric (OG) cancer. Trials of other drugs targeting the vascular endothelial growth factor (VEGF) pathway have met with mixed results but this remains an important pathway for evaluation in OG cancer. Perhaps the most interesting ongoing trials are those which target VEGF in combination with immunotherapy, which have a sound scientific rationale. Given the emerging role of immunotherapy in OG cancer, this is an important area of innovation. This review aims to outline targeting VEGF in OG cancer, the rationale behind the continued interest in this mechanism and possible future directions in combination with immunotherapy.

Keywords: gastroesophageal cancer, vascular endothelial grow factor, immunotharapy, ramucirumab, biomarkers, bevacizumab, tyrosine kinasa inhibitor

## INTRODUCTION

Oesophagogastric (OG) cancer consists of esophageal, gastro-esophageal junctional (GOJ), and gastric cancer and is associated with a poor prognosis. Gastric and esophageal cancers are the third and sixth leading causes of cancer related death worldwide with an estimated 723,000 and 400,000 deaths in 2012, respectively (1). A SEER cancer statistics review revealed an increase in 5 year survival in OG cancers from 1975 to 2014, from 15.2 to 32.1% in gastric cancers and 5.0–21.1% in esophageal cancers (2), although this continues to be poor and the median overall survival (mOS) remains less than a year. Histologically, OG cancers are divided in to adenocarcinoma and squamous cell carcinoma with most esophageal cancers (72%) and nearly all gastric cancers (96%) being adenocarcinoma (3).

### CURRENT TREATMENT PARADIGM FOR OG CANCER

Two thirds of Western patients present with advanced inoperable disease and for these patients, median overall survival is short (4). First line palliative chemotherapy for OG adenocarcinoma involves a platinum and fluoropyrimidine based doublet or triplet regimen with the addition of trastuzumab if Human Epidermal Growth Factor Receptor 2 (HER2) positive. Median overall survival is 3 months with best supportive care (BSC), less than a year with palliative chemotherapy and just over a year with the addition of trastuzumab in selected patients (**Figure 1**). Even with doublet and triplet chemotherapy, median survival of over a year is only achieved with the addition of the first biologic to be approved in this disease, trastuzumab. There is considerable geographical variation in survival, with improved survival in Japanese patients compared to western patients being well-documented (13–15).

Half of patients receiving first line chemotherapy can be expected to proceed on to second line chemotherapy on progression, although this figure varies considerably across the world, and there may be a role for sequential therapy for those who can tolerate it (16–18). Second-line chemotherapy with a taxane (docetaxel, paclitaxel) or irinotecan is recommended for patients who are of a good performance status (14–16, 19, 20). Such treatment has been shown to be superior to BSC by a number of studies with a 37% reduction in the risk of death (16, 21). However, the actual benefit remains limited, with mOS 3.8 months with BSC vs. 5.3 months with salvage chemotherapy and improved therapeutics are required (20). More recently, data has emerged to support the targeting of VEGF in the second line setting with the use of ramucirumab either as a single agent or in combination with paclitaxel. High level evidence for treatment in the third line setting is lacking but there is a role for immunotherapy emerging.

### TARGETING VASCULAR ENDOTHELIAL GROWTH FACTOR (VEGF)

Angiogenesis is mediated by the interaction between VEGFs and their tyrosine kinase receptors, VEGFRs. This mechanism can be targeted by monoclonal antibodies as well as by small molecule tyrosine kinase inhibitors. Inhibition of this pathway can be achieved at different levels using various mechanisms; with monoclonal antibodies to VEGFA or its receptor, with recombinant fusion protein to VEGF and with various multitargeted tyrosine kinase inhibitors (TKIs). About half of gastric cancers overexpress VEGF and this is associated with a poor prognosis (22, 23).

As a hallmark of cancer, angiogenesis logically stands out as a potential target (24, 25). The hypothesis that malignant tumor growth is dependent upon angiogenesis has been demonstrated in multiple studies and this pathway has been successfully exploited across many tumor types (26, 27). Bevacizumab, aflibercept, ramucirumab, and regorafenib are all FDA approved for use in metastatic colorectal cancer. In addition, bevacizumab is approved for multiple tumor types, including non-small cell lung cancer, ovarian cancer, metastatic renal cell carcinoma (RCC), and glioblastoma. The TKIs have various indications, including the treatment of metastatic RCC, hepatocellular cancer, medullary thyroid cancer, and sarcoma.

The REGARD and RAINBOW studies, published in 2014, led to the approval of ramucirumab in the second line setting in OG cancer (28). Whilst other VEGF targeting drugs (namely bevacizumab and aflibercept) have been investigated in this field, ramucirumab is the only targeted therapy to have FDA and EMA approval in the advanced setting after chemotherapy. There are a number of studies looking in to new combinations of VEGF targeting with both conventional chemotherapy and with immunotherapy. As immunotherapy is likely to be licensed for pre-treated OG cancer this highlights an important area of innovation in this disease. This review aims to outline targeting VEGF in OG cancer, the rationale behind the continued interest in this mechanism and possible future directions in combination with immunotherapy.

#### RAMUCIRUMAB

Ramucirumab is a fully humanized monoclonal antibody to VEGFR-2, a subtype of the VEGFR which is thought to mediate all known vascular endothelial responses to VEGF (29).

REGARD (30), a randomized phase III placebo controlled trial, investigated ramucirumab in patients who had progressed after first line chemotherapy. One hundred and seventeen patients were randomized to placebo plus BSC and 238 were randomized to ramucirumab plus BSC. Whilst response rates were only 4% with ramucirumab, the rate of stable disease in the treatment arm was 45%, compared with 21% with placebo, giving a disease control rate (DCR) of 45 vs. 21%. Ramucirumab monotherapy increased mOS from 3.8 to 5.2 months [Hazard Ration (HR) 0.776, 95 % CI 0.603–0.998, p = 0.047] and median progression free survival (mPFS) from 1.3 to 2.1 months (HR 0.483, 95% CI 0.376–0.620, p < 0.0001; **Table 1**). The treatment was well-tolerated. As expected, rates of hypertension were higher in the treatment than the placebo group (16 vs. 8%) but otherwise there were similar rates of adverse events. This represents a potential treatment option for those patients who are keen to avoid the toxicity of chemotherapy.

RAINBOW (31), another randomized phase III placebo controlled trial, subsequently investigated combining ramucirumab with paclitaxel in patients with advanced OG adenocarcinoma who had disease progression on or within 4 months of first line chemotherapy. In the study, 335 patients were randomized to paclitaxel with placebo, 330 patients to paclitaxel with ramucirumab. The results demonstrated a significant increase in OS with the combination of ramucirumab with paclitaxel of 9.6 vs. 7.4 months (HR 0.807, 95% CI 0.678–0.962, p = 0.017; **Table 1**). PFS was improved to 4.4 vs. 2.9 months with placebo (HR 0.635, 95% CI 0.536–0.752, p < 0.0001). The objective response rate (ORR) was also improved to 28% (vs. 16% with placebo) and the disease control rate was 80% (vs. 64% with placebo). The study reported higher rates of grade 3 or 4 toxicity in the group treated with the combination, although this did not result in higher rates of treatment-related mortality which was 2% in both groups.

In light of the known geographical differences in OG cancer outcomes, the data for RAINBOW was analyzed for Asian and Non-Asian patients as two cohorts. Whilst OS was not significantly improved for Asian patients, the mPFS was. The HRs for OS were 0.73 and 0.99 for non-Asian and Asian patients, respectively and the HRs for PFS were 0.64 and 0.63 for non-Asian and Asian patients. It has been suggested that these differences may be as a result of higher use of third line treatment in Asian populations (almost 70 vs. almost 40%).

FIGURE 1 | Median Overall Survival in Advanced OG Adenocarcinoma with selected first line therapy. Cisplatin + Capecitabine + Trastuzumab (5); Epirubicin + Oxaliplatin + Capecitabine (6); Epirubicin + Cisplatin + Capecitabine (6); mFOLFOX (7); FOLFIRI (8); Capecitabine + Cisplatin (9); Docetaxel + Cisplatin + Fluorouracil (10); Epirubicin + Oxaliplatin + Fluorouracil (6); Epirubicin + Cisplatin + Fluorouracil (6); Irinotecan + Fluorouracil (11); Cisplatin + Fluorouracil (10); Best Supportive Care (12).



A subsequent subgroup analysis of the safety and efficacy of ramucirumab in Japanese and Western patients in RAINBOW (13, 35) noted safety profiles of the ramucirumab plus paclitaxel arm were similar between populations, though there was a higher incidence of grade 3 neutropenia in Japanese patients (66.2 vs. 25.4%). The analysis also reported improved PFS, ORR and 6-month survival rates in the Japanese population compared with the Western population (**Table 2**). Again post discontinuation therapy rates were much higher in the Japanese than in the Western patients (75 vs. 37%) and it was postulated that this masked any OS benefit.



Following the publication of the REGARD and RAINBOW studies, the FDA granted approval for single agent ramucirumab for the treatment of advanced OG adenocarcinoma that had progressed following 1st line therapy in 2014. Later that year, the FDA then approved the use of ramucirumab combined with paclitaxel to treat advanced OG adenocarcinoma following failure of first line therapy. Ramucirumab has since also been approved by the EMA in both indications although local reimbursement for ramucirumab is variable across different countries.

These studies have validated targeting the VEGF pathway in the 2nd line setting. In the 1st line setting, results have been less encouraging. In 2016, a randomized phase II study failed to show a benefit when adding ramucirumab to FOLFOX in the first line setting for advanced OG adenocarcinoma (36). The multicentre study, involving 168 patients, failed to meet its primary end point of improving PFS [6.4 vs. 6.7 months, HR 0.98 (95% confidence interval 0.69–1.37)]. Objective response rates were also similar between both arms (45.2 vs. 46.4%). The investigators felt that the difference in outcome between this study and REGARD and RAINBOW was likely multifactorial. Firstly, it was postulated that disease biology may be different in 1st and 2nd line settings. Secondly, they noted a higher rate of discontinuation in those treated with ramucirumab and FOLFOX compared to FOLFOX alone. Thirdly, it was noted that this study had a higher proportion of esophageal rather than junctional or gastric tumors than REGARD and RAINBOW and a preplanned subgroup analysis indicated some benefit in gastric and junctional tumors over esophageal tumors. In gastric, junctional and cardia tumors, mPFS was 8.7 months for the ramucirumab arm vs. 7.1 months in the placebo arm (HR = 0.77) compared to patients with a primary esophageal tumor where mPFS was 5.6 vs. 6.1 months (HR = 1.30).

RAINFALL (NCT02314117), a global phase III study, included 616 patients with advanced gastric, or GOJ adenocarcinoma with tumors of the esophagus excluded. The patients were randomized to either first line treatment with fluoropyrimidine and cisplatin (CX) alone or to CX plus ramucirumab. The study completed in December 2017. The findings revealed a statistically significant 25% reduction in the risk of disease progression or death for the primary endpoint of PFS. PFS was 5.7 months in the intervention arm vs. 5.4 months in the placebo arm (HR 0.75, 95% CI 0.61–0.94, p = 0.011). There was no difference in mOS between the ramucirumab and placebo arms (11.17 vs. 10.74 months; HR 0.96, 95% CI 0.80–1.16; p = 0.68). There was also no significant difference between ramucirumab and placebo in the ORR (41 vs. 36%; p = 0.17) or the DCR (82 vs. 77%; p = 0.10) (32). Based on these findings, ramucirumab will not play a role in front line therapy in unselected patients.

The role of ramucirumab in the maintenance setting is currently being explored in the PLATFORM study (NCT02678182). This study will recruit 770 patients to evaluate the efficacy of various maintenance therapies following completion of standard first-line chemotherapy in patients with locally advanced/metastatic HER-2 positive/HER-2 negative OG adenocarcinomas. One of the arms will investigate maintenance capecitabine in combination with ramucirumab.

### BIOMARKERS AND RAMUCIRUMAB

Despite the use of anti-angiogenics across multiple indications in cancer there are as yet no robust predictive biomarkers to guide patient selection. Using samples from the REGARD and RAINBOW studies amongst others, attempts have been made to find a predictive biomarker for ramucirumab. Tumor biomarkers such as VEGFR2 and HER2 expression were studied but were not statistically significantly associated with ramucirumab efficacy (37). Serum markers studied include VEGF-C and -D, soluble VEGFR1, 2 and 3, angiopoietin-2, platelet derived growth factor but again baseline levels were not associated with ramucirumab efficacy (37, 38).

Studies have also been conducted in Korean and Japanese patients specifically, investigating trends in potential biomarkers as well as baseline levels. In the Korean study tissue molecular characteristics [Epstein Barr Virus (EBV), Mismatch Repair (MMR), HER2, epidermal growth factor receptor-1 (EGFR-1), hepatocyte growth factor receptor (C-MET) etc.] and circulating biomarkers [VEGF, sVEGFR2, Hepatocyte growth factor (HGF), neuropillin-1, IL-8, and placental growth factor (PIGF)] were assessed. A higher disease control rate with ramucirumab was found in patients with high EGFR expression tumors (2+/3+) compared with low expression tumors (0/1+) (87.5 vs. 50%, p = 0.02). A longer PFS was seen in patients with higher level of pre-treatment circulating VEGFR2 (4.1 vs. 2.3 months; p = 0.01) and lower level of pre-treatment serum neuropillin-1 (4.1 vs. 2.4 months; p = 0.02) (39). The Japanese study focused on dynamic changes in circulating biomarkers. Lower than median Day8/baseline ratios of VEGF-A were significantly associated with a longer PFS (6.3 vs. 2.4 months; p = 0.004) and patients with early disease progression had higher Day8/baseline ratios of VEGF-C, Angiopoietin 1, and lower baseline NRP1 levels (40). Both of these studies were small (n = 55 and 25, respectively) and these findings require validation but suggest a predictive biomarker for ramucirumab may yet be found.

Evaluation of predictive biomarkers with the use of ramucirumab has also been evaluated in other tumor types. The RAISE study investigated the use of ramucirumab or placebo in combination with FOLFIRI in second line metastatic colorectal cancer and found a significant improvement in Butters et al. Targeting Angiogenesis in OG Cancer

OS and PFS with the use of ramucirumab. A subsequent biomarker analysis identified VEGF-D as a potential marker, noting improved median OS in those patients with high levels of VEGF-D compared to low levels and investigators are currently developing an assay for further testing in clinical practice (41).

#### BEVACIZUMAB

Bevacizumab is a humanized monoclonal antibody to VEGF-A. This has not demonstrated the same benefit as ramucirumab in OG cancer despite encouraging phase II studies and a proven role in other tumor types. Bevacizumab has been investigated both in the advanced setting and in the peri-operative setting with 2 phase III studies of bevacizumab in the advanced setting, AVAGAST and AVATAR, and one in the peri-operative setting, ST03.

The STO3 study (42) was a multicentre randomized phase II/III trial investigating the addition of bevacizumab to conventional perioperative chemotherapy. Five hundred and thirty three patients received chemotherapy alone and 530 patients received chemotherapy plus bevacizumab. Three-years overall survival was 50% (95% CI 45.5–54.9) in the chemotherapy alone group and 48% (43.2–52.7) in the chemotherapy plus bevacizumab group (HR 1.08, 95% CI 0.91–1.29; p = 0.36). With bevacizumab there were increased rates of wound healing complications (12 vs. 7%) and anastamotic leaks in patients who underwent oesophagogastrectomy (24 vs. 10%). The investigators suggested that bevacizumab may have a prolonged effect that delays wound healing. The results of this trial do not support the use of bevacizumab with chemotherapy in the peri-operative setting in unselected patients.

AVAGAST (33) was a global phase III trial to investigate the addition of bevacizumab to 1st line therapy for advanced gastric cancer. Three hundred and eighty seven patients were randomized to doublet therapy of cisplatin with fluoropyrimidine therapy (FC) and 387 patients received FC plus bevacizumab (total 774 patients). Whilst the study did not meet its primary end point with no significant improvement in median OS, 12.1 months with bevacizumab plus FC and 10.1 months with FC alone (HR 0.87, 95% CI 0.73–1.03, p = 0.1) there was a trend toward improved survival with bevacizumab. Further, there was a significant improvement in PFS (6.7 vs. 5.3 months, HR 0.80, 95% CI 0.68–0.93, p = 0.004) and ORR (46 vs. 37.4%, p = 0.0315; **Table 1**).

Subgroup analysis of AVAGAST revealed that the effect of the addition of bevacizumab varied with geographical location. OS was improved in the pan-America population in comparison to the European and Asian populations (**Table 3**). The reason for the variability in OS is not clear the authors suggested that it may be as a result of differences in the burden of disease (Asian patients having fewer liver metastases and fewer GOJ tumors) or different patterns of treatment (Asian patients more commonly receive second and further lines of therapy).

A pre-planned biomarker analysis following AVAGAST also shed some light on biological differences between the patient


populations (43). Markers evaluated included plasma VEGF-A and tumor expression of VEGF-A, VEGFR-1 and−2, neuropilin-1, EGFR-1 and HER2. Plasma VEGF-A levels were higher at baseline in the non-Asian patients whereas neuropilin-1 levels were higher in the Asia-Pacific patients. Both had potential prognostic effects, with high baseline plasma VEGF-A and low tumor neuropilin-1 being associated with worse outcomes. Further, high baseline plasma VEGF-A levels and low tumor neurophilin-1 expression were identified as potential predictive biomarker candidates for bevacizumab efficacy in non-Asian patients although further studies are required to confirm this role. Plasma levels of Angiopoietin-2 (Ang-2) have also been studied in this cohort and again a differential expression was noted between Asian and non-Asian patients. Ang-2 was also associated with a worse OS and the presence of liver metastases but was not predictive for response to bevacizumab and these findings require further validation (44).

As the AVAGAST study only included 12 Chinese patients the AVATAR study was conducted to establish if the geographical effects demonstrated in AVAGAST held true in this population. The study recruited 202 patients who were randomized to capecitabine and cisplatin in combination with bevacizumab or placebo. Again there was no improvement in mOS with bevacizumab and here PFS was similar in both treatment arms (45).

Based on these results it is difficult to see a role for bevacizumab in OG cancer at present, although there are currently phase 1 trials investigating bevacizumab in combination with the anti PDL-1 monocolonal antibody atezolizumab in solid tumors, including esophageal, and gastric cancers (NCT02715531, NCT01633970). Biomarkers of response or resistance remain elusive but bevacizumab's role may be revisited should a robust biomarker be found.

### AFLIBERCEPT

Aflibercept is a recombinant fusion protein consisting of human VEGF receptor domains fused with the Fc portion of human Immunoglobulin G (IgG). It binds with circulating VEGF, preventing it from interacting with the VEGFR on endothelial cells and has a high affinity for VEGF-A, VEGF-B and PIGF subtypes. Aflibercept has demonstrated some efficacy in a range of tumor types and is now approved for use in metastatic colorectal cancer in combination with FOLFIRI in the second line setting (46, 47). However, a multicentre randomized phase II trial comparing FOLFOX with either placebo or aflibercept in patients with chemotherapy-naïve metastatic OG adenocarcinoma did not meet its primary endpoint of improved PFS at 6 months (48). Only 64 patients were enrolled and 6 month PFS was found to be 60.5% in the aflibercept arm, compared to 57.1% in the placebo arm (p = 0.8) and median PFS was 9.9 vs. 7.3 months, respectively (p = 0.69). There are no further on-going studies of aflibercept in OG cancer and it appears unlikely that this drug will be developed further in this setting.

### MULTI-TARGETED TYROSINE KINASE INHIBITORS (TKIs)

Multi-targeted TKIs inhibit angiogenesis via the VEGF pathway and have demonstrated benefit in other tumor types, including GIST, RCC, and NSCLC. Sunitinib, sorafenib, pazopanib, regorafenib, and apatinib have all been investigated in the context of OG cancer either as single agents or in combination with chemotherapy The majority of these studies have been disappointing, with apatinib and regorafenib being the notable exceptions (**Table 4**).

As outlined in **Table 4**, the only TKI to be investigated in phase III trials is apatinib, where an Asian study of 267 patients with advanced gastric or GOJ adenocarcinoma demonstrated a significantly improved median OS with apatinib compared with placebo. It also noted that in this heavily pre-treated population the drug was well-tolerated with an acceptable safety profile (34). Grade 3 to 4 events occurred more frequently in the treatment arm (8.5 vs. 0%) and included hypertension, proteinuria, and neutropenia.

There are three ongoing phase III trials further investigating apatinib in the treatment of advanced OG cancer. Firstly, as the aforementioned phase III trial (34) was in an Asian population, the ANGEL study (NCT03042611), a phase III double blind randomized controlled trial, is investigating apatinib vs. placebo in patients with advanced gastric cancer in Asian as well as European and North American populations. The second is investigating apatinib as maintenance therapy after 1st line chemotherapy (NCT02537171), given that it has been shown to be well-tolerated. The third is investigating the use of apatinib in combination with XELOX chemotherapy as adjuvant treatment for resected gastric cancer (NCT03355612). Given the previous results of bevacizumab in the adjuvant setting in the STO3 study, it will be interesting to see how targeting the VEGF receptor using a different approach fares here.

### FUTURE COMBINATIONS WITH IMMUNOTHERAPY

Targeting the VEFG pathway in OG adenocarcinoma, through various mechanisms, has been well-investigated with both positive and negative studies as discussed. Given its proven role in the second line setting, there remains a considerable interest in the further development of ramucirumab in OG adenocarcinoma and there are a number studies ongoing. Perhaps the most topical area of study is the combination of ramucirumab and other anti-angiogenics with immunotherapy.

### IMMUNE CHECKPOINT BLOCKADE IN OG CANCER

The use of immune checkpoint inhibitors has brought about a paradigm shift in the treatment of a number of solid tumors and multiple trials of checkpoint inhibitors and other novel drugs targeting various aspects of the immune system are underway in OG cancer. Nivolumab and pembrolizumab have both demonstrated activity in OG cancer when used alone (53, 54), as detailed below. Current studies investigating their combination with ramucirumab are underway. Other immune checkpoint inhibitors investigated in oesophagogastric cancer include avelumab, atezolizumab, durvalumab, and tremelimumab. It is beyond the scope of this article to describe all of these studies but they have recently been well-reviewed by Taieb et al. (55).

Nivolumab, a PD-1 inhibitor, demonstrated a statistically significant improvement in OS in the treatment of advanced chemo-refractory gastric and GOJ cancer in the large phase III ATTRACTION-2 trial involving 493 Asian patients (53). OS was 5.26 months (95% CI 4.60–6.37) in the nivolumab arm and 4.14 months (3.42–4.86) in the placebo arm (hazard ratio 0.63, 95% CI 0.51–0.78; p < 0.0001). This study led to a license being granted by the Japanese Ministry of Health, with other regulatory authorities currently reviewing the data.

Nivolumab has also been studied in Western patients in the phase I/II CheckMate-032 study (NCT01928394). Nivolumab was investigated as a single agent and in combination with ipilimumab, a monoclonal antibody against CTLA-4, in the first line setting in patients unselected for PD-L1 status (n = 160). Here OS was 6.2 months (95% CI 3.4, 12.4) with nivolumab alone, 6.9 months (95% CI 3.7, 11.5) with nivolumab 1 mg/kg and ipilimumab 3 mg/kg and 4.8 months (95% CI 3.0, 8.4) with nivolumab 3 mg/kg and ipilimumab 1 mg/kg. Using a cut-off of more than or equal to 1% staining for PD-L1 as positive, OS was unchanged in the nivolumab monotherapy cohort and slightly increased in the nivolumab plus ipilimumab cohorts in this group (56). Additional studies of Nivolumab with or without ipilimumab in oesophagogastric cancer are on-going (e.g., NCT02872116, NCT03044613).

Pembrolizumab is another PD-1 inhibitor which has demonstrated activity in gastric and GOJ adenocarcinoma. The KEYNOTE-012 (NCT01848834) phase I trial (n = 39) investigated the efficacy of pembrolizumab in patients with advanced solid tumors, including recurrent or metastatic PD-L1 positive gastric cancer (∼40% of all gastric cancers). A 22.1% ORR was observed, with 6 month PFS and OS being 24 and 69%, respectively. The authors noted that pembrolizumab demonstrated manageable toxicity and promising anti-tumor activity in this setting (54). Five (13%) patients had grade 3/4 treatment-related adverse events with no treatment related deaths. There were two cases of grade 3 fatigue, one case each of grade 3 pemphigoid, grade 3 hypothyroidism, and



TABLE 4 | Selected studies of TKIs in gastric cancer.

grade 3 peripheral sensory neuropathy, and one case of grade 4 pneumonitis.

The subsequent KEYNOTE-059 study (57), a global phase II open-label study, recruited 259 patients with advanced gastric or GOJ cancer who had previously received at least 2 lines of treatment, unselected for PD-L1 status. Single agent pembrolizumab demonstrated promising activity with an ORR of 11.6% in all patients (95% CI, 8.0–16.1%; 30 of 259 patients). The duration of response in these heavily pre-treated patients varied from 1.6 to 17.3+ months (median 8.4 months). Both ORR and duration of response were higher in the PD-L1 positive patients (15.5 vs. 6.4% and 16.3 and 6.9 months, respectively) as was OS, at 5.8 months (95% CI, 4.5–7.9) vs. 4.9 (95% CI, 3.4–6.5) months. Just fewer than 20% patients experienced 1 or more grade 3–5 treatment-related adverse events with 2 patient deaths attributed to treatment.

Based on the KEYNOTE-059 results, the FDA granted accelerated approval to pembrolizumab in Sept 2017 for patients with recurrent locally advanced or metastatic, gastric or GOJ adenocarcinoma whose tumors express PD-L1 as determined by an FDA-approved test. Patients needed to have had disease progression on or after two or more prior specified systemic therapies. This extended the existing tumor agnostic license for pembrolizumab in patients with unresectable or metastatic, microsatellite-instability–high or mismatch-repair– deficient solid tumors, which would apply to ∼4–5% gastric tumors. This decision was made as in KEYNOTE-059 55% patients (n = 143) had PD-L1 positive tumors and either microsatellite stable (MSS), or undetermined microsatellite instability (MSI) or mismatch repair (MMR) status. In this group the ORR was 13.3% (95% CI: 8.2, 20.0) with over 50% having a response lasting over 6 months and ∼25% having a response lasting over a year and these patients would have been ineligible for treatment with the existing license (58).

However, the KEYNOTE-061 study (59) (n = 592) has just reported that pembrolizumab did not significantly improve OS in the second line setting for patients with PD-L1 positive oesophagogastric cancer when compared to paclitaxel. Median overall survival was 9·1 months (95% CI 6.2–10.7) with pembrolizumab and 8.3 months (7.6–9.0) with paclitaxel (HR 0.82, 95% CI 0.66–1.03; one-sided p = 0.0421) but responses were more durable in the pembrolizumab group than in the paclitaxel group, with a median response duration of 18.0 months (95% CI 8.3–not estimable) vs. 5.2 months (3.2–15.3) and pembrolizumab had a better safety profile than paclitaxel. Additional studies are underway looking at the combination of pembrolizumab with various agents in this disease (e.g., NCT02494583, NCT03382600).

As discussed above, to date a number of studies of checkpoint blockade have provided promising results of activity in OG cancer, although additional randomized trials against chemotherapy are required. Further rationally designed combination studies are also needed to try to maximize the benefit of this approach in appropriately selected patients. The combination of checkpoint blockade and VEGF inhibition is one such option.

#### CHECKPOINT BLOCKADE IN COMBINATION WITH VEGF INHIBITION

There is increasing pre-clinical evidence to support VEGF inhibition and immunotherapy as a viable combination strategy. Inhibiting the VEGF pathway may improve the efficacy of checkpoint blockade through both direct effects on the vasculature and through inhibiting VEGF's immunosuppressive functions (**Figure 2**). There may also be a reciprocal positive impact on the efficacy of anti-angiogenics by vascular changes brought about by immunotherapy.

As reported across multiple tumor types, those patients who respond well to immunotherapy often have an immunologically "hot" tumor containing multiple tumor infiltrating lymphocytes (TILS), whilst those patients with fewer TILs, "cold" tumors, or those with TILS restricted to the margin of the tumor microenvironment (TME), "excluded" tumors, tend to have a lesser response (60–62). For TILs to enter the TME angiogenesis is required to provide blood vessels to deliver them. Cancers are associated with dysregulated angiogenesis, tortuous abnormal blood supplies and resulting hypoxia, high interstitial fluid pressures, and an acid pH (63, 64). Such a hypoxic TME is associated with the recruitment of regulatory T-cells, tumor associated macrophages switching to their immunosuppressive M2 phenotype, a direct inhibition of effector T cells and an accumulation of immunosuppressive metabolites (65).

It may be possible to use anti-angiogenic drugs to normalize tumor vasculature and thereby alleviate this immunosuppressive hypoxia. However, as anti-angiogenic drugs may also destroy blood vessels within tumors rather than normalizing them, this approach will have to be carefully conceived. From animal studies it appears that the effect of anti-VEGF therapy on the vasculature may be dose dependent and as such lower "vascular-normalizing" doses may be required rather than the treatment doses with which we are familiar (66).

In addition to promoting an immunosupportive TME through vascular normalization, anti-VEGF therapy may also reduce direct immunosuppression caused by VEGF. VEGF has various immunosuppressive functions on dendritic cells (DCs), effector T cells, regulatory T cells, myeloid derived suppressor cells (MDSCs) and in facilitating immune evasion through the induction of FAS antigen ligand in endothelial cells and resulting in a barrier to infiltrating CD8 +ve T cells (**Table 5**) (76). Blockade of VEGF signaling has been shown to reverse these systemic immunosuppressive effects in animal models (65).

Preclinically, whilst studies in animal models of OG cancer do not exist, a synergistic effect of VEGF inhibition in combination with immunotherapy has been demonstrated in a number of other tumor types. For example a murine study using Colon-26 adenocarcinoma demonstrated that simultaneous treatment with anti-PD-1 and anti-VEGFR2 monoclonal antibodies resulted in


a synergistically increased inhibition of tumor growth compared with either therapy alone without excess toxicity (77). Using a different immunotherapy approach, adoptive T cell transfer, in a mouse model of melanoma the addition of anti-VEGF therapy resulted in significantly increased anti-tumor activity when compared to the immunotherapy alone (78).

A set of experiments with murine models of breast cancer, pancreatic neuroendocrine carcinoma and glioma demonstrated that anti-PD-L1 therapy can sensitize tumors to anti-angiogenic therapy and prolong its efficacy. Further, the experiments also showed the converse, that anti-angiogenic therapy can improve anti-PD-L1 treatment by generating intratumoural high endothelial venules (HEVs) that facilitate enhanced CTL infiltration, activity, and tumor cell destruction in the breast and neuroendocrine but not the glioma models (79).

This combination approach has now been taken forward into clinical trials investigating the use of various antiangiogenics with immunotherapeutic approaches including checkpoint blockade, vaccination and cell therapies (80). A phase I study of ipilimumab and bevacizumab in patients with advanced melanoma reported a disease control rate of 67% with 24% patients experiencing grade 3/4 toxicity. Tumor biopsies revealed intense infiltration by CD8+ T cells and DCs within the tumor vasculature, with less infiltration seen in those patients treated with ipilimumab alone (81). In colorectal cancer the PD-L1 inhibitor atezolizumab has been investigated in combination with bevacizumab and chemotherapy in a phase Ib study with no unexpected toxicities and a positive signal of activity (82). More advanced studies have reported for RCC and lung cancer.

In RCC a phase II study of bevacizumab and atezolizumab reported encouraging activity in the first line setting in PD-L1 positive patients (83) and the subsequent phase III IMmotion151 study (NCT02420821) study. This study randomized 915 patients with advanced untreated RCC to either a combination of atezolizumab and bevacizumab or sunitinib monotherapy and patients were stratified according to PD-L1 status. The study demonstrated an improved PFS for the combination arm vs. sunitinib in both the intention to treat population [11.2 (95% CI 9.6, 13.3) vs. 8.4 (95% CI 7.5, 9.7) months, HR 0.83, p = 0.0219], and the PD-L1 positive population [11.2 (95% CI 8.9, 15) vs. 7.7 (95% CI 6.8, 9.7) months, HR 0.74, p = 0.0217]. The combination arm was well-tolerated with a safety profile in keeping with the individual drugs and quality of life was improved, measured as an increased time to interference with activities of daily living in the atezolizumab and bevacizumab combination (11.3 vs. 4.3 months, HR 0.56, 95% CI 0.46, 0.68) (84). Overall survival data were immature and are awaited but this study provides early support for this approach in RCC.

The Phase III IMpower150 study (NCT02366143) assessed the combination of atezolizumab plus carboplatin and paclitaxel with or without bevacizumab vs. carboplatin, paclitaxel, and bevacizumab in patients with advanced Non-Squamous Non-Small Cell Lung Cancer (NSCLC) (85). The addition of atezolizumab to the triplet of carboplatin, paclitaxel and bevacizumab improved OS in the wild type genotype cohort (n = 692) from 14.7 to 19.2 months [HR 0.78 (95% CI 0.64, 0.96)]. Grade 3/4 treatment-related adverse events occurred in 55.7% patients with the addition of atezolizumab vs. 47.7% without and were consistent with known toxicity for the drugs involved.

In OG cancer there are several ongoing clinical trials investigating the combination of immune-checkpoint inhibitors with anti-angiogenic therapy (**Table 6**).

The JVDF study is a multicentre phase I study of ramucirumab plus pembrolizumab in patients with advanced gastric or GOJ adenocarcinoma, NSCLC, TCC of the urothelium or biliary tract cancer. The trial is split in to 2 phases, the 1st phase determining the safety and tolerability of treatment and the second phase assessing the efficacy of treatment in cohorts of each tumor type. The study is ongoing, however preliminary data from the cohort of patients with advanced gastric or gastro-esophageal junction adenocarcinoma has been presented (86, 87). As of July 2017, 28 treatment naïve OG adenocarcinoma patients had been treated in this study and 68% were PD-L1 positive, assessed by DAKO PD-L1 22C3 IHC pharmDx assay with staining of ≥1% being positive. Treatment-related adverse events occurred in 96% of patients; with 61% experiencing grade 3 adverse events, most commonly hypertension (14%) and diarrhea (11%). No grade 4–5 treatment related events occurred. An objective response was demonstrated in 25% (7/28) of patients with 6 of those responding being positive and 1 negative for PD-L1 expression. The disease control rate was 68%, mPFS 5.3 months (95% CI 3.2–11) and median duration of response was 10 months (95% CI 9.7–10.3). Median OS has not yet been reached (87). These results suggest encouraging activity for the combination in this setting. Activity has also been demonstrated in the second or subsequent line setting in GOJ cancer in another cohort of the JVDF study with a DCR of 46% and a 6 month OS of 51.2% (95% CI, 33.9–66.1) (88).

The phase I clinical trial (NCT02572687) investigating durvalumab, another PD-1 inhibitor, with ramucirumab has also recently had interim results presented. This study enrolled patients with advanced OG adenocarcinoma who had progressed on 1 or 2 lines of systemic therapy. As of May 2017, there were 29 patients in this cohort of whom 48% had PD-L1 ≥25% expression in tumor or immune cells and 3.5% were MSIhigh. Seventy two percentage of patients experienced grade 3–4 treatment adverse events. Treatment related adverse events of any grade occurring in over 10% of patients were as expected



and included hypertension (34%), fatigue (31%), headache (24%), diarrhea (21%). In this interim analysis 17% of patients achieved a confirmed partial response, including 1 MSI-high patient. For patients with a PD-L1 expression of over 25% the overall response rate was 36%. Progression free survival was 2.6 months (95% CI, 1.45 to 6.28) (89). The final results of this study and JVDF, as well as those detailed in **Table 6**, are awaited.

#### CONCLUSION

Targeting angiogenesis through the VEGF pathway has been demonstrated to be a viable approach in OG cancer using two different methods, in the form of a monoclonal antibody with ramucirumab and a TKI, apatinib. However, these treatments provide a limited benefit for unselected patients and combination strategies and robust predictive biomarkers are required.

Immune checkpoint blockade has also demonstrated activity in this disease but only for a limited number of patients. Again robust predictive biomarkers are needed as well as methods to convert immunologically "cold or excluded tumors" to "hot tumors" to allow more patients to benefit from this approach.

Combination therapy with anti-angiogenics and immunotherapy may theoretically solve some of these problems, and the scientific rationale is compelling, but a number of hurdles remain. The dose and scheduling of the anti-angiogenic therapy will require careful consideration to ensure the optimum reduction in immunosuppression with vascular normalization, without risking worsening hypoxia or excessive toxicity. A sequencing approach may also be considered.

Biomarkers are required to enable selection of the patients who may respond to each drug and also to inform clinicians as to when optimum vascular normalization has occurred. As discussed biomarkers for anti-angiogenic therapy remain elusive. For immune checkpoint blockade there are multiple biomarkers under investigation, including PD-L1 for which the optimum assay, cut-off, staining pattern, and significance are yet to be established for OG cancer. Other features such as tumor mutational load, microsatellite instability, and an Interferonγ-related mRNA profile have also been suggested as putative predictive biomarkers for immunotherapy but again additional work is required here (90–93).

Further clarification of the biological differences underlying the geographical variability in response to treatment with antiangiogenics is also needed to ensure rational drug combinations in different populations. Increased understanding of the microbiome in OG cancer may play a role here, both in understanding geographical differences in treatment response as well as in explaining individual variations in response to immunotherapy. Finally, strategies to overcome resistance, which inevitably develops with targeted therapies and may develop with checkpoint inhibition over time, will be required.

As we further elucidate the role of VEGF and other angiogenic pathways, alongside the immunobiology of OG cancer, it is highly possible that these hurdles will be overcome in this rapidly evolving field and such combinations may become part of the treatment paradigm for this disease in the future.

#### ETHICS STATEMENT

This article does not contain any studies with human or animal subjects performed by any of the authors.

#### AUTHOR CONTRIBUTIONS

OB and KY contributed equally to this work and wrote the first draft of the manuscript. DC, IC, and NS contributed to manuscript revision, read, and approved the submitted version.

#### ACKNOWLEDGMENTS

The authors acknowledge support from the National Institute for Health Research Royal Marsden/Institute of Cancer Research Biomedical Research Center.

#### REFERENCES


metastatic colorectal cancer which has progressed following prior oxaliplatinbased chemotherapy: a critique of. Pharmacoeconomics. (2015) 33:457– 66. doi: 10.1007/s40273-015-0257-z


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Butters, Young, Cunningham, Chau and Starling. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Crosstalk Between Long Non-coding RNAs, Micro-RNAs and mRNAs: Deciphering Molecular Mechanisms of Master Regulators in Cancer

Eduardo López-Urrutia<sup>1</sup> , Lilia P. Bustamante Montes <sup>2</sup> , Diego Ladrón de Guevara Cervantes <sup>2</sup> , Carlos Pérez-Plasencia1,3 and Alma D. Campos-Parra<sup>3</sup> \*

<sup>1</sup> Unidad de Biomedicina, FES-IZTACALA, Universidad Nacional Autónoma de México, Tlalnepantla de Baz, Mexico, <sup>2</sup> Decanato, Ciencias de la Salud, Universidad Autónoma de Guadalajara, Zapopan, Mexico, <sup>3</sup> Laboratorio de Genómica, Instituto Nacional de Cancerología (INCan), Mexico City, Mexico

#### Edited by:

Laurence A. Marchat, National Polytechnic Institute, Mexico

#### Reviewed by:

Marco Ragusa, University of Catania, Italy Lihua Wang, Second Affiliated Hospital of Harbin Medical University, China

\*Correspondence:

Alma D. Campos-Parra adcamposparra@gmail.com

#### Specialty section:

This article was submitted to Cancer Molecular Targets and Therapeutics, a section of the journal Frontiers in Oncology

> Received: 24 May 2019 Accepted: 09 July 2019 Published: 25 July 2019

#### Citation:

López-Urrutia E, Bustamante Montes LP, Ladrón de Guevara Cervantes D, Pérez-Plasencia C and Campos-Parra AD (2019) Crosstalk Between Long Non-coding RNAs, Micro-RNAs and mRNAs: Deciphering Molecular Mechanisms of Master Regulators in Cancer. Front. Oncol. 9:669. doi: 10.3389/fonc.2019.00669 Cancer is a complex disease, and its study requires deep understanding of several biological processes and their regulation. It is an accepted fact that non-coding RNAs are vital components of the regulation and cross-talk among cancer-related signaling pathways that favor tumor aggressiveness and metastasis, such as neovascularization, angiogenesis, and vasculogenic mimicry. Both long non-coding RNAs (lncRNAs) and micro-RNAs (miRNAs) have been described as master regulators of cancer on their own; yet there is accumulating evidence that, besides regulating mRNA expression through independent mechanisms, these classes of non-coding RNAs interact with each other directly, fine-tuning the effects of their regulation. While still relatively scant, research on the lncRNA-miRNA-mRNA axis regulation is growing at a fast rate, it is only in the last 5 years, that lncRNA-miRNA interactions have been identified in tumor-related vascular processes. In this review, we summarize the current progress of research on the cross-talk between lncRNAs and miRNAs in the regulation of neovascularization, angiogenesis and vasculogenic mimicry.

Keywords: cancer, long non-conding RNAs, miRNAs, mRNAs, angiogenesis, vascularization, vasculogenic mimecry

## INTRODUCTION

Cancer is a serious worldwide health problem that affects the health of all human cultures. Prostate and breast cancer rank as top prevalent cancer types in men and women, respectively (1). Cancer has been defined as a complex, heterogeneous, and multifactorial disease that occurs by the presence of driver mutations that leads to the activation of proto-oncogenes and corresponding inactivation of tumor suppressors. This provokes a switch in cell functions that ultimately leads to the hallmarks of cancer (2).

In addition to mutations in protein-coding genes, recent advances in molecular oncology have described the aberrant expression of non-conding RNAs such as micro-RNAs (miRNAs) and longnon-coding RNAs (lncRNAs) (3, 4). Both molecules are well-established as master regulators of multiple protein-coding genes (5). Among other functions, lncRNAs can act as molecular decoys, sequestering miRNAs, and consequently, inhibiting their interaction with their target messenger RNAs (mRNA) (6, 7). This way, lncRNAs regulate a wide range of biological processes through their crosstalk with miRNAs that, in turn, regulate mRNAs (8). Since these crosstalking molecules

**36**

are so closely related, abnormal expression of lncRNAs interferes with mRNA expression patterns creating a dysregulation that can culminate in cancer development (9). In the present review, we summarize the recent studies on the lncRNA-miRNA-mRNA crosstalk in order to provide insight into the complexity of the molecular mechanism that underlies neovascularization, angiogenesis, and vasculogenic mimicry.

#### MicroRNAs

Micro-RNAs (miRNAs), are small single-stranded 18–25 nucleotide RNAs. They play key roles in biological processes such as development, stem cell differentiation, and tissue identity through negative regulation of mRNA transcripts (10). Twentysix years after their discovery, the number of studies that describe their role in cancer is still increasing, so they have earned their place as diagnostic, prognostic, and therapeutic biomarkers (5).

The earliest report on miRNAs was made by the Ambros lab (11). Lin-4 is a 22-nucleotide RNA with sequence complementarity to a region of the 3′UTR in the lin-14 mRNA, which inhibits lin-14 mRNA from being translated. However, it was not until 2001 that Ambros coined the term miRNA when describing a number of small RNAs with a role in gene regulation that had been recently identified in C. elegans (12).

Most miRNAs are transcribed in the form of a primary miRNA (pri-miRNA) by RNA polymerase II (Pol II), then processed by the nuclear microprocessor (comprised by the Ribonuclease II DROSHA, and DGCR8) to form the pre-miRNA, which is later exported to the cytoplasm by means of an Exportin-5-Ran-GTP-shuttle protein. In the cytoplasm, DICER binds to the pre-miRNA and cleaves it to its final 22 nt mature form that associates with AGO 2 to form the RNA-induced silencing complex (RISC). MiRNAs function through sequence complementary: within the RISC, the miRNA binds the target mRNA 3′UTR and, based on the degree of complementarity, leads to full mRNA degradation or blocking of the ribosomal machinery, both result in gene silencing (13).

The first reported miRNAs contributing to cancer were miR-15/16 in Chronic Lymphocytic Leukemia (CLL). Under normal conditions, both miRNAs repress antiapoptotic Bcl-2 protein, which is overexpressed in CLL (14). Since then, several miRNAs associated with cancer have been described. Ongoing research on miRNAs and their role in cancer development shows their great potential as biomarkers, therapeutically targets or even as potential therapies, restoring function of tumor suppressor miRNAs (10).

#### LONG-NON-CODING RNAs

Transcripts that do not encode proteins and are more than 200 nucleotides in length, are termed long non-coding RNAs (lncRNA) (15). Many of them resemble mRNAs in aspects such as being 5 ′ capped, spliced, and polyadenylated; but differ in a shorter overall length, fewer but longer exons, and lower expression levels (16).

Transcription of lncRNAs is similar to other eukaryotic RNAs, transcribed by RNA Pol II from bidirectional promoters (15). These promoters are often enriched in H3K27ac, H3K4me3, and H3K9ac modified histones and are repressed by remodeling complexes such as Swr1, lsw2, Rsc, and Ino80; therefore, SWI/SNF complex activity is needed to promote transcription initiation. After being transcribed, their structure is unstable, and they are subject to nuclear exosome or cytosolic non-sensemediated decay, so their half-life is short (<2 h) compared to miRNA (48-h half-life). It is still unknown whether this mechanism is followed by all lncRNAs (17).

LncRNA classification relies on the empirical attributes originally used to detect them such as size, localization, and function (18) although it is yet to reach a universally recognized consensus. The latest classification by the genomic consortium GENCODE categorizes them according to their genomic location in five groups: (1) Antisense RNAs: encompasses RNAs that are transcribed from the antisense strand near an exon of a proteincoding locus; (2) Long intergenic non-coding (LincRNA): includes RNAs that are transcribed from intergenic loci; (3) Sense overlapping transcripts: transcripts that comprehend a coding gene inside an intron on the same strand, (4) Sense intronic transcripts: comprises transcripts that are encoded in introns of coding genes, (5) Processed transcripts: RNAs that do not contain an ORF and cannot be otherwise classified (19).

Due to their ability to interact with DNA, RNA, and proteins, lncRNAs are able to regulate very diverse cellular processes such as chromatin modification, transcription, post-transcriptional modifications, scaffolding, and post-transcriptional mRNA regulation. Consequently, lncRNAs can be found in equally diverse subcellular locations: nucleus, subnuclear domains, and cytoplasm (6, 7).

The existence of lncRNAs was first reported in the early 1990s with the discovery of H19 and Xist in mouse (20, 21). Subsequently, novel lncRNAs candidates were identified and their true relevance in human biology and disease was revealed (22, 23). A role in cancer for lncRNAs was only suggested last decade, when HOTAIR (24) and H19 (25) were found to modify the transcriptional landscape through chromatin modification. Since then, many reports have concurrently established a role for lncRNAs in cancer development (26). Moreover, they are uniquely promising cancer biomarkers since they are easily detectable in body fluids (27, 28).

#### LncRNA-miRNA INTERACTION

Besides the regulation that both miRNAs and lncRNAs alone exert on mRNAs, it has been reported that they interact with one another, further modulating their influence in the transcriptome. These interactions lead to miRNA-triggered RNA decay, competition between miRNAs and lncRNAs for the same mRNA target, miRNA generation from lncRNAs, and lncRNAs acting as decoys for miRNAs [extensively reviewed in (29)].

Multiple reports show that the latter is the most prevalent lncRNA-miRNA interaction in cancer. LncRNAs that bind miRNAs and prevent their interaction with their target are mRNA translation is allowed.

regarded to as competitive endogenous RNAs (ceRNAs), decoys or sponges (30); since they prevent miRNAs from completing their regulatory function, lncRNAs acting as sponges are, effectively, positive regulators of mRNA transcripts (**Figure 1**). Interestingly, most lncRNAs capture miRNAs using regions close to their 3′ end named miRNA Response Elements (MRE), which are complementary with the Ago binding sites present in most miRNAs (31). It is relevant to mention that, while most RNA-RNA interaction reports come from strictly controlled experiments, the exact relationship between the plethora of RNAs in the cell—and thus the efficiency of competitive endogenous interactions—remains to be entirely understood in pathological models, which often present strong dysregulation of specific competing endogenous RNAs (32).

Prediction of these mechanisms has gained importance in the latest years due to the broad impact of the lncRNA-miRNA regulation. This has led to the development of bioinformatic tools such as MechRNA (33), RNAHybrid (34), RNADuplex (35), and RNAcofold (36) among others, that aim to elucidate lncRNA-miRNA interactions. Likewise, searchable repositories of lncRNA-miRNA interactions such as miRcode (37) are working to facilitate the study of RNA regulation through information. At the time of writing, experimental validation of lncRNA-miRNA interactions is necessary (38).

The role of lncRNAs is certainly complex. For instance, it was recently reported that UCA1 binds the 3′UTR of mRNAs to prevent their degradation by miRNAs, constituting a RNA-based regulatory signaling, which regulates cancer-linked pathways (39).

In the following sections, we review experimentally validated lncRNA-miRNA interactions with a role in tumor development processes. The path toward a full understanding of the ncRNA regulation networks is still long, but we are convinced that this is an exciting time to study regulatory RNAs.

#### ANGIOGENESIS

Angiogenesis is the process that generates capillary networks from pre-existing blood vessels in response to the need of nutrients in a given tissue region (40). It occurs throughout development and adult life, precisely controlled by a network of angiogenesis activators such as VEGF and inhibitors such as VASH2 (41). Tumor cells demand nutrients and thus modulate angiogenesis to their advantage altering the delicate activator-inhibitor balance (42). In the reviewed literature, we found that the VEGF-A mRNA participates in at least four lncRNA/miRNA/mRNA axes, albeit in different cancers. The TUG1/miR-299/VEGF-A axis increased angiogenesis in glioblastoma (43); LINC00668/miR-297/VEGF-A axis led to increased cell proliferation in oral squamous cell (44); and AK131850/miR-93-5p/VEGF-A promoted differentiation, migration and tube formation of endothelial progenitor cells (45).

Interestingly, miR-199a regulates both VEGF-A and its activating transcription factor, HIF-1a; thus, both of them are upregulated by Snhg1 lncRNA when it blocked miR-199a in a dual action Snhg1/miR-199a/VEGF-A&HIF-1a axis in bone marrow microvascular endothelial cells, promoting their proliferation (46). A somewhat similar mechanism was observed in HUVEC cells, where MALAT1 lncRNA antagonized miR-320a and upregulated the transcription factor FOXM1 (47), which also activates VEGF-A transcription. More studies are still needed to confirm whether VEGF upregulation is a common mechanism, attained by different lncRNAs in different tumors or this regulation has a high degree of redundancy and each of the investigated lncRNAs are active in other tumors as well.

Some other angiogenesis-related signaling proteins are upregulated by lncRNAs as well. For instance, VASH2 has been shown promote angiogenesis in tumors (48) and H19 lncRNA—highly expressed in glioma cells—upregulates VASH2 through the H19/miR-29a/ VASH2 axis. Zheng et al. (49) found that ANGPT2, a pro-angiogenesis signaling molecule is targeted by miR-26b, and upregulated in HUVEC cells by the sponge activity of PVT1 over miR-26b (50). This same PVT1-miR-26b interaction results in the upregulation of CTGF, a pro-inflammatory mediator with a role in promoting angiogenesis (51).

Interestingly, lncRNA-driven upregulation of angiogenesis has been observed in at least one non-tumoral context. The WTAPP/miR-3120-5p/MMP-1 axis, promotes angiogenesis in endothelial progenitor cells (52). Since, MMP-1 has an established role in cancer development (53), it is likely that WTAPP1 also promotes angiogenesis in tumors.

#### NEOVASCULARIZATION

Neovascularization is a mechanism through which new blood vessels are made from preexistent ones, this process is

coordinated by angiogenesis inductors and inhibitors following endothelial cell proliferation and migration (54). Developing tumors obtain the required nutrients and oxygen from neighboring blood capillaries; nonetheless, since the diffusion distance of oxygen is 100–200µm, the generation of new blood vessels is necessary for tumors larger than 1–2 mm (55). In this hypoxic environment, the HIF-1 induces the expression several growth factors (e.g., HGF) and VEGF to promote hypervascularization (56). An important distinction between neovascularization and angiogenesis is that the latter is a requirement for tumor progression accelerating the tumor development (57). In the reviewed literature, we found only 5 papers published from 2015 to 2017, describing the lncRNAmiRNA-mRNA crosstalk orchestrating this mechanism.

Deng et al. determined the role of the CCAT1/Let-7/c-myc axis in hepatocellular carcinoma. High expression of CCAT1 was associated with larger tumor size, microvascular invasion and alpha fetoprotein (58). Both HMGA2 and c-myc are let-7 targets; however, only c-myc was observed up-regulated while CCAT1 was stably overexpressed in SMMC-7721 cells. Deng et al. concluded that CCAT1 regulates let-7 and this, in turn regulates c-myc in order to coordinate proliferation and migration events in hepatocarcinoma (58).

Dong et al. through in vitro and in vivo analysis demonstrated the participation of TUG1/miR-34a-5p/VEGF-A axis in hypervascularity and hepatoblastoma progression (59). In a xenograft model, TUG1 knockdown lead to a significant tumor reduction up to 28% compared to the control group. Significantly diminished VEGF-A levels indicated that miR-34a-5p is a miRNA target of TUG1. At the same time, VEGF-A was a mRNA target of miR-34a-5p (59). Thus, the TUG1/miR-34a-5p/VEGF-A axis contributes to unusual hypervascularity in hepatoblastoma.

Glioma is a well-studied model for neovascularization (60). Significant H19 overexpression in microvessels from glioma specimens vs. normal brain microvessels leads to enhanced proliferation, migration, and tube formation with major tubule length and number of branches in H19 overexpressed glioma-associated endothelial cells. Besides, H19 overexpression decreased the miR-29a level and promoted the VASH2 overexpression. H19 acts a sponge for miR-29a; moreover, H19 knockdown promoted miR-29a overexpression and decreased VASH2 protein level in consequence diminished proliferation, migration and tube formation, establishing the H19/miR-29a/VASH2 axis (60). In another report from glioma cells, cell growth was arrested by H19 expression inhibition. MiR-140 was detected as a H19 miRNA-target, as suggested when H19 overexpression and miR-140 downregulation were determined and was corroborated by luciferase assay. Simultaneously, it was determined that iASPP—previously reported to promote cancer cell growth—was a direct target of miR-140 (61). Also, it was reported that PVTI lncRNA and miR-186 expression were inversely correlated in glioma. Functional analyses showed that PVTI stable transfection of glioma vascular cells lines favored proliferation, migration and tube formation. Likewise, miR-186 knockdown supported proliferation, migration and tube formation of glioma vascular endothelial cells; miR-186 inhibits expression of ATG7 and Beclin I, essential proteins to autophagy-lysosome formation. The authors suggested that PVTI and miR-186 could be deliver new objectives for glioma anti-angiogenic therapy (62). Together, these reports strongly suggest and important role for H19 in neovascularization in glioma through at least three lncRNA/miRNA/mRNA axes.

#### VASCULOGENIC MIMICRY

Vasculogenic mimicry (VM) was first described by Maniotis et al. They defined it a vascular-like structure which can mimic the embryonic vascular network (microcirculatory channels comprised of extracellular matrix) to sustain tumor tissue providing it with plasma and red blood cells (63). An important distinguishing characteristic is that vasculogenic mimicry resembles the embryonic vasculogenesis processes, suggesting that tumor cells can be converted back to an undifferentiated, embryonic-like phenotype to provide nutrients that ensure tumor growth in hypoxic environment (64). This mechanism has been observed in several tumors such as melanoma, ovarian, breast, prostate, osteosarcoma, bladder, colorectal, and lung cancers, where it plays an important role in invasion and metastasis; thus, patients with VM have a worse prognosis (65).

Several key molecules have been reported associated with this process including Notch1, MMP-2, MMp-9, vimentin (66), VE-cadherin, EphA2, FAK, PI3-Kinase (67), VEGF, endostatin, TGF-ß1 (68), Dickkopf-1 (69), maspin (70), laminin, CD44, thrombospondin 1, and cyclin E2 (64), among others. The participation of master regulators such as miRNAs and lncRNAs has not gone unnoticed, although our literature review yielded only five papers on the miRNAs/lncRNAs/mRNAs cross-talk and VM regulation.

Gao et al. observed HOXA-AS2 overexpression in glioma cell lines and tissues. HOXA-AS2 knockdown lead to underexpression of MMP-9, MMP-2 and VE-cadherin proteins and, consequently to VM inhibition; HOX-AS2 turned out to sponge miR-373, which, interestingly, did not target MMP-9, MMP-2, or VE-cadherin but EGFR. Furthermore, HOXA-AS2 knockdown favored miR-373 expression and EGFR downregulation in U87 and U251 cell lines. Xenograft and orthotopic models further demonstrated that HOXA-AS2 knockdown plus pre-miR-373 produced the smallest tumors, the longest survival time and the lowest VW densities (71).

We found that TWIST1 has an important role in VM, as it participates in at least two lncRNA-miRNA axes. In glioma, in is upregulated by LncRNA LINC00339 via miR-539-5p. Functional analysis revealed that overexpression of miR-539-5p inhibited the viability, migration, invasion and tube formation of the cell lines by downregulating TWIST1. Moreover, TWIST1 binds to the promoter of MMP-2 and MMP-14, both involved in VM formation. In xenograft models with knockdown LINC00339 and pre-miR-539-5p, smaller tumors and longer overall survival supported the LINC00339/miR-539-5p/TWIST1 axis (72). In triple-negative breast cancer (TNBC), the regulation of TWIST1 is through miR-430-3p which, in turn, is regulated by TP73- AS1. Both an inverse correlation between TP73-AS1 and miR-430-3p expression, and the interaction between miR-490-3p and TWIST1 were found in MDA-MB-231 cells. Interestingly, it was observed that the enforced expression of TWIST1 and the inhibition of miR-430-3p increased VM formation (73).

Zhao et al. reported that lncRNA n339260 overexpression was associated with the presence of metastasis, shorter overall survival and with MV in hepatocellular carcinoma (HCC) patients (74). LncRNA n339260 resulted critical to induce stemlike characteristics and VM formation; also, its expression was correlated with c-Myc, SOX2 and Nanog expression, which are pluripotency-maintaining molecules. Interestingly, the target miRNAs of n339260 were miR-31-3p, miR-30e-5p, miR-519c-5p, miR-520c-5p, miR-29b-1-5p, and miR-92a-1-5p, which were detected by microarray in HepG2 cells transfected with this lncRNA (74).

The MALAT1/miR-145-5p/NEDD9 axis was described in lung cancer: MALAT1 sponges miR-245-5p to amplify NEDD9 expression. Interestingly, MALAT1 is induced by the ERβ, a novel role for this receptor in lung cancer progression in female patients. NEDD9 also plays an important role in metastasis through TGFβ signaling pathway. This axis was analyzed in xenograft models and it was observed that ERβ promoted metastasis via MALAT1/miR-145-5p/NEDD9 signal (75).

#### CONCLUSION AND PERSPECTIVES

Angiogenesis, neovascularization and VM, as tumor progression and metastasis mechanisms, are becoming more important as sources of biomarkers and therapeutic targets, as the authors of several of the reviewed papers point out. On the other hand, the nuances of lncRNA/miRNA/mRNA regulation are not analyzed when ncRNA expression profiles are sought (76), and global analyses of this regulation mechanisms are still scarce [e.g., (77)]. So we considered it important to summarize current knowledge on the lncRNA/miRNA/mRNA axis regulation regarding angiogenesis, neovascularization, and VM, as it is still limited and deserves further scrutiny, perhaps due to the high methodological requirements. Upon analyzing the PubMed-listed papers, we found few studies that address lncRNA/miRNA/mRNA axis regulation of these nutrient supply processes.

So far, available information shows that lncRNA H19 is involved in angiogenesis and neovascularization, although in diverse manners. The sharing of the H19/miR-29a/VASH2 axis by both angiogenesis and neovascularization hints at a master regulation role for H19 and VASH2 (**Figure 2**). Interestingly, vasculogenic mimicry did not share any lncRNA/miRNA/mRNA axes with angiogenesis or neovascularization, which makes it reasonable to speculate that this is a specific molecular process and suggests pivotal role for it in aggressive tumors.

Our review has shown us the important role of lncRNA/miRNA/mRNA regulation in cancer development, an open area of opportunity that grants broader and deeper exploration in the following years.

#### AUTHOR CONTRIBUTIONS

AC-P, CP-P, and EL-U contributed to the conception of the article. AC-P, CP-P, EL-U, LB, and DL wrote and revised the final manuscript and agreed on its submission to this journal.

#### REFERENCES


#### FUNDING

This study was partially supported by the Council for Science and Technology (CONACyT) (SALUD-2015-1-262044).

#### ACKNOWLEDGMENTS

We thank Alan Mario García Mendoza for their contribution to documental research.

of their gene structure, evolution, and expression. Genome Res. (2012) 22:1775–89. doi: 10.1101/gr.132159.111


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 López-Urrutia, Bustamante Montes, Ladrón de Guevara Cervantes, Pérez-Plasencia and Campos-Parra. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Fact or Fiction, It Is Time for a Verdict on Vasculogenic Mimicry?

Andrés Valdivia<sup>1</sup> , Gabriel Mingo<sup>1</sup> , Varina Aldana<sup>1</sup> , Mauricio P. Pinto<sup>2</sup> , Marco Ramirez <sup>3</sup> , Claudio Retamal <sup>4</sup> , Alfonso Gonzalez <sup>4</sup> , Francisco Nualart <sup>5</sup> , Alejandro H. Corvalan2,6 and Gareth I. Owen1,2,6,7 \*

<sup>1</sup> Faculty of Biological Sciences, Pontificia Universidad Católica de Chile, Santiago, Chile, <sup>2</sup> Faculty of Medicine, Pontificia Universidad Católica de Chile, Santiago, Chile, <sup>3</sup> Faculty of Medicine, Universidad de Chile, Santiago, Chile, <sup>4</sup> Faculty of Medicine and Science, Center of Cellular Biology and Biomedicine (CEBICEM), Universidad San Sebastian, Santiago, Chile, <sup>5</sup> Faculty of Biological Sciences, Universidad de Concepcion, Concepción, Chile, <sup>6</sup> Advanced Center for Chronic Diseases (ACCDiS), Santiago, Chile, <sup>7</sup> Millennium Institute on Immunology and Immunotherapy, Santiago, Chile

The term vasculogenic mimicry (VM) refers to the capacity of certain cancer cells to form fluid-conducting structures within a tumor in an endothelial cell (EC)-free manner. Ever since its first report by Maniotis in 1999, the existence of VM has been an extremely contentious issue. The overwhelming consensus of the literature suggests that VM is frequently observed in highly aggressive tumors and correlates to lower patient survival. While the presence of VM in vivo in animal and patient tumors are claimed upon the strong positive staining for glycoproteins (Periodic Acid Schiff, PAS), it is by no means universally accepted. More controversial still is the existence of an in vitro model of VM that principally divides the scientific community. Original reports demonstrated that channels or tubes occur in cancer cell monolayers in vitro when cultured in matrigel and that these structures may support fluid movement. However, several years later many papers emerged stating that connections formed between cancer cells grown on matrigel represented VM. We speculate that this became accepted by the cancer research community and now the vast majority of the scientific literature reports both presence and mechanisms of VM based on intercellular connections, not the presence of fluid conducting tubes. In this opinion paper, we call upon evidence from an exhaustive review of the literature and original data to argue that the majority of in vitro studies presented as VM do not correspond to this phenomenon. Furthermore, we raise doubts on the validity of concluding the presence of VM in patient samples and animal models based solely on the presence of PAS+ staining. We outline the requirement for new biomarkers of VM and present criteria by which VM should be defined in vitro and in vivo.

Keywords: vasculogenic mimicry (VM), angiogenesis, endothelial, model, in vivo–in vitro

#### INTRODUCTION

All cells within our bodies require a continuous supply of blood that contains oxygen and nutrients if they are to thrive. In order to ensure this, a subset of cells may synthesize and secrete Vascular Endothelial Growth Factor (VEGF) in response to certain conditions such as low oxygen levels (a condition called hypoxia). Secreted VEGF then mobilizes and activates pre-existent endothelial cells (ECs) that form new blood vessels in a process called angiogenesis. In normal tissues,

#### Edited by:

César López-Camarillo, Universidad Autónoma de la Ciudad de México, Mexico

#### Reviewed by:

Jose Javier Bravo-Cordero, Icahn School of Medicine at Mount Sinai, United States Venugopal Thayanithy, University of Minnesota, United States

> \*Correspondence: Gareth I. Owen gowen@bio.puc.cl

#### Specialty section:

This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology

> Received: 26 April 2019 Accepted: 10 July 2019 Published: 02 August 2019

#### Citation:

Valdivia A, Mingo G, Aldana V, Pinto MP, Ramirez M, Retamal C, Gonzalez A, Nualart F, Corvalan AH and Owen GI (2019) Fact or Fiction, It Is Time for a Verdict on Vasculogenic Mimicry? Front. Oncol. 9:680. doi: 10.3389/fonc.2019.00680

**44**

angiogenesis plays a key role in fetal development and tissue repair. As a consequence, this process is highly conserved among mammals.

As occurs with other physiological processes, cancer cells can hijack angiogenesis in order to potentiate their survival and propagation. Indeed, "tumor angiogenesis" was described 80 years ago and has been extensively confirmed in a variety of experimental models, demonstrating that tumor growth is accompanied by the formation of new blood vessels. Based on these findings, in 1971 Judah Folkman hypothesized that the inhibition of angiogenesis in cancer cells could be therapeutic, coining the term "anti-angiogenesis." In recent years, several compounds with anti-angiogenic activity have been tested in cancer patients with disparate results; in many cases a favorable initial response is followed by tumor recurrence.

Evolutionary biology teaches us that a selection pressure can generate a resilient system via the natural selection, as such cancer cells (like any other cell) exposed to anti-angiogenic drugs may develop a number of strategies to circumvent the suppression of angiogenesis with these therapies. These strategies include: use of alternative angiogenic pathways, vessel co-option and vasculogenic mimicry (VM) among many others [for a full review see (1)].

In lay terms, VM occurs when a subset of cells within a tumor modify their expression profile/phenotype and form ECfree (i.e., non-angiogenic) tubular structures that supply oxygen and nutrients to cancer cells. Although the existence of VM in tumor samples has been extensively demonstrated there are a number of controversies surrounding published in vitro and in vivo VM models. Here, we review and discuss the available evidence and controversial issues around VM, seeking to provide a critical assessment of the current literature and a final verdict on the validity of these models.

#### WHAT IS VASCULOGENIC MIMICRY (VM)?

As pointed above, the term VM (also reported as vascular mimicry) was originally used to describe the process by which tumor cells formed a network of tubular structures with the ability to conduct fluids, thereby "mimicking" the vasculogenic process of ECs during angiogenesis. Several studies have reported VM both in vivo and in vitro (see **Supplementary Table SII**). As explained, the proposed functions of VM are: oxygen supply, nutrient transport, and the elimination of cell waste. These are all critical functions required at the early stages of invasive tumorigenesis that may not be fully accounted by conventional angiogenesis. More recently, the term VM has been expanded to incorporate any EC-free fluid-conducting structure (i.e., not a blood vessel). This came after a study in knockout mice demonstrated that macrophages suffer a phenotypic change acquiring the ability to form fluid-conducting structures (2). Furthermore, prior to the formation of the placenta, trophoblast cells infiltrate the uterine walls, and form EC-free tubular structures that resemble VM (3), suggesting VM may be responsible for blood and nutrient supply in the early stages of pregnancy.

But exactly how can we explain this "phenotypic switch" that allows the formation of vascular structures without ECs? An answer to this question may lie in the vessel structure of Amphioxus (Branchiostoma lanceolatum), an invertebrate cephalochordate with a body plan similar to that of vertebrates. Like vertebrates, Amphioxus vessels are lined by an extracellular matrix (ECM) however the endothelial basement membranes in vertebrates display some differences in their molecular composition (4). These studies not only provide some hints on the evolutionary origins of VM but also demonstrate that EC-free vasculatures are not exclusive to malignancy. In fact, this might be yet another example of an existing or ancient physiological pathway being hijacked by cancer cells.

Although angiogenesis, lymph vessel formation and VM share the same goal of establishing fluid-conducting structures within a tissue, they display some notorious differences. **Figure 1** shows a comparative diagram of traditional blood vessels (formed by vasculogenesis or angiogenesis), lymph vessels, and VM vessels. In traditional blood vessels (left), a single layer of ECs lines the lumen: an external continuous inner-basement membrane surrounds ECs in these vessels. Similarly, lymph vessels have a central inner layer of ECs; however, their basement membrane is non-continuous (**Figure 1** center) (5). Our current understanding of VM vessels suggests that cancer cells sit on top of a glycoprotein rich membrane (matrix) which surrounds a central lumen (**Figure 1** righthand panel) (6). As observed in the basement membrane on traditional blood vessels, these studies suggest that VM vessels also have a glycoprotein-rich inner coating composed by collagens and laminin, among other proteins (7, 8). In summary, traditional (or conventional) blood vessels and VM vessels can be identified and distinguished based on structural and composition differences as indicated in **Figure 1**. These features have been systematically used in the literature to identify VM in cancer patient samples.

### VM IN THE CLINIC: WHAT IS THE EVIDENCE IN CANCER PATIENTS?

Since its first publication, many authors have embraced the existence of VM, while others have disputed it. The latter argue this is only a remote phenomenon that occurs within tumors and may be open to misinterpretation (9–11). The basal membranes of both blood and lymph vessels contain a variety of mucinous proteins (glycoproteins) that stain positive for the Periodic Acid– Schiff (PAS, mucosubstance stain) (5). Throughout the literature, the existence of VM vessels is inferred by the presence of PAS+ vessel–like structures within tumors in the absence of EC markers such CD34 and CD31, among others. Hence, authors have postulated VM as an angiogenesis-independent alternate tumor perfusion pathway for tumors. Indeed, human tumor biopsies have shown the presence of red blood cell (RBC) containing PAS+ vessels that stain negative for EC markers.

Originally described in uveal melanoma, VM is now reported in >20 malignancies (**Supplementary Table SI**). VM critics such as Professor McDonald have claimed that this is nothing more than an "artifact" consequence of the erratic structure

of the tumor endothelium and the accumulation of blood, derived from microhaemorrhages (11). Indeed, this has been a recurrent argument among critics claiming that these structures are merely "blood pools" brought about by the process of tissue acquirement (see **Figure 2A**) (12). While plausible, this argument does not take into consideration that a trained pathologist can easily distinguish a "blood pool" from RBCs trapped within a tubular structure. Moreover, if these were indeed blood pools, then RBCs would not be enclosed within a PAS+ structure. As an example, **Figure 2B** shows RBCs surrounded by melanoma cells (black spots are melanin) with black arrows indicating a continuous covering of a tubular structure; this may be interpreted as a basal membrane. However, the field of VM may have itself to blame for the current controversy as several inconsistencies among VM reports have generated skepticism. For example, some studies postulate the presence of VM based on a luminal space in a carcinoma cross-section, however no PAS+ border is present (16). Similarly, weak PAS staining always leaves the doubt of whether a membrane is present or the structure is in fact a blood pool (17–20). In contrast, several reports from the group of Sun and colleagues clearly demonstrate the presence of PAS+/CD31- structures that contain RBCs in both Hepatocarcinoma and Gastro Intestinal Stromal Tumor (GIST) patients (14, 21). Encouragingly and as a proof of concept, these reports demonstrate the presence of both VM (PAS+/CD31–) and blood vessels (PAS+/CD31+) within the same field (shown in **Figure 2C**). Similar evidence is reported in uveal melanoma, where a fluorescent dye was injected into the patient and tracked through to the eye (13). An exhaustive analysis of glioblastoma by Scully and colleagues showed the presence of CD31, CD34, and/or Vascular Endothelial (VE)-Cadherin+ positive (and thus endothelial) and negative (potentially VM) luminal structures. This study also demonstrated that endothelial confirmed vessels presented alpha

smooth muscle actin (SMA, a pericyte marker) while potential VM structures did not (22). Taken collectively these publications demonstrate the existence of non-endothelial blood containing vessels in human tumors.

### IS THE COMBINATION OF PAS+ AND ABSENCE OF EC MARKERS A DEFINITIVE PROOF OF VM?

Not exactly, we agree that the confirmed observation of a PAS+/CD31– lumen containing RBCs maybe indicative of VM. However, we believe this is not a definitive proof. **Figure 2D** shows thread-like PAS+ structures commonly reported throughout the literature as "patterned structures" (13, 23, 24). As we will describe later, cancer cells secrete large amounts of mucoproteins that stain PAS+, however this does not imply these are forming tubular structures. Another example of patterned structures is shown in **Figure 2C**; where strands of PAS+ structures can be observed over a "true" blood vessel (RBC containing CD31+ tubular structure). Furthermore, PAS+ "patterned structures" have also been reported in medulloblastoma where potential VM structures are suggested to connect to the EC vasculature (25). However, electron microscopy by Maniotis et al. of pattern structures does suggest that blood components can be present in the vessel interior (26). Thus, the jury is still out on whether all "pattern structures" can be classified as VM. A further problem in the reporting of the presence of VM occurs when no imagery is shown; without physical evidence it is difficult to draw conclusions (27). Similarly, small images in black and white do not allow the reader to be convinced of the presence of PAS+/CD31– structures (28). While these publications may be validly reporting the presence of VM, without a standardized

method of reporting this phenomenon it is difficult to verify any conclusion on incidence and function.

In summary, in the absence of a reliable VM biomarker the combination of PAS+ and absence of classic EC markers like Von Willebrand factor, CD34 or CD31, plus RBCs in a clearly defined lumen should be the standard for reporting VM+ status across the literature.

#### WHY SHOULD WE CARE ABOUT VM?

Because a large number of studies indicate that VM+ is associated to a decrease in cancer patient survival, measured as OS or as progression-free survival (PFS) (13). **Supplementary Table SI** and **Figure 3** compare OS levels in VM+ vs. VM- tumors across 20 cancer types. Overall, 19 out of 20 reports confirm that VM+ associates with a decrease in OS; with the exception in synovial sarcoma (29). **Supplementary Table SI** summarizes all current literature reporting occurrence rates and OS in pathology observed cancers. Strikingly, reports in ovarian and colorectal cancers classified as VM+ showed lower survival time in the magnitude of years compared to VM- tumors (30, 31). Similar differences were observed in orbital rhabdomyosarcoma and adrenocorticoid carcinomas (32, 33). Gastric cancer patients with PAS+ structures were prone to present higher histological grade, metastasis, distant recurrence, and 12 months less cumulative OS (34). Similarly, VM+ prostate cancer patients correlated with Gleason score, preoperative prostate-specific antigen (PSA) levels, pathological stage and both lymph node and distant metastasis. Studies to date have come principally

from the Chinese population, although isolated reports have been published from European, Japanese, North American, and Thai populations. However, as observed in cancer incidence, the frequencies of cancer type and the mutational burden within each classification vary according to region and further studies need to be performed to get a clearer picture of prognostic value of VM presence within a specific population. In summary, the overwhelming consensus of the literature suggests that VM is frequently observed in highly aggressive tumors and correlates with poor prognosis. Therefore, the elucidation of specific treatments targeting this subset of aggressive cells may have offer a benefit for cancer patients in terms of survival.

#### IS VASCULOGENIC MIMICRY A "HALLMARK OF CANCER"?

Or in other words do all cancer cells undergo VM? A short answer to this question would be no. Reports indicate the percentage of VM+ tumors (by PAS+/CD31–) varies wildly from 5 to 65% depending on the cancer type and the pathologists' inclusion criteria. Among the studies that assessed tumor-based data the average VM incidence is about 29%. As shown in **Supplementary Table SI**, glioblastoma has the highest incidence among tumor types (65.9%), with the lowest incidence to date reported in melanoma (5.25%) (29). There is a notable heterogeneity in the reporting of VM as demonstrated in glioblastoma patients. Han et al. reported a 65.9% of VM+ in glioblastoma patient samples (35), however two similar studies reported 26% (36) and 16% (24) also in glioblastoma. Evidently, patient and/or tumor characteristics such as tumor stage or histological grade, could be responsible, however, to differences in the reporting criteria for VM+ could also be attributed, illustrating the need for a standard classification.

### THE IN VIVO CONTROVERSY: AN ANIMAL MODEL OF VM

As in all in vivo models of cancer the mouse xenograft has been the standard for VM research. Initial studies of breast cancer cell xenografts were assessed for VM by Hematoxylin & Eosin (H&E) stain and investigators acknowledge that by using this technique alone, a pathologist could misinterpret VM as blood pools caused by internal tumor hemorrhages (37, 38). Later the same year utilizing LnCaP prostate cell xenografts stained by H&E and prostate specific membrane antigen (PSMA) demonstrated structures that were CD31– yet positive for platelet aggregates and fibrin (39). The first report that used the PAS+/CD34– combination came in a model of B16 melanoma cells injected into C57Bl/6 mice (40). This pioneering study demonstrated the presence of PAS+ non-EC structures that contained RBCs within their lumen (40). Following this study, several authors reported PAS+/CD31+ (blood vessel) or PAS+/CD34– (VM) structures, however, in some cases low quality or low-resolution images failed to prove CD31– status or presence of RBCs (37, 41–43). In contrast, a number of studies have provided solid evidence of PAS+/CD31– stained structures that also contain RBCs in their lumen (29, 39, 44–46).

The current tools to identify VM in vivo are clearly deficient! PAS+ staining alone does not guarantee VM presence and thus novel biomarkers that discriminate between VM and blood pools are urgently needed. As potential biomarkers, Bajesy et al. used 3D Z-stack reconstructions to identify intratumoral structures that were both laminin+ and CD34 in metastatic uveal melanoma samples (47). A recent study used a pan-laminin antibody along with an EC-binding lectin to identify VM structures in xenografted human glioblastoma cells (48). The authors demonstrate the presence of lectin+ and lectin- tubular structures. These results suggest the mucoprotein content and composition of these tubular structures may vary substantially. Hence, future studies could aim to identify specific mucoproteins within CD31- vessels, perhaps specific lectins or other ECM components that will improve current VM identification methods.

#### THE IN VITRO CONTROVERSY: THE PRINCIPLE PROBLEM

The presence of an in vitro model is potentially the most controversial aspect of the VM field. To understand this fully and to trace the errors that have occurred within our scientific discipline, in the following paragraph we examine the origin of the in vitro model and speculate how the majority of the papers in medical literature may be erroneously presenting conclusions based on an assay that is not measuring VM.

During the 2001–2002 period Mary Hendrix's group published several articles providing the first evidence suggesting that VM structures contained a lumen, lined by a glycoproteinrich membrane (12, 49–55). This process only occurred in a 3D matrix (Matrigel) and after several days in culture. A study by Sanz et al. (56) was the first to present an in vitro assay claiming that intercellular connections formed within 1 day of cancer cell culture in Matrigel represented VM. These structures initially were thicker that those observed in the classic tube forming assays using endothelial cells (classic angiogenesis assay using HUVECs or EC lines) however, this study failed to proof these were functional lumen containing structures (i.e., could conduct fluid). Furthermore, the study proposed a quantification method based on cellular connections (56). This could have been a turning point in VM research, as these structures (and structures which were slightly thinner and more similar to those seen in angiogenesis assays) became adopted as an accepted in vitro representation of VM. A subsequent study by Vartanyan et al. described side by side EC and cancer structures claiming that both were lumen containing and that the VM was a representation in vitro of the blood filled CD31- vessels seen in histological cross-sections of tumors (57). Perhaps the greatest contributor to the current controversy came in 2011 when Francescone et al. published a paper entitled "A Matrigel-based tube formation assay to assess the vasculogenic activity of tumor cells." This has been cited as a reference validating the concept that intercellular connections represent VM ever since (58). Although there have been notable exceptions, most of the VM research in vitro has utilized intercellular connections formed between cancer cells to report the presence and mechanisms of this phenomenon. Thus, the field of VM, at least in vitro, has continued to be shrouded in controversy, leading to divided opinions in the scientific community.

### BACK TO BASICS: THE HENDRIX MODEL REVISITED

Initial representations shown by the Hendrix group of VM in vitro demonstrated tubular structures that formed after numerous days in culture, that where lumen containing and importantly were capable of fluid conduction (6). Herein, we suggest that this model, with improvements, should be the standard for in vitro assays of VM.

To elaborate upon this idea and to demonstrate to the reader that intercellular connections or a congregation of cells do not represent fluid containing vessels, **Figure 4** depicts representative imagery complementing previous results presented by our group and in line with the initial representations shown by the Hendrix (6). In this figure there are two cell lines that demonstrate structures reported to be VM in the literature. **Figure 4A** shows the HEY cancer cell line forming intercellular connections at day 1 in Matrigel culture, which become develop into elevated structures above a cell monolayer at day 4. However, the appearance of intercellular connections on the first day does not necessarily mean that VM structures will occur at a later date. With the aim of demonstrating cell lines that form intercellular connections but do not produce a hollow lumen or conduct fluid, **Figure 4B** shows the formation of network structures at day 1 and 4 in the MeT5A and U87 cell lines. We observed that if we inject a fluorescent dye into 1 and 4 day-old structures there is no dye movement; the dye stays diffusely only around the individual cell that receives the injection (actually it almost below detectable levels, hence the black image). However, dye movement is observed in Day 4 cultures of the HEY cell line. As was shown in a previous publication, injecting the dye into individual cells of the HEY cell monolayer does not result in dye movement and furthermore, injecting dye into structures spanning clusters of cells in other cancer cell lines (UCI101 and A2780) also fails to show presence of a fluid conducting tube [this can be observed in Figure 4c of (6)].

These results suggest that although there appear to be tubular structures, only intercellular connections or cellular aggregates are present, and thus the majority of structures presented as VM in the literature may not in fact contain a lumen and

are thus incapable of fluid conduction. Although an argument could be made that some of the published intercellular structures shown at day 1 may develop into VM tubular structures, the authors cannot be sure of this claim and thus we suggest that the model is not valid. In our own work on primary cultures we often saw initial intercellular connections during the first day in culture that subsequently disappear after several days (6). Following this line of thinking, a future area of controversy may be the report of intercellular connections formed after 12–24 h in Matrigel of cell lines at that have been previously reported to produce fluid conducting structures at latter time points. While this may be currently acceptable, it is dangerous to assume that anything that inhibits tubular structures at day 1 is specific to the pathway required for the process of VM. Any tested compound or pathway component may in fact be representing toxicity to the cell, an inhibition of cell cycle or a change in cytoskeleton that will inhibit all movement related biological processes such as migration and invasion. We recommend that assays examining the process of VM be followed to the formation of undeniable fluid conducting structures.

Valdivia et al. Opinion on Vasculogenic Mimicry

In **Supplementary Table S1** we have divided the publications in the field of VM into those that either demonstrate or fail to show the presence of a lumen and/or conduction of fluid. This analysis reveals that of the 357 published papers reporting VM in vitro, only 49 (13.7%) convincing demonstrate a tubular structure. Although, this does not mean that all reports of tubular structures within the first 24 h (intercellular connections) will not eventually form VM structures, it is impossible based on this assay to distinguish between merely intercellular connections or the process of VM with the presence of fluid conducting tubes. A universally accepted model of VM that demonstrates a lumen or fluid conduction is required for research in this field to advance. Furthermore, conclusions based on assays that do meet these criteria should be interpreted skeptically.

#### PRESENTATION OF A STANDARDIZED IN VITRO MODEL OF CANCER VM

In our opinion only a few in vitro studies have convincingly demonstrated a functional lumen in tubular structures (12, 49–55). Building upon these pioneering studies, our research group recently published an in vitro model demonstrating (we believe convincingly) that cancer cells grown in Matrigel form tubular structures with a central lumen lined by glycoproteinrich borders (**Figure 5**). After several days in culture, cancer cells originate PAS+ structures that appear to be atop cancer cell monolayers. These PAS+ structures may reach up to 200µm in diameter (**Figures 5A–C**). The movement of microinjected trypan blue dye along these structures confirms they contain a functional lumen (**Figure 5D**). Confocal microscopy and IMARIS (Microscopy Image Analysis Software) reconstruction further confirm the presence of a lumen and a glycoproteinrich layer flanked by cancer cells (see VM vessels in **Figure 1**). Our data indicate these structures can be obtained in Matrigel cultures derived from cancer cell lines, primary tumors or from patient ascites (6). In 13 advanced ovarian cancer patient samples, only 38.5% (5 out of 13) of samples were capable of producing tubular structures in vitro. Previous studies report that 29–43% of ovarian cancers samples analyzed by immunohistochemistry present PAS+ and endothelial marker negative structures, thus we speculate that the ability of a tumor cell population to undergo VM may be retained in vitro (59).

### VM QUANTIFICATION: IS PAS A GOOD MARKER?

No, as we explained above PAS+ along with absence of EC markers allows VM identification but it is not sensitive enough to allow quantification. The literature on in vitro VM models contains several attempts for a quantification method. Such studies have employed a variety of methods including: tubule length, number of structures, tubular structure connections, or PAS+ levels (16, 17, 60–62). However, as explained above most studies have failed to demonstrate these tubular structures are indeed functional (i.e., have a fluid-conducting lumen) therefore the validity of these methods remains questionable.

Historically, PAS has been used as a staining method to identify mucosubstances such as glycoproteins, polysaccharides, and glycolipids (63). While VM channels clearly display a strong PAS+ stain (6, 13, 64), our micro-CT analyses [shown in **Figure 6** and also in **Supplementary Video 1** and Figure 2 of Racordon et al. (6)] demonstrate that in many cases PAS+ structures do not contain a lumen. Using this micro-CT technique, we observe tubular-like structures along with flatter areas that also stain heavily for glycoproteins (PAS+). In this technique, white areas denote air-containing structures. In **Figure 6B** we can observe that the flatter less tubular elevated structures do not possess a hollow structure. Alternatively, the **Figure 6C**, demonstrate rounded structures that clearly contain a lumen (white area). Hence, PAS staining in some cases may just represent glycoprotein-rich areas around aggregations of cancer cells. Accordingly, PAS+ structures obtained on a glioblastoma cell culture in culture (**Figure 6D**) are not able to conduct fluids. The lack of a lumen in these structures is further confirmed by confocal microscopy reconstruction (**Figure 6F**).

Thus, as we move toward a standardization of non-endothelial vessels/VM, until a unique biomarker has been identified, the use of PAS+ staining alone should be viewed cautiously and the reporting of "pattern structures" (shown in **Figure 2D**) should be replaced by PAS+ straining accompanied by the absence of an EC marker and preferably the presence of RBCs in a luminal structure.

### IN SEARCH OF THE SIGNALING PATHWAY LEADING TO VM FORMATION

Beyond the controversy over the nature, definition, and identification of non-endothelial vascular structures, a number of articles have sought to define a mechanism for tubule formation. Most studies used an in vivo approach, double stain PAS+/CD31– or PAS+/CD34– for VM+ and then correlated these structures with molecular markers (12, 65–68). Other studies have used pharmacological inhibitors on in vitro models (66, 69, 70). Using our criteria for true VM structures: PAS+/CD31– or PAS+/CD34– and presence of a lumen for in vivo and in vitro studies we elaborated a list of 93 articles that fulfilled these criteria and also postulated a VM mechanism based on molecular pathways (**Supplementary Table SIII**). We found that signaling/molecular pathways across all relevant literature could be grouped into 4 specific areas:

### Matrix Metalloproteases and Extracellular Matrix Components

A number of reports have suggested a role of matrix metalloproteases (MMPs) in VM. Sood et al., were the first to demonstrate a correlation between VM+ and expression of metalloproteases (MMPs)-1, MMP-2, MMP-9, MMP-14 in ovarian cancer samples (12). These studies also reported an association with Laminin-5 g-2. T ECM rearrangements and the secretion or incorporation of laminin subunits. A subsequent report showed that MMPs and Laminin-5 g-2 were required for the formation of VM in melanoma (51). In prostate cancer,

VM+ correlated with laminin and integrin α6β1 (52) and in mesothelial sarcomas and alveolar rhabdomyosarcomas with the presence of collagen IV fibers (29). In 2008, Demou reported that VM+ was associated to the presence of integrin α3 subunit (71). As it is established that in vitro VM only occurs upon an ECM substitute (Matrigel), it may be reasonable to assume the process requires ECM remodeling by MMPs. Future experiments will need to elucidate whether ECM is the source of the glycoproteinrich lined lumen observed in tubular structures in vitro or if this glycoprotein is secreted by the cancer cells.

#### PI3K-AKT Pathway

A study by Hess et al. (72) was the first of several studies to implicate the phosphoInositide-3 kinase (PI3K)-AKT pathway in VM (72–74). Subsequently, the same research group presented evidence for a role of focal adhesion kinase (FAK), an upstream component of the PI3K pathway and important component of the integrin signaling pathway (75). Two related studies demonstrated VM structures were associated to AKT (76) or correlated to MMPs, PI3K and FAK (68) in melanoma and gallbladder cancer, respectively, adding to the possibility that the integrin-FAK and PI3K-AKT signaling pathway are also involved. In our opinion, the PI3K pathway has provided the most solid evidence to date for a role in VM formation; this could also offer the opportunity for a therapeutic intervention in the future.

### Angiogenesis Signaling Pathways

As both VM and angiogenesis result in tubular fluid-conducting structures, it would appear logical that they have signaling pathways in common. However, the relationship between VM and angiogenesis is a controversial topic. Many authors have reported that the angiogenesis signaling pathway plays a role in VM, with a correlation between VM+ and either VEGF or PDGFRβ expression in cancer samples (22, 37, 54). Another factor associated to VM is the Hypoxia Inducible Factor (HIF)- 1α, its presence is also widely linked to the stimulation of proangiogenic pathways (65, 77–80). However, in sharp contrast, some reports demonstrate that antiangiogenic therapies, such as treatments against VEGF or its receptors have no impact upon VM, demonstrating the inconsistencies across the VM literature (48, 81, 82). Indeed, several studies speculate VM is a key process that allows tumor irrigation and growth even in the presence of anti-angiogenic therapy (1, 68, 83). Evidently, the lack of a

Committee approval and written patient consent from the Clinical Hospital of the University of Chile, Santiago, Chile. Cell culture was as described previously in Racordon et al. (6). (D) Light microscopy imagery of primary cultured cells grown on matrigel, with an image of the cells grown in plastic in the inlay. Size bar represents 500µm. (E) Primary cultured glioblastoma cells presented elevated structures over the cell monolayer that stained for PAS. Size bar represents 500µm. (F) Confocal 3D reconstruction using ZEN 2012 demonstrates that the PAS positive structures observed in (E) are elevated over the cell monolayer but do not possess a lumen.

consensus on the criteria to report VM may explain why the role of proangiogenic factors on VM remains unclear.

#### Other Signaling Pathways

Complementing the abovementioned studies, further reports have speculated on key components of VM formation. VM presence and poor patient prognosis has been reported with Tissue Factor Pathway Inhibitor-1 (TFPI-1) and TFPI-2. Antibody inhibition experiments revealed that TFPI-2 was required for VM in vitro, and that the blockade of TFPI-2 suppressed MMP2 activation (41). Whether this suggests that the coagulation cascade is involved in VM, or a non-homeostatic role of these proteins is responsible, has still to be evaluated.

Given the presence of fluid conducting tubular structures, VE-Cadherin has also been commonly associated to VM (84). VE-Cadherin is a cell-adhesion transmembrane protein classically expressed in ECs (85). Hendrix et al. described the presence of VE-Cadherin in melanoma cells undergoing VM (86). Furthermore, in melanoma VE-Cadherin has been reported to promote VEGFR-1 signaling, that in turn promotes the signaling of the PI3K/PKC pathway, which is critical for VM (87, 88). However, despite isolated reports it is still open for investigation to determine whether the process of VM is using similar pathways to that of angiogenesis or vasculogenesis.

The Wnt signaling pathway and EMT, commonly implicated in cancer, angiogenesis, and development have also been implicated in VM formation (89, 90). Wnt3a and β-Catenin are shown to increase formation of tubular structures in colon cancer (91), while essential EMT proteins Slug, Snail, and Twist, have been correlated with the presence of tubular structures (92, 93). While it may appear logical that developmental signaling process and pathways would be implicated in VM formation, the abovementioned publications, together with numerous others, further demonstrate that the true mechanism of VM formation is still to be defined.

### CONCLUDING REMARKS AND FUTURE DIRECTIONS

It was 1971 when Judah Folkman first postulated that the inhibition of tumor angiogenesis could be therapeutic, coining the term "anti-angiogenesis" to refer to the suppression of tumor blood supply (94). At the time, the rationale behind tumor irrigation seemed quite simple. However, over time we have learned that cancer cells (like any cell) have the ability to adapt and evade treatment regimens by systematically activating pathways and tools already present within our genome to ensure continuous self-propagation. Indeed, cancer cells can develop a number of strategies to compensate for angiogenesis and/or circumnavigate the inhibition of specific angiogenesis pathway by using alternative/compensatory pathways, vessel cooption or VM (1). We speculate that VM plays a key role in both bourgeoning tumors and in the evasion of antiangiogenic treatments. A standardization of assays for VM detection and quantitation in clinical samples along with reliable in vitro VM models will allow the development of biomarkers, drug discovery, and more effective treatments for antiangiogenic refractory patients.

Regarding VM biomarkers, the evidence suggest PAS alone may not serve as an effective biomarker (6). Novel, more specific biomarkers are required to discriminate endothelial vs. nonendothelial structures. Furthermore, it is critical to determine if "pattern structures" represent structures with a true lumen or merely polls of glycoproteins secreted by tumor cells. For now, we suggest pathology-based VM reports should demonstrate: PAS+, absence of EC markers, and a lumen containing RBCs.

Regarding the elucidation of a VM mechanism the interpretation of the literature is arbitrary, at best. In our opinion, most studies that provide a VM mechanism of action are based on in vitro assays that unfortunately need to be discarded, or at best treated with skepticism. To date, mechanistic data have come almost exclusively from in vitro models that wrongfully interpret intercellular connections as formation of VM and therefore should be assessed with caution. On the other hand, VM studies based on immunohistochemistry of tumor sections cannot deliver mechanisms, only association for example enrichment of EMT-related proteins or HIF-1α expression (65, 77, 78). Studies to date have failed to provide a gain-of-function/loss-of-function system for VM either by chemical inhibition or gene silencing.

In summary, reliable in vitro and in vivo VM models are urgently required and need to be universally adopted by the scientific community in order to identify, quantitate, and elucidate the mechanisms behind this phenomenon. The delivery of a clinical marker for VM could serve as a marker for anti-angiogenic treatment refractory patients. Finally, reliable VM models may identify actionable targets and thus finally accomplishing Judah Folkman's dream of total suppression of tumor irrigation.

### DATA AVAILABILITY

All datasets generated for this study are included in the manuscript and/or the **Supplementary Files**.

#### ETHICS STATEMENT

This study was carried out in accordance within the guidelines and recommendations of the Ethics and Bioethics committees of the Servicio de Salud Metropolitana Oriente (15122015) and the Pontificia Universidad Catolica de Chile (resolution 13-226, FONDECYT 1180241, 2018). All subjects gave written informed consent in accordance with the Declaration of Helsinki.

### AUTHOR CONTRIBUTIONS

AV, GM, VA, MR, CR, and FN performed literature searches, experiments, and assisted in preparing the manuscript and figures. MP, AG, AC, and GO designed and wrote the manuscript.

### FUNDING

CONICYT FONDAP-1513001, Millennium Institute on Immunology and Immunotherapy IMII P09/016-F and FONDECYT 1180241, 1191928, and 1181243. Programa de Apoyo a Centros con Financiamiento Basal AFB 170005 and AFB170004 from CONICYT and FONDECYT 1181907.

### ACKNOWLEDGMENTS

We acknowledge the patients who donated their tumor biopsies for research contributing to this paper. Also, the participating medical staff at the University of Chile and the Pontifical Catholic University of Chile. We would also like to acknowledge the Support Team for Oncological Research and Medicine (STORM) in Santiago for their reading and comments on this manuscript.

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fonc. 2019.00680/full#supplementary-material

Supplementary Table SI | Presence of vasculogenic mimicry and cancer patient survival. Papers presenting overall survival cancer patient survival were used for analysis. Patient survival was converted to months to unify the data.

Supplementary Table SII | Vasculogenic mimicry in the literature. In vivo model: Y = Yes; N = No. In vitro model: Y = Yes; N = No. In vivo characterization: 1 = H&E/PAS only; 2 = H&E/PAS-CD31; 3 = H&E/PAS-CD34; 4 = Other (leptin marker, laminin marker, electron microscopy, etc); A = Lumen containing structure only; B = Lumen containing structure with red blood cells; C = VM Pattern (PAS Stain). In vitro characterization: 5 = only light microscopy; 6 = light microscopy/PAS; 7 = electron/confocal microscopy; 8 = lumen presence demonstration; 9 = functionality demonstration (microinjection); W = Others; X = cells projection; Y = tubular structures; Z = Not Shown. Xenograft model: Y = Yes; N = No.

Supplementary Table SIII | Summary of the literature presenting mechanisms of action for vasculogenic mimicry. To be considered to represent a valid mechanism of VM formation the publications had to include the criteria of in vivo and/or in vitro characterization of 2-A, 3-A, 4-A, 2-B, 3-B or 4-B; or 5-Y, 6-Y, 7-Y, 8-Y, 9-Y or W-Y. In vivo model: Y = Yes; N = No. In vitro model: Y = Yes; N = No. In vivo characterization: 1 = H&E/PAS only; 2 = H&E/PAS-CD31; 3 = H&E/PAS-CD34; 4 = Other (leptin marker, laminin marker, electron microscopy, etc); A = Lumen containing structure only; B = Lumen containing structure with red blood cells;

#### REFERENCES


C = VM Pattern (PAS Stain). In vitro characterization: 5 = only light microscopy; 6 = light microscopy/PAS; 7 = electron/confocal microscopy; 8 = lumen presence demonstration; 9 = functionality demonstration (microinjection); W = Others; X = cells projection; Y = tubular structures; Z = Not Shown. Xenograft model: Y = Yes; N = No.

Supplementary Video 1 | Dye Microinjection of Different Structures Formed by the U87 and Hey Cell Line.


required for mimicry of embryonic vasculogenesis by aggressive melanoma. Cancer Res. (2001) 61:6322–7. Retrieved from: Retrieved from: http:// cancerres.aacrjournals.org/content/61/17/6322.long


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Valdivia, Mingo, Aldana, Pinto, Ramirez, Retamal, Gonzalez, Nualart, Corvalan and Owen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Vasculogenic Mimicry: Become an Endothelial Cell "But Not So Much"

Mónica Fernández-Cortés † , Daniel Delgado-Bellido† and F. Javier Oliver\*

CSIC, CIBERONC, Instituto de Parasitología y Biomedicina López Neyra, Granada, Spain

Blood vessels supply all body tissues with nutrients and oxygen, take away waste products and allow the arrival of immune cells and other cells (pericytes, smooth muscle cells) that form part of these vessels around the principal endothelial cells. Vasculogenic mimicry (VM) is a tumor blood supply system that takes place independently of angiogenesis or endothelial cells, and is associated with poor survival in cancer patients. Aberrant expression of VE-cadherin has been strongly associated with VM. Even more, VE-cadherin has constitutively high phosphorylation levels on the residue of Y658 in human malignant melanoma cells. In this review we focus on non-endothelial VE-cadherin and its post-translational modifications as a crucial component in the development of tumor VM, highlighting the signaling pathways that lead to their pseudo-endothelial and stem-like phenotype and the role of tumor microenvironment. We discuss the importance of the tumor microenvironment in VM acquisition, and describe the most recent therapeutic targets that have been proposed for the repression of VM.

#### Edited by:

Laurence A. Marchat, National Polytechnic Institute, Mexico

#### Reviewed by:

Gabriele Multhoff, Technical University of Munich, Germany Fahd Al-Mulla, Genatak, Kuwait

\*Correspondence: F. Javier Oliver joliver@ipb.csic.es

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology

> Received: 21 June 2019 Accepted: 07 August 2019 Published: 22 August 2019

#### Citation:

Fernández-Cortés M, Delgado-Bellido D and Oliver FJ (2019) Vasculogenic Mimicry: Become an Endothelial Cell "But Not So Much". Front. Oncol. 9:803. doi: 10.3389/fonc.2019.00803 Keywords: vasculogenic mimicry, tumor microenvironment, metastasis, VE-cadherin, anti-angiogenesis therapeutic failure, cell plasticity

## BACKGROUND

The concept of neovascularization was described for the first time in 1787 in the context of developmental biology. The term angiogenesis was coined early in the 20th century but was not applied to tumor biology until decades later (1). Angiogenesis can arise in a variety of forms during cancer, namely sprouting angiogenesis, intussusceptive microvascular growth, and glomeruloid microvascular proliferation (2). Emerging studies have shown that a few tumors can grow without need of angiogenesis even in hypoxic conditions, while other tumors display both angiogenic and non-angiogenic regions (3). Vasculogenic mimicry refers to the ability of cancer cells to organize themselves into vascular-like structures for the obtention of nutrients and oxygen independently of normal blood vessels or angiogenesis.

Research in VM has often been surrounded by skepticism to a certain extent. The reason for this controversy is usually the difficulty to distinguish VM channels from endothelial blood vessels or blood lakes in vivo. Similarly, in vitro research presents inherent inconveniences, given the need to develop 3-D models where tumor cells can develop capillary-like structures. Most in vitro research is based on cell cultures using matrigel. However, the tubular structures observed in this model might not always represent in vivo VM. **Table 1** shows a few articles where VM is thoroughly described in various cancer types in vivo and in vitro (4–10, 12, 13).

The unique arrangement of VM grids simulates embryonic vasculogenesis, suggesting that malignant tumor cells acquire an embryonic-like phenotype. Gene expression analysis showed that aggressive tumor cells capable of VM display a diversified gene profile, expressing genes from multiple cell types such as epithelial cells, fibroblasts and endothelial cells (16, 17). Nevertheless, the molecular mechanisms that give rise to VM remain largely unknown.

**58**

TABLE 1 | Main tumor types where VM has been reported and main inhibitors described against VM.


Vascular endothelial cadherin (VE-cadherin), also known as cadherin 5 or CD144, is a cell-cell adhesion protein commonly expressed by endothelial cells. Phosphorylation of VE-cadherin at a number of residues can modulate endothelial junction stability and permeability in different contexts (18). VE-cadherin was believed to be specific for endothelial cells (ECs), though in the past few decades it has been found in a wide variety of tumors, where it crucially contributed to aggressiveness and VM. In fact, VE-cadherin was found in highly aggressive tumor cells but it was not expressed by their poorly aggressive counterparts. Moreover, VE-cadherin down-regulation leads to the inhibition of VM formation (19). In this mini-review, we focus on the new findings on the role of extravascular VE-cadherin and its different localizations in the acquisition of the VM phenotype. In addition, we discuss the influence of the tumor microenvironment, stromal cells and mural cells in the development of VM.

#### VE-CADHERIN IN NON-ENDOTHELIAL CONTEXT

VE-cadherin is a single pass transmembrane protein typically found in endothelium, where it takes part in adherents junctions (20). VE-cadherin has been extensively studied with reference to vascular adhesion, but its function during VM in aggressive tumor cells is not fully understood. The structure of VEcadherin contains five extracellular calcium-dependent domains (aminoacid residues 46-599) which can establish cis homodimers with another VE-cadherin molecule. Similarly, VE-cadherin can form trans dimers binding through the ECDI-ECDIV domains. The intracellular domain of VE-cadherin can undergo several post-translational modifications and 13 different residues of VE-cadherin have been reported to undergo phosphorylation in humans. Of these, residues Y658, S665, Y685, and Y731 have drawn most of the attention. They have been implicated in the intracellular dynamics of the cytoskeleton that control endothelial permeability. In particular, phosphorylation of VE-cadherin can trigger junctional changes via VE-cadherin internalization. As a result, vascular permeability is increased, allowing intravasation and extravasation of different cell types, including tumor cells (21). Recent investigations demonstrated that focal adhesion kinase (FAK) can phosphorylate VE-cadherin at Y658 in tumor-associated ECs, pinpointing the importance of FAK in regulating EC barrier function and hence tumor metastasis (22). FAK is a cytoplasmic tyrosine kinase co-activated by vascular endothelial growth factor receptor (VEGFR) 2 and integrin in the control of vascular permeability (23). Recently, we reported that human aggressive melanoma cells have a constitutively high FAK-dependent phosphorylation of VEcadherin at Y658 (pY658-VEC). pY658-VEC interacts with p120 catenin and the transcriptional repressor kaiso in the nucleus. The inhibition of FAK led to the release of kaiso, promoting its recruitment to kaiso binding sites and therefore repressing kaiso target genes. Moreover, the repression of kaiso target genes CCDN1 and WNT11 abrogated VM. In that line, uveal

melanoma cells genetically deficient for VE-cadherin (either through CRISPR/Cas9 technology or after silencing of VEcadherin) lost the ability to develop VM. Even more, the rescue of WT-VE-cadherin reverted the ability to form VM; in contrast, expression of the non-phosphorylated Y658F-VEcadherin blunted in vitro VM (24) (see **Figure 1**).

The intracellular domain V of VE-cadherin is necessary to bind vascular endothelial protein tyrosine phosphatase (VE-PTP) (25), in a plakoglobin (γ-catenin)-dependent way (26, 27). VE-PTP is an endothelial receptor-type phosphatase which was first described in relation to its implications in embryonic vasculature. VE-PTP-deficient mice undergo vasculogenesis but still die at the embryonic stage due to angiogenesis malfunction (28). VE-PTP may have many other implications, such as ocular vascular pathology (29), blood vessels development (30), breast cancer vasculature and metastatic progression (31). VE-PTP is also involved with the TIE1/2-Angiopoietin pathway. Different laboratories have shown that targeting VE-PTP with a specific inhibitor (AKB-9978, Aerpio Pharmaceuticals) activates TIE2 and stabilizes the ocular vasculature in ischemic/inflammation models (29). AKB-9778 induced TIE2 phosphorylation, directly as well as via ANG1. Furthermore, other signaling components of the TIE2 pathway, such as ERK, AKT, and eNOS (see **Figure 1**), also display increased phosphorylation (29). In 2015, Gong et al. reported that hypoxia increased the expression of VE-PTP in acute lung injury in a HIF-2α-dependent manner (32). All of these functions of VE-PTP have always been studied in endothelial models, though the implications of the VE-PTP/VE-cadherin axis in a VM context remain unknown. In view of the anomalously high levels of phospho-VE-cadherin in cells undergoing VM and the role of VE-PTP in the exquisite maintenance of VE-cadherin phosphorylation, new studies should address the function of this phosphatase in regulating vasculogenic mimicry.

### TUMOR MICROENVIRONMENT AND VM

The complexity of tumors has been increasingly acknowledged in the past decades, to the point where numerous articles published in the cancer research field are no longer focused exclusively on cancer cells. On the contrary, the different components of the tumor microenvironment have received ever greater attention.

Low oxygen concentration in tumors, commonly known as tumor hypoxia, has been repeatedly associated with malignancy, metastasis and therapy resistance in cancer (33). Hypoxia has been linked to VM by many research groups as well (15). In hepatocellular carcinoma, hypoxia promoted VM through transcriptional co-activation of Bcl-2 and Twist1 (34). Nuclear co-expression of Bcl2 and Twist1 correlated with VE-cadherin expression in tumor cells. In fact, VE-cadherin gene expression can be induced by hypoxia, specifically by hypoxia inducible factor (HIF) 2α (35). Though it was first described in ECs, hypoxia-driven expression of VE-cadherin has been reported in a large number of tumor types too (36–38), where it is always involved in a promotion of the VM phenotype.

Hypoxia has been shown to promote VM through other signaling pathways apart from VE-cadherin. For instance, reactive oxygen species (ROS)-mediated stabilization of HIF1α activated the met proto-oncogene, which induced in vitro tube formation on matrigel in melanoma cells (39). Moreover, HIF1 and HIF2α promoted in vitro tube formation on matrigel through the upregulation of vascular endothelial growth factors (VEGF) C and D, as well as VEGF receptor (VEGFR)3 (40). In triple negative breast cancer, hypoxia increased the subpopulation of CD133<sup>+</sup> cells (commonly regarded as cancer stem cells) through a Twist1-mediated mechanism. This population shift seemed to enhance tube formation, since CD133<sup>+</sup> cells were found to line the VM-like tubes (41). In addition, HIF1α could promote tube formation in hepatocellular carcinoma by up-regulating lysyl oxidase like 2 (LOXL2) (42, 43), which is involved in collagen cross-linkage during extracellular matrix (ECM) remodeling.

ECM per se can play a fundamental role in regulating VM. The NC11 domain in collagen XVI could trigger tube formation in oral squamous cell carcinoma, since it could induce VEGFR1/2 expression (43). On the contrary, the presence of collagen I altered the vascular potential of pancreatic ductal adenocarcinoma (PDAC) CSCs, decreasing the secretion of proangiogenic factors and the expression of VEGFR2, altogether hindering VM formation in PDAC (44). Furthermore, Velez et al. showed that ECM architecture can influence VM; in particular, collagen matrices with small pores and short fibers induced β-integrin expression and hence VM (45).

The importance of non-cancer cells within the tumor stromal is slowly gaining attention in the study of VM as well. Tumor-associated macrophages (TAMs) seemed to promote VM formation in glioblastoma multiforme, namely increasing the expression of cyclooxygenase 2 in the tumor cells (46). Cancer-associated fibroblasts (CAFs) can be determinant in VM formation too. In a recent study, vasculogenic murine melanoma cells were injected in mice carrying a CAF-specific deletion for the matricellular protein CCN2. As a result, the absence of fibroblast-derived CCN2 reduced tumor vasculature, including VM (47). Finally, a recent publication by Thijssen et al. (48) showed that PAS<sup>+</sup> tissues in human cutaneous melanoma stained positive for pericyte marker α-smooth muscle actin (αSMA) within the ECM networks lined by tumor cells. Furthermore, when VM<sup>+</sup> tumor cells were co-cultured with pericytes, there was a stabilization of the VM networks for up to 96 h. Pericyte recruitment to VM networks was shown to be dependent on PDBF-B signaling, whereas the addition of STI-571 (imatinib mesylate) to inhibit PDGF receptor hindered VM as well as tumor growth.

#### TARGETING VM AND PERSPECTIVES

A meta-analysis of 22 clinical studies derived from data concerning VM and 5-year survival of 3,062 patients across 15 cancer types showed that tumor VM is correlated with poor prognosis (49). Anti-angiogenic therapies (preferably, antibodies against VEGF receptor bevacizumab and related) against tumor development have had limited results so far. Therefore, the development of novel anti-tumor neovascularization strategies is of vital importance, expanding the targets from conventional angiogenesis to all the alternative mechanisms recently discovered, such as VM (50).

A unique small molecular compound with particular interest in anti-VM cancer treatment is CVM-1118, which is currently undergoing clinical trials (NCT03582618). CVM-1118 is classified as a phenyl-quinoline derivative, whose core structure displays potent anti-neoplastic and anti-mutagenic properties (51).

As mentioned above, pharmacological inhibition of activity of FAK/Y658 VE-cadherin with PF-271 may represent a new therapeutic opportunity in the repression of genes involved with VM promotion in cancer cells. Similarly, inhibition of the VE-PTP/TIE-2 pathway with AKB-9778 could open new ways to control the capacity to form pseudo-vessels by vascular mimicry cells. Finally, new treatments targeting mural cells, such as pericytes, could also have therapeutic value. It is the case of targeting PDGF-B axis with STI-571, which proved useful in VM mice models. Targeting VM with specific molecular compounds combined with front-line therapies may represent the best approach to obtain a good prognosis in patients in the future.

#### AUTHOR CONTRIBUTIONS

MF-C and DD-B designed and wrote the review. FO designed, wrote and coordinated the review.

#### FUNDING

This work was supported by the grants from the Spanish Ministry of Economy and Competitiveness SAF2015-70520-R, Fundación Domingo Martínez and the Spanish Ministry of Science and Technology RTI2018-098968-B-I00 and CIBERONC ISCIII CB16/12/00421 to FO.

## REFERENCES


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Fernández-Cortés, Delgado-Bellido and Oliver. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Integrated Transcriptome Analysis Reveals KLK5 and L1CAM Predict Response to Anlotinib in NSCLC at 3rd Line

Jun Lu1†, Qin Shi 2†, Lele Zhang1†, Jun Wu<sup>3</sup> , Yuqing Lou<sup>1</sup> , Jie Qian<sup>1</sup> , Bo Zhang<sup>1</sup> , Shuyuan Wang<sup>1</sup> , Huimin Wang<sup>1</sup> , Xiaodong Zhao<sup>4</sup> \* and Baohui Han<sup>1</sup> \*

<sup>1</sup> Department of Pulmonary Medicine, Shanghai Chest Hospital, Shanghai Jiao Tong University, Shanghai, China, <sup>2</sup> Department of Oncology, Baoshan Branch of Shuguang Hospital, Shanghai University of Traditional Chinese Medicine, Shanghai, China, <sup>3</sup> School of Life Science, East China Normal University, Shanghai, China, <sup>4</sup> Shanghai Center for Systems Biomedicine, Shanghai Jiao Tong University, Shanghai, China

#### Edited by:

César López-Camarillo, Universidad Autónoma de la Ciudad de México, Mexico

#### Reviewed by:

Alma D. Campos-Parra, National Institute of Cancerology (INCan), Mexico Daniel Sotelo, Autonomous University of Guerrero, Mexico

#### \*Correspondence:

Xiaodong Zhao xiaodongzhao@sjtu.edu.cn Baohui Han 18930858216@163.com; xkyyhan@gmail.com

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology

Received: 16 June 2019 Accepted: 27 August 2019 Published: 11 September 2019

#### Citation:

Lu J, Shi Q, Zhang L, Wu J, Lou Y, Qian J, Zhang B, Wang S, Wang H, Zhao X and Han B (2019) Integrated Transcriptome Analysis Reveals KLK5 and L1CAM Predict Response to Anlotinib in NSCLC at 3rd Line. Front. Oncol. 9:886. doi: 10.3389/fonc.2019.00886 The oral multi-targeted tyrosine kinase inhibitor (TKI) anlotinib is effective for non-small cell lung cancer (NSCLC) in clinical trials at 3rd line. However, a fraction of patients remains non-responsive, raising the need of how to identify anlotinib-responsive patients. In the present study, we aimed to screen potential biomarkers for anlotinib-responsive stratification via integrated transcriptome analysis. Comparing with the anlotinib-sensitive lung cancer cell NCI-H1975, we found 1,315 genes were differentially expressed in anlotinib-resistant NCI-H1975 cells. Among the enriched angiogenesis-related genes, we observed high expression of KLK5 and L1CAM was mostly associated with poor clinical outcomes in NSCLC patients through Kaplan-Meier survival analysis in a TCGA cohort. Moreover, an independent validation in a cohort of ALTER0303 (NCT02388919) indicated that high serum levels of KLK5 and L1CAM were also associated with poor anlotinib response in NSCLC patients at 3rd line. Lastly, we demonstrated that knockdown of KLK5 and L1CAM increases anlotinib-induced cytotoxicity in anlotinib-resistant NCI-H1975 cells. Collectively, our study suggested serum levels of KLK5 and L1CAM potentially serve as biomarkers for anlotinib-responsive stratification in NSCLC patients at 3rd line.

#### Keywords: KLK5, L1CAM, anlotinib, non-small cell lung cancer, transcriptome

### INTRODUCTION

Biomarkers play an important role in therapies of non-small cell lung cancer (NSCLC). Genomic features, such as gene amplification, point mutations, gene over-expression, and chromosomal translocation, have been identified as biomarkers in NSCLC (1). NSCLC, as the leading cause of cancer mortality worldwide, has greatly benefited from biomarker investigations. Precision therapies have dramatically improved progression free survival (PFS) and overall survival (OS) of NSCLC patients whose tumors harbor positive driver gene mutations, such as EGFR (19 Del and L858R) (2), rearranged ROS1 (3), or translocated ALK (4). Furthermore, immune checkpoint inhibitors have significantly prolonged PFS and OS in specific advanced NSCLC patients, due to the assessment of PD1/PDL1 expression and tumor mutation burden (TMB) (5–7). Therefore, biomarkers for drug-responsive stratification play crucial roles in NSCLC precision therapy.

**64**

Anlotinib is an oral multi-targeted tyrosine kinase receptor inhibitor (TKI) that was recently discovered (8–12). Anlotinib has exhibited its efficacy in various tumor cell line-derived xenograft animal models (9, 11). Clinical trials have revealed that anlotinib is a potent inhibitor for NSCLC therapy at 3rd line (8, 10, 12). Moreover, mechanistic studies indicated that anlotinib is actively involved in anti-angiogenesis and may selectively inhibit VEGFR (2/3), PDGFR (α/β), and FGFR (1– 4), and other targets (9–11). Our and other recent studies have revealed some potential biomarkers for anlotinib stratification (13–17). However, due to the complex architecture of angiogenic signaling, the biomarkers for anlotinib-responsive stratification remain further exploration. With the aim of screening potential biomarkers for anlotinib-responsive stratification, in this study we performed integrated transcriptome analysis on anlotinibresistant NCI-H1975 cells and NSCLC patients both in a TCGA cohort, and examined the stratifying effects in an anlotinib clinical trial cohort (NCT02388919).

### MATERIALS AND METHODS

#### Cell Culture

Human NSCLC cell lines NCI-H1975, PC-9, HCC-827, and A549 were obtained from the ATCC: The Global Bioresource Center (https://www.atcc.org/). All cell lines were validated to exclude mycoplasma contamination using a TransDetect PCR

evaluated by flow cytometry. (E) Early apoptosis and total apoptosis ratios were analyzed based on the results of flow cytometry. Bars = mean ± SD, n = 3, \*\*\*P < 0.001. (F,G) Anlotinib-resistant NCI-H1975 cells were exposed to anlotinib (4µg/ml) for 24 h. Apoptotic processes were examined by flow cytometry. Analysis of early apoptosis and total apoptosis were performed based on flow cytometric detection. Bars = mean ± SD, n = 3. (H) A transwell assay was performed to evaluate NCI-H1975 cell invasion with or without anlotinib (2µg/ml) for 24 h. (I) A transwell assay was performed to evaluate anlotinib-resistant NCI-H1975 cell invasion with or without anlotinib (2µg/ml) for 24 h. (J) Statistical analysis of invasion ratios on NCI-H1975 cells and anlotinib-resistant NCI-H1975 cells. Bars = mean ± SD, n = 5, \*\*\*P < 0.001.

Mycoplasma Detection Kit (TransGen, China). The cells were cultured in RPMI 1640 medium (Gibco, USA) supplement with 10% FBS (Gibco, USA), 0.1 mg/ml streptomycin and 100 U/ml penicillin. All cells were incubated at 37◦C and 5% CO<sup>2</sup> in a humidified incubator.

### Cell Viability Analysis

In total, 1,500 cells per well were cultured in 96-well plates. After incubating with culture medium overnight, the cells were then exposed to anlotinib for 24 h. CCK8 (Dojindo, Japan) was used to evaluate cell viability according to the manufacturer's protocol. The absorbance was measured at 450 nm using a spectrophotometric plate reader (Bio-Tek, USA). Cell viability was performed according to our previous studies (18, 19).

### Establishment of an Anlotinib-Resistant NCI-H1975 Cell Line

As our previous study described (20), as shown in **Figure 1A**, 10<sup>7</sup> NCI-H1975 cells were exposed to 100 mg/ml ENU (Sigma, USA) for 24 h. Anlotinib administration was performed to screen anlotinib-resistant NCI-H1975 cells. For the first 5 days, NCI-H1975 cells were exposed to anlotinib (4µg/ml) and the medium was changed every day. Then, anlotinib (6, 8, 10, and 12µg/ml) treatments were performed over the next two months. The resulting cells (approximately 100 cells) showed viability when exposed to anlotinib (12µg/ml). After approximately 1 month of culture, the anlotinib-resistant NCI-H1975 cells were used in functional assays.

### Cell Apoptosis Analysis

In total, 5 × 10<sup>5</sup> cells per well of NCI-H1975 or anlotinibresistant NCI-H1975 were cultured in six-well plates for 24 h. Then, the cells were exposed to anlotinib for 24 h. To assess the apoptosis rate, an Annexin V-FITC/PI Apoptosis kit (Zoman Biotechnology Co., Ltd, China) was used to determine the phosphatidyl serine and membrane integrity of each cell. Briefly, the anlotinib-treated and anlotinib-untreated cells were stained with annexin V-FITC and PI simultaneously and then detected by flow cytometry (BD LSRFortessa, USA). The ratio of early apoptosis and total apoptosis were analyzed by FlowJo 7.6 (BD, USA).

#### Cell Invasion Analysis

Cell invasion was evaluated by transwell assay. One day before the experiment, all cells were incubated in RPMI 1640 (Gibco, USA) for starvation. 5 × 10<sup>4</sup> NCI-H1975 cells or anlotinib-resistant NCI-H1975 cells per well were then seeded on the top pre-coated chamber in 100 µl RPMI 1640. Five hundred microliter RPMI 1640 containing 15% FBS was added into the lower chamber. After 24 h of incubation, the non-invasive cells were cleaned, and the invasive cells were fixed with 4% PFA for 30 min. The invasive cells were stained with 0.1% crystal violet (Sigma, USA), and photographed using fluorescence microscopy (Nikon, Japan).

### RNA-seq Library

The preparation of RNA-seq library was performed according to our previous studies (18, 21, 22). Briefly, NCI-H1975 cells or anlotinib-resistant NCI-H1975 cells were cultured in 10 cm dishes. Then, 1 ml Trizol reagent (Life Technologies, Inc., USA) was used to lyse the cell samples, followed by total RNA isolation using standard procedures. An Oligotex mRNA Mini Kit (Qiagen, Germany) was used to purify mRNA. Approximately 100 ng mRNA of each sample was used for reverse-transcription, followed by end repair, ligation, using NEBNext Ultra Directional RNA Library Prep Kit (NEB, USA) and PCR amplification (12 cycles) using Q5 High-Fidelity DNA Polymerase (NEB, USA). Lastly, the PCR products were subjected to Illumina sequencing by Next 500 (Illumina, USA). All raw data were deposited at EMBL database under accession number E-MTAB-5997 and E-MTAB-7068.

### Bioinformatics Analysis

Raw sequencing data were mapped to a reference genome (hg38) by Tophat. Cufflinks was used to determine the differential transcription pattern. Kilo-base of per million reads mapped (RPKM) was used to define gene expression level. To screen significant differential genes, we filtered the genes whose gene expression levels were no more than a 2-fold change. Log<sup>2</sup> (Fold Change) > 1 represented at least 2-fold up-regulation, and log<sup>2</sup> (Fold Change) < −1 represented at least 2-fold down-regulation. Gene ontology (GO) analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis were performed using a public bioinformatics resource platform (DAVID, https://david. ncifcrf.gov/) by uploading the differential gene lists.

### Quantitative Real-Time PCR

Total RNA extraction and reverse transcription reactions were performed according to our previous studies (18, 21, 22). Briefly, the mRNA levels of the genes of interest were detected by quantitative real-time PCR (RT-qPCR) using ABI step one plus (Applied Biosystems, USA). GAPDH was used as a control gene for normalization. The relative levels of mRNA were calculated as 211Ct. All primer sequences used for RT-qPCR are listed in **Table S1**.

### Transcriptome Analysis of the TCGA Cohort

RNA-seq data and clinical data for NSCLC patients [including lung adenocarcinoma (LUAD) and lung squamous carcinoma (LUSC)] were downloaded from the TCGA portal (https:// cancergenome.nih.gov/). The RNA-seq data of normal controls were excluded based on TCGA barcode principle (https://wiki. nci.nih.gov/display/TCGA/TCGA\$\pm\$barcode). After filtering the unqualified samples, 503 LUAD patients and 494 LUSC patients were used for survival analysis. The method of raw data collection was described by the Cancer Genome Atlas Research Network. The correlation analysis of RPKM values and overall survival was performed by R package (survival, version. 2.41- 3). Best cutoff value was determined using the "Ward method." Briefly, to determine the P-value, we detected the correlations between each mRNA level and OS. The cutoff value was defined as the lowest P-value.

#### RNA Interference

RNA interference was performed according to our previous studies (18, 22). NCI-H1975 cells and anlotinib-resistant NCI-H1975 cells were transfected with KLK5 siRNA (5′ - GCAUGUUCUCGCCAACAAUTT-3′ ) or L1CAM siRNA (5′ - CAGCAACUUUGCUCAGAGGTT-3′ ) when they reached 50% confluence, using the Lipofectamine 3000 reagent (Invitrogen, USA). An unrelated, scrambled siRNA was used as a negative control (5′ -UUCUCCGAACGUGUCAGGUTT-3′ ).

### Detection of Serum KLK5 and L1CAM Levels

Twenty-eight peripheral blood samples from patients with refractory advanced NSCLC (time since prior anlotinib treatment: 2 weeks; Registered No. NCT02388919) were provided by Chia-tai Tianqing Pharmaceutical Co Ltd, Jiangsu Province, China. These samples were selected from 294 anlotinib clinical trial participants randomly. The participants received anlotinib as 3rd line or after 3rd line therapy. For each cycle of medication, the patients received anlotinib (12 mg/day) for 2 consecutive weeks and then discontinued for 1 week. Anlotinib treatment was terminated at disease progression or if intolerable toxicity occurred. The patients with stable disease or partial response lasting 80 days were defined as responders while those patients with disease progression ≤80 days were defined as non-responders. The patients harboring any driver mutations (detected by standard methods afforded by participant hospitals), such as EGFR, ROS1, and ALK, were defined as positive. All enrolled patients were followed up every 8 weeks until death. The ELISA kit for KLK5 detection was purchased from Abcam. Serum L1CAM levels were determined using

genes detected by RT-qPCR in NCI-H1975 cells and anlotinib-resistant NCI-H1975 cells.

the DRG Diagnostics ELISA kit (Marburg, Germany). All experimental procedures were performed according to the manufacturer's protocols.

### Specificity and Sensitively Analysis

For the TCGA cohort of NSCLC patients, the receiver operator characteristic (ROC) curve for predicting OS was generated by the cutoff value of the mRNA level using GraphPad Prism (GraphPad software, version 5, USA). For anlotinib response prediction, the ROC curves for predicting PFS and OS were generated by the cutoff value of the serum protein level.

### Statistical Analysis

There were at least three biological replicates, excluding RNAseq analysis, for each sample. PFS and OS were summarized as median values and were analyzed using the Kaplan-Meier method. The Mantel-Cox test was used to perform Meier survival analysis in GraphPad Prism 5. Log-rank test, two-tailed Student's t-test, or one-way ANOVA with post-hoc Bonferroni correction were used to examine the raw data. Differences were considered significant at <sup>∗</sup>P < 0.05, ∗∗P < 0.01, and ∗∗∗P < 0.001.

## RESULTS

#### Anlotinib-Induced Cytotoxicity Disappeared in Anlotinib-Resistant NCI-H1975 Cells

To verify the anti-tumor effects of anlotinib, we administered anlotinib to NSCLC cell lines (including NCI-H1975, PC-9, HCC827, and A549). After exposure to anlotinib (8µg/ml) for 24 h, the cell viabilities of those NSCLC cell lines decreased to different degrees (**Figure 1B**). Among the various lines of NSCLC cells, the NCI-H1975 cells underwent the most cytotoxicity. To investigate the effect of anlotinib resistance, we established anlotinib-resistant NCI-H1975 cells in vitro (see methods). We treated NCI-H1975 cells and anlotinib-resistant NCI-H1975 cells with anlotinib simultaneously and then examined the cell viability, cell apoptosis, and cell invasion activity. Under the anlotinib (6µg/ml) stress, the viability of NCI-H1975 cells decreased remarkably, and the similar phenomenon was not observed in anlotinib-resistant NCI-H1975 cells (**Figure 1C**). Furthermore, after exposure to anlotinib (4µg/ml) for 24 h, the apoptosis rate of NCI-H1975 cells increased significantly, while the apoptosis rate of anlotinib-resistant NCI-H1975 cells almost remained unchanged (**Figures 1D–G**). Consistent with the above results, the invasive ability of anlotinib-resistant NCI-H1975 cells was virtually unaffected, although cells were also exposed to anlotinib (2µg/ml) for 24 h (**Figures 1H–J**). These results suggest that anlotinib resistance in NCI-H1975 cells might be attributed to activation/inactivation of tumor survival-related biological processes or signaling pathways.

### Transcriptome Analysis Revealed Anlotinib Resistance in NCI-H1975 Cells Attributed to the Expressions of Angiogenesis-Related Genes

To understand the underlying molecular mechanism of anlotinib resistance in NCI-H1975 cells, we next performed transcriptome profiling analysis on both NCI-H1975 cells and anlotinibresistant NCI-H1975 cells. The analysis flowchart was shown in **Figure 2A**. In total, 14,312 differentially expressed genes were found. After excluding inactive genes (fold change ≤ 2), 595 up-regulated genes and 720 down-regulated genes were obtained for subsequent analysis (**Figure 2B**). Compared with wild type (sensitive cell line), a considerable fraction of genes is differentially expressed in anlotinib-resistant NCI-H1975 cells (**Figure 2C**). GO and KEGG analysis indicated that the up-regulated genes and down-regulated genes are enriched in multiple biological processes (extracellular matrix organization/disassembly, angiogenesis, cell adhesion, and so on) or signaling pathways (ECM-receptor interaction, antigen processing and presentation, viral carcinogenesis, and so on), suggesting that the modulation of these enriched genes may play an important role in the process of anlotinib resistance (**Figure 2D**, **Table S2**). Further analysis suggested that modulation of angiogenesis-related genes (including ANGPTL4, FN1, HSPG2, SRPX2, KLK5, L1CAM, Prr22, FOXJ1, IL24, and TRIM54) potentially contributes to anlotinib resistance (**Figures 2D,E, Tables S3, S4**), as anlotinib is a multi-targeted anti-angiogenesis drug for cancer therapy (8–10, 12, 23).

### High mRNA Levels of KLK5 and L1CAM Are Associated With Poor Clinical Outcomes in NSCLC Patients in the TCGA Cohort

To understand the clinical significances of the angiogenesisrelated genes identified above, we performed survival analysis on NSCLC patients from the TCGA cohort. Kaplan-Meier survival analysis indicated that high mRNA levels of ANGPTL4, FN1, HSPG2, and SRPX2 are associated with poor clinical outcome significantly (**Figure S1**). However, we also found that downregulation of Prr22, FOXJ1, IL24, and TRIM54 is also correlated with poor clinical outcome in NSCLC patients (including LUAD and LUSC) (**Figure S2**). Moreover, our Kaplan-Meier survival analysis showed that high mRNA levels of KLK5 and L1CAM are most significantly associated with poor clinical outcome of NSCLC patients (including LUAD and LUSC) in the TCGA cohort (**Figures 2E**, **3**, **Figures S3A–F**). Collectively, these results indicated that the activation of KLK5 and L1CAM most likely to result in poor clinical outcome in NSCLC patients and the anlotinib resistance in NCI-H1975 cells.

#### Serum Levels of KLK5 and L1CAM Predict Response to Anlotinib in NSCLC Patients

To determine whether serum levels of KLK5 and L1CAM potentially serve as biomarkers for anlotinib-responsive stratification in NSCLC patients at 3rd line, we detected the serum KLK5 and L1CAM levels at baseline in 28 refractory advanced NSCLC patients enrolled in an anlotinib clinical

outcome in NSCLC patients (including LUAD patients and LUSC patients). The cutoff value of L1CAM was determined by the Ward method in NSCLC, LUAD, and LUSC. (D) L1CAM expression is associated with OS in NSCLC patients (including LUAD patients and LUSC patients). NSCLC: n = 997, log rank p < 0.0001; LUAD: n = 503, log rank p = 0.004; LUSC: n = 494, log rank p = 0.003.

trial (NCT02388919), and then performed response analyses. Previous study has revealed that serum levels of L1CAM could be used as an unfavorable prognostic marker in NSCLC patients (24). However, the implications of KLK5 levels vary in different cancers (25–28). Our raw data including the clinical information and levels of KLK5 and L1CAM were shown in **Figure 4A**. Further Kaplan-Meier survival analysis suggested that low levels of serum KLK5 in NSCLC patients had a better response to anlotinib than those patients with a high level of serum KLK5 [Low (n = 11), Median PFS = 166 days vs. High (n = 17), Median PFS = 44 days, P = 0.008] (**Figure 4B**). The NSCLC patients with low levels of serum KLK5 had greater OS benefit from anlotinib treatment [Low (n = 11), Median OS = 315 days vs. High (n = 17), Median PFS = 240 days, P = 0.031] (**Figure 4C**). Furthermore, Kaplan-Meier survival analysis was performed to examine the predictive value of serum L1CAM level at baseline, and the results indicated the NSCLC patients with low serum L1CAM levels had better PFS and OS [PFS: Low

levels in the NSCLC patients treated with anlotinib. n = 28, Cutoff-High: 17 patients, Cutoff-Low: 11 patients; PFS: Log rank p = 0.008, OS: Log rank p = 0.031. (D,E) Kaplan-Meier curves of PFS and OS via stratifying the serum L1CAM levels in advanced refractory NSCLC patients treated with anlotinib. n = 28, Cutoff-High: 11 patients, Cutoff-Low: 17 patients; PFS: Log rank p = 0.002, OS: Log rank p = 0.038.

(n = 17), Median PFS = 166 days vs. High (n = 11), Median PFS = 44 days, P = 0.002; OS: Low (n = 17), Median OS = 259 days vs. High (n = 11), Median OS = 163 days, P = 0.038] (**Figures 4D,E**). The sensitivity and specificity analysis also confirmed that serum KLK5 and L1CAM levels at baseline had preferable predictive value for anlotinib response (**Figure S3**).

#### Knockdown of KLK5 or L1CAM Increases the Sensitivity of NCI-H1975 Cells and Anlotinib-Resistant NCI-H1975 Cells to Anlotinib

To further investigate the roles of KLK5 and L1CAM in anlotinib resistance, we performed RNA interference assays to evaluate anlotinib-induced cytotoxicity in anlotinibresistant NCI-H1975 cells. When anlotinib was administered, knockdown of KLK5 or L1CAM significantly decreased the cell viabilities of anlotinib-resistant NCI-H1975 cells (**Figure 5A**). Meanwhile, anlotinib-induced apoptosis increased significantly, with combined knockdown of KLK5 or L1CAM (**Figures 5B,C**). Consistent with these results, the invasive ability of anlotinib-resistant NCI-H1975 cells decreased remarkably, after anlotinib administration and knockdown of KLK5 or L1CAM were performed simultaneously (**Figures 5D,E**). These data indicated that anlotinib-induced cytotoxicity was partially recovered in anlotinib-resistant NCI-H1975 cells after KLK5 or L1CAM knockdown.

### DISCUSSION

Previous studies have demonstrated that anlotinib prolongs PFS and OS in refractory advanced NSCLC patients in clinical trials and indicated that anlotinib may play an important

role in anti-angiogenesis and proliferation inhibition (10, 12, 23). These anti-tumor effects may be attributed to anlotinib selectively inhibiting tyrosine kinase receptors, including VEGFR (2/3), PDGFR (α/β), FGFR (1–4), etc. (9–12). Our and other recent studies have been revealed some potential biomarkers for anlotinib stratification (13–15, 17). However, the underlying biomarker for predicting anlotinib-responsive NSCLC patients remain further exploration because of the complex architecture of angiogenic signaling. To address this issue, in this study we sought to screen valuable biomarkers via integrated transcriptome analysis.

Drug resistance is inevitable in the last stage of all anti-tumor drug-related therapeutic regimes (29). Cancer cells can acquire resistance to the anti-tumor drugs by various mechanisms, including over-expression or mutation of the drug target, activation of pro-survival pathways, and eliminative induction of cell death (30). For example, studies have demonstrated the mechanisms of acquired resistance to 1st generation TKIs in NSCLC patients with a positive EGFR mutation, including EGFRT790M mutation, MET amplification, HER-2 mutation, HGF over-expression, etc. (31). In other words, NSCLC patients are not suitable for 1st generation TKI therapy when primary tumors harbor resistant mutations or over-expression. These genomic alterations have been used as biomarkers for antitumor drug-responsive stratification (32, 33). Similarly, acquired resistance to anlotinib has been observed in our clinical trials (8, 23). Here, we established anlotinib-resistant NCI-H1975 cells and then demonstrated the resistant effects in vitro. Investigation of anlotinib-resistant NCI-H1975 cells may contribute to screening for biomarkers for anlotinib-responsive stratification in NSCLC patients at 3rd line.

Biomarkers play an important role in precision therapy for NSCLC patients. According to gene mutation types, tumor driver gene-derived inhibitors (including EGFR inhibitor, ROS1 inhibitor, and ALK inhibitor) have been screened and used for stratifying treatments in NSCLC patients (2–4). Furthermore, positive PD1/PD-L1 expression and TMB will be used as biomarkers for guiding treatment with immune checkpoint inhibitors in advanced NSCLC patients (5–7). Next generation sequencing (NGS) provided the platform for screening the above biomarkers (6, 7). Our transcriptome analysis suggested that up-regulation of angiogenesis-related genes contributed to anlotinib resistance. Kaplan-Meier survival analysis in the TCGA cohort indicated that the NSCLC patients harboring high mRNA levels of angiogenesis-related genes (including ANGPTL4, FN1, HSPG2, and SRPX2) have poorer prognosis, suggesting that those patients may be unsuitable for anlotinib therapy.

KLK5 and L1CAM play important roles in cancer progression (including cell proliferation, migration, angiogenesis, invasion, and metastasis) (34, 35), and their expression levels are associated with prognosis. KLK5 not only regulates KRT19 expression to increase the malignancy of ovarian cancer cells strongly (36), but also induces miRNA-mediated anti-oncogenic pathways in breast cancer (37). However, KLK5 plays different roles in different cancers (38, 39). The analysis of correlation between KLK5 expression and prognosis indicated that higher KLK5 mRNA level could sever as indicator for predicting unfavorable prognosis in ovarian cancer patients (25, 26) and breast cancer patients (28) and sever as indicator for predicting favorable prognosis in prostate cancer patients (27) and testicular cancer patients (39). L1CAM has been characterized as an important pro-angiogenesis molecular via regulating metalloproteinase expression (40). More important, higher serum L1CAM levels have been described as an unfavorable prognostic marker in NSCLC patients (24). Our data indicated that knockdown of KLK5 or L1CAM contributes to increased anlotinibinduced cytotoxicity upon anlotinib-resistant NCI-H1975 cells. Furthermore, our results indicated that up-regulated mRNA levels of KLK5 and L1CAM are simultaneously associated with anlotinib resistance in NCI-H1975 cells and poor prognosis in NSCLC patients. Although the two cohorts (TCGA and ALTER0303) there may be differences in the population profile, but, here we found that low serum levels of KLK5 and L1CAM at baseline are favorable biomarkers for anlotinib-responsive stratification in NSCLC patients (ALTER0303 cohort) at 3rd line.

Collectively, our integrated transcriptome analysis revealed that high mRNA levels of KLK5 and L1CAM are candidate biomarkers for predicting OS in NSCLC patients. High serum KLK5 and L1CAM levels are potentially associated with poor anlotinib response in NSCLC at 3rd line. Knockdown of KLK5 and L1CAM contributes to increasing sensitivity to anlotinib upon anlotinib-resistant NCI-H1975 cells. Collectively, this study suggested serum levels of KLK5 and L1CAM have the potential for clinical application for anlotinibresponsive stratification.

#### REFERENCES


## DATA AVAILABILITY

The datasets generated for this study can be found in the EMBL database under accession number E-MTAB-5997 and E-MTAB-7068.

### AUTHOR CONTRIBUTIONS

Experiments were conceived and designed by BH, XZ, and JL. Cell assays were performed by JL, QS, BZ, JQ, SW, YL, and LZ. Bioinformatics analysis and statistical analysis were performed by LZ, JW, and JL. The manuscript was written by JL and revised by HW, XZ, and BH.

## FUNDING

This work was supported by Shanghai Leading Talents Program (2013), a programme at the system biomedicine innovation centre of Shanghai Jiao Tong University (Project No. 15ZH4009), a key programme of translational medicine from Shanghai Jiao Tong University School of Medicine (Project No. 15ZH1008), and a project for the Science and Technology Commission of Shanghai Municipality (Project No. 16140902700).

### ACKNOWLEDGMENTS

The authors would like to thank Prof. Liming Lu for their helpful discussion and careful proofreading.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fonc. 2019.00886/full#supplementary-material


non-small cell lung cancer therapy. Eur Respir J. (2018) 53:1801562. doi: 10.1183/13993003.01562-2018


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Lu, Shi, Zhang, Wu, Lou, Qian, Zhang, Wang, Wang, Zhao and Han. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Angiogenesis in Gynecological Cancers: Role of Neurotrophins

Maritza P. Garrido1,2, Ignacio Torres <sup>1</sup> , Margarita Vega1,2 and Carmen Romero1,2 \*

<sup>1</sup> Laboratory of Endocrinology and Reproductive Biology, Hospital Clínico Universidad de Chile, Santiago, Chile, <sup>2</sup> Departamento de Obstetricia y Ginecología, Facultad de Medicina, Universidad de Chile, Santiago, Chile

Angiogenesis, or generation of new blood vessels from other pre-existing, is a key process to maintain the supply of nutrients and oxygen in tissues. Unfortunately, this process is exacerbated in pathologies such as retinopathies and cancers with high angiogenesis as ovarian cancer. Angiogenesis is regulated by multiple systems including growth factors and neurotrophins. One of the most studied angiogenic growth factors is the vascular endothelial growth factor (VEGF), which is overexpressed in several cancers. It has been recently described that neurotrophins could regulate angiogenesis through direct and indirect mechanisms. Neurotrophins are a family of proteins that include nerve growth factor (NGF), brain-derived growth factor (BDNF), and neurotrophins 3 and 4/5 (NT 3, NT 4/5). These molecules and their high affinity receptors (TRKs) regulate the development, maintenance, and plasticity of the nervous system. Furthermore, it was recently described that they display essential functions in non-neuronal tissues, such as reproductive organs among others. Studies have shown that several types of cancer overexpress neurotrophins such as NGF and BDNF, which might contribute to tumor progression and angiogenesis. Besides, in recent years the FDA has approved the use of pharmacologic inhibitors of pan-TRK receptors in patients with TRKs fusion-positive cancers. In this review, we discuss the mechanisms by which neurotrophins stimulate tumor progression and angiogenesis, with emphasis on gynecological cancers.

Keywords: gynecological cancers, angiogenesis, VEGF, BDNF, NGF

## INTRODUCTION: ANGIOGENESIS IN GYNECOLOGICAL MALIGNANCIES

Gynecological neoplasms belong to a group of malignances that include ovarian, cervical, uterine, fallopian tubes, vulvar, vaginal cancer and gestational trophoblastic neoplasms. The following sections of this review will be focused on the first two types, which are the most frequent (1). Gynecological neoplasms are characterized by exacerbated angiogenesis (which is defined as the generation of new blood vessels from pre-existing ones) and vascular endothelial growth factor (VEGF) is the most widely studied angiogenic factor in the context of cancer. VEGF is secreted by most tumor cells, mainly in response to hypoxia and low nutrient concentrations (2), and promotes angiogenesis through its receptors expressed in endothelial cells. This antecedent has been crucial for the development of new drugs as bevacizumab, a humanized monoclonal antibody directed against human VEGF. Unfortunately, this drug has shown modest results (3), because ovarian and uterine cells may overexpress other molecules that can act as angiogenic factors, such as neurotrophins (NTs) and their receptors (4–7).

#### Edited by:

Laurence A. Marchat, National Polytechnic Institute, Mexico

#### Reviewed by:

Rafael Roesler, Federal University of Rio Grande Do Sul, Brazil Caroline Brunetto De Farias, Children's Cancer Institute (ICI), Brazil

### \*Correspondence:

Carmen Romero cromero@hcuch.cl

#### Specialty section:

This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology

Received: 21 June 2019 Accepted: 02 September 2019 Published: 19 September 2019

#### Citation:

Garrido MP, Torres I, Vega M and Romero C (2019) Angiogenesis in Gynecological Cancers: Role of Neurotrophins. Front. Oncol. 9:913. doi: 10.3389/fonc.2019.00913

**74**

NTs are a group of molecules widely present in the central and peripheral nervous system. They have a key role in developmental neurobiology, by regulating neuronal survival, differentiation, neurites growth, and synthesis of neurotransmitters (8). NTs not only display key roles in neuronal tissues, but also in several non-neuronal tissues, such as mammary glands (9, 10) and gynecological organs (11–13). During the neoplastic processes, NTs and their receptors are overexpressed by tumoral cells, promoting progression and angiogenesis in several cancer models. For instance, the expression of NTs predicts poor survival rates in breast and ovarian cancer patients (14–16) and NTs have been proposed as potential therapeutic targets in these neoplasms (4, 17, 18).

Angiogenesis is a key process to supply nutrients and oxygen to tumor cells, as well as a way for cells to leave or enter to the circulation (19). In fact, tumors that have a high microvascular density could be more aggressive and generate distant metastasis (20). The term angiogenesis was first used by the British surgeon John Hunter in 1787; however, the study of vascular morphology in animal and human tumors began only in the first half of twentieth century (21).

Endothelial cells, a baseline membrane and pericytes are the minimal components of vasculature. Endothelial cells form a barrier that controls the trans-endothelial flux of soluble components and most cell types (22). During angiogenesis, there are several important steps: a detection of humoral paracrine signals or angiogenic factors, resulting in the sprouting of endothelial cells, followed by an orchestrated increase of endothelial cell proliferation, migration, and differentiation (23). Activation of endothelial cells is accompanied by pericytes detachment, proliferation, and migration into the vessel interstitium to envelop the surface of the vascular tube. In addition, fibroblasts and endothelial cells build and remodel the new extracellular matrix (23, 24). All of these changes are necessary to generate new capillary vessels.

### TUMOR ANGIOGENESIS

Tumor growth has two phases: an avascular stage (when tumors are constrained at diameters of 1–2 mm) and a posterior vascular stage (25), in which tumor cells need to secrete soluble factors to promote an increase of angiogenesis and continued growing (26).

In the normal vasculature, endothelial cells are stable; rarely they sprout or divide and they are associated to mural cells (pericytes) in a basal membrane. However, in the case of the tumor vasculature several chromosomal abnormalities arise (27–29), as well as variations of size and thickness, irregular shape, and big trans-cellular holes and fenestrae (30, 31). These characteristics produce a decrease of blood flow and drug delivery, and increase the interstitial fluid pressure, the extravasation of blood components and the intravasation of tumor cells (30, 32). Particularly in gynecologic neoplasms, angiogenesis plays a key role, since the ovary and uterus cyclically regulate the angiogenesis during the ovarian cycle involving blood vessel growth and regression, with a fine regulation (33–35). Therefore, angiogenesis is undoubtedly crucial in gynecological cancers, but this process is uncontrolled. Given that angiogenesis is a complex process that involves different cell types, in vivo experiments constitute the ideal condition to evaluate it. Some examples of in vivo assays are: the chick embryo chorioallantoic membrane (CAM) assay (36), zebrafish embryo assay (37, 38), corneal micropocket assay (39, 40), and matrigel plug assays (41). Moreover, there are some experimental approaches in vitro to evaluate the angiogenic potential of cells, which may have some advantages, such as the reproducibility and low cost to perform these assays (42). However, it is considered that in vitro assays evaluate vasculogenesis or de novo formation of vasculature-like structures and usually involve only endothelial cells and extracellular matrix. Examples of this are tubular formation assays in matrigel (43, 44) and the recently developed microfluidic cell culture systems (45). Nevertheless, in vitro assays are widely used, because they are a cheap and reproducible method to evaluate the angiogenic potential (46).

#### VEGF: CLASSICAL ANGIOGENIC FACTOR IN CANCER

There are many known angiogenic factors, among which VEGF is the most widely studied in the context of cancer. VEGF genes include VEGF-A to VEGF-E and another related gen, placental growth factor (PLGF) (47–50). VEGF-A (from now referred as VEGF) has the most important effect in the formation of blood vessels during development or in pathological conditions as cancer (51). At the same time, VEGF undergoes alternative exon splicing (52, 53), leading to several transcripts that include VEGF121, VEGF145, VEGF165, VEGF189, and VEGF206, which give origin to VEGF peptides of 121, 145, 165, 189 and 206 amino acids, respectively (54). Besides, VEGF<sup>121</sup> is totally secreted and VEGF<sup>165</sup> is partially secreted from cells (55, 56). In ovarian, endometrial and cervical cancers, VEGF<sup>121</sup> and VEGF<sup>165</sup> are the most dominantly expressed (57–60).

#### ROLE OF NEUROTROPHINS IN GYNECOLOGICAL CANCER ANGIOGENESIS: NGF/TRKA AND BDNF/TRKB

#### Neurotrophins and Its Functions in Reproductive Tissues

NTs belong to a family of homodimeric polypeptide growth factors that promote neuronal survival and differentiation, and display important functions in non-neuronal cells (13, 61). Members of the NTs family include nerve growth factor (NGF) that was first described by Dr. Levi-Montalcini in 1956 (62), brain derived neurotrophic factor BDNF, neurotrophin-3 (NT-3), and neurotrophin-4/5 (NT-4/5) (63). Among them, NGF and BDNF are the most important NTs studied in the context of reproduction and cancer. NTs bind with different affinity to Tropomyosin kinase (TRK) receptors and produce the dimerization and transphosphorylation of its tyrosine kinase domains, activating PI3K/AKT, MAPK/ERK, and PLCγ/PKC signaling pathways (64). NGF binds with high affinity to TRKA receptor, while BDNF binds preferentially to TRKB receptor (PMID: 1649702, PMID: 2927393), as shown in **Figure 1**.

Both NGF/TRKA and BDNF/TRKB are expressed in reproductive tissues as the ovary and uterus (13, 65). These NTs are involved in the control of early follicular growth and ovarian function (66–70). NGF increases cell proliferation of granulosa and thecal cells and promotes the expression of Follicle Stimulating Hormone (FSH) receptor in rat and human granulosa cells (68, 71, 72), while BDNF/TRKB are required for the growth of newly formed follicles and are involved in the maturation of human oocytes and their developmental competence after fertilization (70, 73, 74). In addition, BDNF levels in follicular fluid (75) and plasma (76) have been studied as possible predictors of in vitro fertilization outcome. BDNF and NGF seem to have a positive correlation with oocyte maturation and pre-implantation and with embryonic development in various mammalian species, including humans (73, 77–80).

On the other hand, NGF expression is present in epithelial and stromal cells in the rabbit uterus (81), as well as in human uterus (82), but its expression is lower than in the ovary (13). In addition, NGF expression seems to be necessary to ensure maternal tolerance in healthy pregnancies in mice, but an excess of NGF results in fetal rejection due to exacerbated inflammation (83). BDNF levels in menstrual blood are higher than in peripheral blood, and this factor is also present in the endometrium in both follicular and luteal phases (65). Furthermore, BDNF levels in menstrual blood of fertile women are higher than in anovulatory women (65). All these findings show that NGF and BDNF play a key role in the homeostasis and function of tissues in the context of female reproduction.

### Roles of Neurotrophins as Direct and Indirect Angiogenic Factors

One of the first evidence of the angiogenic role of NGF comes from the expression of TRKA receptors in human umbilical vein endothelial cells (HUVEC): when using a VEGF-neutralizing antibody, NGF-induced HUVEC proliferation was not observed (84). In another work, NGF from different biological sources (mouse, viper and cobra) was tested in a CAM assay (85), and an increased rate of angiogenesis in a dose-dependent fashion and comparable with recombinant VEGF effects was described. Additionally, one study performed in matrigel plugs in immunedeficient mice shows that NGF strongly increases invasion, cord formation and the monolayer permeability of endothelial cells (86). Furthermore, a recent work shows that NGF increases cell proliferation, migration and differentiation of the human endothelial cell line EA.hy926 in a dose-dependent manner (87). In fact, **Figure 2** shows that NGF increases inter-cellular contact structures (junctions) and polygonal structures (meshes) of EA.hy926 cells, evaluated by Image J Angiogenesis Analyzer (88). Additionally, it has been reported that NGF increases the angiogenic score of EA.hy926 cells, the effect being several times lower compared with VEGF (87).

In a comparable way, BDNF displays direct angiogenic effects in other types of tissues. For example, in a model of BDNF null mice, the survival of endothelial cells in intramyocardial arteries and capillaries in the early postnatal period is impaired (89). Additionally, BDNF increases angiogenic tube formation of the endothelial cells in HUVEC (90). Besides, the overexpression of BDNF in a mouse endothelial cell line promotes endothelial cell proliferation, migration, invasion and survival (91). This evidence indicates that BDNF/TRKB exhibits a direct role in the angiogenic process and can partially explain that the anti-angiogenic therapy with Bevacizumab (neutralizing antibody against VEGF) is not optimal in the cancer context.

On the other hand, both NTs (NGF and BDNF) have an indirect angiogenic role, mediated by VEGF modulation in different cellular models. It is described that NGF and BDNF induce VEGF expression in MAPK/ERK 2-dependent pathways in granulosa cells (92) and osteoblasts (93), respectively. Besides, NGF promotes VEGF expression in neuronal superior cervical ganglia (94), while BDNF increases VEGF expression in human chondrosarcoma (95) and neuroblastoma cells (96). Another key point is that plasmatic levels of VEGF are lower in deficient BDNF animals compared to wild type animals (97). All these antecedents indicate that NTs not only act directly in vascular cells, but also affect several cell types by increasing VEGF expression and therefore their angiogenesis potential.

### Role of NGF/TRKA in the Ovarian Cancer Angiogenesis

Ovarian Cancer is the most lethal gynecological malignancy in developed countries (98–100). It is characterized by nonspecific symptoms and therefore is diagnosed at later stages, resulting in poor survival rates (101, 102). Approximately 80% of them are Epithelial Ovarian Cancer (EOC) (101) which is characterized by its high extent of angiogenesis that facilitates rapid tumor growth and dissemination (103). NGF and its high affinity receptor TRKA are found in very low levels or are absent in normal ovarian surface epithelium, whereas they are highly expressed in EOC (60). Another study shows that significantly higher levels of NGF, total TRKA, and phospho-TRKA (active receptor) are present in poorly differentiated EOC vs. normal ovary (4). In addition, NGF/TRKA stimulates cellular proliferation of EOC cells, by the activation of MAPK/ERK and AKT pathways, increasing Bcl2/Bax ratio and c-Myc (104), indicating the importance of NGF/ TRKA in EOC progression and suggesting that they could be considered as a potential tumor markers. As previously shown, several studies performed in in vitro and ex vivo models support the direct angiogenic role of NGF in EOC (105). It is relevant to point out that the TRKA receptor is present in endothelial cells from EOC biopsies (4), supporting the idea that the endothelium can respond to NGF stimulation.

On the other hand, an indirect angiogenic role of NGF has been described through the modulation of VEGF expression in EOC. In fact, in EOC explants, NGF increases in a dosedependent manner the mRNA of VEGF121, VEGF165, and VEGF<sup>189</sup> (60). Equivalent results were obtained in in vitro models, where NGF increases VEGF expression and protein levels in the culture supernatants of the EOC cell line (4).

### Role of BDNF/TRKB in the Ovarian Cancer Angiogenesis

It has been reported that TRKB displays a key role in ovarian development, which gives proliferative signaling in granulosa cells during the beginning of mammalian ovary development (70). Increased TRKB levels can promote the increase of cell proliferation, invasion and angiogenesis, suppression of anoikis and decreased chemotherapy response and apoptosis in different cancer cell lines, including ovarian cancer cells (5, 106–112). Observational studies show that high TRKB expression in ovarian cancer is correlated with poor survival in ovarian cancer patients (5), and that TRKB is overexpressed in metastatic lesions compared with the corresponding primary lesions (113). In addition, BDNF treatment enhances cell invasion and migration of ovarian cell lines and TRKB-silenced cells increase the percentage of apoptotic cells (5). This evidence indicates that BDNF/TRKB may contribute to ovarian cancer progression.

In agreement with other authors, our group has found that TRKB receptor is present in stroma and in transformed epithelia of human ovary. The active TRKB receptor is upregulated in serous adenocarcinomas and its immunodetection is almost absent in the epithelia from functional ovaries or ovarian serous adenomas (114)

Interestingly, in ovarian cell lines, the silencing of TRKB receptor reduces VEGFR-2 mRNA by 70% (5), which suggests that BDNF could regulate the expression of VEGF receptors. In addition, a positive correlation between TRKB expression and lymph vessel density has been described in ovarian cancer (113). These results are consistent with other studies, in which BDNF promotes VEGF-C-dependent lymphangiogenesis in chondrosarcoma cells (95) and TRKB expression is associated with the expression of VEGF-C and VEGF-D in oral squamous cell carcinoma (115). These findings suggest that BDNF could be implicated in ovarian cancer progression and modulate angiogenesis and/or lymphangiogenesis by the increase of different VEGF isoforms.

#### ROLE OF NTs IN CERVICAL CANCER AND UTERINE PATHOLOGIES

Cervical cancer is the fourth most frequent cancer in women. Approximately 90% of deaths from cervical cancer occur in low-income and middle-income countries, in which strategies of prevention, early diagnosis, effective screening, and treatment programs are less common (116).

In the context of cervical cancer, BDNF/TRKB are perhaps the best studied NTs. It has been described that BDNF and TRKB expression are significantly higher in cervical cancer tissues than in normal tissues and that their presence is higher in advanced stages of this neoplasm (6, 7). In addition, BDNF levels are positively associated with lymph node metastasis (7) in cervical cancer patients. In cervical cancer cell lines, BDNF/TRKB increases cell proliferation (7, 117), apparently involving ERK and AKT signaling pathways (118). TRKB downregulation in cervical cancer cells suppress the activation of epithelial mesenchymal transition (EMT) by downregulation of N-cadherin and vimentin, among other proteins, and strongly diminishes cell proliferation, migration and invasion (117, 118).

Considering that the activation of ERK signaling pathway by BDNF/TRKB was associated with an increase of VEGF expression in osteoblasts (93), and given that TRKB can activate PI3K and ERK signaling pathways which regulate VEGF expression in several models (119–121), it is plausible that the VEGF expression could be increased by TRKB in cervical cancer, similarly to ovarian cancer.

There is no direct evidence that overexpression of NTs and its receptors are involved in the physiopathology of endometrial cancer. However, antecedents suggest that NTs could contribute to this pathology, since their expression increase in endometriosis (122–124), a condition that has been associated with higher risk of ovarian and endometrial cancer (125–127). The endometriosis is an estrogen-dependent inflammatory disease, characterized by the presence of endometrial-like tissue outside the uterine cavity (128). An important characteristic of this pathology is that angiogenesis is deregulated. In endometriosis, the VEGF expression is increased and promotes the spreading of new blood vessels at the endometriotic lesions and surroundings, which contributes to the survival of lesions (129). A recent study has shown that drospirenone, a drug used for endometriosis treatment, significantly decreases inflammatory cytokines and NGF expression, as well as VEGF expression in human endometriotic stromal cells (130). Similarly, Ginsenoside (a ginseng-derivate extract) decreases both VEGF and BDNF in rat endometriotic implants (131). These antecedents suggest that NTs could contribute not only to the pelvic chronic pain typical of endometriosis, but also to pathological angiogenesis, probably by the increase of VEGF levels.

### PHARMACOLOGIC INHIBITORS OF NEUROTROPHIN RECEPTORS

Since the TRK receptors (TRKA, TRKB, and TRKC) are implicated in the progression of different kind of neoplasms, several drugs have been developed to target tumors that overexpress TRK receptors or present chromosomal rearrangements of TRK genes. For instance, in 2018, the Food and Drug Administration (FDA) approved Larotrectinib (Vitrakvi) for treatment of adult and pediatric patients with solid tumors that have TRK gene fusions (132). This was based in promissory results of 3 clinical trials (NCT02122913, NCT02637687, and NCT02576431) with Larotrectinib that showed an objective response rate of 75% in pediatric patients, with good tolerability and safety (133, 134). Larotrectinib is a small molecule that binds to NTs receptors, thereby preventing neurotrophin-TRK interaction and TRK activation, which results in the induction of cellular apoptosis and the inhibition of cell growth (135). It is important to point out that Larotrectinib was one of the first "tissue-agnostic drug" approved by FDA,

concept that refers to a substance to treat cancer based on genetic and molecular features of tumor cells, regardless of the cancer type or origin (136).

Additionally, Entrectinib (Rozlytrek), a potent and selective ATP-competitive inhibitor, was approved by the FDA in 2019 for adults and pediatric patients above 12 years old with solid tumors (as ovarian cancer) that have a TRK fusion without a known acquired resistance mutation (137). The first results of phase I/II studies show promising results: for example, an objective response rate of 57.4% was obtained in 54 adults with advanced or metastatic TRK fusion-positive solid tumors (138). Unfortunately, some patients have reported resistance to TRK inhibition with this drug considered as first generation of TRK inhibitors (139), probably due to mutations in TRK domain (140, 141). To improve this aspect, a next-generation TRK-targeted agent is under study. For example, Loxo-195 is a recently developed drug, which phase 1/2 of the study started in 2017 in patients with TRK-positive solid tumors and TRK fusion-positive cancers (clinical trials NCT03215511 and NCT03206931). This drug could become an alternative treatment for tumors with acquired resistance to first-generation TRKtargeted agents (142). VMD-928 is another specific TRK inhibitor which is under phase 1 of the study since 2018 for treatment of advanced adult solid tumors or lymphoma (NCT03556228).

Because TRK overexpression is present in gynecological cancers, and particularly TRK fusion has been described in cervical and uterine cancer (143, 144), the use of TRK inhibitors could be beneficial in these kinds of neoplasms. However, it is necessary to continue the studies to determine their effectiveness in gynecological cancers.

#### CONCLUSIONS

NGF/TRKA and BDNF/TRKB are the main NTs studied in the context of cancer. These NTs and their receptors are over-expressed in gynecological neoplasms, such as ovarian and cervical cancers, in which they promote the progression

#### REFERENCES


of these diseases. Furthermore, NTs are involved in uterine pathologies such as endometriosis, which suggests that they could contribute to endometrial cancer progression, however this has not been elucidated yet. NTs are indirect angiogenic factors, acting through the induction of VEGF expression in ovarian cancer cells; besides, it is possible that NTs could display the same effect in other cancer cells such as cervical and endometrial. In addition, NTs exhibit a direct angiogenic role, mainly studied in endothelial cells that express NTs receptors, and respond by increasing endothelial cell proliferation, migration and differentiation. Moreover, NTs increases angiogenesis both in in vitro and in vivo models. Consequently, NTs and their receptors may be considered as important angiogenic factors, mostly in the context of anti-angiogenic therapy against VEGF, where overexpression of NTs could increase the angiogenesis independent of VEGF levels and contribute to therapy failure. Since NTs and TRK receptors are drivers of a wide variety of adult and pediatric cancers as gynecological neoplasms, the FDA has recently approved pan-TRK inhibitors for the treatment of TRK fusion-positive solid tumors. Because TRK fusion has been described in several gynecological cancers, the recently developed TRK inhibitors emerge as a new therapeutic approach for the treatment in this subtype of neoplasms. Given that angiogenesis is a key feature in gynecological neoplasms, and NTs acts as direct and indirect angiogenic factors, it may be relevant to study whether TRK inhibitors could improve the efficacy of antiangiogenic drugs as bevacizumab, which was not elucidated yet.

#### AUTHOR CONTRIBUTIONS

MG, IT, and CR: conceptualization. MG: writing original draft. MV and CR: writing, review, and editing.

#### FUNDING

National Fund for Scientific and Technological Development (FONDECYT) #1160139.


Endocrinology. (1995) 136:4681–92. doi: 10.1210/endo.136.10. 7664689


protein forms are encoded through alternative exon splicing. J Biol Chem. (1991) 266:11947–54.


development of oocytes into preimplantation embryos. Proc Natl Acad Sci USA. (2005) 102:9206–11. doi: 10.1073/pnas.0502442102


and increases sensitivity to paclitaxel. Mol Cancer Ther. (2012) 11:2556–65. doi: 10.1158/1535-7163.MCT-12-0463


human endometriotic stromal cells. J Reprod Immunol. (2017) 119:44–8. doi: 10.1016/j.jri.2016.12.002


pathway activation. Nat Med. (2019) 25:1422–7. doi: 10.1038/s41591-019- 0542-z


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Garrido, Torres, Vega and Romero. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Non-coding RNAs as Regulators of Lymphangiogenesis in Lymphatic Development, Inflammation, and Cancer Metastasis

Ming-xin Cao1†, Ya-ling Tang2†, Wei-long Zhang<sup>2</sup> , Ya-Jie Tang3,4 \* and Xin-hua Liang<sup>1</sup> \*

<sup>1</sup> State Key Laboratory of Oral Diseases & National Clinical Research Center for Oral Diseases, Department of Oral and Maxillofacial Surgery, West China Hospital of Stomatology, Sichuan University, Chengdu, China, <sup>2</sup> State Key Laboratory of Oral Diseases & National Clinical Research Center for Oral Diseases, Department of Oral Pathology, West China Hospital of Stomatology, Sichuan University, Chengdu, China, <sup>3</sup> State Key Laboratory of Microbial Technology, Shandong University, Qingdao, China, <sup>4</sup> Hubei Key Laboratory of Industrial Microbiology, Key Laboratory of Fermentation Engineering (Ministry of Education), Hubei Provincial Cooperative Innovation Center of Industrial Fermentation, Hubei University of Technology, Wuhan, China

#### Edited by:

César López-Camarillo, Universidad Autónoma de la Ciudad de México, Mexico

#### Reviewed by:

Alessandra Ghigo, University of Turin, Italy Prasanna Ekambaram, University of Pittsburgh, United States

#### \*Correspondence:

Ya-Jie Tang yajietang@sdu.edu.cn Xin-hua Liang lxh88866@scu.edu.cn

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology

Received: 15 March 2019 Accepted: 03 September 2019 Published: 20 September 2019

#### Citation:

Cao M, Tang Y, Zhang W, Tang Y-J and Liang X (2019) Non-coding RNAs as Regulators of Lymphangiogenesis in Lymphatic Development, Inflammation, and Cancer Metastasis. Front. Oncol. 9:916. doi: 10.3389/fonc.2019.00916 Non-coding RNAs (ncRNAs), which do not encode proteins, have pivotal roles in manipulating gene expression in development, physiology, and pathology. Emerging data have shown that ncRNAs can regulate lymphangiogenesis, which refers to lymphatics deriving from preexisting vessels, becomes established during embryogenesis, and has a close relationship with pathological conditions such as lymphatic developmental diseases, inflammation, and cancer. This review summarizes the molecular mechanisms of lymphangiogenesis in lymphatic development, inflammation and cancer metastasis, and discusses ncRNAs' regulatory effects on them. Therapeutic targets with regard to lymphangiogenesis are also discussed.

Keywords: non-coding RNA (ncRNA), lymphangiogenesis, lymphatic development, inflammation, cancer metastasis

## INTRODUCTION

Lymphangiogenesis are termed lymphatics deriving from preexisting vessels, and generally progress through a number of stages: establishment of lymphatic endothelial cell (LEC) identity, formation of the primary lymphatic structures, maturation, and remodeling of the lymphatic vessels (1, 2). The normal growth and development of lymphatics contribute to their non-negligible roles in the cardiovascular system, including maintaining tissue fluid homeostasis, directing the trafficking of immune cells during immunosurveillance, and absorbing dietary lipids from the digestive tract (3, 4). However, in developmental diseases such as lymphedema, lymphangiectasia, and vascular malformations, or in inflammatory conditions such as infectious diseases, and after surgical intervention, lymphatic function is impaired which might lead to swellings and edema (5). In other pathological conditions, such as cancer, it might be essential to inhibit lymphangiogenesis, thus preventing cancer metastasis (6). Our comprehension of lymphangiogenesis regulation is mainly based on understanding the functions of proteins and their interactions, while it is widely known that only 3–5% of our genome encodes proteins and protein-target therapy may cause drug resistance (7). Therefore, the clinical regulation of lymphangiogenesis further requires other types of targets for successful intervention.

**84**

Currently, emerging studies have implicated connections between lymphangiogenesis and non-coding RNAs (ncRNAs), especially the well-known microRNAs (miRNAs), long noncoding RNAs (lncRNAs), and the newly discovered circular RNAs (circRNAs), which are transcribed from the vast majority of human genome. ncRNAs, though most do not encode proteins, contain genetic information or have functions in the biological process of cells. ncRNAs include structural RNAs such as tRNAs and rRNAs, which are abundant and have well-defined roles in translation, and regulatory RNAs such as miRNAs, lncRNAs, and circRNAs. These regulatory RNAs contain physiological and pathological functions, by altering protein expressions, interacting as signaling partners with specific proteins or acting as scaffolds for protein complexes to change signaling pathways (8, 9). Preclinical studies and increased success rates of ncRNAtarget therapy provide a possibility of targeting ncRNAs in lymphangiogenesis-related disorders (10, 11).

Just as the old Chinese saying goes that "one stone with three birds," understanding the underlying mechanisms important for ncRNAs targeting lymphangiogenesis in lymphatic developmental diseases, inflammation and cancer metastasis will help build new therapeutics when more than one disorder exists. Here, we review the molecular mechanisms of lymphangiogenesis in lymphatic development, inflammation and cancer metastasis, emphasize the ncRNAs' regulation on them, and hope to harness this knowledge for translational medicine.

### LYMPHATIC DEVELOPMENT

#### Lymphatic Development

In vertebrates, the first definitive sign of lymphatic development is the presence of endothelial cells with the expression of PROX1, considered to be the master regulator of LEC fate specification in cardinal veins (12, 13). Prox1 deletion in mice led to a complete absence of the lymphatic vasculature. PROX1-positive LECs bud, sprout and migrate away from both the cardinal and intersomitic veins to form the primary lymphatic plexus and sacs (14–16). Once exiting the veins, primitive LECs exhibit LEC identity including podoplanin, and increased levels of VEGFR-3 and NRP2 (14–16). This exiting process is absolutely dependent on VEGF-C, which acts via its tyrosine kinase receptor VEGFR-3 and the non-signaling co-receptor NRP2 (16–18). PROX1 positive LECs do not egress from the veins in embryos when deficient in key regulators of VEGF-C/VEGFR-3 signaling, such as collagen and calcium binding EGF domains 1 (CCBE1) (16, 19, 20). As for the initiation of PROX1 expression in venous endothelial cells, transcription factor SOX18 (21) and NR2F2/COUP-TFII (22) are thought to be critical. Intriguingly, PROX1 expression is induced in the cells of dorsolateral aspect of veins, while its inducer/co-regulator, SOX18, and COUP-TFII, are expressed throughout the cardinal vein endothelial cells (12, 23). Current explanation for this polarization involves retinoic acid signaling (24), Notch signaling (25) and BMP2 signaling pathway (26). They are all researched to regulate the emergence of lymphatic endothelial progenitor cells from the veins.

After LECs migrating and forming primary lymphatic vascular structures, major events involved in lymphatic development includes formation of lympho-venous valves, induction of platelet aggregation in valve regions, and remodeling of the initial lymphatic plexus to form a hierarchical lymphatic vascular tree (27, 28). Upregulation of FOXC2, together with high levels of PROX1 and GATA2, exist in the clusters of cells destined to form valves (29, 30). FOXC2 and PROX1 coordinately control expression of the gap junction protein connexin37 and activation of calcineurin/NFAT signaling, which are required for the assembly and delimitation of lymphatic valve territory during development (31). And cell surface molecules including the planar cell polarity pathway members, CELSR1 and VANGL2 (32), signaling pathways including Notch (33), BMP (34), and Semaphorin/Neuropilin/Plexin axes (35), and mechanical stimuli including disturbed flow are also important for valve development (31). As for platelet aggregation, signaling induced by podoplanin on the surface of LECs can bind to platelet C-type lectin-like receptor2 (CLEC2) to prevent blood entering into the lymphatics (27, 36). Both valves and platelet thrombi are crucial for separating the blood and lymphatic vascular compartments. In addition, signaling pathways such as angiopoietin/Tie signaling (37, 38), EphrinB2 signaling (39), and Reelin signaling (40), are significant for primitive lymphatic plexus remodeling to form initial and collecting vessels.

#### miRNAs and Lymphatic Development

miRNAs (19–24 nucleotides) are endogenous, non-proteincoding small RNAs that serve as post-transcriptional gene regulators (41, 42). According to the miRBase (version 21.0), over 60% of the protein-coding genes in human are targeted by miRNAs (43). Recent studies have defined the critical roles of miRNAs in lymphangiogenesis. Kazenwadel et al. demonstrated that miR-181a, the first discovered miRNA that targets PROX1, could bind to the Prox1 3′ -UTR, resulting in translational inhibition and transcript degradation (44). Increased miR-181a in primary embryonic LECs led to the substantial reduced levels of PROX1 and resulted in reversion of LEC identity toward a blood vascular phenotype. Inhibition of endogenous miR-181a in blood endothelial cells (BECs) leads to elevated PROX1 expression, therefore promoting the acquisition of LEC identity. miR-31, as a novel BEC-specific miRNA, inhibited lymphatic lineage-specific differentiation in BECs, at least in part by repressing PROX1 in vitro, and impaired lymphatic development and venous sprouting in vivo (45). miR-31 candidate targets also include FOXC2, which is required for specification of lymphatic capillaries vs. collecting lymphatic vessels (46, 47), and RAMP2, which triggers lymphangiogenesis in response

**Abbreviations:** ncRNA, non-coding RNA; PROX1, Prospero homeobox 1; BMP2, bone morphogenetic protein 2; FOXC2, Forkhead box C2; IL-1β, Interleukine-1 beta; VEGFR-3, vascular endothelial growth factor receptor-3; NSCLC, non-small cell lung cancer; VEGF-A, vascular endothelial growth factor A; VEGF-C, vascular endothelial growth factor C; NF-κB, nuclear transcription factor-κB; OSCC, oral squamous cell carcinoma; NRP2, Neuropilin 2; Flt4, Fms-related tyrosine kinase 4; CXCL12a, Chemokine (C-X-C motif) ligand 12a; FoxO1, Forkhead box O1; TNF-α, Tumor Necrosis Factor alpha; LYVE-1, Lymphatic vessel endothelial hyaluronan receptor 1; WISP-1, WNT1-inducible signaling pathway protein-1; ANRIL, antisense non-coding RNA in the INK4 locus; C21orF96, Chromosome 21 open reading frame 96.

to adrenomedullin signaling (48). Recent evidence has shown that BMP2 signaling, the negative modulator for lymphatic fate during development, could promote both miR-181a and miR-31 in a SMAD-dependent manner, thus negatively regulating PROX1 expression at the post-transcriptional level. BMP2 signaling is therefore essential for constructing therapeutic manipulation of lymphangiogenesis in development (26).

miR-182, which is induced by JunB and attenuates FoxO1 expression in zebrafish, is crucial for the formation of parachordal lymphangioblasts and the thoracic duct. This JunB/miR-182/FoxO1 axis is regarded as a novel key player in governing lymphangiogenesis (49). A recent study has identified miR-126a as a director of LEC migration and lymphatic assembly. In vivo studies by Chen et al. reported that VEGF-C/FLT4 signaling acted as a cooperator of miR-126a, allowing modulation of lymphangiogenic sprout formation. miR-126a upregulated CXCL12a by targeting its 5′ -UTR, then inducing chemokine signaling, resulting in parachordal lymphangioblast extension along a horizontal myoseptum (50, 51). Subsequent research confirmed miR-126 as a conserved modulator of lymphatic development in vivo and in vitro, and put forward the potential of miR-126 in preventing lymphedema, the most recognized aspect of lymphatic system malfunction as a result of genetic cause (52) (**Figure 1**).

### INFLAMMATION

#### Lymphangiogenesis in Inflammation

Inflammation is a common feature of various conditions, characterized by pathological neovascularization, including hemangiogenesis and lymphangiogenesis. Hemangiogenesis refers to the new outgrowth from pre-existing blood vessels, and is an important pathological aspect of chronic inflammatory diseases (53). Lymphangiogenesis in inflammation is often induced by factors produced by macrophages and dendritic cells, and its existence involves tissue edema reduction, immune response initiation, and antigen clearance (54). Macrophages, especially CD11b<sup>+</sup> macrophages, play a pivotal role in the inflammatory lymphangiogenesis mediated by VEGF ligands (55, 56). The VEGF family consists of five members that bind to and activate three distinct receptors. VEGF-A binds to VEGFR-1 and VEGFR-2; placental growth factor (PlGF) and VEGF-B bind only to VEGFR-1; and VEGF-C and VEGF-D bind to VEGFR-2 and VEGFR-3. Generally speaking, ligation of VEGF-A to VEGFR-2 induces only hemangiogenesis, while interactions of VEGF-C/D and VEGFR-3 mediate lymphangiogenesis (57– 59). However, recent observations contradicted this notion and found that there was some crosstalk between them. VEGF-C produced by activated macrophages can induce local proliferating and sprouting of preexisting lymphatic cells (60, 61), while VEGF-A, expressed by activated leucocytes in inflammatory context, can recruit VEGFR-1-expressing macrophages, which are known to release VEGF-C/D, thus inducing inflammatory lymphangiogenesis (62). Maruyama et al. showed that VEGFR-3-expressing CD11b<sup>+</sup> macrophages could directly transdifferentiate into lymphatic endothelial cells (LECs), forming cell aggregates that gradually developed into sprouting lymphatic vessels (63). In addition to the two ways of macrophages supporting lymphangiogenesis, dendritic cells activated by IL-1β and TNF-α in inflammation milieu can also migrate to lymphatic vessels, express VEGF-C, promoting lymphangiogenesis (64).

#### miRNAs and Inflammatory Lymphangiogenesis

The first example of the regulatory role of miRNA in inflammatory lymphangiogenesis is miR-1236, which is expressed in LECs and binds to VEGFR-3. Jones et al. demonstrated that miR-1236, induced by IL-1β, could negatively regulate VEGFR-3 expression and VEGFR-3-dependent signaling Akt and ERK1/2, and attenuate VEGF-C/VEGFR-3 mediated LECs migration and tube formation in primary human LECs in vitro. They also found that miR-1236 could reduce lymphangiogenesis in vivo (65, 66). Considering the fact that IL-1β contributes to initial lymphangiogenesis by inducing VEGFs and also induces miR-1236, which in turn suppresses VEGFR-3-dependent signaling, modulation of VEGFR-3 levels using miR-1236 may be a promising approach for the treatment of inflammatory diseases. Additionally, studies by Chakraborty et al. revealed that miR-9 expressed on inflamed LECs, which was induced by TNF-α, could inhibit NF-κB-mediated inflammation, increase VEGFR-3 and induce LEC proliferation and tube formation to activate VEGFR-3-mediated lymphangiogenesis (67). Studies concerning inflammatory lymphangiogenesis usually involved models of corneal lymphangiogenesis, as cornea exhibited alymphatic feature under normal condition and lymphangiogenesis under pathologic insults such as inflammation. miR-466, miR-184, and miR-199a/b-5p have all been reported to be significantly downregulated in corneal inflammatory lymphangiogenesis, and accordingly, inhibit corneal lymphatic growth in vivo and suppress LECs functions of adhesion, migration, and tube formation in vitro. This offers a clinical potential for lymphangiogenesis interference and lymphatic-related diseases treatment (68–70). Wang et al. revealed that miR-132 isolated from exosomes of VEGF-Ctreated adipose-derived stem cells could be directly transferred to LECs and promote LECs proliferation, migration, and tube formation by targeting Smad-7 and activating TGF-β/Smad signaling, thus reversing acute to chronic inflammatory processes in inflammatory bowel disease (71).

Recently, circular RNA (circRNA) cZNF609 could serve as a sponge for miR-184 and subsequently elevated miR-184-target gene heparanase in inflamed corneas, which could significantly elevate VEGF-C expression and facilitate lymphangiogenesis in vitro and in vivo. However, whether cZNF609 intervention could act as a therapeutic strategy in preventing inflammation-induced lymphangiogenesis and treating ocular inflammatory diseases remains unknown, therefore requires further investigation (72) (**Figure 2**).

#### CANCER METASTASIS

#### Lymphangiogenesis in Tumor

Cancer metastasis, the dissemination of cancer cells from primary tumors to distant organs, is considered to be the primary cause of cancer-related deaths. The majority of epithelial cancers

FIGURE 1 | Identified ncRNAs regulating lymphangiogenesis in lymphatic development. The development of lymphatic vascular network starts with the cells of cardinal vein losing blood endothelial characteristics and acquiring lymphatic endothelial cell (LEC) identity. LECs then bud off the cardinal vein and form lymphatic sacs and plexus. Subsequently, remodeling of the primitive lymphatic vasculature begins, and becoming a hierarchical network. We described ncRNAs, mostly miRNAs identified to-date, which influence different steps of developmental lymphangiogenesis.

and dendritic cells to express VEGF-C/D, leading to inflammatory lymphangiogenesis. We described ncRNAs, mostly miRNAs identified to-date, which influence inflammatory lymphangiogenesis.

firstly develop metastasis through spreading via lymphatic vessels (73). Tumor hypoxia microenvironment stimulates tumor cells, tumor stroma cells, and tumor-infiltrating inflammatory cells to express a series of lymphangiogenic factors, including the well-known VEGF family, especially VEGF-A/C/D (74), and other mediators such as PDGF-BB (75), IGF1/2 (76), FGF2 (77–79), HGF (80, 81), angiopoietin-2 (82), sphingosine-1 phosphate (83), adrenomedullin (84), and IL-7 (85, 86). In response to these factors, lymphangiogenesis can start from existing lymphatic vessels via sprouting, LEC proliferation, and formation of intra- and peri-tumoral lymphatics. Additionally, it can also derive from precursor LEC and bone marrow-derived cells (87, 88). After disseminating into sentinel lymph nodes (SLNs, the first tumor draining LN), lymphangiogenic factors induce LN lymphangiogenesis prior to the arrival of cancer cells. Besides inducing new lymphatic vessels, tumors can co-opt existing lymphatics at the primary site (73).

Crosstalk between tumors and lymphatic vessels are bidirectional. In addition to being influenced by tumors mentioned above, lymphatic vessels in return can contribute to cancer metastasis by secreting chemokines CCL21 (89) or CXCL12 (90), which bind to CCR7 or CXCR4 receptors, respectively, expressed in invading cancer cells, thus recruiting cancer cells toward lymphatic vessels. Lymphatic vessels can also provide a cancer stem cell niche (91) and modulate antitumor immune responses (92, 93), affecting metastatic tumor cells.

#### miRNAs and Lymphatic Metastasis

The most established lymphangiogenic factor, VEGF-C, can be targeted by miR-128 in human non-small cell lung cancer (NSCLC) cells and human umbilical vein endothelial cells (HUVECs). Hu et al. demonstrated that miR-128 could directly suppress VEGF-C and simultaneously decrease VEGF-A, VEGFR-2, and VEGFR-3 indirectly, thus reducing the phosphorylation of downstream VEGFR signaling pathways extracellular signal-regulated kinase (ERK1/2), phosphatidylinositol 3-kinase(AKT), and p38, resulting in tube formation inhibition in vitro. Furthermore, by analyzing immunohistochemical staining with anti-LYVE-1 antibodies of tumor tissues, they found out that miR-128 could suppress lymphangiogenesis of tumor xenografts in vivo, suggesting the therapeutic significance of miR-128 in NSCLC (94). VEGF-C can also be indirectly targeted by miR-206. Keklikoglou et al. revealed that, in pancreatic adenocarcinoma, miR-206 suppressed lymphangiogenesis through abrogating the expression of VEGF-C. Also, there was a striking reduction in the number of capillary-like tubes and intratumoral lymphatics coverage in the existence of miR-206, indicating that miR-206 based therapy might have important translational implications in pancreatic adenocarcinoma treatment (95). In chondrosarcoma, a series of studies have indicated that miR-381, miR-507, miR-27b, miR-624-3p, and miR-186 contributed to the inhibition of VEGF-C-dependent lymphangiogenesis with different mechanisms, all of which provided information on the potential miRNA-based molecular diagnosis and treatment for VEGF-Cmediated lymphangiogenesis in chondrosarcoma (96–100). In oral squamous cell carcinoma (OSCC) cells, Lin et al. found that decreased miR-300, which was suppressed by WNT1-inducible signaling pathway protein-1 (WISP-1), could contribute to VEGF-C-dependent lymphangiogenesis (101). And inhibited miR-195-3p, targeted by chemokine CCL4, could also induce VEGF-C and lymphangiogenesis in OSCC cells (102).

Besides VEGF-C, another member of VEGF family, VEGF-A, could induce lymphangiogenesis apart from angiogenesis, and accelerate nodal metastasis in OSCC (103). Research showed that miR-126 negatively regulated VEGF-A, and thus decreased lymphatic vessel density in OSCC specimens (104). Neuropilin-2 (NRP2), another important regulator of lymphangiogenesis, was directly suppressed by miR-486-5p in colorectal carcinoma cells, leading to the reduction of peritumoral lymphatic microvessels in vivo, and thus demonstrating the suppressor role of miR-486- 5p in colorectal carcinoma (105). miR-93 was reported to inhibit angiogenesis and lymphangiogenesis by targeting angiopoietin2, and thus suppressed malignant pleural effusion, a sign of an advanced tumor stage (106).

Conversely, there were also pro-lymphangiogenetic miRNAs in cancer metastasis. miR-7 in gastric cancer cells promoted p65-mediated aberrant NF-κB activation and its downstream metastasis-related molecules including VEGF-C, and thus facilitated metastasis by alleviating hemangiogenesis, lymphangiogenesis, and inflammatory cells infiltration (107). miR-548k acted as a pro-lymphangiogenic miRNA in esophageal squamous cell carcinoma (ESCC) via promoting VEGF-C secretion and stimulating lymphangiogenesis, highlighting its crucial role as a new diagnostic and prognostic marker of ESCC (108). miR-27a could be induced in LECs by co-culturing with colon cancer cells, and promoted lymphangiogenesis via targeting SMAD4, a pivotal member of the TGF-β signaling and a tumor suppressor (109). Additionally, exosomes secreted from cancer cells could mediate lymphangiogenesis. A recent study showed that cervical squamous cell carcinoma (CSCC) secreted exosomal miR-221-3p could transfer into LECs to promote lymphangiogenesis and lymphatic metastasis through downregulation of VASH1, representing a novel diagnostic biomarker and therapeutic target for metastatic CSCC patients in early stage (110). Furthermore, circulating miR-10b and miR-373 were shown to be potential biomarkers in detecting lymph node metastasis of breast cancer (111). There have been some studies focusing on miRNAs associated with lymphangiogenesis in various cancers, such as gastric cancer (112), lung cancer (113), and papillary thyroid cancer, of which the underlying mechanisms need to be further elucidated (114, 115).

### LncRNAs and Lymphatic Metastasis

Although the functional roles of miRNAs in lymphangiogenesis are now established, relatively less is known about the regulatory roles of lncRNAs (>200 nts) (116, 117). Known as the "transcriptional noise," lncRNAs rarely code for proteins, but are regulated like that of protein coding RNAs, being subjected to transcriptional regulation or even splicing (118, 119). Unlike miRNAs acting mainly as post-transcriptional repressors, functional lncRNAs can regulate gene expression at various levels, such as chromatin modification, transcriptional and posttranscriptional processing (120, 121). A number of findings have indicated the contribution of lncRNAs in cancer metastasis (122), while the question as to whether these lncRNAs are involved in lymphangiogenesis and lymph node metastasis is still being studied.

Recent evidence has shown that antisense non-coding RNA in the INK4 locus (ANRIL), a kind of lncRNA, induced lymphangiogenesis and lymphatic metastasis in colorectal cancer. ANRIL expression was correlated with the increased expressions of VEGF-C, VEGFR3, LYVE-1, and tube formation in both colorectal cancer cell lines and surgical specimens. ANRIL downregulation reduced lymphatic metastasis rate, lymphatic microvessel density (LMVD), and the expressions of VEGF-C, VEGFR3, LYVE-1, representing the potential role of ANRIL as a therapeutic target in colorectal cancer (123). In addition, a lncRNA termed Lymph Node Metastasis Associated Transcript 1 (LNMAT1), upregulated in lymph node-positive bladder cancer and associated with lymph node metastasis and prognosis, could epigenetically activate CCL2 expression and recruit macrophages into the tumor, which promoted lymphangiogenesis via VEGF-C secretion. LNMAT1 may represent a potential therapeutic target for clinical intervention in lymph node-metastatic bladder cancer (124). Other lymphangiogenesis-related lncRNAs still need further functional studies to verify their roles. For example, C21orF96 was overexpressed in positive lymph node and gastric cancer tissues, and promoted tubular formation in gastric cancer cell lines, while its pathogenesis was less well-characterized (125). MALAT-1 (126), UCA1 (127), HOTTIP (128), and HOTAIR (129) have all been proven to be associated with lymph node metastasis in various cancers and might serve as novel predictors, but well-designed studies are awaited to explain the mechanisms underlying it for uncovering better therapeutic strategies (**Figure 3**).

miRNAs identified to-date, which influence different steps of lymphangiogenesis in cancer metastasis.

### CONCLUSIONS AND FUTURE PERSPECTIVES

In summary, ncRNAs, the dark matter of the genome, account for >80% of total mature RNA and have many crucial, but as yet, undefined roles in regulating lymphangiogenesis concerning lymphatic developmental disorders, inflammatory diseases, and cancer metastasis (130). The two major types of regulatory ncRNAs, miRNAs, and lncRNAs, modulate interrelated steps and mediators of lymphangiogenesis, therefore exert their influence on lymphatic developmental disorders, inflammatory diseases or cancer metastasis, if not all of them. The identification of key pro-lymphangiogenic and antilymphangiogenic ncRNAs is currently the aim of investigation and will underpin the generation of novel therapeutic targets, as well as potential targets on diagnosis, prognosis and response prediction (**Table 1**).

The clinical potential function of ncRNAs as new targets has been carried out. For example, antisense oligonucleotide therapy can be applied to correct aberrant splicing (131, 132); via replacing or inhibiting ncRNAs especially miRNAs, affecting levels or functions of ncRNAs (133). Loss of MALAT1 with antisense oligonucleotide provided a potential therapeutic approach to prevent lung cancer metastasis via regulating gene expression, but not alternative splicing (134, 135). Though breakthroughs in targeted therapy have involved ncRNAs, major challenges exist with limited examples and acquired resistance.

Lymphangiogenesis has the potential to become the therapeutic target (44), since lymphatic vessels are mostly quiescent in adults and LEC identity is more plastic during adulthood than during embryo. However, some problems still need to be addressed. Interfering with tumor lymphangiogenesis can decrease or prevent lymphatic metastasis through blocking lymphatic drainage, while it may also lead to tissue fluid accumulation and cause lymphedema. Is there a target to solve both the metastasis and lymphedema? Furthermore, in a radical operation that resects primary or metastatic tumor, concurrent anti-lymphangiogenesis therapy could postpone wound healing, as lymphangiogenesis does good to inflammation in some conditions (136). That necessitates TABLE 1 | ncRNAs that mediate lymphangiogenesis in lymphatic development, inflammation, and cancer metastasis.


ncRNA, noncoding RNA; NSCLC, non-small cell lung cancer; OSCC, oral squamous cell carcinoma; ESCC, esophageal squamous cell carcinoma; CSCC, cervical squamous cell carcinoma; PROX1, Prospero homeobox 1; FOXC2, Forkhead box C2; VEGFR-3, vascular endothelial growth factor receptor-3; VEGF-A, vascular endothelial growth factor A; VEGF-C, vascular endothelial growth factor C; DDR1, Discoidin domain receptor 1; NRP2, Neuropilin 2; Flt4, Fms-related tyrosine kinase 4; CXCL12a, Chemokine (C-X-C motif) ligand 12a; FoxO1, Forkhead box O1; LYVE-1, Lymphatic vessel endothelial hyaluronan receptor 1; WISP-1, WNT1-inducible signaling pathway protein-1; ANRIL, antisense non-coding RNA in the INK4 locus; LNMAT1, Lymph Node Metastasis Associated Transcript 1; C21orF96, Chromosome 21 open reading frame 96.

the identification of tumor lymphatics-specific markers. Recently, therapies aimed at blocking the VEGF-C/VEGF-D/VEGFR-3 signaling axis have entered clinical trials in some types of tumors, while these could not block metastasis completely (137, 138). Olmeda et al. have reported that tumors actually could induce lymphangiogenesis in distant organs rather than just SLNs before tumor cells colonization by secreting MIDKINE, and the molecular profiles and functions of LECs in distant and local lymphatic vessels might vary depending on which tissue they were in Olmeda et al. (139). As ncRNAs also have tissue-specificity expressions, their serving as a target to orchestrate lymphangiogenesis, even in more than one pathological conditions when required, is worth validation.

### AUTHOR CONTRIBUTIONS

MC, YT, Y-JT, and XL designed the subject of the review. MC and Y-JT were involved in material collection. MC, YT, and WZ were responsible for writing of the manuscript. Y-JT, WZ, and XL were responsible for figure drawing. All authors reviewed the manuscript.

#### FUNDING

This work was supported by National Natural Science Foundation of China grants (Nos. 81572650, 81672672, 81972542, and 81772891), and by State Key Laboratory of Oral Diseases Special Funded Projects.

### REFERENCES


with an indirect mechanism of action. Blood. (2006) 107:3531–6.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Cao, Tang, Zhang, Tang and Liang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Mechanisms of Vasculogenic Mimicry in Ovarian Cancer

Lízbeth Ayala-Domínguez 1,2, Leslie Olmedo-Nieva2,3, J. Omar Muñoz-Bello<sup>2</sup> , Adriana Contreras-Paredes <sup>2</sup> , Joaquín Manzo-Merino<sup>4</sup> , Imelda Martínez-Ramírez <sup>2</sup> and Marcela Lizano2,5 \*

<sup>1</sup> Programa de Doctorado en Ciencias Biomédicas, Universidad Nacional Autónoma de México, Mexico City, Mexico, <sup>2</sup> Unidad de Investigación Biomédica en Cáncer, Instituto Nacional de Cancerología-Instituto de Investigaciones Biomédicas, Universidad Nacional Autónoma de México, Mexico City, Mexico, <sup>3</sup> Programa de Doctorado en Ciencias Bioquímicas, Universidad Nacional Autónoma de México, Mexico City, Mexico, <sup>4</sup> Cátedras CONACyT-Instituto Nacional de Cancerología, Mexico City, Mexico, <sup>5</sup> Departamento de Medicina Genómica y Toxicología Ambiental, Instituto de Investigaciones Biomédicas, Universidad Nacional Autónoma de México, Mexico City, Mexico

#### Edited by:

Laurence A. Marchat, National Polytechnic Institute, Mexico

#### Reviewed by:

Kwok-Ming Yao, The University of Hong Kong, Hong Kong Maja Cemazar, Institute of Oncology Ljubljana, Slovenia

> \*Correspondence: Marcela Lizano lizano@unam.mx

#### Specialty section:

This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology

Received: 22 June 2019 Accepted: 17 September 2019 Published: 27 September 2019

#### Citation:

Ayala-Domínguez L, Olmedo-Nieva L, Muñoz-Bello JO, Contreras-Paredes A, Manzo-Merino J, Martínez-Ramírez I and Lizano M (2019) Mechanisms of Vasculogenic Mimicry in Ovarian Cancer. Front. Oncol. 9:998. doi: 10.3389/fonc.2019.00998

Solid tumors carry out the formation of new vessels providing blood supply for growth, tumor maintenance, and metastasis. Several processes take place during tumor vascularization. In angiogenesis, new vessels are derived from endothelial cells of pre-existing vessels; while in vasculogenesis, new vessels are formed de novo from endothelial progenitor cells, creating an abnormal, immature, and disorganized vascular network. Moreover, highly aggressive tumor cells form structures similar to vessels, providing a pathway for perfusion; this process is named vasculogenic mimicry (VM), where vessel-like channels mimic the function of vessels and transport plasma and blood cells. VM is developed by numerous types of aggressive tumors, including ovarian carcinoma which is the second most common cause of death among gynecological cancers. VM has been associated with poor patient outcome and survival in ovarian cancer, although the involved mechanisms are still under investigation. Several signaling molecules have an important role in VM in ovarian cancer, by regulating the expression of genes related to vascular, embryogenic, and hypoxic signaling pathways. In this review, we provide an overview of the current knowledge of the signaling molecules involved in the promotion and regulation of VM in ovarian cancer. The clinical implications and the potential benefit of identification and targeting of VM related molecules for ovarian cancer treatment are also discussed.

Keywords: vasculogenic mimicry, ovarian cancer, signaling molecules, angiogenesis, anti-angiogenic therapy

## INTRODUCTION

Ovarian cancer is the second most common and lethal gynecological cancer (1). Among ovarian cancer types, the epithelial ovarian cancer accounts for almost 90% of such malignancy (2), which is usually diagnosed in advanced aggressive stages due to its asymptomatic nature. Extensive tumor invasion, peritoneal metastases, and treatment failure are frequent in advanced epithelial ovarian cancer (3).

The normal physiology of the ovary is characterized by increased permeability of blood vessels during follicular development, ovulation, and subsequent formation of the corpus luteum, with cyclic changes in the formation, differentiation, and regression of ovarian vasculature (4).

**95**

These vascular processes are deregulated in ovarian cancer, which is characterized by intense neovascularization (5, 6). Neovasculature in ovarian cancer is formed not only from endothelial cells but also from endothelial progenitor cells and/or cells from the tumor itself, allowing the supply of blood and nutrients to the tumor with great efficiency (7).

The versatility of the vascularization processes in ovarian cancer could partially explain its aggressive nature and the limited efficacy of anti-angiogenic therapies (8). An alternative vascularization process, vasculogenic mimicry (VM), has been shown to increase after anti-angiogenic treatment with bevacizumab, in preclinical models of ovarian cancer (9). This finding suggests that VM could be a strategy for escaping antiangiogenic treatment, highlighting the importance to study the mechanisms involved in vascular remodeling.

In this review, we provide an overview of the current knowledge of the mechanisms and signaling molecules involved in the promotion and regulation of VM in ovarian cancer, its clinical implications and the potential benefit of therapeutic approaches based on the identification and targeting of VM related molecules.

#### TUMOR VASCULARIZATION PROCESSES IN OVARIAN CANCER

The study of the vascularization processes in solid tumors has gained importance due to its implication in growth and metastasis, as well as its possible implication for antiangiogenic treatment resistance (10). The most studied tumor vascularization process is angiogenesis, although tumor tissue has the capacity to generate its own vasculature from alternative mechanisms such as vasculogenesis, vessel co-option, and VM (11–13).

#### Angiogenesis

Angiogenesis is a highly regulated process aimed to produce new blood vessels with a key role in development and postnatal life; it is also involved in invasion, growth, and metastasis of solid tumors (14, 15). The onset of angiogenesis occurs in response to hypoxia or ischemia where pro-angiogenic signals overcome anti-angiogenic signals. The vascular endothelial growth factor A (VEGF-A) is the master regulator of angiogenesis, both in physiological and pathological conditions (16). During angiogenesis activation, a complex signaling cascade begins, leading to the proliferation of endothelial cells (ECs) that assemble new vascular networks from the pre-existing vessels, increasing permeability and leakage, and restoring the supply of oxygen and nutrients toward the tumor mass (15, 17).

Angiogenesis is essential for the growth of ovarian cancer cells and their spreading to the peritoneum. VEGF-A has been associated with peritoneal ECs proliferation, migration, and formation of tube-like structures (18). The inhibition of VEGF-A does not revert these processes, suggesting that another proangiogenic factors secreted by surrounding ovarian cancer cells or their microenvironment could be involved in the angiogenic activation of peritoneal ECs during metastasis (18, 19). A high level of pro-angiogenic signals has been associated with the formation of ascites, a frequent feature of advanced ovarian cancer (20, 21).

#### Vasculogenesis

Vasculogenesis, a de novo vessel formation process, is distinguished by the in situ differentiation of ECs from myeloid cells or endothelial progenitor cells (EPCs). This process takes place at the beginning of vascular development and during post-natal life (11, 22). Myeloid cells and EPCs are recruited by pro-angiogenic or pro-inflammatory factors to the tumor vascular bed, where they differentiate into ECs and give place to neovasculature (23–25). Vasculogenesis has a modest impact on tumor vascularization when the angiogenesis pathway is active, however, it is recognized as an important rescue process when this pathway is blocked (10, 26). For instance, when angiogenesis is inhibited after anti-angiogenic treatment or radiotherapy, myeloid cells, and EPCs are recruited by the stroma-derived factor 1 (SDF-1) in response to an increased level of hypoxia-inducible factor 1α (HIF-1α) (10, 26).

Vasculogenesis has an important role in ovarian cancer. It has been related to treatment resistance as a consequence of the overexpression of matrix metalloproteinase 2 and 9 (MMP-2 and MMP-9) after radiotherapy (27). Furthermore, CD34+ EPCs from peripheral blood incorporate into vasculogenic active sites (25) as well as CD11b+ and CD11c+ myeloid cells, recruited by SDF-1 and β-defensins, that contribute to vasculogenesis (28). β-defensins chemoattract CD11c+ dendritic cell precursors and then VEGF-A induces endothelial-like specialization mediated by VEGF receptor 2 (VEGFR-2); interestingly, recruitment of CD11c+ cells has also been found in ascites (28).

#### Vessel Co-option

Vessel co-option is a process that differs from angiogenesis; instead of inducing the proliferation of ECs, tumor cells grow by adhering to nearby blood vessels (15). Different patterns of vessel co-option have been described in brain, lung, and liver cancers (12). In glioblastoma, CDC42+ CD44+ tumor cells migrate toward a blood vessel in response to a bradykinin gradient created by ECs; when these cells reach the vessel, they fuse with the pericytes or adhere to the basement membrane (12). Vessel cooption has been observed in a mouse model of ovarian cancer (29), where endostatin inhibited vessel co-option by blocking the attachment of ovarian cancer cells to peritoneal vessels through integrins α5β1.

It has been proposed that after the tumor grows by vessel co-option, co-opted vessels regress, and the tumor enters into an avascular phase followed by the induction of peritumoral angiogenesis (30). Vessel co-option facilitates the metastasis of tumor cells since it increases their motility and migration. There is evidence that tumors can switch between angiogenic and non-angiogenic growth during progression and that they can contain angiogenic and non-angiogenic areas (12). The association between vessel co-option and resistance to antiangiogenic treatment is not clear, since vessel co-option could be one cause of the resistance to anti-angiogenic treatment or it could be a consequence of the aggressive nature of cancer cells in response to anti-angiogenic treatment (10, 31).

### Vasculogenic Mimicry

VM is a process by which tumor cells form capillary-like structures, mimicking the embryonic vascular network pattern, without inducing the proliferation of ECs (15). This process increases blood perfusion, allows tumor cells to obtain oxygen and nutrients, and promotes cancer progression (13, 32). It has been proposed that VM is carried out through cancer stem cell (CSC) trans-differentiation into endothelial-like cells (13, 33). Moreover, tumor cells involved in VM resemble mesenchymal cells derived from epithelial to mesenchymal transition (EMT), which is characterized by a down-regulation of epithelial markers (cytokeratin, for example), a loss of cell polarity (E-cadherin, occludin), and the upregulation of mesenchymal markers (vimentin, N-cadherin, fibronectin) (34, 35). Furthermore, these VM cells have an endothelial phenotype. VM has been associated with unfavorable outcome in patients with malignant tumors (36) and has an important participation in tumor invasion and metastasis (37).

Cell-lined vasculature compatible with VM has been observed in ∼30–37% of ovarian cancers (38, 39). The presence of such cell-lined vasculature was associated with a higher histological grade and more aggressive tumors. An increased number of VM channels were found in poorly differentiated ovarian cancer cells (40). The presence of VM, combined with the expression of CD133, was positively associated with poor prognosis in patients with ovarian cancer (41). In a preclinical model of ovarian cancer, an accelerated metastasis was observed together with hypoxia and VM after anti-angiogenic treatment with bevacizumab (an anti-VEGF-A monoclonal antibody) (9). All these findings highlight the importance of identifying the underlying mechanisms and the signaling molecules involved in VM to evaluate their prognostic or predictive value, as well as their use as potential targets for developing more effective therapies (42).

## STRUCTURAL AND FUNCTIONAL DESCRIPTION OF VM

In 1999, Maniotis et al. performed in vitro and in vivo assays in melanoma and found two VM types: a tubular type, and a patterned matrix type (13). The tubular type consists of hollow cords formed by tumor cells that give place to a tubular network. These tubular structures are also connected to other channels that contain red blood cells. Further studies showed that in some cases, a mixture of tumor cells and ECs form those tubular structures (43). The patterned matrix type consists of a network of loops formed by matrix layers that surround clusters of tumor cells. These layers are not uniformly spaced; therefore, the transport of fluid is not uniform around the cell cluster. However, this patterned matrix could provide a greater surface area for diffusion than that provided by a tubular structure (44). The reorganization of tumor cells into cords or clusters, as well as the formation of matrix layers involve mechanisms such as cell-cell adhesion, migration, and extracellular matrix remodeling, where several signaling molecules have been associated with VM.

VM structures have been identified in tissue samples as positive for periodic acid-Schiff (PAS) and negative for EC markers such as CD31 or CD34 (42). PAS+ regions are rich in components of the extracellular matrix, like laminin. Recently, it has been shown that PAS+ regions could also represent non-functional structures unrelated to VM in in vitro studies (45). Moreover, the different vascular structures aimed to conduct fluids within the tumor share several features. Thus, in order to identify the structures that truly correspond to VM as well as to distinguish them from similar structures from the other vascularization processes, it is necessary to assess their architectural and functional features, in addition to their composition. Recently, Valdivia et al. (46) described the architectural and functional features required for differentiating VM from other vascular structures in tumors (46). Whilst blood and lymph vessels are formed by a single line of ECs surrounded by a continuous and noncontinuous basement membrane, respectively, VM structures are formed by cancer cells resting on an inner glycoprotein rich matrix (46). The authors propose that, in addition to the traditional architectural features to identify a VM structure (PAS+ and without EC markers), the presence of red blood cells within the lumen of the structure is an indicator of VM functionality (46).

Early studies in breast cancer and melanoma have shown that tubular and patterned matrix VM structures are capable of conducting plasma and red blood cells in vitro and in vivo (44). Maniotis et al. (47) showed that VM structures formed by aggressive melanoma cells in vitro conducted a tracer by direct microinjection and passive diffusion (47); moreover, the matrix pattern also contained red blood cells. Shirakawa et al. (48) used two breast cancer mice models to evaluate tumor blood flow with micro-magnetic resonance angiography imaging (48). They found that aggressive tumor cells formed VM structures in the center of the tumor, while non-aggressive cells showed necrotic cores. Angiogenic vessels were present in tumor periphery in both types of tumors and blood flow was higher in VM structures than in necrotic cores. Clarijs et al. (49) used a tracer to study perfusion in a melanoma mouse model (49). Tracer distribution suggested that blood vessels could be in contact with VM structures, allowing the perfusion of the latter, mediated by at least three mechanisms: the anastomosis of VM structures to blood vessels (50), an increased leakage from blood vessels (47, 51), and through anticoagulant control exerted by aggressive tumor cells (50).

### MECHANISMS AND SIGNALING MOLECULES INVOLVED IN VM IN OVARIAN CANCER

Several mechanisms are involved in VM, including those related to the capacity of aggressive tumor cells to resemble features of the ECs such as cell adhesion (52), migration (53), extracellular

HIF-1α induces the expression of Twist1, VE-Cadherin (VE-Cad), Slug and Vimentin, which are involved in VM induction; moreover, proteins such as pSTAT3, HCG, and LHR regulate the levels of HIF-1α. CSC markers, including ALDH and CD133 are found in ovarian tissues with VM structures. Different cell signaling pathways are also involved in VM, such as Wnt5a, and RTKs pathways, which strongly correlate with VM formation. Additional molecules that have been proposed in VM regulation in ovarian cancer are Sema4, XAF1, and Mig-7, however the precise mechanisms remain unclear.

matrix remodeling (54), perfusion (50), and maturation of blood vessels (55). Moreover, CSCs promote VM by deregulating pathways involved in embryonic development, such as the transforming growth factor β (TGF-β) (56–58), Wnt (59), Notch (60, 61), Nodal (62–64), and the Hippo pathways (65–67), among others. EMT also plays an important role in VM, and encompasses the pathways previously mentioned as well as transcription factors such as Twist1/2 (68), Snail/Slug (69), and ZEB1/2 (34). Moreover, signaling molecules related to hypoxia (70), inflammation (71–75), and metabolism (76–79) also have an impact on VM. The novel findings regarding these mechanisms and their signaling molecules in the regulation of VM in ovarian cancer are presented in this section and are summarized in **Figure 1**. Additional proteins and signaling pathways identified in other cancer types are shown in **Table 1**.

Vascular endothelial (VE)-cadherin, one of the main participants in cell-cell adhesion in endothelial cells, is strongly associated with VM formation (80). This protein recruits the EC-related kinase Ephrin-A2 receptor (EphA2) to the cell membrane (52), increasing the phosphorylation of the focal adhesion kinase (FAK). Consequently, the activation of extracellular regulatory kinases 1 and 2 (ERK1/2) signaling pathway is promoted, allowing the activation of MMP-14 (81). Then, MMP-14 converts proMMP-2 into active MMP-2. These MMPs degrade extracellular matrix components and facilitate invasion, metastasis, and VM (82). Particularly, MMP-2 and MMP-14 induce the Laminin5γ2 (Ln5γ2) cleavage (53, 83). Although the precise mechanism has not been clearly described, it is known that MMP-2 cleavages Ln5γ2 into two pro-metastatic fragments (Ln5γ2 ′ and Ln5γ2x) (53). Together, these results indicate that the VE-cadherin/EphA2/MMP-2/Ln5γ2 axis is the main regulator of VM.

Interestingly, high expression of VE-cadherin and EphA2 has been found in clinical samples from ovarian cancer patients


TABLE 1 | Signaling molecules and mechanisms that regulate VM in several types of cancer.

VM, vasculogenic mimicry; EMT, epithelial to mesenchymal transition; CSC, cancer stem cell.

that exhibit a highly invasive phenotype (84, 85). Additionally, other studies demonstrated that MMP-2 and MMP-14 are also overexpressed in ovarian cancer samples, which is associated with poor clinical outcome (38, 86). It is worth to mention that those findings strongly correlated with the presence of VM structures, suggesting that these molecules are important players in this process.

The Phosphatidylinositol 3-kinase (PI3K) cell signaling pathway regulates MMPs expression in VM (87). This pathway is activated through FAK phosphorylation (88), impacting in the expression of MMP-14. Moreover, the PI3K pathway is frequently activated in ovarian cancer, probably impacting VM (89).

Another regulator of VM is the urokinase plasminogen activator (uPA), which is required to induce the degradation of the extracellular matrix, impacting in tumor angiogenesis. The overexpression of uPA positively correlates with VM formation in ovarian cancer tissues (54). In addition, it was demonstrated that in SKOV-3 and OVCAR-3 ovarian cancer cells, the ablation of uPA expression results in a decrement of complete VM structures formation and such mechanism involves the participation of AKT/mTOR/MMP-2/Laminin5γ2 signal pathways (54).

VEGF-A also upregulates the expression MMPs. It has been shown that in melanoma, VEGF-A induces VM formation by activating the PI3K/protein kinase C α (PKCα) pathway via VEGFR-1 signaling (90). However, in glioblastoma, VM is induced by the VEGFR-2 signaling (91). In an in vitro model of ovarian cancer using SKOV-3 and OVCAR-3 cells, VEGF-A promoted migration, invasion, and VM by up-regulating MMPs via EphA2 (92). This suggests that VEGF-A interacts with the VE-cadherin/EphA2/MMP-2/Ln5γ2 axis in the regulation of VM in ovarian cancer.

The plasma membrane glycoprotein CD147 plays an important role during tumor progression, invasion and metastasis, regulating metalloproteinases expression in peritumoral stromal cells. Invasion capability was evaluated in two different cell lines derived from ovarian cancer with different invasion activity: CABA I and SKOV3 (93). A correlation of CD147 expression with tumor invasiveness, protease activity (MMP-2 and MMP-9), and vascular channels formation was observed. Interestingly, when high invasive cell line was treated with small interfering RNA against CD147, a suppression of non-EC-lined channels was observed. In addition, when CD147 was overexpressed in a low invasive cell line, those cells exhibited an increase of tumor invasion and vascular channel formation. These data suggest that CD147 plays an important role in VM induction in ovarian tumors and CD147 could be an attractive target for therapeutic intervention (93).

Furthermore, Ln5γ2 activates the endothelial growth factor receptor (EGFR) which promotes the expression of the migration-inducing protein 7 (Mig-7), stimulating invasion and VM (94). A study carried out in ovarian cancer samples revealed an association of VM with VE-cadherin and Mig-7 expression (84). It was observed that ovarian tumors without VM frequently expressed low levels of VE-cadherin compared to those with VM. Meanwhile, Mig-7 expression was increased in tumor samples compared to normal tissues, positively correlating with VM and VE-cadherin expression (84).

Some elements involved in apoptosis have been associated with the formation of VM structures, such as the pro-apoptotic XIAP-associated factor 1 (XAF1). Recently, in vivo xenograft models of ovarian cancer have shown that the overexpression of XAF1 decreases the number of VM structures (39). Moreover, in vitro assays with SKOV3 cells revealed that proliferation, migration and invasion were inhibited, and the levels of VEGF were reduced when XAF1 was exogenously overexpressed (39). Therefore, XAF1 is a potent negative regulator of VM in ovarian cancer.

It has been shown that VEGF-A regulates the expression of the axon guidance factor semaphorin 4D (Sema4D) (95), which has been identified as a promotor of VM in non-small cell lung cancer (96), where the recognition of Sema4D by the plexin B1 receptor activates the small GTPase RhoA, which is implicated cell motility. However, when plexin B1 was inhibited, a disruption of the RhoA/ROCK signaling occurred, suppressing VM formation. Additionally, the presence of VM in clinical specimens correlated with increased levels of Sema4D (96). In an ovarian cancer cell line (A2780), soluble Sema4D promoted angiogenesis and VM via plexin B1 (95); moreover, in clinical samples from patients, a high expression of Sema4D had a positive correlation with the malignant degree of epithelial ovarian cancer. Interestingly, it was observed that VEGFR-2, plexin-B1, and Sema4D control the expression of CD31, MMP-2, and VE-cadherin in ovarian cancer cells, which are the markers and initiators of angiogenesis and VM (95).

CSCs are present in ovarian cancer and are positive for CD133, a unique surface marker of CSCs (97). It is known that CD133+ cells promote VM in several cancer types (41, 91, 98–101). The combined expression of CD133 and VM in samples from patients was associated with high-grade ovarian carcinoma, latestage disease, non-response to chemotherapy and shorter overall survival (41). The trans-differentiation of CD133+ CSCs into ECs may induce VM formation and the expression of EC markers such as VE-cadherin (101) and VEGFR-2 (91). Moreover, it has been shown that in hypoxic environment the subpopulation of CD133+ CSCs is augmented when Twist1 was overexpressed (100). This finding shows that hypoxia may exert an effect on CSCs that probably leads to VM formation.

CSCs can also exhibit a high activity of aldehyde dehydrogenase-1 (ALDH1) (97). The expression of ALDH1 has been evaluated in different types of tumor, including breast cancer, colorectal cancer, and ovarian cancer and strongly correlates with VM, determining an unfavorable clinical outcome (102–104). Although the precise mechanism has not been described, it is known that ALDH1 and VM increase in response to hypoxia (105).

Hypoxia regulates several pathways in cancer, such as angiogenesis, and it has been related to VM in melanoma, glioblastoma, ovarian cancer, and hepatocellular carcinoma (68– 70, 106). Hypoxia induces VM formation by up-regulating VEcadherin expression. The main effectors of this pathways, HIF-1α and HIF-2α, positively regulated VE-cadherin expression; this effect is through the binding of HIF to hypoxia response elements (HRE) located in VE-cadherin promoter in glioblastoma cells (106). Interestingly, it was observed that EMT is promoted in a hypoxic environment and as a result, VM was induced in SKOV3 and OVCAR3 cells (69). In vitro assays showed that hypoxia leads to increased invasion, migration and an enhancement of MMP-2 activity. Therefore, EMT induction as a response to hypoxia is a master regulator of VM in ovarian cancer cells. Moreover, this study demonstrated that in ovarian cancer samples, the levels of HIF-1α were strongly associated with VM formation and the expression of Twist1, Slug, and Vimentin.

Another important regulator of VM under hypoxic conditions are the signal transducer and activator of transcription 3 (STAT3) and the phospho-STAT3 (p-STAT3). It has been suggested that p-STAT3 promote VM, this is due to the binding of pSTAT3 to HIF-1α, which in turn delays its degradation (107, 108). In gastric adenocarcinoma, VM was associated with an increased expression of HIF-1α, STAT3, and p-STAT3 (109). Moreover, STAT3 acts as a transcription factor in VEGF-A transcription (110). Interestingly, in SKOV3 cells p-STAT3 was found in the nucleus, suggesting that was transcriptionally active (111). In addition, when STAT3 was inhibited, the formation of VM structures was completely avoided, suggesting that p-STAT3 is an important regulator of VM in ovarian cancer cells.

The Wnt family members regulate EC differentiation and vascular development (112) and has been associated with VM. In glioma and colon cancer, the canonical Wnt/β-catenin pathway induced VM by increasing the expression of VEGFR-2 and VE-cadherin (59, 113). Interestingly, in ovarian cancer, the non-canonical Wnt signaling is implicated in VM formation. It was found that Wnt5a is overexpressed in tumor samples and is associated with VM (114). Moreover, in vitro analysis revealed that Wnt5a overexpression is linked to PKC pathway activation. Furthermore, it was shown that Wnt5a overexpression induced EMT, increased invasion and motility of SKOV3 cells (114).

An important proangiogenic factor in ovary is the human gonadotropin (HCG). The fifth subunit of β-HCG, CGB5, was shown to promote VM formation in vitro in OVCAR3 cells (115). Additionally, overexpression of CGB5 induced the growth of ovarian cancer cells in a xenograft murine model, as well as VM (116). It was also shown that the activation of luteinizing hormone receptor (LHR), which is the HCG receptor, is required for the promotion of VM formation by CGB5. In another study, it was found that ovarian cancer cells exogenously expressing HCG induced an overexpression of HIF-1α. Importantly, vascular markers such as CD31 and VEGF were also upregulated in those cells (117). Therefore, the HCG/LHR axis induces VM by HIF-1α regulation in ovarian cancer.



VM, vasculogenic mimicry; HCC, hepatocellular carcinoma.

#### MICRO-RNAS AS REGULATORS OF VM IN OVARIAN CANCER

Micro-RNAs (miRNAs) are single stranded and non-coding RNA molecules of 19-25 nucleotides in length that have a post-transcriptional regulatory function (118). Different studies have demonstrated that miRNAs are involved in several physiological processes such as cell proliferation, invasion, migration, differentiation, as well as pathological processes including angiogenesis and VM (119–122). The dysregulation in the expression of these RNA molecules is often observed in numerous types of cancer. Diverse studies demonstrate that miRNAs post-transcriptionally regulate different signaling molecules involved in VM process (123–136); examples of these miRNAs are enlisted in **Table 2**.

A well-described miRNA family is miR-26, which includes miR-26a and miR-26b. Those are commonly downregulated in several types of cancer such as glioma, HCC, and gastric cancer (124, 137, 138). For instance, in gastric cancer miR-26a and−26b suppress angiogenesis by targeting hormone growth factor (HGF) mRNA and consequently affecting HGF/VEGF signaling (138). Moreover, in HCC miR-26b has been identified as tumor suppressor since its down-regulation promotes VM and angiogenesis (123).

Similarly, another cluster of miRNAs belonging to the miR-200 family (miR-141, miR-200a, miR-200b, miR-200c, and miR-429) has been widely studied in several types of cancers (125, 139– 142). It has been shown that miR-141 overexpression inhibits VM formation through directly targeting EphA2 transcript, decreasing EphA2 protein levels in glioma and renal carcinomas (122, 125).

Hitherto, three miRNAs (miR-200a, miR-27b, and miR-765) have been described as VM regulators in ovarian cancer through directly targeting 3′UTRs of VM-related transcripts (85, 126, 128) (**Table 3**). The miR-200a was the first microRNA found in ovarian cancer capable of regulating VM (85). Tumors with low miR-200a expression correlate with the presence of VM structures and poor overall survival. An inverse correlation between mRNA and protein EphA2 levels and miR-200a expression was observed in ovarian cancer samples, suggesting a direct regulation among them. In silico assays revealed a miR-200a binding site at EphA2 3′UTR; this observation was confirmed in SKOV3 ovarian cancer cells, where a direct binding of miR-200a to EphA2 3′UTR was observed through luciferase assays. Consequently, the levels of EphA2 protein and mRNA decreased in this model. In agreement, it was shown that the EphA2 overexpression restores VM in miR-200a expressing cells, indicating that miR-200a inhibits VM by mainly targeting EphA2 (85).

Previously, it has been described that VE-cadherin expression is related to VM formation in different types of cancer. A bioinformatic study identified miR-27b as putative regulator of VE-cadherin by the detection of a binding site at VEcadherin 3′UTR. Concordantly with this result, luciferase assays demonstrated that miR-27b binds to VE-cadherin mRNA 3′UTR in ovarian cancer cells. Furthermore, expression levels of VEcadherin mRNA and protein in different ovarian cancer cell lines negatively correlate with miR-27b expression. Low metastatic cell lines OVCAR3 and SKOV3 express high amounts of miR-27b and low VE-cadherin mRNA, compared to metastatic cells ES2 and Hey1B that exhibit low amounts of miR-27b and high VE-cadherin mRNA. Overexpression of miR-27b on high VEcadherin expressing cells decreases VE-cadherin mRNA and protein levels. When miR-27b is overexpressed in metastatic ovarian cancer cell lines (Hey1B and ES2), the migration, invasion, and VM are decreased in in vivo models (126).

A recent study aimed to determine the set of miRNAs regulated in an early stage before complete VM establishment under hypoxia conditions. It was shown that SKOV3 ovarian cancer cells grown under hypoxia conditions form a higher number of 3D capillary-like structures than those cells grown under normoxia conditions (128). A set of miRNAs involved in the regulation of tumorigenesis-related pathways, as well as several genes involved in VM and angiogenesis was found. Among them, miR-765 was highly downregulated under hypoxia (128). Moreover, its restoration promotes a dramatic inhibition of 3D capillary-like structures and down-regulates VEGF-A expression. Importantly, it was demonstrated that VEGF-A mRNA is a direct target of miR-765, since it binds to VEGF-A 3'UTR. Additionally, low levels of miR-765 and high levels of VEGF-A were associated with low overall survival from a cohort of 1,485 ovarian cancer patients (128).

Although only three miRNAs have been directly associated with VM in ovarian cancer, several signaling pathways, and proteins controlling this mechanism are regulated by miRNAs (143–164); therefore, these non-coding transcripts could have a potential role on VM regulation. **Table 3** shows the common VM targets in ovarian cancer that are regulated by miRNAs.

#### TABLE 3 | VM related miRNAs in ovarian cancer.


VM, vasculogenic mimicry; OC, ovarian cancer.

#### CLINICAL IMPLICATIONS OF THE SIGNALING MOLECULES OF VM IN OVARIAN CANCER

Anti-angiogenic therapies have shown limited effects against cancer progression, due to alternative vascularization processes, such as VM, triggered by aggressive tumor cells (10). The knowledge of the mechanisms and signaling molecules involved in VM may lead to the development of novel anti-vascularization therapies that overcome the limitations found in conventional therapies. Therefore, it is necessary to explore the possible therapeutical strategies that could improve the clinical outcome of ovarian cancer patients.

Therapies targeting VM have not been developed in ovarian cancer so far. However, some inhibitory molecules of VM elements have been studied and have shown promising anti-VM effects (165–174). These inhibitor molecules are summarized in **Table 4**.

Studies using pancreatic cancer cells showed that Ginsenoside Rg3, a tetracyclic triterpenoid saponin, reduces VM in xenograft mice models. Moreover, the expression of VE-cadherin, EphA2, MMP-2, and MMP-9 was also down-regulated after the treatment (174). Ginsenoside Rg3 has been proved in ovarian cancer derived cells restraining HIF-1α expression by activating the ubiquitin-proteasome pathway. This effect efficiently blocked migration and EMT in in vitro and in vivo ovarian cancer models, promising a novel anti-VM therapeutic agent (175, 176).

It has been shown that PARP inhibition sensitizes for chemo and radiotherapy in different types of tumors. In melanoma cells that were treated with PARP inhibitors (PJ-34, Isoquinolinone, or Olaparib) a reduction of pro-metastatic and VM markers was observed (177). PARP I inhibitors, such as Olaparib


TABLE 4 | Inhibitor molecules that target VM-related proteins.

The different molecules with a potential VM-therapeutic effect that has been tested in different types of cancer.

VM, vasculogenic mimicry; HCC, hepatocellular carcinoma.

and Rupaparib, have been approved for the treatment of recurrent BRCA-associated ovarian cancer by the Food and Drug Administration (FDA); while Niraparib is used as maintenance therapy following chemotherapy for recurrent ovarian cancer (178). Nevertheless, to date there is no information about their effect on VM in ovarian cancer.

Thalidomide is an immunomodulatory agent with strong antiangiogenic properties and has been proved in ovarian cancer, glioblastoma, hepatocellular carcinoma, and multiple myeloma in diverse clinical trials. Induction therapy with thalidomide significantly improved the overall response rate, progression free survival and overall survival (179). Previously, it has been shown that thalidomide suppresses tumor growth and angiogenesis in murine models (180). Interestingly, in a xenograft mouse model of melanoma, it was observed that mice treated with Thalidomide induced necrosis in melanoma cells. In addition, VM and tumor growth were significantly reduced compared to non-treated specimens. This effect could be related to the down-regulation of NF-kappaB signaling pathway (181). However, further studies are required to elucidate this statement.

A monoclonal antibody has been developed to target VM, unfortunately it has not been introduced for ovarian cancer treatment. This antibody targets the outer-membrane immunoglobulin-like domains of VE-cadherin, blocking receptor function. In lung cancer cells, it was observed that this antibody functions as an anti-VM agent for cancer treatment, since it inhibited the activation of the VE-cadherin-related pathway in VM (182). Due to the advantages that monoclonal antibody therapies imply, its application in ovarian cancer as an anti-VM agent is promising.

Other molecules implicated in VM in ovarian cancer, such as miR-200a, miR-27b, and miR-765 represent potential candidates for anti-tumoral therapies (85, 126, 128). Importantly, the current strategies are focused in the reduction of cancer through restoring the expression of down-regulated miRNAs, also known as miRNA replacement therapy. There are several ways to harness miRNAs in cancer cells for therapeutic purposes, including introduction of synthetic miRNA mimics, miRNA expressing plasmids, and small molecules that epigenetically alter endogenous expression of miRNAs (183). Such anti-VM strategies could represent an opportunity to venture into the study of new molecules for therapeutic purposes in ovarian cancer. Further studies will be required to prove the effectiveness of such molecules for treatment purposes.

#### CONCLUDING REMARKS

Ovarian cancer is a common gynecological cancer and it is usually diagnosed in advanced stages where therapeutic success is limited. This type of tumors exhibit an aggressive phenotype characterized by a high rate of metastasis, invasion, and poor treatment response. These features are highly associated with the development of neovasculature formed by both endothelial and tumor cells. Particularly, MV is a process that may be influencing ovarian cancer poor prognosis and limited efficacy of anti-angiogenic strategies. Nevertheless, the mechanisms underlying VM formation in ovarian cancer remains unclear and deserves further studies. Recently, molecules that regulate cellular adhesion, hypoxia and EMT have been identified as key regulators of VM. Additionally, it has been shown an important post-transcriptional regulation mediated by microRNAs, that impact on the expression of VM-related proteins such as VEcadherin, EphA2, and VEGF. Furthermore, this information has allowed the development of strategies with therapeutic potential directed against VM formation. However, subsequent studies will be necessary to elucidate the mechanisms that allow the development of conventional anti-angiogenic therapies combined with the novel anti-VM targets that improve the clinical outcomes of ovarian cancer patients.

#### AUTHOR CONTRIBUTIONS

LA-D, LO-N, JM-B, AC-P, JM-M, IM-R, and ML performed the bibliographic review, wrote, and critically revised the manuscript. ML conceived and directed the manuscript.

#### FUNDING

This work was partially supported by CONACyT grant CB-251497, PAPIIT-UNAM IN103219, and Instituto Nacional de Cancerología Ref. 018/051/1B1/CE1/1294/18.

### REFERENCES


### ACKNOWLEDGMENTS

LA-D and LO-N are students from the following programs: Doctorado en Ciencias Biomédicas and Doctorado en Ciencias Bioquímicas, respectively, at the Universidad Nacional Autónoma de México and are recipients of scholarships from CONACyT, México (221487 and 289892, respectively).


mediated by angiopoietins and VEGF. Science. (1999) 284:1994–8. doi: 10.1126/science.284.5422.1994


promoting Bcl-2/Twist1 cooperation. Exp Mol Pathol. (2015) 99:383–91. doi: 10.1016/j.yexmp.2015.08.009


of lung cancer cells. Oncol Rep. (2018) 39:2837–44. doi: 10.3892/or.2018. 6374

183. Hosseinahli N, Aghapour M, Duijf PHG, Baradaran B. Treating cancer with microRNA replacement therapy: a literature review. J Cell Physiol. (2018) 233:5574–88. doi: 10.1002/jcp.26514

**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Ayala-Domínguez, Olmedo-Nieva, Muñoz-Bello, Contreras-Paredes, Manzo-Merino, Martínez-Ramírez and Lizano. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# IL27Rα Deficiency Alters Endothelial Cell Function and Subverts Tumor Angiogenesis in Mammary Carcinoma

Annika F. Fink <sup>1</sup> , Giorgia Ciliberti <sup>2</sup> , Rüdiger Popp<sup>2</sup> , Evelyn Sirait-Fischer <sup>1</sup> , Ann-Christin Frank <sup>1</sup> , Ingrid Fleming<sup>2</sup> , Divya Sekar <sup>1</sup> , Andreas Weigert <sup>1</sup> \* and Bernhard Brüne<sup>1</sup> \*

<sup>1</sup> Faculty of Medicine, Institute of Biochemistry I, Goethe-University Frankfurt, Frankfurt, Germany, <sup>2</sup> Faculty of Medicine, Institute for Vascular Signalling, Goethe-University Frankfurt, Frankfurt, Germany

#### Edited by:

Laurence A. Marchat, National Polytechnic Institute, Mexico

#### Reviewed by:

Prashant Trikha, Nationwide Children's Hospital, United States Aman Sharma, ExoCan Healthcare Technologies Pvt Ltd, India

#### \*Correspondence:

Andreas Weigert weigert@biochem.uni-frankfurt.de Bernhard Brüne b.bruene@biochem.uni-frankfurt.de

#### Specialty section:

This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology

Received: 12 July 2019 Accepted: 23 September 2019 Published: 04 October 2019

#### Citation:

Fink AF, Ciliberti G, Popp R, Sirait-Fischer E, Frank A-C, Fleming I, Sekar D, Weigert A and Brüne B (2019) IL27Rα Deficiency Alters Endothelial Cell Function and Subverts Tumor Angiogenesis in Mammary Carcinoma. Front. Oncol. 9:1022. doi: 10.3389/fonc.2019.01022 IL-27 regulates inflammatory diseases by exerting a pleiotropic impact on immune cells. In cancer, IL-27 restricts tumor growth by acting on tumor cells directly, while its role in the tumor microenvironment is still controversially discussed. To explore IL-27 signaling in the tumor stroma, we used a mammary carcinoma syngraft approach in IL27Rα-deficient mice. Tumor growth in animals lacking IL27Rα was markedly reduced. We noticed a decrease in immune cell infiltrates, enhanced tumor cell death, and fibroblast accumulation. However, most striking changes pertain the tumor vasculature. Tumors in IL27Rα-deficient mice were unable to form functional vessels. Blocking IL-27-STAT1 signaling in endothelial cells in vitro provoked an overshooting migration/sprouting of endothelial cells. Apparently, the lack of the IL-27 receptor caused endothelial cell hyper-activation via STAT1 that limited vessel maturation. Our data reveal a so far unappreciated role of IL-27 in endothelial cells with importance in pathological vessel formation.

Keywords: IL-27 cytokine, endothelial cell, mammary cancer, cytokine, angiogenesis

#### INTRODUCTION

Interleukin 27 (IL-27) is a heterodimeric cytokine of the IL-12 family, composed of IL-27p28 and Epstein–Barr virus (EBV)-induced gene 3 (EBI3). It is mainly expressed and secreted by antigen presenting cells. IL-27 signals via a receptor complex, consisting of IL27Rα and the signaltransducing glycoprotein 130 (gp130) (1, 2). Gp130 is found in a number of receptor complexes, including the IL-6 receptor. Therefore, specificity of IL-27 signaling depends on IL27Rα. IL27Rα is expressed on many immune and stromal cells, whereas it is nearly absent on B cells and neutrophils. Once IL-27 binds to its receptor complex, mainly janus kinase (JAK) and downstream signal transducer and activator of transcription (STAT) are activated (3, 4).

IL-27 regulates inflammation by acting, among others, on T cells. Its pleiotropic functions are shaped by a given inflammatory environment. IL-27 can enhance Th1 immunity by suppressing Th2/Th17 cell development (5, 6), but also acts immune-suppressive, e.g., by upregulating inhibitory immune checkpoint receptors, such as PD-L1 and CTLA4 (7, 8). Consequently, IL-27 affects a number of diseases. IL27Rα-deficient mice treated with a high dose of dextran sulfate sodium (DSS) elevated Th17 cell activity, translating into aggravated colitis. In contrast, IL-27 application in an acute colitis model attenuated disease outcome (9, 10). Moreover, IL-27 delayed the onset of experimental autoimmune encephalomyelitis (EAE), which was attributed to enhanced IL-10 expression and downstream suppression of IL-17 production (11). Indeed, the absence of IL27Rα aggravated EAE outcome, with increased Th17 cell numbers (12).

Also, the impact of IL-27 on tumor development revealed divergent effects. IL-27 overexpressing C26 colon carcinoma cells induced interferon γ (IFNγ) expression in splenic cells, promoting antitumor activity by augmenting CD8<sup>+</sup> T cells (13). In addition to potential immune-stimulatory effects, IL-27 directly inhibited proliferation and tumorigenicity of human prostate cancer cells (hPCa) in vitro (14), as well as in vivo in a xenograft mouse models with hPCa cells or human multiple myeloma cells (14, 15). Immune cell independent effects were also suggested when IL-27 inhibited the growth of subcutaneously implanted B16-F10 melanomas, in wildtype (WT) as well as IFNγ-deficient or NOD-SCID mice. In this setting, IL-27 restricted B16-F10 pulmonary metastasis by inducing the production of the antiangiogenic chemokines CXCL10 or CXCL9 from HUVECs (16). However, a tumor-promoting role of IL-27 has also been proposed. IL-27 induced immune-suppressive molecules in stromal cells, including immune checkpoint molecules and CD39 (17, 18). To further explore the role of IL-27 in tumor stromal cells, we used a mammary carcinoma cell syngraft approach in IL27Rα-deficient mice. While our data confirm a tumor-promoting role of IL-27 in the tumor stroma, we uncovered an unexpectedly strong impact of IL-27 signaling on the tumor vasculature. The absence of IL-27 signaling severely limits the formation of functional blood vessels and thus, tumor angiogenesis.

### MATERIALS AND METHODS

#### Reagents

Epigallocatechin gallat (EGCG), Stattic and lipopolysaccharide (LPS) were purchased from Sigma-Aldrich (St. Louis, USA). IFNγ was from BioVision (Milpitas, USA). IL-4 was from Peprotech (Hamburg, Germany). IL-27 was obtained from Biolegend (Koblenz, Germany), IL-27 neutralizing antibody was from Invitrogen (Carlsbad, USA), and the IgG2a istotype control was from BioXCell (West Lebanon, USA). Macrophage colony-stimulating factor (M-CSF) and granulocyte-macrophage colony-stimulating factor (GM-CSF) were from ImmunoTools (Friesoythe, Germany). All reagents were dissolved according to the manufacturer's instructions.

#### Cell Culture

The murine endothelial cell line bEnd5 was obtained from the HPA Culture Collections via Sigma-Aldrich in August 2018. Experiments with these cells were completed within 3 months and the cells were therefore not authenticated again. bEnd5 cells were cultured in DMEM (Thermo Fisher Scientific, Waltham, USA) containing 1% sodium pyruvate (Sigma-Aldrich) and 1% non-essential amino acids (Sigma-Aldrich). Fibroblast 3T3 cells were cultured in DMEM/F-12 medium (Thermo Fisher Scientific). Murine breast cancer cells (PyMT) were cultured in DMEM containing 1% sodium pyruvate, 1% non-essential amino acids, and 10 mmol/L HEPES (Sigma-Aldrich). Media was supplemented with 10% FCS (Capricorn Scientific, Epsdorfergrund, Germany), 100 U/ml penicillin, and 100µg/ml streptomycin (PAA laboratories, Cölbe, Germany).

### Animal Experiments

Murine breast cancer cells derived from a mouse expressing the polyoma virus middle T oncoprotein (PyMT) under the mouse mammary tumor virus promoter were transplanted into four mammary glands of IL27Rα wildtype (WT) and knockout (KO) mice. Tumor growth was monitored for up to 31 days until tumors reached a diameter of 1.5 cm in WT animals. Tumor volume was calculated as follows: volume = 0.5 × (length × width<sup>2</sup> ). After 21 or 31 days, mice were euthanized followed by cardiac perfusion with 0.9% NaCl solution and tumors were harvested. Animal experiments followed the guidelines of the Hessian animal care and use committee (approval No. FU/1106).

#### Flow Cytometry

Single suspensions of tumors were generated using the mouse tumor dissociation kit and the gentleMACS dissociator (Miltenyi Biotec, Bergisch Gladbach, Germany). Single cell suspensions were stained with fluorochrome-coupled antibodies and analyzed by flow cytometry using an LSRII Fortessa cell analyzer (BD Biosciences, Heidelberg, Germany). Data were analyzed using FlowJo software VX (Treestar, Ashland, USA). Antibodies were titrated to determine optimal concentrations. For single-color compensation CompBeads (BD Bioscience) were used to create multi-color compensation matrices. Cells were blocked with 2% Fc Receptor Binding Inhibitor (Miltenyi) in PBS for 10 min on ice. Afterwards, cells were stained for either analyzing the immune cell composition, or for characterizing endothelial cells. To discriminate immune cell subsets in tumors the following Abs were used: anti-CD3-PE-CF594 (BD); anti-CD4- BV711 (BD); anti-CD8-BV650 (Biolegend); anti-CD11b-BV605 (Biolegend); anti-CD11c-BV711 (BD); anti-CD19-APC-H7 (BD); anti-CD25-PE-Cy7 (BD); anti-CD44-AlexaFluor700 (BD); anti-CD45-VioBlue (Miltenyi Biotec); anti-CD326- BV711 (BD); anti-GITR-FITC (Biolegend); anti-F4/80-PE-Cy7 (Biolegend); anti-Ly-6C-PerCP-Cy5.5 (BD); anti-Ly-6G-APC-Cy7 (BD); anti-NK1.1-BV510 (BD). To define endothelial cell (EC) populations the following Abs were used: anti-CD45-AlexaFluor700 (BD); anti-CD326-BV711 (BD); anti-CD31-PE-Cy7 (eBioscience); anti-CD204-PE (Miltenyi); anti-LYVE-1-PE (R&D system); anti-CD90.2-PE (Miltenyi); anti-CD146-AlexaFluor488 (BD); anti-ICAM1(CD54)-BV421 (BD); anti-CD62P(P-selectin)-BV510 (BD); anti-CD62E(Eselectin)-BV650 (BD); anti-CD109(VCAM1)-PerCP-Cy5.5 (Biolegend); anti-CD141(Thrombomodulin)-APC (Novus, Wiesbaden, Germany).

#### Histology and Immunohistochemistry

Tumors and lungs were zinc fixed and paraffin-embedded. Tumor sections were stained using the Opal staining system and analyzed with InForm software using the phenotyping tool according to the manufacturer's instructions (PerkinElmer, Rodgau, Germany). Tumor sections were stained with the following antibodies: cleaved caspase (Cell Signaling, Cambridge, U.K.); Ki67 (abcam, Cambridge, U.K.); hypoxia-inducible factor 1-alpha (HIF1α) (Novus); panCytokeratin (abcam); CD31 (BD); alpha smooth muscle actin (αSMA) (Sigma-Aldrich); spectral DAPI (PerkinElmer); neural/glial antigen 2 (NG2) (R&D systems, Minneapolis, USA). For metastases at least nine independent sections of each lung were stained with Mayer's hemalum (Merck, Darmstadt, Germany) and analyzed. Secondary antibody controls for each antibody species were routinely included (**Supplementary Figure 1**).

### BSA-FITC Vessel Permeability Assay

FITC labeled BSA (50 mg/kg) (Sigma-Aldrich) was injected i.p. 90 min prior to sacrificing mice. FITC-dependent fluorescence was visualized together with CD31 as indicated above. The FITCpositive area was analyzed using ImageJ.

### Isolation and Generation of Bone Marrow Derived-Macrophages (BMDM)

For the generation of BMDMs, femur and tibia of WT and KO mice were extracted. BM cells were plated in RPMI 1640 medium containing 20 ng/ml GM-CSF and 20 ng/ml M-CSF. Cells were incubated for 7 days. Afterwards cells were exposed to 100 ng/ml LPS, 10 ng/ml IFNγ, 20 ng/ml IL-4 or directly co-cultured with PyMT cells.

### RNA Isolation and Quantitative Real-Time PCR

RNA from tumor samples were isolated using the PeqGold RNAPureTM protocol (Peqlab Biotechnologie, Erlangen, Germany) and transcribed into cDNA using Fermentas Reverse Transcriptase Kit (Thermo Fisher Scientific). Quantitative Real-Time PCR was performed using the SYBR green and the MyIQ real-time PCR system (Bio-Rad, Munich, Germany). The following primers were used from Biomers (Ulm, Germany): mouse ubiquitin-40S ribosomal protein S27a (Rps27a) forward 5 ′ -GACCCTTACGGGGAAAACCAT-3′ , reverse 5′ -AGACAA AGTCCGGCCATCTTC-3′ ; mouse Ki67 forward 5′ -ACCGTG GAGTAGTTTATCTGGG-3′ , reverse 5′ -TGTTTCCAGTCC GCTTACTTCT-3′ ; mouse proliferating cell nuclear antigen (Pcna) forward 5′ -TTTGAGGCACGCCTGATCC-3′ , reverse 5 ′ -GGAGACGTGAGACGAGTCCAT-3′ ; mouse collagen type 1 alpha 1 chain (Col1a1) forward 5′ -GCTCCTCTTAGGGGC CACT-3′ , reverse 5′ -CCACGTCTCACCATTGGGG-3′ ; mouse collagen type 3 alpha 1 chain (Col3a1) forward 5′ -AAGGCT GCAAGATGGATGCT-3′ , reverse 5′ -GTGCTTACGTGGGAC AGTCA-3′ ; mouse alpha smooth muscle actin (Acta2) forward 5 ′ -CCCAGACATCAGGGAGTAATGG-3′ , reverse 5′ -TCTATC GGATACTTCAGCGTCA-3′ ; mouse fibronectin 1 (Fn1) forward 5 ′ -TCAGAAGAGTGAGCCCCTGA-3′ , reverse 5′ -AAGATT GGGGTGTGGAAGGG-3′ ; mouse mannose receptor C-type 1 (Mrc1) forward 5′ -GGAGTGATGGAACCCCAGTG-3′ , reverse 5 ′ -CTGTCCGCCCAGTATCCATC-3′ ; mouse arginase 1 (Arg1) forward 5′ -GTGAAGAACCCACGGTCTGT-3′ , reverse 5′ -CTG GTTGTCAGGGGAGTGTT-3′ ; mouse transglutaminase 2 (Tgm2) forward 5′ -AGAGTGTCGTCTCCTGCTCT-3′ , reverse 5 ′ -GTAGGGATCCAGGGTCAGGT-3′ ; mouse inducible nitric oxide synthases (Nos2) forward 5′ -ACCCTAAGAGTCACA AAATGG-3′ , reverse 5′ -TTGATCCTCACATACTGTGGA CG-3′ ; mouse IL27Rα forward 5′ -GGACCAGGAAACCAT TGGAGT-3′ , reverse 5′ -GTTGAGCTTGTCCAGGCTGTC-3′ ; mouse IL-27p28 forward 5′ -CAGGGCTATGTCCACAGCTT-3′ , reverse 5′ -CGAAGTGTGGTAGCGAGGAA-3′ .

Primers for mouse vascular endothelial growth factor A (Vegf), Il10, and tumor necrosis factor α (Tnf-α) were from QuantiTect (Qiagen, Hilden, Germany).

#### siRNA Transfection

To analyze the impact of the IL27Rα chain on endothelial cells, bEnd5 cells were transfected either with IL27Rα siRNA or control siRNA (Dharmacon, Lafayette, USA) using HiPerfect (Qiagen) according to the manufacturer's instructions.

### Generation of Tumor Supernatants

Tumors were crushed with mortar and pestle in liquid nitrogen. Two times 2 × PBS of the tumor weight was added to the crushed tumors and the suspension was incubated for 3 h at 4◦C under rotation. After centrifugation, the supernatant and the cell pellet were used for further experiments.

### Cytokine Quantification

To analyze cytokines in bEnd5 cell culture supernatants and tumor extracellular fluid (19), the LEGENDplex Mouse cytokine panel 2 was used (Biolegend) according to the manufacturer's instructions. Samples were acquired by flow cytometry and analyzed using FlowJo VX.

### Immunoblotting

Tumor cell pellets were sonified in HIF-lysis buffer (6.65 M Urea, 10% glycerol, 1% SDS, 10 mM Tris; pH 7.4), 100 ng protein per sample was loaded on SDS polyacrylamid gels together with SDS loading buffer (0.5 M Tris, pH 6.8; 2% SDS, 20% glycerol, 0.002% bromphenol blue, 5 mM DTT). Proteins were blotted on a nitrocellulose membrane, incubated with β-actin (Sigma-Aldrich), phospho-STAT (pSTAT1) (Cell Signaling), total STAT1 (tSTAT1) (Cell Signaling), pSTAT3 (Cell Signaling), tSTAT3 (Cell Signaling), and visualized by IRDye 680- and IRDye 800-coupled secondary Abs using the Li-Cor Odyssey imaging system (LICOR Biosciences, Bad Homburg, Germany).

### Enzyme-Linked Immunosorbent Assay

An ELISA for VEGF (R&D systems) was used to quantify VEGF in tumor supernatants. Tumor supernatants were generated as descripted above and diluted 1:50. ELISA was performed according to the manufacturer's instructions.

#### Aortic Ring Assay

The aortic ring sprouting assay was performed as previously (20). Briefly, aortas were harvested from 8 to 10 weeks old mice and washed with DMEM/F14 medium (Gibco, Carlsbad, USA) supplemented with 100 U/mL penicillin, and 100µg/mL streptomycin. The dissected aortas were subsequently cleaned, sectioned in 12–16 rings of 1 mm length, and embedded in collagen type 1 (Corning, New York, USA). After polymerization of the collagen gel, microvascular endothelial cell growth medium (PeloBiotech, Planegg, Germany) supplemented with 100 U/mL penicillin, 100µg/mL streptomycin, and 2% murine serum (BD) was added into the well. Tube-like structures were allowed to develop over 7 days. Thereafter, samples were fixed in 4% PFA and endothelial cells were visualized using antibodies against CD31 (Dianova, Hamburg, Germany) and VE-Cadherin (R&D), while NG2 (Merck, Darmstadt, USA) staining was employed to detect pericytes. The total volume of vascular and perivascular sprouting in each explant was calculated trough the IMARIS-BITPLANE 9.3 software. Additionally, total sprout length was measured with ImageJ.

### Wound Healing Assay

To study endothelial cell migration, a wound healing assay using bEnd5 cells was performed. Cells were grown until they reached confluence. Afterwards the wound was created with a 10 µl pipette tip. To analyze the impact of IL-27 signaling, an IL-27 neutralizing antibody (1 ng/ml), an IgG2a istotype control (1 ng/ml), Stattic (50 ng/ml), Epigallocatechin gallat (EGCG) (10 ng/ml), siControl, or IL27Rα siRNA were used. Images were taken 16 and 24 h after wound generation and analyzed using the wound healing tool in ImageJ.

### Proliferation Assay

The IncuCyte <sup>R</sup> S3 live-cell analysis system (Sartorius, Göttingen, Germany) was used to study proliferation of bEnd5 endothelial cells. Images were taken every 4 h and the doubling time of cells was calculated as follows: doubling time = (duration + log2)/(log(final concentration) – log(initial concentration)).

### Statistics

Data are presented as means ± SEM. Statistical comparisons between two groups were performed using the Mann Whitney test, or paired/unpaired two-tailed Student's t-test as indicated. Data were pre-analyzed to determine normal distribution and equal variance with D'Agostino–Pearson omnibus normality test. Statistical analysis was performed with GraphPad Prism v8. Differences were considered significant at p < 0.05. No statistical test was used to predetermine sample size, and all samples were included in the analysis. Details on statistical tests used in individual experiments are found in the figure legends.

## RESULTS

### Stromal IL-27 Signaling Promotes Mammary Tumor Growth and Reduces Immune Cell Infiltrates

To analyze IL-27 signaling during breast cancer development, murine breast cancer cells derived from a polyoma middle T oncogene-driven primary tumor were transplanted into mammary glands of IL27Rα WT or KO mice (**Figure 1A**). Tumor growth was monitored up to 31 days (**Figure 1B**). Tumors transplanted into WT mice started to appear within the first week following transplantation, whereas the growth of tumors transplanted into mammary glands of IL27Rα KO mice was delayed. Moreover, tumor progression in IL27Rα KO mice was strongly reduced from day 21 onwards (**Figure 1C**). To explore mechanisms, mice were sacrificed at day 21 to analyze early stage tumors or at day 31, when first tumors in WT mice reached a pre-defined ethical end-point (tumor diameter of 1.5 cm). Analyzing earlier time points was not feasible due to low amounts of available tumor material. Correlating with reduced tumor growth, we also observed a lower number of pulmonary metastasis in IL27Rα KO mice with 31 days old tumors (**Figure 1D**).

The impact of IL-27 in tumors was so far attributed to direct suppression of tumor cells, or an altered immune cell infiltrate. Since tumor cells did not differ in their IL27Rα expression in both groups, we initially focused on immune cells. IL-27 augments the generation of cytotoxic T lymphocytes (CTL), blocks proliferation of CTL, activates natural killer (NK) cells and limits Th17 generation, but also promotes Treg expansion and/or activation (7, 8). To analyze changes in immune cell composition, tumor single cell suspensions were analyzed using multicolor FACS staining (**Supplementary Figure 2**). We observed a clear reduction of the overall immune cell infiltrate in tumors growing in IL27Rα KO compared to WT mice. This was apparent at early stage, as well as late stage tumors (**Figure 1E**) and affected all major immune cell subsets, with the notion that predominantly myeloid cells were affected at early and lymphocytic cells were affected at late stage (**Figure 1F**). Overall, a decrease in immune cell abundance in tumors was observed during tumor development, which is due to a decline in the acute response toward a transplanted tumor and an increase in tumor cells due to rapid proliferation. Importantly, we did not observe an increase of CTL or Tregs in tumors of IL27Rα KO mice. Quantitative PCR analysis revealed a minor increase in IL-17 mRNA in late stage tumors of IL27Rα KO mice, while a major increase was observed in early stage tumors of IL27Rα KO mice (**Figure 1G**), confirming an impact of IL-27 on Th17 generation. For control reasons, we evaluated the presence of IL-27 in tumors. IL-27 was expressed in tumors and there was an increase in the amount of IL-27 p28 mRNA in early stage tumors growing in IL27Rα KO mice, which was, however, not observed at protein level (**Figures 1H,I**). There was no change in the expression of IL-27 p28 mRNA in late stage tumors between WT and KO animals. The general decrease of IL-27 p28 mRNA expression in late stage tumors can be explained by a reduced number of infiltrating immune cells, which are the main producers of IL-27. In conclusion, the overall reduction in immune cell infiltrates made it unlikely that specific lymphocyte subsets account for the altered tumor growth in IL27Rα KO mice.

## IL-27 and the IL-27 Receptor Do Not Directly Affect Macrophage Polarization

Tumor-associated macrophages (TAM) were the major immune cell population in tumors (**Figure 1F**). They decreased in early stage tumors of KO mice, following the decrease in their progenitors, i.e., monocytes (21). In late stage tumors of KO mice, they remained unaltered although monocyte numbers were still lower, indicating an uncoupling from recruitment into the tumors. This may be due to local proliferation (21). TAM may either support or restrict tumor growth based on

FIGURE 1 | Tumor growth, metastasis, and immune cell composition are reduced in tumors of IL27Rα knockout KO mice. (A) PyMT breast cancer cells were transplanted into four mammary glands of IL27Rα wildtype (WT) and knockout (KO) mice, respectively. (B) Tumor onset and progression were time-dependently observed. (C) Tumor growth slope from day 21 to 31 was analyzed (WT n = 13, KO n = 17). (D) Lungs were harvested after 31 days and analyzed by immunohistochemistry (Mayer's hemalum staining) for metastasis occurrence. Nine section from independent regions of one lung lobe per animal were analyzed. Quantification shows the mean of these regions for each animal (WT n = 5, KO n = 4). (E–G) Immune cell composition of tumor single cell suspensions was analyzed using multicolor FACS analysis. (E) CD45<sup>+</sup> immune cells are shown and (F) major immune cell subsets are displayed (WT day 21 <sup>n</sup> <sup>=</sup> 4, KO day 21 <sup>n</sup> <sup>=</sup> 4, WT day 31 n = 8, KO day 31 n = 8). (G) Quantitative real time PCR from whole tumor RNA for Il-17 is given relative to Rps27a (n = 4). (H) Quantitative real time PCR from whole tumor RNA for IL-27p28 is given relative to Rps27a (WT day 21 n = 4, KO day 21 n = 4, WT day 31 n = 6, KO day 31 n = 6). (I) IL-27 cytokine production within early stage tumor was analyzed using Legendplex (WT day 21 n = 4, KO day 21 n = 4). Data are means ± SEM, p-values were calculated using one-sample t-test; \*p < 0.05, \*\*p < 0.01, \*\*\*p < 0.001; n.d., not detected.

their polarization state. Inflammatory M1-like macrophages show anti-tumor potential, whereas anti-inflammatory M2-like macrophages promote tumor development (22, 23). We explored polarization of macrophages by analyzing the expression of different macrophage markers at mRNA level. Within the bulk tumor mRNA, we detected a trend toward higher expression of M2-like macrophage markers in tumors of IL27Rα KO mice at early and late tumor stages (**Supplementary Figure 3A**). This pattern fits to a potential increase in proliferation, since macrophage proliferation was triggered by M2 stimuli (24). Expression of the classical M1 marker Nos2 was significantly decreased in tumors of IL27Rα KO mice, which may suggest a reduced anti-tumor potential of TAM in tumors of IL27Rα KO mice. To study whether this was linked to IL-27-signaling in macrophages, we generated BMDM from WT and IL27Rα KO mice and induced classical activation with LPS/IFNγ, alternative activation with IL-4, or directly co-cultured them with PyMT cells to induce tumor-like conditions. All stimuli were applied with or without the addition of IL-27. Afterwards quantitative PCR analysis for alternatively activated macrophage/M2 markers (Tgm2, Arg1, Mrc1), the classically activated macrophage/M1 marker (Nos2), as well as the cytokines Il10 and Tnf-α was performed (25, 26) (**Supplementary Figures 3B–G**). Stimulation of BMDM with LPS/IFNγ increased Nos2, Tnf-α, and Il-10 expression, which was unaltered in KO BMDM or upon IL-27 addition. Stimulation with IL-4 enhanced expression of Tgm2, Arg1, and Mrc1, which was again independent of IL-27. Coculturing BMDM with PyMT cells upregulated Arg1 and strongly suppressed Tnf-α expression, again with no impact of IL-27. Thus, IL27Rα-deficiency may restrict M1-like polarization in tumors, although tumors growing in IL27Rα KO mice were smaller. Therefore, these alterations were likely secondary and not due to IL-27 signaling directly in macrophages. In conclusion, we excluded macrophages as major players in reducing tumor growth in IL27Rα KO mice.

### Cancer-Associated Fibroblasts in Late Stage Tumors of IL27Rα KO Mice

Since immune cells did not explain reduced tumor growth in IL27Rα KO mice, we focused on other stromal cells. Tumor sections were stained for the tumor fibroblast marker αSMA. An unexpectedly large difference of αSMA-expressing cells between late stage, but not early stage tumors, was detected. In late stage tumors of KO mice significantly more αSMA positive fibroblasts were observed compared to WT mice (**Supplementary Figures 4A,B**). To analyze a potentially direct impact of IL-27 on fibroblasts, 3T3 murine fibroblasts were stimulated with transforming growth factor β (TGFβ) to induce a cancer-associated fibroblast phenotype (27), with or without the addition of IL-27, and several fibroblast activation and proliferation markers were analyzed (**Supplementary Figure 4C**). The mRNA expression of Col1a1, Acta2, as well as Fn1 was upregulated after TGFβ treatment. Stimulation with IL-27 did not alter expression of these genes, although Col3a1 was slightly decreased upon IL-27 stimulation. To analyze the effect of IL-27 toward fibroblast proliferation we analyzed the proliferation markers Ki67 and Pcna (**Supplementary Figure 4C**), but did not observe changes in proliferation when stimulating with TGFβ and/or IL-27. Apparently IL-27 signaling did not increase fibroblast numbers or activation in vitro.

### Increased Hypoxia in Tumor of IL27Rα KO Mice

To understand reduced tumor growth in IL27Rα KO mice, we next analyzed proliferation and apoptosis of tumor cells. Tumor sections of late stage tumors were stained for proliferating (Ki67) and apoptotic tumor cells (cleaved CASP3). There was a tendency toward decreased proliferation of tumors growing in IL27Rα KO mice, and a major increase in apoptotic tumor cells in IL27Rα KO compared to WT mice, both in early and late-stage tumors (**Figures 2A,B**). Cancer cells can deregulate proliferation signals and become hyper-proliferative (28), which requires constant nutrient and oxygen supply. Oxygen availability in solid tumors is often limited. As a consequence, the α-subunits of hypoxiainducible factors (HIF1 and 2) are stabilized. HIF transcription factors then induce a selected set of target genes to increase blood supply and restore oxygen levels (29). To explore this connection, late stage tumor sections were stained for HIF1α (30) (**Figure 2C**). We noticed significantly more hypoxic cells in tumors of IL27Rα KO mice, compared to WT mice (**Figure 2D**). This pattern was confirmed at the level of HIF1α target genes including Bcl-2/adenovirus E1B 19 kDa interacting protein 3 (Bnip3), which was significantly increased in late stage tumors in IL27Rα KO mice, while glucose transporter 1 (Glut1) increased in early and late stage tumors of IL27Rα KO mice (**Figure 2E**). As tumors of IL27Rα KO mice were more hypoxic, an impaired oxygen supply may account for reduced growth and increased tumor cell death.

### Altered Endothelial Cell Numbers and Vessel Structure in Tumors of IL27Rα KO Mice

To better understand increased HIF1α expression and HIFresponses in tumors of IL27Rα KO mice, we analyzed the number and morphology of tumor blood vessels using immunofluorescence and multi-spectral FACS. Staining tumor sections of early and late stage tumors for the endothelial marker CD31 revealed marked differences in vessel architecture (**Figure 3A**). WT tumors contained well-structured vessels with a lumen, whereas vessels of IL27Rα KO mice were smaller, without luminal structures (arrows, **Figure 3A**). Often, single scattered CD31 positive cells were detected in tumors of KO mice (arrows, **Figure 3A**). Quantitative analysis by FACS showed that CD31<sup>+</sup> ECs, both CD31+CD146<sup>+</sup> blood endothelial cells (BEC) and CD31+CD90+LYVE1<sup>+</sup> lymphatic endothelial cells (LEC), were increased in early stage, but not in late stage tumors of IL27Rα KO mice (**Figure 3B**). While EC infiltration was increased, immune cell infiltration was markedly decreased. Since immune cells interact with activated blood endothelial cells to infiltrate into tissues, EC activation in tumors was investigated using FACS (31). EC activation is characterized by cell-surface

molecules, e.g., CD54 or CD106 (32). No major differences in EC subsets were detected, e.g., activated CD54+CD106<sup>+</sup> cells, activated CD54<sup>+</sup> or resting double negative (DN) cells (**Figure 3C**). However, activated CD54+CD106<sup>+</sup> ECs of early IL27Rα KO tumors showed significantly more P-selectin (CD62P) and a tendency for increased E-selectin (CD62E) expression at the cell surface, which, however, was lost in late stage tumors (**Figure 3D**). Both molecules are essential for leukocyte recruitment. Their enhanced expression would be expected to increase immune cell interactions with ECs and cause immune cell recruitment, which did not correlate with our tumor phenotype. CD54 single positive cells and double negative resting cells showed no significant changes in P-selectin or Eselectin expression (**Figures 3E,F**). To understand the increase of EC in early stage IL27Rα KO tumors, we analyzed the expression of vascular endothelial growth factor-α, a HIF1α target gene and major pro-angiogenic growth factor (33). At protein level, we detected increased VEGF amounts in tumor supernatants of late stage tumors, but no changes between tumors growing in IL27Rα KO or WT mice (**Figure 3G**). These data suggested that altered VEGFA levels do not explain increased EC infiltration.

wildtype (WT) and knockout (KO) mice, respectively. Tumors were harvested after 21 or 31 days. (A) Representative immunohistochemical stainings of tumor sections for endothelial cells (CD31). (B–F) EC abundance (B) and activation status (C–F) were analyzed by flow cytometry. CD54 and CD106 expression (C) and the expression of cell surface markers CD62P, CD62E, and CD141 in CD54<sup>+</sup> CD106<sup>+</sup> (D), CD54<sup>+</sup> CD106<sup>−</sup> (E) and double negative (DN) (F) EC populations were analyzed (WT day 21 n = 4, KO day 21 n = 4, WT day 31 n = 7, KO day 31 n = 7). (G) VEGF protein amount in whole tumor protein lysates analyzed by ELISA (WT day 21 n = 4, KO day 21 n = 4, WT day 31 n = 6, KO day 31 n = 10). Data are means ± SEM; p-values were calculated using one-sample t-test; \*p < 0.05.

Since differences in vessel architecture could be observed, we next investigated vessel integrity within tumors. FITClabeled BSA was injected i.p. 90 min before sacrificing tumorbearing mice. Afterwards tumor sections were stained for CD31 by immunohistochemistry, combined with analysis of FITC fluorescence resulting from BSA leakage through the blood vessels into the tumor. In WT mice, BSA-FITC was mainly observed within tumor vessels, whereas in tumors of IL27Rα KO mice, FITC-BSA leaked into the tumor area surrounding vessels (arrows, **Figure 4A**). A significant increase in both, count and average size of BSA-FITC positive areas was apparent in tumors of IL27Rα KO compared to WT mice (**Figure 4B**). Vessel integrity is, among others, determined by coverage of the extraluminal side of EC with pericytes (34). Given the leakiness of vessels in tumors of IL27Rα KO mice, we analyzed expression of the pericyte marker neural/glial antigen 2 (NG2) in tumor sections. As NG2 is also expressed by tumor cells, CD31 and NG2 were co-stained to determine double-positive cells. Pericytes were

then discriminated from NG2+ tumor cells by CD31 expression and morphology using the phenotyping tool of the Inform software (arrows, **Figure 4C**). Double positive cells were reduced in tumors of IL27Rα KO mice as was the ratio of CD31+NG2<sup>+</sup> pericytes to CD31<sup>+</sup> ECs (**Figure 4D**). These findings indicate reduced vessel maturation in tumors of IL27Rα KO compared to WT mice.

## Loss of IL-27 Signaling Enhances EC Sprouting, Proliferation, and Migration

To analyze if a direct impact of IL-27 on vessels may explain the phenotype in IL27Rα KO mice, we first analyzed EC sprouting using aortic rings from IL27Rα WT and KO mice ex vivo. Sprouted aortic rings were stained for CD31, VE-cadherin and NG2. To analyze microvascular sprouting, Z-stacks were merged and the total volume of sprouted CD31 and VE-Cadherin expressing ECs and NG2 positive pericytes was determined (**Figures 5A,B**). Sprouting of IL27Rα KO ECs was significantly enhanced compared to WT ECs, whereas no differences in pericyte outgrowth occurred. Besides microvascular sprouting, the endothelial sprout length was significantly enhanced in aortic rings lacking IL27Rα (**Figures 5C,D**).

To explain alterations in EC sprouting, we next analyzed EC proliferation and migration. For this, the endothelial cell line bEnd5 was used. These cells constitutively produce IL-27 and are therefore suitable for IL-27 neutralization approaches (**Figure 5E**). Cells remained either untreated, were transfected with IL27Rα-specific siRNA compared to a nontargeting control (**Supplementary Figure 5**), received an IL-27 neutralizing antibody compared to an isotype control antibody,

ImageJ is displayed (n = 5). (E) Cytokine production by bEnd5 endothelial cells was analyzed using Legendplex (n = 4). (F–I) bEnd5 endothelial cells were controls, treated with IL-27 neutralizing Ab, IgG2a, or IL-27 (F,G), or transfected with siIL27Rα, or siControl (siCont) (H,I). Proliferation was monitored for up to 72h. Time kinetics (F,H) and doubling time (G,I) are displayed (siCont n = 6, siIL27Rα n = 6, unst n = 3, IgG2a n = 4, IL-27 neutralizing Ab n = 4, IL-27 n = 4). Data are means ± SEM; p-values were calculated using one-sample t-test; \*p < 0.05, \*\*p < 0.01.

or were supplemented with IL-27 (**Figures 5F–I**). Proliferation was followed over a time course of up to 72 h and differences were analyzed at the endpoint. Untreated cells showed a proliferation slope of ∼14.5 h and doubling time of 18.8 h. IL-27 treated cells showed a proliferation slope of 13.6 and a doubling time of 19.3 h (**Figures 5F,G**). This suggested mildly impaired proliferation upon IL-27 treatment. Interfering with IL-27 signaling by adding the IL-27 neutralizing antibody

FIGURE 6 | IL-27 signaling restricts EC migration. (A–E) bEnd5 endothelial cells were controls, transfected with siIL27Rα, siControl (siCont), or treated with a IL-27 neutralizing Ab (neutr), IgG2a, VEGF, or IL-27 and subjected to a wound healing assay. Quantification of wound closure after 16 and 24 h (A–D) and representative images (E) are shown (n = 4). (F,G) bEnd5 endothelial cells were controls, treated with DMSO, the STAT1 inhibitor Epigallocatechin gallat (EGCG), or the STAT3 inhibitor Stattic. Representative images (F) and quantification of wound closure after 16 and 24 h (G) are shown (n = 4). (H) bEnd5 endothelial cells were controls, transfected with siIL27Rα, siControl (siCont), or treated with a IL-27 neutralizing Ab, or IgG2a. Protein expression of phospho-STAT1 (pSTAT1) vs. total STAT1 (tSTAT1), and phospho-STAT3 (pSTAT3) vs. total STAT3 (tSTAT3) were determined after 24 h of the scratch assay using Western analysis (cropped blots, siCont n = 8, siIL27Rα n = 8, IgG2a n = 3, IL-27 neutralizing Ab n = 3; DMSO n = 4; Stattic n = 4; unst, n = 4; EGCG n = 4). Data are means ± SEM; p-values were calculated using one-sample t-test; \*p < 0.05.

promoted proliferation compared to the IgG2a isotype control (**Figures 5F,G**). Differences in proliferation when IL-27 signaling was absent were stronger when siIL27Rα treated cells were used (**Figures 5H,I**). This may be due to the fact that the neutralizing antibody interferes with IL-27 p28, which by itself (then designated IL-30) can signal through the IL-6 receptor (35). Thus, the neutralizing antibody is less specific compared to IL27Rα-specific siRNA.

To analyze migration, we used a wound assay and monitored wound closure over time. Neutralizing IL-27 signaling significantly promoted wound closure compared to the IgG2a isotype control after 16 h, while the presence of IL-27 slowed wound closure. Wound areas treated with IL-27 neutralizing Ab were closed to roughly ∼90% after 24 h, whereas wound closure in IgG2a isotype control samples reached only ∼75% (**Figures 6A,E**). The difference in migration between IL-27 neutralizing Ab and IgG2a-treated samples was again stronger when an siRNA approach was used. A knockdown of IL27Rα significantly enhanced migration at 16 and 24 h compared to cells treated with a non-targeting control (**Figures 6B,E**). Additional stimulation with VEGF showed no further effect on wound closure (**Figures 6C,D**). IL-27 signals mainly via STAT1 and STAT3. To question whether these signaling pathways enhanced migration, the wound assay was performed in bEnd5 cells with STAT1 and STAT3 inhibitors. Stattic was used as a STAT3 inhibitor, while EGCG inhibits STAT1 (36). In the presence of EGCG wound closure was similarly enhanced compared to the situation seen in siIL27Rα treated cells (**Figures 6F,G**), while Stattic reduced EC migration compared to the control (**Figures 6F,G**). Importantly, STAT1 phosphorylation was reduced in cells treated with siIL27Rα, IL-27 neutralizing Ab and EGCG (**Figure 6H**). Our findings suggest STAT1 as a likely signaling pathway that attenuates EC migration downstream of the IL-27 receptor in ECs, and furthermore indicate that IL-27 may restrict functional angiogenesis by limiting EC migration, proliferation and sprouting.

### DISCUSSION

In order to grow and survive, tumor cells show a high demand for nutrients and oxygen, and, thus, need to provoke angiogenesis. Tumor angiogenesis is considered to be fundamentally different from physiological angiogenesis. In tumors, excessive sprouting and vessel branching generates convolute and leaky vessels (28, 37). Anti-angiogenic therapy using vascular endothelial growth factor (VEGF)-targeting agents alone, or in combination with chemotherapy, normalized a disordered tumor vasculature rather than disrupting it (38). Rather, we suggest a third scenario in tumors, where compromised tumor vessels can be rendered even more dysfunctional, to again restrict tumor growth. In support of this hypothesis two recent studies showed that a loss of deltalike 4 (Dll4) increased in non-functional and convolute vessels, thereby reducing tumor growth (39, 40). We observed a similar phenomenon in our study when depleting IL27Rα in stromal cells. It would have been of interest to reduce dysfunctional angiogenesis to a certain degree to prove causality in our system, i.e., by neutralizing VEGF to normalize vessels. However, our data did not establish a functional interplay between IL-27 and VEGF signaling. Therefore, we refrained from testing this hypothesis. Importantly, our study in accordance with the studies of Noguera-Troise et al. and Ridgway et al. suggests that tipping the balance of angiogenesis in tumors toward both directions, vessel maturation or a loss of function might be suitable during tumor therapy.

A role of IL-27 in EC function and angiogenesis is currently underappreciated. Only one study demonstrated that IL-27 reduced tumor angiogenesis in a melanoma model and an in vivo angiogenesis assay. In this study, IL-27 elicited the production of CXCL9 and CXCL10 in human ECs (16). Although CXCL9 and CXCL10 are described as anti-angiogenic chemokines, their impact on the tumor vasculature under conditions of IL-27 treatment was not tested (16). We did not observe altered expression of CXCL9 or CXCL10 in endothelial cells isolated from WT or IL27Rα KO tumors (data not shown). We show that rather depletion of IL-27 signaling disturbs tumor angiogenesis. IL27Rα deficiency increased EC proliferation, migration and sprouting, which corroborates that IL-27 limits angiogenesis. It remains to be determined whether acute inhibition of IL-27 signaling would phenocopy the effects on the vasculature seen in IL27Rα KO mice. Functional vessels are needed for immune cells to infiltrate tumors (41, 42). If blood vessel are disturbed, immune cells, such as T cells, B cells, or NK cells are unable to enter the tumor. Dysfunctional vessels in IL27Rα KO mice may explain reduced immune cell numbers, particularly lymphocyte numbers in late stage tumors of KO mice, rather than reduced proliferation of these cells in tumors.

Blocking IL-27 signaling was without consequences regarding the number of activated or quiescence/resting ECs. However, activation of CD54+CD106<sup>+</sup> and CD54<sup>+</sup> ECs was increased in tumors of early stage KO mice. EC activation is a defined twostage process. Type I EC activation occurs immediately after stimulation, when endothelial adhesion molecules, such as Pselectin emerge at the cell surface. Type II EC activation is a delayed process, whereupon E-selectin is induced at the cell surface and chemokines are released (43). If one of these two steps is uncontrolled, ECs can undergo morphological changes or become dysfunctional. In early stage tumors of KO mice, we detected increased markers for both activation states in CD54+CD106<sup>+</sup> and CD54<sup>+</sup> ECs. Significantly more P-selectin was expressed in CD54+CD106<sup>+</sup> ECs in early stage tumors of KO mice. Also, E-selectin was over-abundant. Overexpression of both EC activation markers suggests unregulated EC activation, which might fit to enhanced migration and proliferation (43). Taken together vessels in tumors of KO mice are poorly perfused, malformed, and leaky, as observed upon FITC-BSA injection. Angiogenesis starts with detachment of pericytes from the vessels and terminates with pericyte recruitment for vessel stabilization and maturation (44, 45). Anti-pericyte treatments in tumor therapy causes vascular regression and inhibits tumor growth (46, 47). We detected less pericytes and an attenuated pericyte to ECs ratio in tumors of KO mice compared to WT mice. This points to vessel leakiness under these conditions.

EC proliferation and migration are important for angiogenesis (48). Within the first phase of angiogenic sprouting a few endothelial cells are selected, which lead the growing sprout (49). These ECs adapt a more invasive and migratory phenotype to migrate toward, e.g., VEGF gradients generated from tumor cells. Leading ECs are followed by a second subset of EC, which proliferate, elongate and form the lumen of new vessels. We show that the absence of IL-27 signaling enhanced endothelial and microvascular sprouting, whereas pericyte sprouting was unaffected. This suggests an altered ratio of sprouted pericytes relative to ECs, which fits to leaky vessels. When looking at mechanisms that may explain altered sprouting, we noticed enhanced migration and proliferation of ECs treated with a IL-27 neutralizing antibody or lacking IL27Rα. IL-27 signals via STAT1/3 and STAT signaling has been connected to angiogenesis previously. While STAT3 signaling was previously shown to promote angiogenesis, STAT1 is a negative regulator of angiogenesis (50, 51), which fits to our data. Inhibition of STAT3 using Stattic significantly lowered migration. Inhibition of STAT1 using EGCG (36) significantly increased EC migration, similar to the situation when blocking IL-27 signaling. Furthermore, neutralizing IL-27 reduced STAT1 signaling in ECs. These data suggest that IL-27 signaling restricts EC migration and thus, angiogenesis via STAT1. The downstream signals certainly need further investigation. Conclusively, our study reveals a so far unappreciated direct impact of IL-27 signaling on endothelial cells to alter angiogenesis.

#### DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the manuscript/**Supplementary Files**.

#### REFERENCES


#### ETHICS STATEMENT

The animal study was reviewed and approved by Hessian animal care and use committee.

### AUTHOR CONTRIBUTIONS

AW, DS, and BB: conceptualization. AF, GC, RP, ES-F, and A-CF: methodology. AF, GC, RP, and AW: formal analysis. AF, GC, RP, ES-F, and A-CF: investigation. IF and BB: resources. AF, GC, RP, and AW: data curation. AF and AW: writing–original draft. AF, GC, RP, and AW: visualization. IF, DS, AW, and BB: supervision. IF, DS, and BB: funding acquisition. All authors: writing–review and editing.

### FUNDING

The authors are supported by the Faculty of Medicine of Goethe-University, Else Kroner-Fresenius Foundation (EKFS), Deutsche Krebshilfe (70112451), the Landesoffensive zur Entwicklung wissenschaftlich-ökonomischer Exzellenz (LOEWE), LOEWE Center for Translational Medicine and Pharmacology, and Deutsche Forschungsgemeinschaft [SFB815 (TP08, TP16), GRK2336, SFB1039, and FOR2438].

#### ACKNOWLEDGMENTS

We thank Margarethe Mijatovic and Praveen Mathoor for excellent technical assistance.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fonc. 2019.01022/full#supplementary-material

Th2-mediated allergic inflammation. J Immunol. (2007) 179:4415–23. doi: 10.4049/jimmunol.179.7.4415


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Fink, Ciliberti, Popp, Sirait-Fischer, Frank, Fleming, Sekar, Weigert and Brüne. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The Regulatory Roles of Non-coding RNAs in Angiogenesis and Neovascularization From an Epigenetic Perspective

Itzel Alejandra Hernández-Romero† , Lissania Guerra-Calderas † , Marisol Salgado-Albarrán, Tatiana Maldonado-Huerta and Ernesto Soto-Reyes\*

Natural Sciences Department, Universidad Autónoma Metropolitana-Cuajimalpa, Mexico City, Mexico

#### Edited by:

Erika Ruiz-Garcia, National Institute of Cancerology (INCan), Mexico

#### Reviewed by:

Fahd Al-Mulla, Genatak, Kuwait Shao-Chun Wang, China Medical University, Taiwan

#### \*Correspondence:

Ernesto Soto-Reyes esotoreyes@correo.cua.uam.mx

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology

> Received: 20 June 2019 Accepted: 03 October 2019 Published: 24 October 2019

#### Citation:

Hernández-Romero IA, Guerra-Calderas L, Salgado-Albarrán M, Maldonado-Huerta T and Soto-Reyes E (2019) The Regulatory Roles of Non-coding RNAs in Angiogenesis and Neovascularization From an Epigenetic Perspective. Front. Oncol. 9:1091. doi: 10.3389/fonc.2019.01091 Angiogenesis is a crucial process for organ morphogenesis and growth during development, and it is especially relevant during the repair of wounded tissue in adults. It is coordinated by an equilibrium of pro- and anti-angiogenic factors; nevertheless, when affected, it promotes several diseases. Lately, a growing body of evidence is indicating that non-coding RNAs (ncRNAs), such as miRNAs, circRNAs, and lncRNAs, play critical roles in angiogenesis. These ncRNAs can act in cis or trans and alter gene transcription by several mechanisms including epigenetic processes. In the following pages, we will discuss the functions of ncRNAs in the regulation of angiogenesis and neovascularization, both in normal and disease contexts, from an epigenetic perspective. Additionally, we will describe the contribution of Next-Generation Sequencing (NGS) techniques to the discovery and understanding of the role of ncRNAs in angiogenesis.

Keywords: angiogenesis, non-coding RNA, epigenetics, neovascularization, next generation sequencing, miRNAs, lncRNAs, circRNA

#### INTRODUCTION

In the vascular network, blood vessels act as channels for nutrients, oxygen delivery, and metabolic waste evacuation. The growth of new capillary vessels, known as angiogenesis, plays key roles in embryonic development and in tissue homeostasis and remodeling in adults, as well as in cancer initiation and progression (1, 2). The balance between pro- and anti-angiogenic factors (such as VEGF, PDGF, and TSP-1/2) coordinates angiogenesis and other neovascularization mechanisms such as intussusceptive angiogenesis, vasculogenesis, lymphangiogenesis, vessel co-option, and vasculogenic mimicry (3–5).

Over the last few decades, the study of angiogenesis has helped researchers to understand vascular physiology and its implications for several diseases. For instance, in atherosclerosis, ischemia, and retinopathy, excessive or insufficient vascular growth can affect the behavior of endothelial and smooth muscle cells (6, 7). Studies of the neovascularization processes have also provided molecular targets for the development of therapies to delay cancer progression, since it is well-known that angiogenesis is an essential process that is altered in tumors (8).

Nowadays, the study of the molecular mechanisms involved in angiogenesis is being built on different experimental approaches, such as cell migration, proliferation, and metabolic assays or histological and tri-dimensional models, that approach specific stages of angiogenesis; however, only pieces of the puzzle have been elucidated (9). With advances in high-throughput genomic technologies such as microarrays, next-generation sequencing (NGS), and bioinformatic analyses, a genome-wide perspective of the elements involved in the angiogenic process is now being taken. Some of the newest players revealed by these approaches are non-coding RNAs (ncRNAs), which have gained relevance in the field of epigenetics (10–12). Therefore, in this review, we will describe the epigenetic regulatory functions of ncRNAs in physiological angiogenesis and vascular diseases, as well as the contribution of NGS technologies to the discovery of new roles for ncRNAs that are associated with angiogenesis.

#### AN OVERVIEW OF EPIGENETICS

In 1939, the term "epigenetics" was coined by Conrad Hal Waddington (13). Today, one of the most accepted definitions of the term explains that "epigenetics is the study of the heritable changes in gene expression that cannot be explained by alterations in the DNA sequence" (14). Among the epigenetic components that coordinate nucleus organization and gene transcription are DNA methylation, histone post-translational modifications (PTMs), and histone positioning, but recently, ncRNAs have been incorporated as epigenetic modifiers, because many of these can function as scaffolding elements to transport proteins with epigenetic functions (15). Each of these processes is stimulated by the signals derived from a dynamic epigenetic code that is established on the chromatin depending on the physiological and extracellular context. The writers, readers, and erasers of this code are proteins that place, recognize, or remove chemical modifications of DNA nucleotides and within the amino-terminal regions of histones. Most chromatin "writers" are methyltransferases that catalyze the transfer of methyl groups. DNA methylation occurs predominantly in regions enriched in CpG sites. The occurrence of methylation at the promoter regions of genes is associated with gene silencing. PTMs alter the regulation of gene transcription by changing the structure of chromatin depending on the particular residue that is modified (16, 17). The "readers" are proteins that recognize and associate with the epigenetic modifications, interpret them, and, in many cases, promote the assembly of protein complexes. The erasers remove the modifications and, therefore, alter signaling components that contribute to the regulation of gene expression. Recently, it has been reported that ncRNAs can mediate the binding of epigenetic proteins to their target sequences. Though they do not function alone as "classic" epigenetic modifiers, they play a vital role in both the recruitment and transcriptional regulation of epigenetic modifiers (18). In fact, multiple chromatin-remodeling enzymes have been shown to directly contact ncRNAs, including Enhancer of Zeste Homolog 2 (EZH2) and Suppressor of Zeste 12 Protein Homolog (SUZ12) (writer and eraser within the Polycomb repressive complex 2/PRC2, respectively), and nuclear architectural proteins like Yin Yang 1 and CTCF, among others (19–22). The incorporation of ncRNAs as epigenetic elements has opened up new fields of study in which they have been shown to regulate gene expression. In the following pages, we will provide an overview of the ncRNAs involved in angiogenesis, focusing on those involved in epigenetic processes.

### MiRNAs AND THEIR EPIGENETIC TARGETS IN NEOVASCULARIZATION AND ANGIOGENIC PROCESSES

MicroRNAs (miRNAs) are short ncRNAs with a length of 19– 23 nucleotides that are conserved in animals, plants, and some viruses (23–25). MiRNAs are transcribed as long pri-microRNAs (pri-miRNA) and are subsequently processed to ∼70-nucleotide precursor hairpins (pre-miRNA) by the RNase Drosha (26). PremiRNAs are then exported to the cytoplasm and recognized by the RNase DICER, which removes the loop linking the 3′ and 5′ ends of the hairpin, producing a ∼20-nucleotide miRNA duplex (27). Later, one of these strands is fused into the RNA Induced Silencing Complex (RISC), where both the miRNA and its messenger RNA (mRNA) target interact (28).

MiRNAs have two main functions: post-transcriptional gene regulation and RNA silencing. They act by pairing bases with a complementary sequence located in the 3′UTR region of target mRNA (29, 30). Consequently, these mRNAs are regulated by one or more mechanisms that include the inhibition of mRNA translation to proteins by ribosomes and by mRNA strand cleavage into two fragments and poly(A) tail shortening that results in mRNA disruption (29, 31). In the last 10 years, the field of miRNA biology has ignited, revealing amazing functions in angiogenesis. These miRNAs have been termed angiomiRs, and they target key angiogenesis molecular drivers, such as metalloproteinases, hypoxia inducible factor 1 (HIF1), cytokines, and growth factors, such as EGFL7, FGF11, PDGFRB, and the vascular endothelial growth factor (VEGF) family (32–34).

MiRNAs are not considered epigenetic components, but some of them are modulated by epigenetic mechanisms. This mainly affects their regulatory region through the incorporation of DNA methylation, repressive histone marks, or the loss of transcriptional factors, as has been reported for miR-125b1 and miR-124 (35, 36). Others, known as Epi-miRNAs, can also regulate the gene expression of epigenetic elements, DNA methyltransferases (DNMTs) (such as miR-152, miR-30, and miR-148a/b), histone deacetylases (HDACs) (such as miR-140,

**Abbreviations:** BDNF, Brain-Derived Neurotrophic Factor; BRG1, Brahma related gene-1; CAD, Coronary Artery Disease; circRNA, circular RNAs; DEGs, Differentially Expressed Genes; DNMT, DNA methyltransferase; EIciRNAs, Exonintron circular RNAs; EPCs, Endothelial Progenitor Cells; EZH2, Enhancer of Zeste Homolog 2; GRO-Seq, Global run-on sequencing; HDAC, Histone deacetylase; HDL, High-density lipoprotein; HF, Heart Failure; HIF1, Hypoxia Inducible Factors 1; HUVEC, Human Umbilical Vein Endothelial Cells; IH, Infantile hemangioma; lincRNAs, intergenic lncRNAs; lncRNAs, long noncoding RNAs; LOXL2, Lysyl oxidase-like 2; MeCP2, Methyl-CpG-binding protein 2; miRNAs, microRNAs; MMP, Matrix metalloproteinase; mRNA, messenger RNA; ncRNAs, non-coding RNAs; NGS, Next-Generation Sequencing; PB-EPCs, Peripheral Blood EPCs; PRC, Polycomb Repressive Complex; pre-miRNA, precursor hairpin miRNA; REST, Repressor Element-1 Silencing Transcription; RNA-seq, RNA sequencing; SIRT1, NAD-dependent deacetylase sirtuin1; smRNAseq, small RNA-seq; SNPs, Single Nucleotide Polymorphisms; SUZ12, Suppressor of Zeste 12 Protein Homolog 2; TF, Transcription Factor; TGF-β, Transforming Growth Factor; TSS, Transcription Start Sites; UC-EPCs, Umbilical Cord EPCs; UHRF1, E3 ubiquitin ligase with PHD and RING finger domain 1; VASH1, Angiogenesis inhibitor vasohibin 1; VEGF, Vascular Endothelial Growth Factor.

miR-1, and miR-449a), and the Polycomb Group of genes (such as miR-101 and miR-26a) (37–44), and some of them have been considered angiomiRs (39, 40). MiRNAs and their identified epigenetic targets in angiogenesis are listed in **Table 1**.

#### MiR-30a-3p

Transforming Growth Factor (TGF-β) is a relevant cytokine that functions in the process of vascular homeostasis and is involved in the vascular development of endothelial cells. It has been reported that the administration of TGF-β to endothelial cells leads to decreased miR-30a-3p expression. The absence of this microRNA results in increased levels of methyl-CpG-binding protein 2 (MeCP2), a protein associated with silencing of SIRT1 (45). SIRT1 is necessary for the migration of endothelial cells to occur throughout sprouting angiogenesis, and the loss of this enzyme induces abnormal angiogenesis in vivo (52). Conversely, increased levels of miR-30a-3p expression lead to the activation of SIRT1 expression (**Figure 1A**). Further experiments revealed that MeCP2 enhanced the methylation status of the SIRT1 promoter, probably by DNMT1 recruitment, leading to a reduction in SIRT1 expression and endothelial angiogenic defects (53).

#### MiR-101

The microRNA miR-101 acts as a tumor suppressor, promoting apoptosis and inhibiting cell proliferation, angiogenesis, invasion, and metastasis. MiR-101 performs its regulatory functions by targeting an abundant range of epigenetic molecular effectors, such as DNMT3A, EZH2, and HDAC9 (54, 55). In endothelial cells, high levels of VEGF are associated with the downregulation of miR-101, allowing an increase in EZH2 (46). EZH2 is associated with the formation of heterochromatin and can affect multiple target genes such as Vasohibin 1 (VASH1), which functions as a negative feedback modulator of angiogenesis in vascular endothelial cells (56, 57) (**Figure 1B**). The overexpression miR-101 leads to EZH2 repression and the activation of VASH1 transcription. This evidence, taken together, suggests that miR-101 is involved in multiple processes such as cellular growth attenuation, migration, and invasion mechanisms and the ability of endothelial cells to form capillary-like structures in glioblastomas (47).

#### MiR-20a

MiR-20a belongs to the miR-17-92 cluster and has been linked to breast cancer cells with a high angiogenic profile. High levels of miR-20a are correlated with complex vascular structures and larger vessels, suggesting that miR-20a could be used as a potential new angiogenic target (58). Additionally, overexpression of miR-20a affects the mRNA stability of the lysine acetyltransferase, p300. In mouse myocardium cells, p300 is a key factor that regulates angiogenic and hypertrophic programs, influencing the expression of many related genes, such as Hif,1 Vegfc, Vegfa, Angpt1, and Egln3. Interestingly, high p300 levels induce an increase in the expression of miR-20a, providing a feedback inhibition loop for p300 that prevents its pro-angiogenic effects (48).

### MiR-137

MiR-137 has a tumor suppressor gene function that has been reported for several neoplasms (49, 59, 60). It was also reported that this miRNA can inhibit angiogenesis and cell proliferation by EZH2 downregulation in glioblastomas. Overexpression of miR-137 reduces the mRNA and protein levels of EZH2, while downregulation of miR-137 is associated with poor prognosis in affected patients (49).

#### MiR-124

The miRNA miR-124 is highly conserved, from nematodes to humans. Three human genes encoding miR-124 have previously been characterized (miR124a-1, miR-124a-2, and miR-124a-3) and the majority have been shown to be deregulated in neoplasms (61). Also, it has been shown that expression of miR-124 is elevated after treatment with certain drugs such as niclosamide. In this case, it is associated with the inhibition of vasculogenic mimicry formation, particularly by reducing levels of phosphorylated STAT3 (62).

Some reports propose that miR-124 suppresses the E3 ubiquitin ligase with PHD and RING finger domain 1 (UHRF1) expression, a factor involved in the recruitment of epigenetic components in bladder cancer tissues. Also, UHRF1 is known to enhance malignancy, inducing cellular proliferation, migration, and angiogenesis (63). MiR-124 overexpression resulted in UHRF1 suppression through the competitive binding of its 3'- UTR region. In addition, miR-124 overexpression attenuated tumor growth and cell proliferation in vivo and invasion, migration, and vasculogenic mimicry in vitro. Further, it reduced VEGF protein levels and levels of the matrix metalloproteinases MMP-2 and MMP-9. A matrigel assay in a three-dimensional culture revealed reductions in tubular channel formation when miR-124 was over-expressed in bladder cancer cell lines compared to the control group, suggesting that miR-124 indirectly regulates vasculogenic mimicry in bladder cancer (44).

### MiR-214

Originating from intron 14 of the Dynamina-3 gene (DNM3), the primary transcript of miR-214 produces four different miRNAs (miR-199-3p, miR-199-5p, miR214-3p, and miR-214-5p) (64). During the endothelial differentiation of embryonic stem cells, the Brain-Derived Neurotrophic Factor (BDNF) promotes angiogenesis, in vitro and in vivo, by increasing levels of miR-214. The miR-214 inhibits EZH2 at the post-transcriptional level, leading to reductions in EZH2 occupancy at the NOS3 promoter (50). Also, miR-214 controls the BDNF-mediated upregulation of neuropilin 1, VEGF-R, and Crk-associated substrate kinase (50, 65). Thus, miR-214 is a downstream player within the BDNF signaling pathway that regulates important angiogenic targets.

#### MiR-200b

miR-200b is part of the miR-200 family, which is organized into two main groups according to seed sequence. The miRNAs of group A are miR-141 and miR−200a, while the miRNAs in group B are miR-200b, miR−200c, and miR−429 (66). Particularly, miR-200b has been indicated to have a role in the process of angiogenesis. Studies of malignant neoplasms demonstrated that


miR-200b controls the epithelial to mesenchymal transition by downregulating p300 (67–70). In addition, p300 activates HIF1, which is a transcriptional regulator of VEGF-A, and triggers the development of abundant blood vessels (71–73). Since miR-200b negatively regulates p300, this miRNA has antiangiogenic properties (51).

In sum, these studies suggest that miRNAs have the capacity to indirectly affect epigenetic pathways in endothelial cells and influence the angiogenic response. This opens up the possibility of considering miRNAs as therapeutic targets or biomarkers, an exciting prospect since therapies for both vascular diseases and cancer are needed. In several diseases, miRNAs have proven to be excellent biomarkers as a result of their high circulating levels. Indeed, analysis of oncogenic and suppressor miRNAs that are found in primary tumors against non-neoplastic cells revealed exosome-mediated sorting mechanisms related to cancer progression (74, 75). It is unknown whether similar mechanisms could be utilized by Epi-miRNAs during the evolution of vascular diseases. Recently, the attention of the scientific community has been focused on other, widely-studied ncRNAs known as long non-coding RNAs (lncRNAs), which have master regulatory functions in angiogenesis.

#### LONG NON-CODING RNAs AS SCAFFOLDS FOR EPIGENETIC PARTNERS IN NEOVASCULARIZATION

LncRNAs are all ncRNAs larger than 200 nucleotides and are classified according to their proximity to protein-coding genes as intergenic, intronic, bidirectional, sense, and antisense lncRNAs. Massive analyses have revealed that lncRNAs are originated using the same mechanisms as protein-coding genes; however, contrary to protein-coding genes, lncRNAs show a preference for having two-exon transcripts, and most of them lack any protein codingpotential. Also, lncRNAs show tissue-specific expression patterns and are predominantly located in the nucleus rather than the cytoplasm. In fact, there are several lines of evidence that suggest that lncRNAs are significantly more enriched in chromatin than miRNAs (76).

LncRNAs can indirectly modulate DNA methylation at CpG sites, which in turn regulates gene transcription. For example, Tsix recruits DNMT3a to methylate and silence the XIST promoter. XIST is an important effector involved in the inactivation of the X chromosome (77). Likewise, the lncRNA Kcnq1ot1 recruits the de novo DNA demethylase DNMT1 to control the methylation status of ubiquitously imprinted genes during mouse development (78). LncRNAs can act as guides or scaffolds, facilitating interaction between several proteins, such as those that are part of chromatin-modifying complexes, causing gene activation or repression, depending on the interaction partners involved (79, 80). The polycomb repressive complexes PRC1 and PRC2, the transcriptional repressor element-1 silencing transcription factor REST, its cofactor (REST/CoREST), other epigenetic components like the mixed lineage leukemia protein and the H3K9 methyltransferase G9a, physically interact with lncRNAs (78, 80, 81). In addition, many lncRNAs such as HOTAIR, Xist, Kcnq1ot1, and Breaveheart interact with PRC2, implying that these ncRNAs play a role in recruiting this complex through its subunits (EZH2, SUZ12, EED, RBBP4, and AEBP2) or through a bridging protein (such as JARID2) to their target genes (82, 83). Likewise, the expression of many angiogenesis-related genes involved in the VEGF signaling pathway is regulated through lncRNAs (such as H19, MEG3, and HOTAIR), and recently, researchers discovered that some of them perform their regulatory function by influencing the expression and activity of several epigenetic modulators (20, 22). LncRNAs and their identified epigenetic targets in angiogenesis are listed in **Table 2**.

in blue panels, and lncRNAs are represented in green.

### MANTIS

MANTIS is a recently discovered lncRNA required for endothelial cell function and proper angiogenesis. MANTIS is induced in the endothelium of glioblastoma tumors and is overexpressed during vascular regeneration in atherosclerosis regression. It alters angiogenic sprouting, tube formation, and epithelial cell migration. Loss of MANTIS expression is reported during pulmonary arterial hypertension, and its downregulation also led to the reduced expression of many angiogenesis-related mRNAs (80).

In endothelial cells, MANTIS is upregulated following the knockdown of the histone demethylase JARID1B. JARID1B loss triggers increased H3K4me3 levels at transcription start sites (TSS) of the MANTIS gene, facilitating gene expression. Interestingly, in patients with idiopathic pulmonary arterial hypertension, a disease characterized by endothelial dysfunction, MANTIS expression is downregulated, while JARID1B is upregulated (80).

Novel studies have revealed that MANTIS functions as a scaffold and regulates the activity of Brahma related gene-1 (BRG1), the catalytic subunit of the SWI/SNF chromatin remodeling complex. The MANTIS-BRG1 interaction allows for increased binding of BAF155, which is a core component of the SWI/SNF complex, enhancing BRG1 ATPase activity and chromatin relaxation at the TSS of the transcription factor COUP-TFII, which, in turn, recruits RNA Pol II binding and transcription of the pro-angiogenic genes SOX18 and SMAD6. The knockdown of MANTIS reduces BRG1 ATPase activity (80) (**Figure 1C**).

#### ANRIL

ANRIL is an antisense lncRNA from the INK4 locus. It encodes two cyclin-dependent kinase inhibitors, p15 (INK4b) and p16 (INK4a), and a protein known as ARF. All of the genes cooperate in tumor suppressor networks. When these genes are silenced, proatherosclerotic cellular mechanisms are enhanced,


such as increased adhesion and diminished apoptosis (86). In fact, ANRIL expression is correlated with the risk of some vascular diseases such as coronary atherosclerosis and carotid arteriosclerosis (87).

It has been shown that ANRIL recruits PRC2 or PRC1 to different target genes by directly interacting with their subunits EZH2, SUZ12, and CBX7 (86, 88, 89). In a diabetic retinopathy cellular model, high glucose levels upregulated ANRIL and VEGF expression. In turn, ANRIL positively regulated EZH2, EED, and p300 levels. Furthermore, ANRIL recruits EZH2 and histone acetyl-transferase p300 to the VEGF promoter, enhancing its expression and angiogenic effects. It was shown that ANRIL silencing prevented the formation of capillarylike structures in spite of the angiogenic influence of high glucose levels (84) (**Figure 1D**). Moreover, ANRIL silencing also promoted miR-200b expression, a previously described miRNA that has been shown to be involved in regulating VEGF (90).

#### GATA6-AS

GATA6-AS is the hypoxia-regulated long non-coding antisense transcript of GATA6 and promotes angiogenesis by negatively regulating lysyl oxidase-like 2 (LOXL2). LOXL2 catalyzes the oxidative deamination of lysines and hydroxylysines, which results in the generation of non-methylated H3K4 and gene silencing. Thus, GATA6-AS silencing leads to increased LOXL2 activity and transcriptional repression. In the nucleus, the physical interaction between GATA6-AS and LOXL2 positively regulates the expression of several angiogenesis- and hypoxia-related genes, such as periostin and cyclooxygenase-2. It has been shown that GATA6-AS silencing in epithelial cells significantly prevented TGF-β2-induced endothelial to mesenchymal transition and augmented angiogenic sprouting in xenograft models in vivo (85).

Like epi-miRNAs, the epi-lncRNAs are excellent candidates biomarkers due to their easy collection and tissue specificity. Although there are few examples of epi-lncRNAs in angiogenesis, the implications behind these interactions provide an interesting view of the mechanisms in which lncRNAs regulate not only the recruitment but also the activity of chromatin modifiers. Another layer of complexity is added if we consider that lncRNAs have many alternative splice forms, including the non-linear, circular RNAs (circRNAs).

### CIRCULAR RNAs IN NEOVASCULARIZATION

Circular RNAs (circRNA) are single-stranded RNAs that are widely conserved in all life domains and form a covalent closed loop (91). The discovery of this type of RNA has occurred fairly recently, and before their discovery, the RNAs were considered the result of errors within the process of gene transcription. These circRNAs are produced by a back-splicing process of pre-mRNA, in which a downstream splice donor is linked to an upstream acceptor (92, 93). The splice forms can circularize from exonic, intronic, or a combination of both regions (EIciRNAs) (94).

In cancer-derived cell lines, it has been reported that changes in DNMTs and the hypermethylation of the CpG islands of some genes that host circRNA can induce gene silencing of both linear RNA and circRNA, suggesting an epigenetic mechanism that produces two molecular "hits" (95). Because circRNA lack 5 ′ and 3′ ends, these cannot be degraded by exoribonucleases. Instead, circRNA levels may be regulated by endonucleases and exosomal deportation (96). These molecules are stable, abundant and specific to certain cell types, having distinct transcriptional patterns for specific tissues and multiple isoforms in eukaryotic cells (97). CircRNAs have been linked to different biological processes, including cell proliferation, senescence, and apoptosis, among others. The study of circRNA has increased in recent years, since they have been shown to be related to both physiological and pathological processes (98). In fact, circRNAs have been proposed as potential biomarkers for neurological disorders, infectious diseases, cancer, and preeclampsia as a result of their availability in circulating body fluids (99–102).

The circRNAs have transcriptional and post-transcriptional regulatory functions. EIciRNAs such as EIF3J associate with ribonucleoproteins like U1 and the Pol II at the promoters of their parental genes to enhance their own expression (94). Similar to EIciRNAs, some circRNAs (such as ciANKRD52) can positively regulate their own expression through interaction with the Pol II complex (103). Other circRNAs regulate alternative splicing or serve as sponges to bind, store, or sequester miRNAs and other protein complexes containing transcription factors and RNA binding proteins (94, 104, 105). Due to the ability of cirRNA to bind to miRNAs, they have been referred to as miRNA sponges (106). Despite their recent discovery, some evidence suggests that circRNAs are implicated in angiogenesis (e.g., circRNA-MYLK) and many cardiovascular diseases, such as atherosclerosis (e.g., circR-284), myocardial infarction (e.g., ciRS-7), and coronary artery disease (CAD) (e.g., circ\_0124644), among others (107, 108). However, to our knowledge, no study has shown that circRNAs have an epigenetic regulatory role in angiogenesis. Similar to the lncRNA ANRIL, a circularized and anti-sense splice variant of the INK4/ARF locus (cANRIL) has been associated with atherosclerotic vascular disease (109). Moreover, in the cytoplasm, the binding of circANRIL to the rRNAprocessing machinery impairs its function and causes nucleolar fragmentation and stress signaling (110). These findings suggest that, just like their longer-sized isoform, the variant cANRIL may have a role in the epigenetic regulation of vascular disease.

The study of ncRNA has opened up a new research field, and this has been extended to the genome scale. This type of experimental approach has become common practice in both the research laboratory and at the clinical level. Therefore, along with a growing array of genomic analysis machinery, bioinformatics platforms have also been developed, thus generating a new set of tools for the study and analysis of ncRNA.

### CONTRIBUTION OF NGS TECHNOLOGIES TO THE DISCOVERY OF NEW ncRNAs

In recent years, increasing quantities of data have been obtained from NGS technologies such as mass RNA sequencing (RNAseq), small RNA-seq (smRNA-seq), and single-cell RNA-seq, among others. These technologies have revealed that the human genome encodes for more than 90,000 non-coding RNAs and that these play an important role in several diseases (111). Using publicly available genomic information, it is now possible to discover and characterize novel disease-associated ncRNAs. In the next section, we will describe some of the key discoveries that have been made thanks to NGS data, in which ncRNAs are shown to have roles in angiogenesis and neovascularization processes.

The study of the ncRNAs involved in molecular processes associated with neovascularization and angiogenesis in several diseases can be carried out by using RNA-seq approaches, especially where angiogenesis or neovascularization is one of the causes, risk factors, or consequences of the disorders. Some of the diseases studied in this manner have been ischemia stroke, CAD, hemangioma, and heart failure (HF). Furthermore, angiogenesis and neovascularization are strongly related to endothelial functioning and the transcriptional programming of endothelial progenitor cells (EPCs). Thus, the study of the molecular mechanisms involved in the regulation of EPCs is of great interest. Nevertheless, only a few studies have been conducted on human umbilical vein endothelial cells (HUVEC) or other endothelial models to understand the role of ncRNAs using NGS technologies. In this section, we will provide a compilation of some studies aiming to identify or characterize ncRNAs involved in vascular processes.

First, in 2012 Cheng et al. performed smRNA-seq on umbilical cord blood EPCs (UC-EPCs), which was known for its enrichment in EPCs, and compared the expression profiles against EPCs derived from peripheral blood in adults (PB-EPCs) to understand the underlying mechanisms involved the functional differences between these two models. They identified specific patterns of miRNAs (miRNome) in UC-EPCs and PB-EPCs in which 54 miRNAs were overexpressed in UC-EPC and 50 miRNAs were overexpressed in PB-EPCs. For instance, UC-EPCs expressed miRNAs involved in angiogenesis such as miR-31 and mir-18a, while PB-EPCs are enriched in tumor-suppressive miRNA expression such as that of miR-10a and mir-26a (112).

A study performed by Wang and colleagues in 2014 revealed that there was cooperation between VEGF and miRNAs in CAD progression. They performed smRNA-seq and identified EPCspecific miRNome that was related to angiogenic processes, which suggests that miRNAs in EPCs with a poor capacity to enhance angiogenesis might have higher levels of miRNAs targeting VEGF. Indeed, they identified anti-VEGF miRNAs such as miR-361-5p that were enriched in EPCs and in the plasma of patients with CAD (113).

Also, atherosclerosis appears to be one of the factors leading to CAD. In 2018, Mao and colleagues conducted a study to identify miRNAs linked with carotid atherosclerosis. They performed a differential expression analysis to identify genes that were specifically associated with either primary or advanced atherosclerotic plaque tissues. Using public databases, they predicted 23 miRNAs that targeted the differentially expressed genes, such as miR-126, miR-155, miR-19A, and miR-19B, which can play a regulating role in neovascularization and angiogenesis (114).

Furthermore, a study from Liu et al. (115) identified differentially expressed ncRNAs that were predicted to be involved in the regulation of high-density lipoprotein (HDL) metabolism, the deregulation of which is believed to be one of the main causes of CAD. To this end, they treated HUVEC cells with HDL from healthy subjects and patients with CAD and hypercholesterolemia. After RNA-seq analysis, 41 ncRNAs were identified, and researchers were able to show that the ncRNAs, along with protein-coding genes such as DGKA and UBE2V1, have critical functions in vascular cells (115).

Additionally, it is well-known that endothelial cell metabolism is sensitive to hypoxia, which is an adverse effect of atherosclerotic lesions in humans. In 2018, Moreau et al. investigated the lncRNA profiles of HUVEC cells using global run-on sequencing (GRO-Seq). GRO-seq is a sequencing method that measures active transcription, identifying newly synthetized RNA, and providing sufficient resolution to map the position and orientation of transcripts detected. This group aimed to discover changes in the expression patterns of lncRNAs in HUVEC cells exposed to hypoxia and demonstrated that hypoxia affects the transcription of ∼1,800 lncRNAs. Among the most relevant lncRNAs identified were MALAT1, HYMAI, LOC730101, KIAA1656, and LOC339803, which were differentially expressed in human atherosclerotic lesions compared to normal vascular tissue (116).

In contrast, heart and circulatory system diseases often involve changes in vascular smooth muscle or cardiac cells. In 2018, Cheng et al. used RNA-seq to identify circRNAs in human aortic valves. They recognized 1,412 specific circRNAs, most of which originated from exons of their host genes. Furthermore, after performing a gene ontology enrichment analysis, they found that the host genes were associated with pathways regulating aortic valve function (ECM-receptor interaction pathway, ErbB signaling pathway, and vascular smooth muscle contraction pathway) (117). In addition, Bell et al. identified novel lncRNAs in human vascular smooth muscle cells in 2014. This work expanded our knowledge of the relevance of lncRNAs in the control of smooth muscle cells. The researchers performed an RNA-seq experiment examining expression patterns in human coronary artery smooth muscle cells. Their analysis revealed 31 novel lncRNAs. They discovered and characterized a novel vascular cell-enriched lncRNA that they named SENCR. They performed RNA-seq after knockdown of SENCR and observed that expression of Myocardin and genes involved in the contraction of smooth muscle were reduced, while expression of other promigratory genes was enhanced (118). These results have enhanced our understanding of vascular cells and should be further studied in order to discern lncRNAs in vascular diseases. Finally, in 2015, Di Salvo et al. analyzed the expression profiles of cells derived from 22 human hearts from patients with Heart Failure (HF) vs. non-HF donor hearts. Initially, they discovered 2,085 lncRNAs, and subsequent analyses revealed 48 differentially expressed lncRNAs in HF patients. Among these, AP000783.2, RP11-403B2.6, and RP11-60A24.3 were identified (119).

Angiogenesis and neovascularization processes affect the prognosis of patients who have suffered from brain stroke ischemia. Thus, the identification of ncRNAs involved in these processes might be useful for their further use as drug targets or biomarkers for the disease. Therefore, Zhang et al. (120) aimed to uncover which ncRNAs have altered expression profiles after cerebrovascular dysfunction in ischemic stroke. Using bulk RNA-seq, they profiled lncRNA signatures in primary brain microvascular endothelial cells after oxygen-glucose deficiency. This approach allowed for the identification of 362 differentially expressed lncRNAs. The top three lncRNAs that were upregulated were Snhg12, Malat1, and lnc-OGD 1006, while the top three downregulated lncRNAs were 281008D09Rik, Peg13, and lnc-OGD 3916 (120).

Another disease model that has been studied in order to identify ncRNAs involved in angiogenesis and neovascularization is infantile hemangioma (IH), which is a type of vascular tumor in infants. Li et al. investigated whether ncRNAs have a role in IH pathogenesis in 2018. The researchers used a bulk RNA-seq approach to examine global ncRNAs expression profiles in IH patients compared to their matched, normal-skin controls. In this study, researchers identified 256 lncRNAs and 142 miRNAs that were differentially expressed. They also found more than a thousand sponge modulators involved in miRNA-, lncRNA- , and mRNA-mediated interactions. These findings suggest the presence of an endogenous ncRNA regulatory network associated with the development of IH and other vascular diseases (121).

Overall, the studies described above have shown that NGS technologies can be very effective in identifying and characterizing ncRNAs. This type of technology has helped researchers to understand the regulatory role of ncRNAs in angiogenic and neovascularization processes. However, studies in this field are just emerging, and additional research will be required to expand our knowledge and translated into clinical use.

### CURRENT APPROACHES USED TO DISCOVER NEW ncRNAs

After the development of NGS technologies, ncRNAs have been discovered, and multiple efforts have been made to organize, collect, provide, and unify all available information regarding ncRNAs so that it can be accessed by the research community. Furthermore, new methods have developed to predict and identify novel ncRNAs. Here we present some of the cuttingedge bioinformatics approaches currently being used to study ncRNAs and give some examples of how they are used in the study of neovascularization processes (**Figure 2**). For a detailed explanation, see the following reference (122).

Transcriptome-wide association studies can be performed to identify expression-trait associations where ncRNAs might be involved. This method can identify single-nucleotide polymorphisms (SNPs) located in transcribed regions of ncRNA genes that can be related to a specific phenotype. A second bioinformatic approach is the use of tools for the prediction of primary, secondary, and tertiary ncRNA structures to obtain information about their potential function. This method has been used for circRNAs, smRNAs, and lncRNAs. The third approach to studying ncRNAs is the use of biological networks. These types of analyses enhance our understanding of the function of ncRNAs by integrating expression, regulatory, and protein–protein interaction networks. NcRNAs are highly connected in these networks and can influence more than one target gene in order to produce a specific phenotype. These approaches can identify disease-specific regulatory modules where ncRNAs play an important role (122).

Though the effective methods described above can be used to discover and understand the biological functions of ncRNAs, they have not been adequately exploited to reveal the roles of ncRNAs in angiogenesis or neovascularization. So far, only a few studies have used advanced bioinformatics tools for this purpose. For example, in 2018, Li et al. detected novel circRNAs related with IH using RNA-seq data. The best experimental approach for the detection of circRNAs is the use of deep sequencing of RNA treated with RNase R (which leaves a circRNA-enriched sample). The availability of tools to predict novel circRNAs from RNA-seq data is of great value, given that RNA-seq data are much more highly available (122). Thus, Li et al. used circRNAFinder, a tool able to predict circRNAs from bulk RNA-seq experiments, and identified 249 circRNA candidates differentially expressed between IH and matched normal skin samples. The circRNAs hsa\_circRNA001885 and hsa\_circRNA006612 where further investigated by this group, providing novel insights about the disease (123).

Frontiers in Oncology | www.frontiersin.org

green panels show the names of the available databases for miRNAs and lncRNAs, respectively.

As shown previously, the development of tools used to predict and identify novel ncRNAs is invaluable. The increasing number of RNA-seq experiments and access to databases will increasingly facilitate the discovery of novel ncRNAs, and the characterization of ncRNAs will become increasingly straightforward. For instance, ANGIOGENES is a database that has been created to store information related to angiogenic processes. It depicts experimental data obtained from RNA-seq experiments in endothelial cells. This allows for the in-silico detection of genes expressed in several endothelial cell types from different tissues. ANGIOGENES uses publicly-available RNA-seq experiments and identifies endothelial cell-specific ncRNAs in human, mouse, and zebrafish. The database facilitates further analyses using GO enrichment terms and is available online (124). In addition to ANGIOGENES, EndoDB is another database that retrieves information about endothelial cells from


TABLE 3 | Databases and tools for the ncRNAs study.

different platforms for several species (125). Other databases are available for the study of ncRNAs; nevertheless, these are not specialized in angiogenesis or neovascular processes. Databases and tools used for the study of ncRNAs are listed in **Table 3**.

We know that endothelial cells are heterogeneous; for instance, they function differently depending on vessel type (162). To uncover the molecular mechanisms controlling this heterogeneity, single-cell RNA sequencing analyses (scRNA-seq) have the potential to enhance our understanding of vascular biology. ScRNA-seq is currently being used to study and assess cellular heterogeneity. Particularly with respect to cancer research, this approach has proved to be valuable (163–165); nevertheless, its use in vascular research is just beginning. Recently published studies have mostly focused on proteincoding genes (166, 167). The participation of ncRNAs, along with epigenetic factors, in regulating the metabolic activities of endothelial cells from a single-cell perspective in vascular development and diseases is not yet clear.

#### CONCLUDING REMARKS

ncRNAs comprise a new frontier in genetic regulation that has impacts on several research areas. Undoubtedly, the study of angiogenesis and neovascularization has been enhanced through the integration of the study of ncRNAs and epigenetics. Further, ncRNAs are involved in the regulation of several angiogenic targets through epigenetic mechanisms. On the basis of this relationship, a new field of opportunity has emerged in which biomarkers and specific therapies may be identified that can improve the treatment of different vascular diseases and cancers. NGS platforms allow for the global analysis of ncRNA expression and can be used to compare different

#### REFERENCES


physiological and pathological processes. Most of the pathways and mechanisms controlling the ncRNA-mediated regulation of angiogenesis remain unexplored. It is likely that new research strategies implementing an epigenetic perspective will facilitate future discoveries.

#### AUTHOR CONTRIBUTIONS

IH-R, LG-C, MS-A, TM-H, and ES-R wrote the manuscript. MS-A and IH-R did the artwork. TM-H and LG-C combined their information to make the tables. All authors contributed to manuscript revision and read and approved the submitted version.

### FUNDING

This work was supported by the Consejo Nacional de Ciencia y Tecnología (CONACyT) through the Fondo Sectorial de Investigación en Salud y Seguridad Social (FOSISS, Grant No. 0261181), Fondo CB-SEP-CONACyT (284748), and UAM-PTC-704. ES-R was supported by the Natural Science Department at UAM Cuajimalpa Unit.

### ACKNOWLEDGMENTS

IH-R and TM-H are masters students and LG-C and MS-A are doctoral students from Programa de Maestría y Doctorado en Ciencias Bioquímicas, UNAM and received a fellowship from CONACyT (IH-R: CVU 886138, TM-H: CVU 924685, LG-C: CVU 588391, and MS-A: CVU 659273). MS-A was also a beneficiary of the German Academic Exchange Service (DAAD Grant No. 91693321).


facilitate malignancy of hepatoma cells. Oncotarget. (2016) 8:17712– 25. doi: 10.18632/oncotarget.10832


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Hernández-Romero, Guerra-Calderas, Salgado-Albarrán, Maldonado-Huerta and Soto-Reyes. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# WMJ-S-001, a Novel Aliphatic Hydroxamate-Based Compound, Suppresses Lymphangiogenesis Through p38mapk-p53-survivin Signaling Cascade

Shiu-Wen Huang1,2†, Hung-Yu Yang3,4†, Wei-Jan Huang<sup>5</sup> , Wei-Chuan Chen<sup>6</sup> , Meng-Chieh Yu<sup>2</sup> , Shih-Wei Wang7,8, Ya-Fen Hsu<sup>9</sup> \* and Ming-Jen Hsu2,5 \*

<sup>1</sup> Department of Medical Research, Taipei Medical University Hospital, Taipei, Taiwan, <sup>2</sup> Department of Pharmacology, School of Medicine, College of Medicine, Taipei Medical University, Taipei, Taiwan, <sup>3</sup> Division of Cardiovascular Medicine, Department of Internal Medicine, Taipei Medical University-Wan Fang Hospital, Taipei, Taiwan, <sup>4</sup> Department of Internal Medicine, School of Medicine, College of Medicine, Taipei Medical University, Taipei, Taiwan, <sup>5</sup> Graduate Institute of Pharmacognosy, Taipei Medical University, Taipei, Taiwan, <sup>6</sup> Graduate Institute of Medical Sciences, College of Medicine, Taipei Medical University, Taipei, Taiwan, <sup>7</sup> Department of Medicine, Mackay Medical College, New Taipei City, Taiwan, <sup>8</sup> Graduate Institute of Natural Products, College of Pharmacy, Kaohsiung Medical University, Kaohsiung, Taiwan, <sup>9</sup> Division of General Surgery, Department of Surgery, Landseed Hospital, Taoyuan, Taiwan

#### Edited by:

Laurence A. Marchat, National Polytechnic Institute, Mexico

#### Reviewed by:

Sathish Kumar Mungamuri, National Institute of Nutrition, India Eduardo Castañeda Saucedo, Autonomous University of Guerrero, Mexico

#### \*Correspondence:

Ya-Fen Hsu yafen0505@gmail.com Ming-Jen Hsu aspirin@tmu.edu.tw

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology

Received: 17 June 2019 Accepted: 21 October 2019 Published: 06 November 2019

#### Citation:

Huang S-W, Yang H-Y, Huang W-J, Chen W-C, Yu M-C, Wang S-W, Hsu Y-F and Hsu M-J (2019) WMJ-S-001, a Novel Aliphatic Hydroxamate-Based Compound, Suppresses Lymphangiogenesis Through p38mapk-p53-survivin Signaling Cascade. Front. Oncol. 9:1188. doi: 10.3389/fonc.2019.01188 Background and purpose: Angiogenesis and lymphangiogenesis are major routes for metastatic spread of tumor cells. It thus represent the rational targets for therapeutic intervention of cancer. Recently, we showed that a novel aliphatic hydroxamate-based compound, WMJ-S-001, exhibits anti-angiogenic, anti-inflammatory and anti-tumor properties. However, whether WMJ-S-001 is capable of suppressing lymphangiogenesis remains unclear. We are thus interested in exploring WMJ-S-001's anti-lymphangiogenic mechanisms in lymphatic endothelial cell (LECs).

Experimental approach: WMJ-S-001's effects on LEC proliferation, migration and invasion, as well as signaling molecules activation were analyzed by immunoblotting, flow-cytometry, MTT, BrdU, migration and invasion assays. We performed tube formation assay to examine WMJ-S-001's ex vivo anti-lymphangiogenic effects.

Key results: WMJ-S-001 inhibited serum-induced cell proliferation, migration, invasion in murine LECs (SV-LECs). WMJ-S-001 reduced the mRNA and protein levels of survivin. Survivin siRNA significantly suppressed serum-induced SV-LEC invasion. WMJ-S-001 induced p53 phosphorylation and increased its reporter activities. In addition, WMJ-S-001 increased p53 binding to the promoter region of survivin, while Sp1 binding to the region was decreased. WMJ-S-001 induced p38 mitogen-activated protein kinase (p38MAPK) activation. p38MPAK signaling blockade significantly inhibited p53 phosphorylation and restored survivin reduction in WMJ-S-001-stimulated SV-LCEs. Furthermore, WMJ-S-001 induced survivin reduction and inhibited cell proliferation, invasion and tube formation of primary human LECs.

Conclusions and Implications: These observations indicate that WMJ-S-001 may suppress lymphatic endothelial remodeling and reduce lymphangiogenesis through

**139**

p38MAPK-p53-survivin signaling. It also suggests that WMJ-S-001 is a potential lead compound in developing novel agents for the treatment of lymphangiogenesisassociated diseases and cancer.

Keywords: hydroxamate, lymphangiogenesis, lymphatic endothelial cells (LECs), p53, p38, survivin

#### INTRODUCTION

Lymphangiogenesis, the formation of new lymphatic vessels, occurs primarily in many physiological processes such as embryonic development, tissue repair and resolution of inflammatory reactions. It also contributes to a variety of pathological events including lymphedema, inflammatory diseases, and tumor metastasis (1). Metastatic spread of tumor cells is the major cause of morbidity and mortality and responsible for ∼90% of cancer-related deaths (2). A variety of mechanisms such as seeding of body cavities, local tissue invasion, invasion into lymphatics and hematogenous spread are involved in metastatic tumor spread. Although both blood and lymphatic systems contribute to tumor progression and metastasis, the dissemination of tumor cells via lymphatic vasculature is the most common route for most carcinomas (3, 4). Increased number of tumor-associated lymphatic vessels has been shown closely correlated with metastasis and poor clinical outcome (5).

Lymphangiogenesis, similar to angiogenesis, is tightly regulated by lymphangiogenic factors and its cognate receptors (6). The member of the vascular endothelial growth factor (VEGF) family, VEGF-C, is currently the best-characterized lymphangiogenic factor. VEGF-C's lymphangiogenic effects is primarily mediated by VEGFR-3 (also known as flt-4). The expression of VEGFR-3 is largely restricted to the lymphatic endothelial cells (LECs) in normal adult tissues (7, 8). It is reported in experimental xenograft models that VEGF-C stimulates tumor lymphangiogenesis, as well as lymph node metastasis (9, 10). In addition, VEGF-C overexpression in tumor tissues significantly correlates with accelerated tumor progression and poor clinical outcome (11). Knocking down VEGF-C expression with siRNA significantly prevented lymphangiogenesis and enhanced chemo-sensitivity in breast cancer cells (12). Therefore, VEGF-C-associated lymphangiogenesis has emerged as a key prognostic marker and represents a promising therapeutic target for cancer intervention (13). To interfere with VEGF-C-VEGFR-3 signaling, a variety of strategies has been reported currently. These includes neutralizing antibodies or peptides that antagonize VEGFR-3 signaling, receptor traps or monoclonal antibodies targeting VEGF-C and small molecule receptor tyrosine kinase inhibitors of VEGFR-3 (14).

The smallest member of the inhibitor of apoptosis protein (IAP) family, survivin, is rarely detected in most terminally differentiated adult tissues with notable exceptions of vascular endothelial or hematopoietic cells (15). However, a high survivin level is found in most common human cancers. Highly expressed survivin positively correlates with tumor progression and poor prognosis (16–18). Survivin not only suppresses cell death, but also participates in a variety of cellular events. These include cell migration, cell cycle progression (16), and angiogenesis, which may promote metastatic spread of tumor cells (17). Cai et al. (19) reported that survivin level is associated with VEGF-C level and the presence of lymphatic invasion in breast cancer. However, the association between endothelial survivin and lymphangiogenesis remains incompletely understood. It is likely that modulating survivin level may provide another means of regulating tumor lymphangiogenesis. Survivin expression is regulated primarily at the transcriptional level. Survivin is up-regulated by transcription factors such as signal transducer and activator of transcription 3 (STAT3) (20), specificity protein 1 (Sp1) (21) or hypoxiainducible factor-1α (HIF-1α). However, p53, a tumor suppressor, may cause survivin reduction (22, 23). Pharmacological targeting of the p53-survivin cascade may be a potential therapeutic strategy in not only causing tumor cell death, but also suppressing lymphangiogenesis and tumor progression.

Hydroxamate, a key pharmacophore, has attracted considerable attention in drug development field due to its diverse pharmacological properties (24). Growing evidence demonstrates the potential use of hydroxamate derivatives as anti-tumor (20, 25), anti-inflammatory (26), or anti-infectious (27) agents. Recently, we synthesized and showed that a novel aliphatic hydroxamate-based compound WMJ-S-001 exhibits anti-tumor (22), anti-inflammatory (28) and anti-angiogenic (20) properties. Given its potential as lead compound for drug discovery, we aimed to investigate WMJ-S-001's antilymphangiogenic effects and its underlying mechanisms in LECs.

#### MATERIALS AND METHODS

#### Reagents

All cell culture reagents including fetal bovine serum (FBS), TrypLETM, DMEM medium and transfection reagent, TurbofectTM were from Invitrogen (Carlsbad, CA, U.S.A.). All chemicals including toluidine blue O and 3-[4, 5 dimethylthiazol-2-yl]-2, 5-diphenyltetrazolium bromide (MTT) were from Sigma-Aldrich (St Louis, MO, U.S.A.). Ribociclib was from MedChemExpress (Monmouth Junction, NJ, U.S.A.). Antibodies against PARP, caspase 3 active form, survivin, ERK1/2 and ERK1/2 phosphorylated at Thr 202/Tyr 204, p38MAPK, p38MAPK phosphorylated at Thr180/Tyr182, p53 phosphorylated at Ser15 and p53 acetylated at Lys379 were from Cell Signaling (Danvers, MA, U.S.A.). Antibodies

**Abbreviations:** IAP, inhibitor of apoptosis protein; LEC, lymphatic endothelial cell; MTT, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide; p38MAPK, p38 mitogen-activated protein kinase; PI, propidium iodide; VEGF, vascular endothelial growth factor.

against α-tubulin and GAPDH, as well as anti-rabbit and anti-mouse IgG conjugated horseradish peroxidase antibodies were from GeneTex Inc (Irvine, CA, U.S.A.). Antibody against Sp1 and normal IgG were from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Immobilon Western Chemiluminescent HRP Substrate was from Millipore (Billerica, MA, USA). All materials for western analysis were from Bio-Rad (Hercules, CA, U.S.A.). Cell Proliferation ELISA, BrdU assay kit was from Roche (Indianapolis, IN, USA). PG13-luciferase construct (p53-luc) with p53 binding sites (Addgene plasmid #16642) and p21/WAF1 promoter luciferase construct (p21 pro-luc, Addgene plasmid #16451) as described previously (29) were kindly provided by Dr. Bert Vogelstein. A Dual-Glo luciferase assay system and renilla-luciferase construct were from Promega (Madison, WI, U.S.A.).

#### Synthesis of WMJ-S-001

WMJ-J compounds were synthesized as described previously (28).

### Cell Culture

The murine LEC line SV-LEC was kindly provided by Dr. J.S. Alexander (Shreveport, LA). SV-LECs were cultured as previously described (30). Primary human lymphatic endothelial cells (HLEC, C-12217), MV2 basal and growth medium were purchased from PromoCell (Heidelberg, Germany). HLECs were maintained in MV2 growth medium in a humidified 37◦C incubator. The MV2 growth medium contains growth supplements including 5% fetal calf serum, 5 ng/ml epidermal growth factor, 10 ng/ml basic fibroblast growth factor, 20 ng/ml insulin-like growth factor, 0.5 ng/ml vascular endothelial growth factor, 1µg/ml ascorbic acid, 0.2 µg/ml hydrocortisone.

#### MTT Assay

We used the colorimetric MTT assay to determine cell viability as described previously (28).

### Cell Proliferation Assay (BrdU Incorporation Assay)

Human lymphatic endothelial cells (HLECs) (2 × 10<sup>4</sup> per well) seeded in 48-well tissue culture plates were starved in MV2 basal medium in the absence of growth supplements for 24 h. After starvation, cells were incubated in MV2 growth medium containing growth supplements with or without WMJ-S-001 at indicated concentrations for another 24 h. A Cell Proliferation ELISA, BrdU (colorimetric) kit (Roche) based on the colorimetric detection of the incorporation of BrdU was used to determine cell proliferation following the manufacturer's instructions.

#### Flowcytometry

SV-LCEs were treated with indicated concentrations of WMJ-S-001 for 24 h. Cells were harvested and fixed in 70% ethanol at 0◦C for another 24 h. After fixation and washed with phosphate-citric acid buffer, cells were stained in the dark for 30 min with staining buffer [(0.1% Triton X-100, 100µg/ml RNase A and 25µg/ml propidium iodide (PI)]. The FACScan and Cellquest program (BD Biosciences, San Jose, CA, U.S.A.) were used to perform flow-cytometric analysis. The percentage of PI-stained cells in the G0/G1, S, G2/M, or subG1 (Apoptosis, Apo) region was analyzed using the FCS Express (De Novo Software, Glendale, CA, U.S.A) or ModFit (BD Biosciences, San Jose, CA, U.S.A.) program.

#### Western Analysis

After treatment, cells were harvested in lysis buffer [0.5% NP-40, 140 mM NaCl, 10 mM Tris (pH 7.0), 0.05 mM pepstatin A, 2 mM PMSF and 0.2 mM leupeptin]. Cell lysate with equal amounts of protein were subjected to SDS-PAGE and transferred onto a NC membrane (Pall Corporation, Washington, NY, U.S.A.). After transfer, membrane was incubated with 5% non-fat milk-containing blocking buffer for 1 h. Specific primary antibodies and horseradish peroxidaseconjugated secondary antibodies were used to recognize target proteins. Enhanced chemiluminescence was used to detect immunoreactivity according to manufacturer's instructions. To obtain the quantitative data, a computing densitometer with a scientific imaging system (Biospectrum AC System, UVP) was used.

### Cell Transfection

SV-LECs (7 × 10<sup>4</sup> cells/well) were transfected with PG13 luciferase (p53-luc) or p21 promoter-luciferase (p21 proluc) plus renilla-luciferase for reporter assay or transfected with negative control siRNA (NC) or survivin siRNA for MTT, flowcytometry, immunoblotting and invasion assay using Turbofect transfection reagent (Invitrogen, Carlsbad, CA, U.S.A.) per manufacturer's instructions.

#### Reporter Assay

After transfection, SV-LECs were treated with vehicle or WMJ-S-001 for another 24 h. A Dual-Glo luciferase assay system kit (Promega, Madison, WI, U.S.A.) was employed to determine the luciferase reporter activity per manufacturer's instructions. The reporter activity was normalized based on renilla-luciferase activity.

#### Survivin Silencing

Target gene silencing in SV-LECs was performed as described previously (31). Negative control scramble siRNA and predesigned siRNAs targeting the murine survivin (BIRC5) were purchased from Sigma-Aldrich (St Louis, MO, U.S.A). The siRNA oligonucleotides were as follows: survivin siRNA, 5′ cgauagaggagcauagaa-3′ and negative control scramble siRNA, 5′ gaucauacgugcgaucaga-3′ .

#### Cell Migration Assay

SV-LECs were seeded in the 12-well tissue culture plates. After growing to confluence, SV-LECS were starved with serum-free DMEM medium for 24 h. Pipette tips were used to create scratch wounds in monolayers of SV-LECs. Cells were washed with PBS, followed by the treatment with WMJ-S-001 at different concentrations in the presence or absence of 10% FBS for another 24 h. Cells were fixed with cold 4% paraformaldehyde and stained with 0.5% toluidine blue. After staining, an OLYMPUS Biological Microscope digital camera (Yuan Li Instrument Co., Taipei,

FIGURE 1 | WMJ-S-001 inhibited cell proliferation and induced apoptosis in SV-LECs. (A) SV-LECs were treated with indicated concentrations of WMJ-S-001 for 24 or 48 h. Cell viability was determined by MTT assay. Each column represents the mean ± S.E.M. of eight independent experiments performed in duplicate (\*p < 0.05, compared with the control group). (B) SV-LECs were starved in serum-free DMEM medium for 24 h. After starvation, cells were treated with vehicle or WMJ-S-001 (10µM) in the presence of 10% FBS for indicated periods. Cell viability was determined by MTT assay. Each column represents the mean ± SEM of five independent experiments performed in duplicate (\*p < 0.05, compared with the control group at the time 0; #p < 0.05, compared with the vehicle-treated control group at the same time point). (C) After starvation as described in (B), cells were treated with vehicle or indicated concentrations of WMJ-S-001 in the presence of 10% FBS for 48 h. Cell viability was determined by MTT assay. Each column represents the mean ± S.E.M. of six independent experiments performed in duplicate (\*p < 0.05, compared with the control group; #p < 0.05, compared with the control group in the presence of 10% FBS). (D) Cells were treated with vehicle or WMJ-S-001 at indicated concentrations for 24 h. The percentage of cells in subG1, G0/G1, S, and G2/M phases was then analyzed by flow-cytometric analysis with PI staining. Each column represents the mean ± S.E.M. of five independent experiments. (\*p < 0.05, compared with the control group) (E) Cells were treated as in (D), the extent of cleavage caspase 3 and PARP were then determined by immunoblotting. Results shown are representative of four independent experiments.

Taiwan) was used to take photographs at 40× magnification. Cell migration rate was determined by calculating the migrated cells in the wound area.

#### Invasion Assay

We performed cell invasion assays as described previously (20). 0.2% gelatin solution was used to coat the lower face of the filter

FIGURE 2 | compared with the control group; #p < 0.05, compared with the group treated with serum alone). (B) A total of 10<sup>4</sup> SV-LECs were seeded in the top gelatin-coated chamber and treated with vehicle or indicated concentrations of WMJ-S-001 using serum (10% FBS) as chemo-attractant. After 16 h, the SV-LECs that invaded through the gelatin-coated membrane were stained and quantified as described in the section Materials and Methods. Each column represents the mean ± S.E.M. of six independent experiments (\*p < 0.05, compared with the control group; #p < 0.05, compared with the group treated with serum alone).

in the transwell plate (Corning, NY, U.S.A.). The lower chambers were filled with containing 10% FBS-containing DMEM medium (SV-LECs) or growth supplements-containing MV2 medium (HLECs). SV-LECs or HLECs (2 × 10<sup>4</sup> cells per chamber) were seeded in the upper chambers in the serum-free DMEM medium or MV2 basal medium with or without WMJ-S-001. After 18 h, the non-invaded cells in the upper chamber were removed by gently scraped with a cotton swab. The invaded cells in the lower face of the filter were fixed, stained with toluidine blue (0.5% in 4% paraformaldehyde) and photographed using an optical microscope (Nikon, Japan) at ×40. The number of stained cells that invaded through the filter were counted. We also quantified cell invasion by dissolving the stained cells in 33% acetic acid and measuring the absorbance at 570 nm.

### Reverse-Transcription-Quantitative Real-Time Polymerase Chain Reaction (RT-qPCR)

After treatment as indicated, cells were harvested for isolation of total RNA and complementary DNA (cDNA) synthesis as previously described (31). We used GoTaq qPCR Master Mix (Promega, Madison, WI, U.S.A.) and StepOne Real-Time PCR systems (Applied Biosystems, Grand Island, NY U.S.A.) to perform RT- qPCR. The cycling conditions were as follows: hot-start activation for 2 min at 95◦C, followed by 40 cycles of denaturation for 15 s at 95◦C, annealing/extension for 60 s at 60◦C. The primers used to transcribe survivin and GAPDH are as follows: human survivin forward, 5′ -gcctttccttaaaggccatc-3 ′ ; human survivin reverse, 5′ -aacccttcccagactcca ct-3′ ; human GAPDH forward, 5′ -gtcagtggtggacctgacct-3′ ; human GAPDH reverse, 5′ -aggggtctacatggcaactg-3′ ; mouse survivin forward, 5′ atcgccaccttcaagaactg-3′ ; mouse survivin reverse, 5′ -tgactgacgggt agtctttgc-3′ ; mouse GAPDH forward, 5′ -ccttcattgacctcaactac-3′ ; mouse GAPDH reverse, 5′ -ggaaggccatgccagtgagc-3′ .

### Chromatin Immunoprecipitation (ChIP) Assay

After treatment as indicated, cells were cross-linked with formaldehyde (1%) for 10 min at 37◦C. Cross-linking was quenched by adding 1.25 M glycine. After harvesting cells with ice-cold PBS, the cell pellet was resuspended in SDS lysis buffer. Samples were sonicated five times (for 15 s each) and centrifuged (10 min) to collect supernatants. An aliquot of each sample was used as "Input." The remainder of the soluble chromatin was diluted in ChIP dilution buffer. Immunoprecipitation was performed by adding normal IgG, anti-p53, or anti-Sp1 antibodies plus protein A-magnetic beads (Millipore, Billerica, MA, U.S.A.) with a gentle rotation at 4◦C for 18 h. The immune complexes were washed sequentially in the following buffers: low-salt, high-salt, LiCl immune complex washing buffer and Tris-EDTA buffer. After last wash, elution buffer (100 µl each) was added twice to elute the immune complex. The cross-linked chromatin complex was reversed by adding 0.2 M NaCl and heating for 4 h at 65◦C. GPTM DNA purification spin columns (Viogene, New Taipei City, Taiwan) were used to purify DNA. Purified DNA was used to perform PCR with PCR Master Mix (Promega, Madison, WI, U.S.A.). To amplify the survivin promoter fragment, the following primers were used: forward, 5′ accgcagcagaaggtacaac-3′ and reverse, 5′ -agacgactcaaacgcaggat-3 ′ . The cycling conditions were as follows: initial denaturation for 5 min at 95◦C, followed by 30-cycles of 30 s at 95◦C, 30 s at 56◦C and 45 s at 72◦C and final extension for another 10 min at 72◦C. PCR products were analyzed by agarose gel electrophoresis (1.5%).

### Tube Formation Assay

The basement membrane matrix, matrigel (Becton Dickinson, Mountain View, CA, USA) was used to perform the tube formation assay as previously described (32). Matrigel was polymerized for 30 min at 37◦C. HLECs were seeded onto the matrigel in MV2 basal medium with or without WMJ-S-001 at indicated concentrations. After 24 h, cells were photographed using an optical microscope (Nikon, Japan) at ×40.

### Blinding and Randomization

We have different people analyzing data (analyst) and conducting experiments (operator) for blinding. The same cell in every single experiment was used to evaluate the WMJ-S-001's effects vs. the related control. Therefore, formal randomization was not employed.

### Data and Statistical Analysis

In this study, the data and statistical analysis comply with the recommendations on experimental design and analysis in pharmacology (33). Results represented as mean ± standard error of mean (S.E.M) (n ≥ 3), where 'n' refers to independent values, and not replicates. Normalization was performed to compare the differences after the treatment to control for unwanted sources of variation and to reveal relevant trends. To reduce the effect of variation from different exposure of immunoblotting, the protein expression levels or the status of protein modification were expressed by normalization that generates control values with no variance (SEM = 0). Such data are not subjected to parametric statistical analysis. SigmaPlot 10 (Build 10.0.0.54; Systat Software, San Jose, CA, U.S.A.) was used to perform statistical analysis. Statistical comparisons between two groups were evaluated by the Mann–Whitney test for non-parametric analysis or

(Continued)

FIGURE 3 | group). (B) Cells were treated with WMJ-S-001 at indicated concentrations for 6 h. The extent of survivin mRNA was determined by Q-RT-PCR as described in the section Materials and Methods. Each column represents the mean ± S.E.M. of five independent experiments (\*p < 0.05, compared with the control group). (C) Cells were transfected with negative control siRNA or survivin siRNA for 48 h. After transfection, cells were harvested and the extent of survivin level was determined by immunoblotting. Each column represents the mean ± S.E.M. of five independent experiments (\*p < 0.05, compared with the negative control siRNA-transfected group). (D) After transfection as described in (C), cell viability was determined by MTT assay. Each column represents the mean ± S.E.M. of five independent experiments performed in duplicate (\*p < 0.05, compared with the negative control siRNA-transfected group). (E) After transfection as described in (C), flow-cytometric analysis was used to determine cell cycle distribution. Each column represents the mean ± SEM of five independent experiments (\*p < 0.05, compared with the negative control siRNA-transfected group). (F) After transfection as described in (C), cells were harvested for cell invasion assay as described in the section Materials and Methods. Each column represents the mean ± S.E.M. of five independent experiments (\*p < 0.05, compared with the negative control siRNA-transfected group in the presence of serum). SF, serum free.

unpaired Student's t-test for parametric analysis. Kruskal-Wallis test followed by Dunn's multiple comparison for nonparametric analysis or one-way analysis of variance (ANOVA) with Tukey's post-hoc test for parametric analysis was used to evaluate the statistical comparisons between more than two groups. A P-value smaller than 0.05 was defined as statistically significant.

#### RESULTS

#### WMJ-S-001 Inhibits SV-LEC Proliferation and Causes Apoptosis

It is still difficult to isolate and propagate LECs from different organs (7, 34, 35). This limits the studies on lymphangiogenesis or signaling mechanisms in LECs. In this study, we selected a SV40 large T-expressed immortalized murine LEC line (SV-LEC), which retain their "lymphatic" endothelial characteristics after repeated passages (30, 36), to overcome these limitations. A MTT assay was used to examine whether WMJ-S-001, a novel aliphatic hydroxamate-based compound, affects SV-LEC viability. As shown in **Figure 1A**, WMJ-S-001 reduced SV-LEC viability in the concentration- and time-dependent manners. We next determined whether WMJ-S-001's inhibitory actions on the cell viability attributes to its anti-proliferative effects. After starvation with serum-free medium for 24 h, SV-LECs were stimulated with serum (10% FBS) for another 6 to 72 h in the presence or absence of WMJ-S-001 (10µM). As shown in **Figure 1B**, WMJ-S-001 significantly reduced serum-induced SV-LEC proliferation. In addition, WMJ-S-001 concentration-dependently inhibited cell proliferation in SV-LECs after 48 h exposure to serum (**Figure 1C**). We next used flow cytometry with propidium iodide (PI) staining to examine whether WMJ-S-001 alters cell cycle progression or induces apoptosis. Treatment of WMJ-S-001 for 24 h significantly reduced the percentage of PI-stained cells in the S region as compared with the control group (**Figure 1D**). This effect was accompanied by a concomitant increase in the percentage of PI-stained cells in the G1 region after treatment of WMJ-S-001 (**Figure 1D**). In addition, WMJ-S-001 at concentrations of 10µM or lower did not significantly cause apoptosis (sub-G1, apoptotic region). However, WMJ-S-001 at 30µM significantly induced cell apoptosis in SV-LECs (**Figure 1D**). We next examined whether WMJ-S-001 activates caspase3, an apoptosis marker. WMJ-S-001 at concentrations higher than 10µM (20 and 30µM) markedly increased the cleaved (active) form of caspase 3 and its substrate, PARP (**Figure 1E**). It suggests that inhibiting LEC proliferation and causing apoptosis may contribute to WMJ-S-001's antilymphangiogenic effects.

### WMJ-S-001 Inhibits Serum-Induced LEC Migration and Invasion

We next examined whether WMJ-S-001 alters cell motility, a pivotal step in lymphangiogenesis, in SV-LECs after serum (10% FBS) exposure. WMJ-S-001 significantly inhibited seruminduced LEC migration as determined by wound-healing migration assay (**Figure 2A**). We also used transwell invasion assay to examine WMJ-S-001's effects on serum-induced cell invasion. As shown in **Figure 2B**, WMJ-S-001 at 10 and 30µM significantly reduced the number of invading cell penetrating the gelatin-coated transwell filter barrier, using serum (10% FBS) as the chemoattractant. These observations indicate that WMJ-S-001 is capable of inhibiting LEC migration and invasion.

### Surivin Reduction Contributes to WMJ-S-001's Inhibitory Effects on SV-LEC Invasion

Growing evidence has demonstrated that suvivin not only regulates mitosis and apoptosis, but also plays a critical role in angiogenesis (37, 38). However, whether endothelial survivin participates in lymphangiogenesis remains unclear. We thus determined whether WMJ-S-001 alters survivin level in SV-LECs. Results from immunoblotting analysis demonstrated that survivin exhibits high expression levels in SV-LECs (**Figure 3A**). However, treatment of cells with WMJ-S-001 for 18 or 24 h concentration-dependently caused survivin reduction (**Figure 3A**). We also determined whether WMJ-S-001 affects survivin mRNA level in SV-LECs. As shown in **Figure 3B**, WMJ-S-001 at 10 or 30µM significantly reduced survivin mRNA expression. It indicates that WMJ-S-001 may reduce survivin expression at the transcriptional level. A survivin siRNA oligonucleotide was employed to determine whether survivin reduction decreases cell viability in SV-LECs. Survivin siRNA significantly reduced the basal level of survivin (**Figure 3C**) and decreased cell viability (**Figure 3D**) in SV-LECs. Results from flow-cytometric analysis further demonstrated that survivin reduction mimics the enhancing effects of WMJ-S-001 in reducing the percentage of PI-stained cells in the S region. Survivin siRNA also caused G2/M

FIGURE 4 | stimulated with WMJ-S-001 at indicated concentrations for another 24 h. Luciferase activity was then determined. Each column represents the mean ± S.E.M. of six independent experiments (\*p < 0.05, compared with the control group). (C) Cells were treated with WMJ-S-001 (10µ M) for 6 h. ChIP assay was then performed as described in the section Materials and Methods. Typical traces representative of three independent experiments with similar results are shown. (D) Cells were treated with vehicle or WMJ-S-001 at indicated concentrations for 6 h. The phosphorylation status of p38MAPK was determined by immunoblotting. Each column represents the mean ± S.E.M. of six independent experiments (\*p < 0.05, compared with the control group). (E) Cells were treated with p38 inhibitor III (1µM) for 30 min followed by WMJ-S-001 (10µM) exposure for another 6 h. The phosphorylation status of p53 was determined by immunoblotting. Each column represents the mean ± S.E.M. of five independent experiments (\*p < 0.05, compared with the control group; #p < 0.05, compared with the group treated with WMJ-S-001 alone). (F) Cells were treated with p38 inhibitor III at indicated concentrations for 30 min followed by WMJ-S-001 (20µM) exposure for another 24 h. The extent of survivin was determined by immunoblotting. Each column represents the mean ± S.E.M. of five independent experiments. (\*p < 0.05, compared with the control group; #p < 0.05, compared with the group treated with WMJ-S-001 alone).

cell cycle arrest and apoptosis in SV-LECs (**Figure 3E**). We examined whether survivin reduction affects SV-LEC motility. As shown in **Figure 3F**, survivin siRNA mimicked the WMJ-S-001's effects in reducing the number of invading cell penetrating the gelatin-coated transwell filter barrier, using serum as the chemoattractant. In contrast, negative control siRNA was without effects (**Figure 3F**). These observations suggest that survivin reduction may contribute to WMJ-S-001's inhibitory effects on SV-LEC motility and subsequent lymphangiogenesis.

#### p38MAPK Mediates WMJ-S-001-Induced p53 Activation and Survivin Reduction in SV-LECs

We investigated the underlying mechanisms of WMJ-S-001 in repressing survivin expression in SV-LECs. Transcription factor p53 participates in various cellular processes such as cell cycle arrest, apoptosis or senescence by regulating diverse target genes (39). It is reported that p53 may suppress survivin expression by counteracting Sp1 binding to the survivin promoter region (40). Therefore, we explored whether WMJ-S-001 induces p53 activation in SV-LECs. Post-translational modifications such as phosphorylation or acetylation play important roles in regulating p53's activity (41). We thus examined WMJ-S-001's effects on p53 phosphorylation and acetylation in SV-LECs. As shown in **Figure 4A**, 6 h exposure to WMJ-S-001 led to concentration-dependent increases in p53 phosphorylation and acetylation. Results from reporter assay showed that treatment of cells with WMJ-S-001 for 24 h significantly increased PG13 luciferase (p53-luciferase) activity. WMJ-S-001 also increased the promoter-luciferase activity of p21, a p53 target gene, in SV-LECs (**Figure 4B**). Moreover, we performed a ChIP analysis to examine whether Sp1 or p53 is recruited to the endogenous survivin promoter region (−236 to −26) containing putative Sp1 and p53 binding sites, in WMJ-S-001-stimulated SV-LECs. As shown in **Figure 4C**, 6 h WMJ-S-001 exposure increased p53 binding to the survivin promoter region (−236/−26). This effect was accompanied by a concomitant decrease in Sp1 binding to the region.

p38-mediated signaling cascade has been shown previously to cause p53 phosphorylation and survivin reduction in cancer cells (23). We thus elucidated whether WMJ-S-001-induced p53 activation involves p38MAPK signaling in SV-LECs. As shown in **Figure 4D**, WMJ-S-001 time-dependently induced p38MAPK phosphorylation. A pharmacological p38MAPK inhibitor, p38 inhibitor III, significantly suppressed WMJ-S-001's effects in inducing p53 phosphorylation (**Figure 4E**) and survivin reduction (**Figure 4F**) in SV-LECs. It appears that WMJ-S-001 induces p38MAPK activation, resulting in p53 activation and survivin reduction in SV-LECs.

#### WMJ-S-001 Suppressed Cell Proliferation, Invasion and Tube Formation of Primary Human LECs

We next examined whether WMJ-S-001 reduces survivin expression and exhibits anti-lymphangiogenic activities in primary human LECs. As shown in **Figure 5A**, WMJ-S-001 at concentrations ranging from 10 to 30µM significantly reduced survivin mRNA levels in human LECs (**Figure 5A**). WMJ-S-001's effects on cell proliferation in human LECs were examined using a BrdU incorporation assay. After starvation with MV2 basal medium for 24 h, human LECs were stimulated by MV2 growth medium in the absence or presence of WMJ-S-001 for another 24 h. The percentage of BrdU-labeled cells increased significantly after a 24 h treatment with MV2 growth medium. However, WMJ-S-001 reduced this increase in a concentration-dependent manner (**Figure 5B**). The effects of WMJ-S-001 on human LEC invasion were also determined using transwell invasion assay. As shown in **Figure 5C**, WMJ-S-001 significantly reduced the number of invading cells penetrating the gelatin-coated transwell filter barrier using MV2 growth medium as the chemoattractant (**Figure 5C**). We also examined whether ribociclib, a potent proliferation inhibitor targeting cyclin-dependent kinases (CDKs) 4/6 (42), affects human LEC invasion. As shown in **Figure 5D**, ribociclib, similar to WMJ-S-001, suppressed human LEC invasion. It is likely that WMJ-S-001's inhibitory effects on LEC invasion may also attribute to its anti-proliferative properties. Another key step of lymphangiogenesis is the tubular formation of LECs. We thus examined WMJ-S-001's effect on this step. Human LECs were seeded on the matrigel surface in complete MV2 growth medium in the presence of vehicle as control or WMJ-S-001. As shown in **Figure 5E**, cells incubated with MV2 growth medium became elongated, and formed capillary-like structure within 16 h. WMJ-S-001, however, concentration-dependently reduced the formation of capillary-like network (**Figure 5E**). Similarly, WMJ-S-001 also significantly reduced VEGF-C-induced tubular formation of human LECs (**Figure 5E**). Furthermore, WMJ-S-001 suppressed VEGF-C-induced ERK1/2 phosphorylation in

were treated with WMJ-S-001 at indicated concentrations for 6 h. The extent of survivin mRNA was determined by Q-RT-PCR as described in the section Materials and Methods. Each column represents the mean ± S.E.M. of six independent experiments (\*p < 0.05, compared with the control group). (B) Primary human LECs (Continued) FIGURE 5 | were starved in MV2 basal medium for 24 h. After starvation, cells were incubated in growth supplements-containing MV2 growth medium in the absence or presence of indicated concentrations of WMJ-S-001 for another 24 h. Cell proliferation was determined as described in the section Materials and Methods. Each column represents the mean ± SEM of ten independent experiments performed in duplicate (\*p < 0.05, compared with the control group; #p < 0.05, compared with the control group in the presence of growth supplements). After starvation as described in (B), cells were seeded in the top chamber in the absence or presence of indicated concentrations of WMJ-S-001 (C) or ribociclib (10µM) (D) using growth supplements as chemo-attractant. After 24 h, invaded cells through the gelatin-coated membrane were stained and quantified. Each column represents the mean ± S.E.M. of six independent experiments. (\*p < 0.05, compared with the control group; #p < 0.05, compared with the control group in the presence of growth supplements). (E) Primary human LECs were seeded on Matrigel in the presence of growth supplements or VEGF-C (50 ng/ml) with or without WMJ-S-001 at indicated concentrations. Cells were then photographed under phase-contrast after 16 h. Bar graphs show compiled data of average sprout arch numbers (n = 6) (\*p < 0.05, compared with the control group; #p < 0.05, compared with the group treated with growth supplements or VEGF-C alone). (F) Cells were treated with WMJ-S-001 at indicated concentrations for 30 min followed by VEGF-C (50 ng/ml) exposure for another 20 min. The extent of ERK1/2 phosporylation was determined by immunoblotting. The complied results shown at the bottom of the blot represents the mean ± S.E.M. of three independent experiments (\*p < 0.05, compared with the control group; #p < 0.05, compared with the group treated with WMJ-S-001 alone).

human LECS (**Figure 5F**). Together these observations suggests that WMJ-S-001 exhibits anti-lymphangiogenic properties through suppressing cell proliferation, invasion and tubular formation of LECs. WMJ-S-001's actions in LECs may also attribute to survivin reduction.

### DISCUSSION

Most cancer-related deaths are caused by tumor metastasis (2). Targeting the major routes for metastatic spread of tumor cells, angiogenesis and lymphangiogenesis, thus represents a promising therapeutic strategy for cancer intervention. To date, the European Medicines Agency (EMEA) or the U.S. Food and Drug Administration (FDA) has already approved several agents targeting angiogenesis including monoclonal antibodies and small molecule inhibitors for the treatment of certain types of cancer (43–45). In addition, some lymphangiogenesis inhibitors have been shown effective in reducing tumor progression and metastasis in solid tumors (46, 47). These observations led to increased efforts in discovering and developing novel agents targeting angiogenesis or lymphangiogenesis. Growing evidence shows beneficial effects of hydroxamate-based compounds in the treatment of cancer (24, 25, 31). Recently, we synthesized and identified a novel aliphatic hydroxamate-based compound, WMJ-S-001, that exhibits anti-tumor (22) and anti-angiogenesis (20) properties in in vitro and in vivo models. In this study, we further demonstrated that WMJ-S-001 also suppressed lymphangiogenesis using LECs as a cell model. We also demonstrated that p38MAPK-p53-survivin signaling might participate in WMJ-S-001's anti-lymphangiogenic actions.

Beyond its anti-apoptotic effects, survivin regulates a variety of cellular events such as cell cycle progression, cell migration and angiogenesis, which may enhance tumor metastasis and progression (16, 17, 37, 48). However, the association between endothelial survivin and lymphangiogenesis, remains to be fully defined. In the present study, we showed that endothelial survivin reduction significantly suppressed LEC invasion, a key step in lymphangiogenesis. Similar to previous studies (22, 40), we noted that WMJ-S-001-induced survivin reduction attributes to the activation of p53 in LECs. WMJ-S-001-activated p53 also led to cell cycle regulator p21 transcriptional activation, which blocks cell cycle machinery. It appears that WMJ-S-001's anti-proliferative effects in LECs may involve additional cell cycle regulatory proteins. In addition to suppressing cell cycle progression, WMJ-S-001 at concentrations higher than 10µM also caused caspase 3 activation and apoptosis. It is likely that WMJ-S-001's anti-lymphangiogenic effects may also attribute to its apoptotic mechanisms when its concentrations is higher than 10 µM.

We showed that p38MAPK mediates p53 phosphorylation and survivin reduction in LECs after WMJ-S-001 exposure. The precise mechanisms underlying WMJ-S-001-induced p38MAPK activation in LECs remains to be established. Chen et al. (49) reported that activating Src homology 2 (SH2) domain-containing protein tyrosine phosphatase-1 (SHP-1)- PP2A-p38MAPK signaling cascade leads to p53 activation and vascular smooth muscle cell death. We have established that SHP-1 activation is involved in WMJ-S-001's anti-angiogenic effects in HUVECs (20). We also noted that SHP-1 signaling blockade by NSC-87877 reduces WMJ-S-001's effects on survivin and p21 levels in both HUVECs and HCT116 colorectal cancer cells (unpublished data). It raises the possibility that WMJ-S-001-induced p53 activation and subsequent cellular events may also involve SHP-1 in LECs.

VEGF-C activation of VEGFR-3 signaling plays a crucial role in lymphangiogenesis (7, 8). In addition to targeting p38MAPK-p53-survivin cascade, we noted that WMJ-S-001 suppresses VEGF-C-induced ERK phosphorylation and tube formation in primary human LECs. It suggests that inhibition of VEGF-C-VEGFR-3 signaling is causally related to WMJ-S-001's anti-lymphangiogenic effects. The inhibitory mechanisms of WMJ-S-001 in VEGF-C-stimulated LECs remain to be further investigated. The importance of protein tyrosine phosphatases (PTPs) in regulating VEGF effects has been highlighted recently in endothelial cells (50). Among these PTPs, density enhanced phosphatase (DEP)-1 (51), PTP1B (52), VE-PTP (53), and SHP-1(20) have been shown to negatively regulate VEGFR-2 signaling. In contrast, PTPs involved in the regulation of VEGFR-3 remains largely unknown. Deng et al. (54) recently showed that VE-PTP regulates VEGF-C-VEGFR-3 signaling in LECs. Together these findings suggest that PTPs may contribute to WMJ-S-001 inactivation of VEGF-C-VEGFR-3 signaling. It is worth to investigate whether certain PTP such as VE-PTP or SHP-1 contributes to WMJ-S-001's antilymphangiogenic actions.

Several lines of evidence demonstrated that angiogenesis inhibitors targeting the VEGF-VEGF-R signaling might enhance the evasive and adaptive resistance to the therapies in tumor cells. The underlying mechanisms by which the cancer cells develop resistance remain incompletely understood. The bevacizumab (a clinical used VEGF monoclonal antibody) adapted tumor cells may switch their dependence to alternative proangiogenic signaling and enhancement of lymphaticmediated metastasis (55). Sunitinib, a multi-targeted tyrosine kinase inhibitor, also induced evasive adaption in cancer cells through an alternative neurophilin 1(NRP-1)-mediated signaling (56). Whereas, sunitinib-adapted tumor exhibits less lymphatic metastasis, as it targets both lymphangiogenesis and angiogenesis (57). It is likely that single-target agents may be inadequate as therapeutics. We have shown in this study that p38MAPK-p53-mediated survivin reduction, at least in part, contributes to WMJ-S-001's anti-lymphangiogenic actions in LECs. Moreover, WMJ-S-001 has additional properties such as anti-tumor (22), anti-angiogenic (20) and anti-inflammatory (28) activities. The precise mechanisms underlying these activities remain to be fully define. Together these findings, however, support WMJ-S-001's potential as a promising lead compound in developing novel agents for oncologic therapy.

### DATA AVAILABILITY STATEMENT

The datasets generated for this study are available on request to the corresponding author.

### AUTHOR CONTRIBUTIONS

Designed the experiments: S-WH, H-YY, W-JH, S-WW, Y-FH, and M-JH. Performed the experiments: S-WH, W-CC, M-CY, Y-FH, and M-JH. Analyzed the data: S-WH, H-YY, W-CC, M-CY, Y-FH, and M-JH. Contributed reagents, synthesized, and WMJ-J compounds: W-JH and S-WW. Wrote the paper: S-WH, H-YY, Y-FH, and M-JH.

### FUNDING

This work was supported by grants (MOST 107-2314-B-706- 001) from the Ministry of Science and Technology of Taiwan, grant (106TMU-WFH-10) from the Taipei Medical University-Wan Fang Hospital, Taipei, Taiwan, grant (LS-2017-05) from the Landseed Hospital, Taoyuan, Taiwan and grant (DP2- 108-21121-01-N-02-04; TMU107-AE1-B01) from the Taipei Medical University.

### ACKNOWLEDGMENTS

We would like to thank Dr. Bert Vogelstein for the kind gift of the constructs of PG13-luciferase with p53 binding sites (p53-luc, Addgene plasmid #16642) and p21/WAF1 promoter luciferase (p21 pro-luc, Addgene plasmid #16451) as described previously (29).

## REFERENCES


pathway. Oncogene. (2011) 30:2098–107. doi: 10.1038/onc.20 10.590


increased local invasion and distant metastasis. Cancer Cell. (2009) 15:220–31. doi: 10.1016/j.ccr.2009.01.027


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Huang, Yang, Huang, Chen, Yu, Wang, Hsu and Hsu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Nanoparticle Delivery and Tumor Vascular Normalization: The Chicken or The Egg?

#### George Mattheolabakis <sup>1</sup> \* and Constantinos M. Mikelis <sup>2</sup> \*

*<sup>1</sup> School of Basic Pharmaceutical and Toxicological Sciences, College of Pharmacy, University of Louisiana Monroe, Monroe, LA, United States, <sup>2</sup> Department of Pharmaceutical Sciences, Texas Tech University Health Sciences Center, School of Pharmacy, Amarillo, TX, United States*

Tumor-induced angiogenesis has been a significant focus of anti-cancer therapies for several decades. The immature and "leaky" tumor vasculature leads to significant cancer cell intravasation, increasing the metastatic potential, while the disoriented and hypo-perfused tumor vessels hamper the anti-tumor efficacy of immune cells and prevent the efficient diffusion of chemotherapeutic drugs. Therefore, tumor vascular normalization has emerged as a new treatment goal, aiming to provide a mature tumor vasculature, with higher perfusion, decreased cancer cell extravasation, and higher efficacy for anti-cancer therapies. Here we propose an overview of the nanodelivery approaches that target tumor vasculature, aiming to achieve vascular normalization. At the same time, abnormal vascular architecture and leaky tumor vessels have been the cornerstone for nanodelivery approaches through the enhanced permeability and retention (EPR) effect. Vascular normalization presents new opportunities and requirements for efficient nanoparticle delivery against the tumor cells and overall improved anti-cancer therapies.

Keywords: nanoparticles, delivery, tumor, vessel, normalization

### INTRODUCTION

Anti-angiogenic therapy has been a major focus area of anti-cancer research for several decades (1). Blocking the immature, disorganized tumor-derived vessels led to significant tumor inhibitory effects in preclinical models and rendered anti-angiogenic therapy as a promising approach for cancer treatment, especially in combination with chemotherapy. A large volume of preclinical data with angiogenesis inhibitors led to the FDA approval and release of anti-angiogenic therapies in the clinic (2, 3). The most characteristic target is vascular endothelial growth factor (VEGF), where anti-VEGF therapy, such as bevacizumab, a humanized monoclonal anti-VEGF antibody, or sorafenib and sunitinib, VEGF receptor tyrosine kinase inhibitors, were incorporated in anti-cancer treatment options either as single agents or adjuvant therapy (3, 4). However, the clinical outcome of anti-angiogenic therapy did not meet the expectations: although progression-free survival was increased in some cases, such as metastatic colorectal (5) and ovarian cancer (6), or renal cell (7) and hepatocellular carcinoma (8), in other cancers, such as breast, melanoma, pancreatic and prostate, progression-free survival and overall survival were not increased (4, 9). The main pitfall of anti-angiogenic treatment is the impaired tumor perfusion, which limits the access to chemotherapeutic agents, impedes the tumoricidal activity of immune cells, and increases hypoxia, further driving tumor aggressiveness and metastasis (**Figure 1**) (4, 10).

The rapid growth of solid tumors induces the secretion of angiogenic factors by the tumor cells to accommodate the needs of their increased proliferation rate. This results to the rapid

#### Edited by:

*Erika Ruiz-Garcia, National Institute of Cancerology (INCan), Mexico*

#### Reviewed by:

*Mikhail Durymanov, Moscow Institute of Physics and Technology, Russia Katarzyna A. Rejniak, Moffitt Cancer Center, United States*

#### \*Correspondence:

*George Mattheolabakis matthaiolampakis@ulm.edu Constantinos M. Mikelis constantinos.mikelis@ttuhsc.edu*

#### Specialty section:

*This article was submitted to Cancer Molecular Targets and Therapeutics, a section of the journal Frontiers in Oncology*

Received: *19 August 2019* Accepted: *28 October 2019* Published: *12 November 2019*

#### Citation:

*Mattheolabakis G and Mikelis CM (2019) Nanoparticle Delivery and Tumor Vascular Normalization: The Chicken or The Egg? Front. Oncol. 9:1227. doi: 10.3389/fonc.2019.01227*

**154**

development of imperfect vascularization in the tumor area, characterized by tortuous and leaky vessels. The imperfections of the rapidly growing vasculature have been identified as porouslike structures of ∼400–600 nm in diameter, leading to the enhanced permeability and retention (EPR) effect (11). The features of the EPR include a size-dependent accumulation of molecules and structures due to the leaky vessels, where particles (such as nanoparticles) and macromolecules will diffuse out of the tumor vessels that bear the imperfections, compared to any healthy tissue, where these imperfections are absent. This can induce an augmented drug concentration in the tumor area, further supported by the impaired lymphatic drainage associated with the abnormal tumor vasculature (12).

The limited clinical outcome of anti-angiogenic therapy has driven the last decade the concept of vascular normalization as a complementary therapeutic approach for anti-cancer treatment. Normalization of tumor vasculature is expected to provide a properly oriented, well-constructed vascular network with reduced vascular density, increased perfusion and limited hypoxia, which will lead to better drug delivery and anti-cancer efficacy (13, 14). An increasing number of studies demonstrate the promising outcome of vascular normalization strategies. Lower doses of current anti-angiogenic therapies, such as bevacizumab, are reported to achieve tumor vessel normalization (4, 13). Even in aggressive tumors, such as glioblastoma, treatment with cediranib, an anti-angiogenic agent, improved vessel perfusion in a subset of patients and increased overall survival (15). Tumor vessel normalization has also been achieved by non-pharmacological approaches; aerobic exercise can drive vascular normalizing effects and improve chemotherapeutic efficacy. The leading player, in that case, is considered to be the shear stress, which, when increased, enhances vascular integrity through secretion of vascular normalization mediators, such as thrombospondin-1 (16–19).

The limitations of vascular normalization approaches follow, to a certain extent, the limitations of anti-angiogenic therapy. The most common is the evasive resistance, the acquired resistance of the endothelium towards anti-angiogenic therapy that targets a growth factor, by upregulating others, which will compensate for its inhibition (4). The main goal for tumor vascular normalization is the improvement of anti-cancer drug delivery. However, the window for anti-cancer therapy is normally short, not easily identifiable, and does not occur uniformly in the patient population (13, 14). The extension and identification of this therapeutic window consist the primary goal of current studies focusing on vascular normalization, and one of the main goals is the identification of markers denoting the potent therapeutic window for vascular normalization. An example is Anterior gradient 2, a plasma protein secreted from tumor cells, which was proposed as a vascular normalization marker during anti-angiogenic treatment (20). Certain approaches to overcome that therapeutic window have been proposed, such as the simultaneous administration of anti-angiogenic and chemotherapeutic drugs through nanoparticle delivery (21).

Nanodelivery methods ensure the selective targeting of a specific tissue, offering improved delivery with significantly fewer side-effects (11). Nanoformulations are known to significantly enhance the effect of certain compounds compared to direct administration (22) and nanodelivery methods are incorporated into therapeutics of multiple diseases, including cancer (23). In this mini-review we summarize the current knowledge from studies, where nanodelivery is utilized for or in conjunction with tumor vascular normalization and discuss advantages, limitations and potentials.

### NANOFORMULATIONS AND CANCER NANOTHERAPEUTICS

Traditional drug delivery systems have been unable to address the complex therapeutic and physicochemical necessities presented by the traditional and new active molecules, including poor aqueous solubility, poor specificity, unfavorable pharmacokinetics and high toxicity (11). Nanotechnology has steadily grown to a promising field of research and application for the diagnosis and treatment of various diseases, among which is cancer. Nanocarriers are colloidal systems used for drug delivery, capable of entrapping, encapsulating and delivering active molecules to tissues and cells (11). As their name suggests, nanocarriers have a particle size at the submicron range (<1µm), though it is generally regarded that nanocarriers used through systemic administration will typically have a size below 200–250 nm. This stems from the natural filtering mechanism of the body, where nanoparticles of larger dimensions are retained and removed from the circulation through splenic filtration (24). Nanotechnology has yielded significant advantages over traditional pharmaceutical formulations, such as: (a) improved drug stability; (b) improved pharmacokinetics/biodistribution; (c) reduced non-specific toxicity; (d) reduction in drug dosage and dosing frequency, and; (e) high drug loading for compounds insoluble in water (11).

The growing field of nanotechnology has yielded new and innovative carriers with distinct and multifaceted properties, while new formulations and approaches are constantly being developed. We provide here a brief overview of the most important aspects of nanotechnology, describing the most frequently studied nanocarriers. Though there are overlaps or combinations of the technological advancements, the classification of the nanocarriers typically relies on their composition, having three major categories: (a) lipid-based; (b) polymer-based, and (c) inorganic nanocarriers (25).

Among the lipid-based formulations, liposomes are the best known and studied nanocarriers. They have achieved broad recognition for their capacity to protect and deliver active compounds, with improved biodistribution profiles and reduced toxicity (11, 25). Liposomes are primarily used for hydrophilic compounds, though their lipid bilayer allows the entrapment of hydrophobic compounds as well. More importantly, liposomal formulations have received FDA approval for use in cancer treatment, i.e., liposomal formulation of doxorubicin—Doxil <sup>R</sup> (26), among others, which constitutes them as a reliable, safe, tested, and thus attractive nanocarrier model for human treatment or new drug development (27, 28).

Solid lipid nanoparticles (SLNs) are lipid-based nanocarriers, commonly prepared by dispersing melted solid lipids in water in the presence of a stabilizing emulsifier (29). Similarly to the oil-in-water (o/w) emulsions, the hydrophobic environment inside the nanocarriers makes them ideal for the entrapment and delivery of molecules with low aqueous solubility, though, in contrast to the o/w emulsions, the hydrophobic core is solid and not liquid (29). Unlike o/w emulsions, which require oils that may present significant toxicity or biocompatibility limitations (30), both the liposomes and the SLNs utilize lipids commonly found in cells (i.e., phospholipids) that can be of natural source or synthetically made/modified. Not surprisingly, liposomes and SLNs are considered biocompatible and biodegradable, with an excellent safety record (31). In fact, synthetic approaches using polymer synthesis and chemical attachment of antibodies or targeting moieties have advanced the development of multifunctional lipids for long-circulating nanocarriers that may actively target a variety of cells, such as macrophages, endothelial or tumor cells (32–34).

Similarly, polymer-based nanocarriers have emerged as promising nanocarriers for drug delivery. The progress on polymer chemistry has allowed the development of new polymer structures with multi-faceted and highly adjustable properties, advancing the development of nanosized micelles, solid-core nanoparticles, polymersomes and dendrimers (11). These carriers have tunable characteristics, defined by the physicochemical properties of the used polymer or combination of polymers, capable of delivering unstable hydrophilic and hydrophobic compounds, or molecules that otherwise would not be capable of crossing the cell membrane, such as nucleic acids (i.e., si/miRNAs and plasmids).

Finally, inorganic nanoparticles are frequently composed of magnetic iron oxide, silica oxide and gold, among other materials. Similar to the other categories, the inorganic nanoparticles can be surface-modified to achieve long-circulating properties, actively target specific cells and tissues, and protect active compounds. Furthermore, inorganic nanoparticles, such as iron oxide/magnetic nanoparticles, can respond to external stimuli, such as magnetism, which permits their detection or active targeting to specific parts of the body, or demonstrate unique optical properties for improved in vivo imaging, such as quantum dots and up-converting nanoparticles, which lipids and polymers cannot provide (35–37).

The efficacy of nanodelivery in different tumors largely varies, guided by the variable tumor vascular characteristics, such as vessel architecture, interstitial fluid and extracellular matrix composition, phagocyte infiltration and presence of necrotic areas. Parameters, such as the extravasation of the nanoparticles from tumor blood vessels, their diffusion through the extracellular tissue and their interaction with the tumor microenvironment constitute the EPR effect, elegantly analyzed by Bertrand et al. (23). The EPR effect in solid tumors was initially described ∼3 decades ago, and was one of the driving forces for the scientific advancements taking place in the field of nanotechnology. The goal of nanotechnology-based treatment is to utilize or enhance the EPR effect in tumors, allowing better pharmacological targeting of the tumor tissue, leading to an increasing build-up of the nanocarriers with the active compound to the tumor area, which is further supported by the impaired lymphatic drainage in solid tumors (38). Alternatively, sonoporation, the combination of ultrasound and microbubbles, has improved liposome accumulation and their penetration through the tumor vasculature into the tumor interstitium (39).

The EPR effect has received criticism recently, regarding its significance in the passive targeting to tumors, its dependency on the stage and the type of tumor (40), and whether it is present in human tumors (41). There is a potential sift on the paradigm on the use of nanoformulations and their drug delivery capacity under rapidly growing vs. slowly growing tumors, as well as the influence of the vascular architectural structure. Below we summarize the up-to-date literature for nanotherapeutics targeting vessel normalization and their potential for antiangiogenic therapies.

### VESSEL NORMALIZATION

The need for vascular normalization has been further highlighted with the recent advances in tumor immunotherapy. Several antibodies targeting the immune checkpoint proteins, such as pembrolizumab, nivolumab and ipilimumab have been approved for clinical practice (42–44), and immune checkpoint inhibitions consist a revolutionary anti-cancer approach for solid tumors (45). However, a subset of patients does not benefit, and the reasons are not known (46). A potential reason for the ineffectiveness of tumor immunotherapy for the non-responding patients could be the inability of the immune cells to sufficiently access the tumor mass, and tumor vascular normalization looks a promising solution (14, 47).

A groundbreaking study for nanodelivery and tumor vasculature normalization was from Rakesh Jain's lab, where they showed that vascular endothelial growth factor receptor-2 (VEGFR2) targeting led to tumor vessel normalization and the subsequent decrease of the intratumoral interstitial pressure, improving nanoparticle delivery. It was also demonstrated that smaller nanoparticles, of 12 nm diameter, are more potent to invade rapidly to the tumor area than the larger ones (48). Although the increasing optimization of surface modifications renders these size constrains not easily applicable in biomedical applications (49), it was later demonstrated that tumor vascular normalization through VEGFR2 inhibition improved accumulation of also larger nanoparticles, of 20 and 40 nm size, in the tumoral bed. However, inside the tumor, smaller nanoparticles presented a more homogeneous distribution (50).

Increased tumor vascularity increases nanoparticle delivery, but increased collagen deposition, which also leads to increased interstitial pressure, is an inhibitory factor (51). For this, recent attempts to induce tumor vessel normalization targeted both the tumor microenvironment, as well as the extracellular matrix (ECM). An example is the coadministration of antibodies targeting vascular endothelial growth factor (VEGF) and transforming growth factor β1 (TGF-β1), which led to a combined vascular and ECM normalization and thus improved intratumoral nanomedicine delivery (52).

Gold nanoparticles have been studied for vascular normalization in several tumor types. Endostatin is an endogenous angiogenesis inhibitor. Gold nanoparticleencapsulated human recombinant endostatin led to a transient tumor vascular normalization in non-small cell lung cancer. Chemotherapy administered during the normalization window was significantly more potent than when administered as a monotherapy (53). Gold nanoparticles have been successfully used to block metastasis in melanoma by increasing tumor vascular normalization (54). Treatment of cediranib, a vascular endothelial growth factor receptor inhibitor, normalized tumor vessels in a breast cancer model, enhancing tumor retention of enzyme responsive size-changeable gold nanoparticles, further demonstrating that combinatorial treatment could be a potent approach for efficient tumor diagnosis and treatment (55).

Epidermal growth factor receptor (EGFR) tyrosine kinase inhibitors, such as erlotinib, are considered effective therapies for EGFR mutation positive non-small cell lung cancers. The promising outcome is often compromised by resistance driven by upregulation of the anti-apoptotic protein survivin, in the cancer cells. A novel approach using chloroquine to normalize the tumor vasculature, combined with anti-EGFR aptamer-mediated delivery of erlotinib and survivin shRNA co-administration significantly hampered tumor growth (56).

Nogo-B is a potent growth factor mediating endothelial cell functions, such as wound healing angiogenesis and chemotaxis, through binding to the Nogo-B receptor (57). Nogo-B receptor knockdown was achieved through nanoparticles with charge convention in the acidic tumor microenvironment, leading to breast cancer vessel normalization in vivo and inhibition of metastatic incidence (58).

Cyclooxygenase-2 (COX-2) is upregulated in several cancerrelated pathways regulating cell proliferation, apoptosis, multidrug resistance and angiogenesis (59). Celecoxib, a clinicallyrelevant COX-2 inhibitor, was reported to normalize the tumor microenvironment, including the tumor vessels, thus improving the uptake of paclitaxel-loaded micelles in xenografts of human lung adenocarcinomas (60).

Brain vascular normalization and blood-brain barrier restoration are important for glioblastoma. Liposomal formulation of the chemotherapeutic drugs irinotecan, doxorubicin and vincristine improved their pharmacokinetic profile and increased their potency in tumor inhibition. Apart from the size, mostly irinotecan- treated tumors led to vascular normalization, characterized by increased perfusion, assessed by Hoechst uptake, decreased extend of the discontinuous basement membrane, increased number of pericyte-covered capillaries and decreased vessel diameter (61).

TABLE 1 | Table summarizing the data regarding vascular normalization, including tumor models, molecular targets, targeting agents, and nanoformulations.


\**Study where nanoparticles were used, not for the transfer of the targeting agent for vascular normalization, but for anti-cancer or imaging purposes.*

miRNAs play a major role in tumor aggressiveness and metastasis. miRNA-200 was initially reported to block epithelialmesenchymal transition (EMT) in tumors through ZEB1 and ZEB2 downregulation (62–64). miRNA-200 blocks tumor angiogenesis through IL-8 and CXCL1 inhibition. Nanoparticlemediated miRNA-200 delivery reduced tumor angiogenesis and induced tumor vessel normalization, leading to tumor growth and metastasis inhibition in ovarian, lung, renal and breast adenocarcinomas (65).

Nanoparticle-based approaches are used not only for the delivery of vascular normalizing agents, but also for their development and evaluation. An example is NGR-TNF, a chimeric protein that couples the tumor homing peptide CNGRCG, which targets aminopeptidase N or myeloid plasma membrane glycoprotein CD13, also expressed in angiogenic vessels, with the N-terminus of the tumor necrosis factor-α (TNF). It is a vascular targeting agent, which presents antitumor effects and is in clinical trials for tumors either as monotherapy or in combination with chemotherapeutic drugs. Low dose treatment inhibited angiogenesis by inducing endothelial cell apoptosis, whereas at the later stages it led to tumor vascular normalization, assessed by the increased pericyte and smooth muscle cell coverage. The CD31 targeting was verified in vivo by coupling of the CNGRCG peptide to fluorescent nanoparticles (quantum dots, described above) (66). The studies are summarized in **Table 1**.

#### DISCUSSION

It is important to note that tumor vessel normalization does not automatically correspond to better distribution of all nanodelivery systems. Tumor vessel normalization induced by imatinib mesylate limited the distribution of large (∼110 nm) but enhanced the distribution of small (∼23 nm) nanoparticles in human lung adenocarcinoma. However, the nanoparticle distribution inside the tumors was overall reduced, compared to that of micelles, and micelle-based delivery of paclitaxel significantly improved its potency (67).

For the concept that vessel normalization is significantly affecting the efficiency for drug delivery using nanoformulations, nanotechnology has undoubtedly allowed the delivery, protection and targeting of compounds that other drug formulations (i.e., implants, microparticles, free drug) are incapable of achieving (68). The controversial EPR effect, along with the vessel normalization approaches, only illustrate the potential of new methodologies, such as smaller nanocarriers and active targeting. It is now widely accepted that mild antiangiogenic therapy leads to tumor vessel normalization and tumor vessel normalization induces the uptake of nanoparticle-based delivery, leading to more potent antitumor activity (48, 69). This process is also accompanied by mathematical models simulating the events and predicting the penetrance of drugs into the tumor area (70). There is significant potential for novel compounds treating the vascular endothelium to be actively delivered by nanocarriers to the tumor area, safely, with reduced toxicity and high specificity, while avoiding in vivo degradation (68). It is the authors' opinion that nanotechnology will play a significant role in the development of these therapies in the future. With the existence of several biological barriers for the successful delivery of active molecules and nanocarriers, some of which we described here, the optimal physicochemical parameters of the nanocarriers will need to be carefully considered, with their size and shape being paramount (48, 50). Finally, the combination of surface modification for cellular specificity and the achieved vascular normalization may enhance and prolong nanocarrier presence in the tumor microenvironment for improved pharmacological activity. Overall, nanoparticlemediated drug delivery targeting both tumor cells and

#### REFERENCES


tumor vessels could be a promising approach for efficient anti-cancer therapies.

#### AUTHOR CONTRIBUTIONS

GM and CM contributed to the conception of the article, wrote and revised the final manuscript, and agreed on its submission to this journal.

#### FUNDING

This work was supported for GM by the College of Pharmacy, University of Louisiana Monroe start-up funding and the National Institutes of Health (NIH) through the National Institute of General Medical Science Grants 5 P20 GM103424-15, 3 P20 GM103424-15S1 and for CM in part by National Institutes of Health Grant (NCI) R15CA231339, Texas Tech University Health Sciences Center (TTUHSC) School of Pharmacy Startup funds and TTUHSC Office of Research grant. The funders had no role in study design, decision to write and preparation of the manuscript.

Adv Drug Deliv Rev. (2007) 59:491–504. doi: 10.1016/j.addr.2007. 04.008


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Mattheolabakis and Mikelis. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# YY1 Promotes Endothelial Cell-Dependent Tumor Angiogenesis in Hepatocellular Carcinoma by Transcriptionally Activating VEGFA

Wendong Yang<sup>1</sup> , Zhongwei Li 1,2, Rong Qin1,2, Xiaorui Wang<sup>3</sup> , Huihui An1,2, Yule Wang4,5 , Yan Zhu4,5, Yantao Liu1,2, Shijiao Cai <sup>1</sup> , Shuang Chen<sup>2</sup> , Tao Sun1,2 \*, Jing Meng1,2 \* and Cheng Yang1,2 \*

*<sup>1</sup> State Key Laboratory of Medicinal Chemical Biology and College of Pharmacy, Nankai University, Tianjin, China, <sup>2</sup> Tianjin Key Laboratory of Early Druggability Evaluation of Innovative Drugs and Tianjin Key Laboratory of Molecular Drug Research, Tianjin International Joint Academy of Biomedicine, Tianjin, China, <sup>3</sup> College of Life Sciences, Nankai University, Tianjin, China, <sup>4</sup> Tianjin State Key Laboratory of Modern Chinese Medicine, Tianjin University of Traditional Chinese Medicine, Tianjin, China, <sup>5</sup> Research and Development Center of TCM, Tianjin International Joint Academy of Biotechnology and Medicine, Tianjin, China*

#### Edited by:

*Laurence A. Marchat, National Polytechnic Institute, Mexico*

#### Reviewed by:

*Nikhlesh Singh, University of Tennessee Health Science Center (UTHSC), United States Youzhi Xu, University of Kentucky, United States*

#### \*Correspondence:

*Tao Sun tao.sun@nankai.edu.cn Jing Meng mengjing0101@163.com Cheng Yang cheng.yang@nankai.edu.cn*

#### Specialty section:

*This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology*

Received: *21 June 2019* Accepted: *21 October 2019* Published: *14 November 2019*

#### Citation:

*Yang W, Li Z, Qin R, Wang X, An H, Wang Y, Zhu Y, Liu Y, Cai S, Chen S, Sun T, Meng J and Yang C (2019) YY1 Promotes Endothelial Cell-Dependent Tumor Angiogenesis in Hepatocellular Carcinoma by Transcriptionally Activating VEGFA. Front. Oncol. 9:1187. doi: 10.3389/fonc.2019.01187* Hepatocellular carcinoma (HCC) is a typical hypervascular solid tumor that requires neoangiogenesis for growth. The vascular endothelial growth factor (VEGF) is the most potent proangiogenic factor in neovascularization. The multifunctional Yin-Yang 1 (YY1) is involved in the regulation of tumor malignancy of HCC. However, the relationship between YY1 and endothelial cell-dependent tumor angiogenesis in HCC remains unclear. In this study, we observed that YY1 is positively correlated with microvessel density (MVD) and poor prognosis in HCC tissues. We further found that YY1 promotes the transcriptional activity of VEGFA by binding its promoter in HCC. The secreted VEGFA from HCC cells activates phosphorylation of VEGFR2 to promotes tube formation, cell migration, and invasion of vascular endothelial cells *in vitro*, and promotes tumor growth and angiogenesis *in vivo*. In addition, upregulation of YY1 enhanced resistance of bevacizumab in HCC cells. These results indicate that YY1 plays essential roles in HCC angiogenesis and resistance of bevacizumab by inducing VEGFA transcription and that YY1 may represent a potential molecular target for antiangiogenic therapy during HCC progression.

Keywords: YY1, angiogenesis, vascular endothelial growth factor A, transcription activation, hepatocellular carcinoma

## INTRODUCTION

Hepatocellular carcinoma (HCC) is the fifth most frequent cancer in the world and the fourth leading cause of cancer-related death (1). HCC is the most common primary malignant liver tumor with abundant tumor vascular network, which provides the evidence for the clinical therapies targeted vascular endothelial growth factor (VEGF) for the treatment of unresectable HCC (2). Angiogenesis is critical to multiple tumor invasion and metastasis (3, 4). Targeted angiogenesis therapy is an important anti-tumor strategy at present and it is particularly important to understand the transcriptional regulation of tumor angiogenesis (5, 6). Angiogenesis involves

**162**

complex signaling pathways (7–9). VEGFA is an important angiogenic factor secreted by both cancer cells and stromal infiltrating cells (10, 11). It is involved in the regulation of metastasis of many solid tumors and their neovasculature (12). VEGFA binds to two tyrosine kinase receptors of endothelial cells: VEGF receptor-1 (VEGFR1/Flt-1) and VEGFR2 (KDR). The function of VEGFR1 remains poorly defined, and VEGFR2 mediates proliferation and survival of endothelial cell (13–15). VEGFR2 is the major mediator of the mitogenic, angiogenic and permeability enhancing effects of VEGF (16). The high affinity between VEGFA and VEGFRs induces the proliferation, migration, and differentiation of vascular endothelial cells. Activated endothelial cells degrade the extracellular matrix, subsequently forming tubular structures and recruiting supporting cells to form stable vessels (17, 18).

Yin-Yang 1 (YY1) is a key transcription factor involved in cancer progression (19). YY1 is well-known for its dual roles in regulating gene expression, either as an activator or repressor, depending on the chromatin remodeling complexes it is recruited to (20, 21). Extensive evidence indicates that YY1 is an oncogene in various cancers, such as colorectal, prostate and breast cancer (22–24). There is reported that CXCR4/YY1 inhibition impairs VEGF network and angiogenesis during osteosarcoma malignancy (25). Competitive binding between Seryl-tRNA synthetase/YY1 complex and NFKB1 at the distal segment results in differential regulation of VEGF promoter activity during angiogenesis (26). In embryonic development, YY1 is responsible for maintaining VEGF in the developing visceral endoderm and that a VEGF-responsive paracrine signal, originating in the yolk sac mesoderm, is required to promote normal visceral endoderm development (27). Our previous studies showed that transcription complexes of YY1 promote malignant progression of hepatocellular carcinoma, and patients with high YY1 expression have poor prognosis (28). Although YY1 is involved in the regulation of tumor malignancy, its role and mechanism in tumor angiogenesis are rarely mentioned.

In our study, we analyzed the correlation between YY1 expression and MVD in HCC tissues and functional role of YY1 in HCC angiogenesis, and examined the underlying mechanism of YY1 regulated angiogenesis and drug sensitivity. This study may provide insights into a new potential therapeutic strategy and antitumor targets for HCC.

### MATERIALS AND METHODS

#### Cell Culture and Transfection

Human umbilical vein endothelial cells (HUVECs), human aortic endothelial cells (HAECs) and HepG2 cells were obtained from the Cell Bank of Shanghai Institute (Shanghai, China), Sciencell Research Laboratories (San Diego, USA) and KeyGen Biotech (Nanjing, China). HepG2 cells were cultured in RPMI1640 medium containing 10% fetal bovine serum (FBS) and 1% penicillin–streptomycin solution. After reaching 60– 80% confluence, then the fresh medium was replenished. The supernatant was collected and centrifuged after 48 h incubation and stored at −80◦C. HUVECs or HAECs were cultured in M-199 medium supplemented with endothelial cell growth supplement, 10% FBS, and 1% penicillin–streptomycin. After reaching 60–80% confluence, the HUVECs or HAECs were stimulated with condition medium (50% HepG2 with different treatment supernatants and 50% M-199) for 48 h and used as induced HUVECs, whereas normal HUVECs were used as control. Cells were maintained at 37◦C in a humidified atmosphere with 5% CO2. All the plasmids were transfected into cells by LipofectamineTM 2000 (Invitrogen, 11668019) in accordance with the manufacturer's instructions. Each experiment was performed in triplicate and repeated at least three times.

### Luciferase Activity Assays

HUVECs were seeded in 96-well plates. After 24 h, a pGL3 promoter vector containing the VEGFA promoter region was cotransfected with the indicated plasmids. The luciferase activities were detected using a dual-luciferase reporter gene assay kit (RG027, Beyotime) after 48 h transfection. Renilla luciferase activity was used as an internal standard. Each experiment was conducted in triplicate.

### Cell Invasion Assays

Matrigel (Corning, 354234) was diluted (1:2) in serum-free media and seeded in a 24-well transwell chamber (JET, TCS013024). After HUVECs or HAECs were incubated with the indicated cell supernatants for 48 h, ∼1 × 10<sup>5</sup> HUVECs or HAECs were seeded on the matrigel, and FBS was added to a 24-well plate located below the chamber to serve as a chemoattractant. After 24 h, invasive cells were stained with 0.1% crystal violet for 10 min. Images were obtained using a phase contrast microscope.

### Wound Healing Assay

HUVECs or HAECs stimulated with condition medium (50% HepG2 with different treatment supernatants and 50% M-199) were seeded in wells for 12 h at 37◦C. A micropipette tip was used to scrape a straight line in each well. After 24 and 48 h, the migration of cells was analyzed by comparing the wound distance ratio at 0 h. Each experiment was performed in triplicate.

### Tube Formation Assay

HUVECs or HAECs suspended in conditioned medium were seeded onto a 48-well plate coated with Matrigel (Corning, 354234) and incubated for 8 h at 37◦C. Tube formation was observed at 3 h post-treatment. The number of tubes for each treatment was quantified. This experiment was independently repeated thrice and four random fields were observed every time.

### Western Blot (WB) Analysis

Cells were washed with phosphate-buffered saline (PBS) and lysed in ice-cold lysis buffer containing protease inhibitor cocktail (Sigma) for 30 min. Lysates were separated by SDS-PAGE and transferred onto a 0.45µm PVDF membrane. After transferring, the membranes were blocked with 5% BSA at room temperature with shaking for 2 h. Membranes were incubated with anti-YY1 (1:1,000, Santa, sc-7341), anti-VEGFA (1:1,000, Affinity, DF7470), anti-pVEGFR2 (1:1,000, Affinity, AF3281), and anti-GAPDH (1:4,000, Affinity, T0004) diluted with 5% BSA overnight at 4◦C. Then, the membranes were washed three times with TBST for 10 min at room temperature and incubated with secondary antibody at room temperature for 2 h. Protein expression was assessed with enhanced chemiluminescent substrate (Millipore, USA) and by exposure to chemiluminescent film.

#### Immunofluorescence

HUVECs or HAECs incubated with the indicated supernatants were grown on glass slides until 70–80% confluent. The cells were washed three times with 1× PBS. They were fixed in 4% PFA at room temperature for 20 min. Subsequently, the cells underwent blocking and permeabilization with 5% BSA containing 0.1% Triton X-100 for 30 min at room temperature. They were incubated overnight at 4◦C with pVEGFR2 antibody (1:200, Affinity, AF3281) and then incubated with TRITC- labeled secondary antibodies (1:50, KeyGEN BioTECH) for 1 h at room temperature. Each step was followed by two 5-min washes in PBS. The prepared specimens were counterstained with DAPI (Solarbio, S2110) for 30 min. Images were acquired using a Leica confocal microscope.

#### qRT-PCR

Total RNAs were extracted from different treatment cells using TRIzol reagent (Invitrogen, 15596026). FastQuant RT kit (TIANGEN, R6906) was utilized to obtain cDNA following the manufacturer's protocol. An SYBR Green Kit (TIANGEN, FP205) was used for transcript quantification with specific primers on QuantStudioTM 6 (Life Technologies, Singapore). The samples were run in triplicate in each experiment, and a housekeeping gene (GAPDH) was used as an internal standard. The 2−11CT method was applied to quantify the relative gene

expression. The sequences of gene-specific primers were as follows: VEGFA: F:5′ -GCCTTGCCTTGCTGCTCTAC-3′ ; R:5′ - TGATTCTGCCCTCCTCCTT CTG-3′ ; GAPDH: F:5′ -GTCCAC TGGCGTCTTCAC-3′ ; R:5′ -CTTGAGGCTGTTGTC ATACTT C-3′ . GAPDH served as loading control.

### Enzyme-Linked Immunosorbent Assay (ELISA)

To detect VEGFA in culture supernatants, ELISA was carried out with ELISA kits (Beyotime, PV963) in accordance with the manufacturer's recommendations.

### Three-Dimensional Minitumor Generation

HepG2 cells with different treatment cocultured with HUVECs or HAECs at a 2:1 mix ratio. To characterize the tumor-like spheroids formed by HepG2 and HUVECs, the cells were stained with DIO (Beyotime, C1038) and DIL (BestBio, BB-441921), respectively, following the manufacturers' instructions. HepG2 and HUVECs were spun down, resuspended, and then divided by 150 µL into wells of a U-shaped 96-well suspension plate (Greiner Bio-One, Stonehouse, UK). The plate was incubated at 37◦C for 48 h to allow for spheroid formation (29). A laser scanning confocal microscope (ZEISS) was used to examine the structural organization of tumor spheroids. The integrated intensity of tumor spheroids was analyzed by ImageJ.

#### ChIP-seq Assay and Analysis

Approximately 1 × 10<sup>7</sup> HepG2 cells were freshly harvested and fixed in 1% formaldehyde/medium buffer for 10 min at room temperature. Fixation was stopped by the addition of glycine to a final concentration of 250 mM. Cell pellets were resuspended in cell lysis buffer containing 1 × Protease Inhibitor Cocktail II and then incubated for 15 min on ice. They were then pipetted for dissociation and pelleted by centrifugation at 800 g at 4◦C for 5 min. Approximately 1 mL of nuclear lysis buffer was added to resuspend the cell pellets. To ensure sonication, bioanalyzer analysis was performed. The chromatin fraction was incubated with the indicated antibody overnight at 4◦C. The protein/DNA complexes were reverse cross-linked to obtain free DNA. Spin columns were utilized to purify DNA and were then quantified by qPCR. The samples were sequenced by Genergy Biotechnology. Chromatin immunoprecipitation sequencing (ChIP-seq) data were obtained from Cistrome Data Browser (http://cistrome.org/ db). IGV software was used to analyze ChIP-seq data and obtain ChIP peak. The primer pair was tested for spanning regions in the VEGFA promoter: F: 5′ -CACTGACTAACCCCGGAACC-3′ ; R: 5′ -GGAGTGACTGGGGTCCTTT G-3′ .

### Xenograft Tumor Model

BALB/c nude mice (weighing ∼20 g, 4–6 weeks) were randomly divided into Ctrl, YY1, siYY1, and YY1 + Beva groups (n = 3 male + 3 female per group). The mice were injected with 1 × 10<sup>6</sup> HepG2 cells or stably overexpressed YY1 subcutaneously in

the mid-dorsal region. When the tumor size reached about 200 mm<sup>3</sup> , the siYY1 group was intratumorally injected with siYY1 loaded in nanoparticles. The YY1 + Beva group was treated with 2 mg/kg bevacizumab twice per week by intraperitoneal injection for 24 days. Solvent buffer at the same volume was used in other groups. The tumor sizes were measured and calculated according to a standard formula every 3 days.

This study was carried out in accordance with the principles of the Basel Declaration and recommendations of International Association of Veterinary Editors guidelines, Nankai University Ethics Committee. The protocol was approved by the Nankai University Ethics Committee.

#### Immunohistochemistry (IHC) Assay

Paraffin sections of human HCC samples and tumor tissues were deparaffinized with xylene and dehydrated with decreasing concentration of ethanol. The endogenous peroxidase activity was blocked with 3% hydrogen peroxide. Microwave antigen retrieval technique was used. Non-specific antigen sites were blocked using normal goat serum at room temperature for 20 min. Primary antibodies, including YY1 (1:100, Santa, sc-7341), pVEGFR2 (1:100, Affinity, AF3281), and CD31 (1:25, abcam, ab9498), were incubated in a humidified chamber overnight at 4◦C. HRP-polymer anti-mouse or rabbit IHC kit (Maixin Biotech, China) was utilized to incubate secondary antibody. Samples were developed with diaminobenzidine reagent and counterstained with hematoxylin. The microvessel density (MVD) were quantified using ImageJ software on the basis of CD31 staining.

#### Patient Samples

HCC tissue contains 26 cases were collected Tianjin Medical General Hospital and Tumor Hospital of Tianjin within 5 years. The donor was completely informed and each specimen from

FIGURE 3 | YY1 binds to VEGFA promoter to enhance VEGFA expression in HCC cells. (A) Genomic tracks for ChIP-seq around VEGFA and location of promoter (pink area). (B) Analysis of motifs enriched in YY1 ChIP-seq. (C) HepG2 cells were treated with YY1 overexpression vectors and YY1siRNA. Cellular extracts were prepared for ChIP assays with anti-YY1. (D) HepG2 cells were transiently transfected with VEGFA-dependent reporter gene plasmids. Luciferase activity was measured when cells were overexpressed with or knocked down of YY1. (E) WB analysis showed the VEGFA expression levels in HepG2 cells overexpressed with or knocked down of YY1. (F) ELISAs were used to determine the VEGFA concentrations in the supernatants of the HepG2 cells transfected with YY1 and siYY1. (G) The mRNA levels of VEGFA in HepG2 cells transfected with YY1 or siYY1were measured by qRT-PCR. (H) VEGFA expression levels in YY1-negative and YY1-positive HCC tissues. (scale bar = 20µm). (I) Correlation analysis between YY1 and VEGFA in TCGA database (*R* = 0.56, *P* = 0). \**P* < 0.05, \*\**P* < 0.01.

co-cultured in a 1:2 ratio and formed three-dimensional spheroids. Images were taken with a laser scanning confocal microscope, scale bar = 50µm. (B) Representative image (left) of the formation of HUVECs and HAECs tubes following an incubation with supernatants collected from the indicated cells. Tube formation quantification were analyzed (right). Scale bar = 50µm. (C) HUVECs and HAECs migration were detected after an incubation with supernatants collected from the *(Continued)*

FIGURE 4 | indicated cells. (D) HUVECs and HAECs invasion were detected following an incubation with supernatants collected from the indicated cells. Scale bar = 20µm. (E) WB analyzed pVEGFR2 expression in HUVECs and HAECs treated with conditioned media. (F) Immunofluorescence of pVEGFR2 expression in HUVECs and HAECs treated with conditioned media. Scale bar = 10µm. \**P* < 0.05, \*\**P* < 0.01.

patients were obtained with hospital and the individual consent. All tissues were harvested under the highest ethical standards.

This study was carried out in accordance with the recommendations of Ethical Review Measures for Biomedical Research Involving Human Beings (Trial Implementation), Nankai University Ethics Committee. The protocol was approved by the Nankai University Ethics Committee. All subjects gave written informed consent in accordance with the Declaration of Helsinki.

#### Statistical Analysis

GraphPad Prism 7.0 software (GraphPad Software, Inc., San Diego, CA, USA) and SPSS v19.0 (IBM, Armond, NY, USA) were utilized to perform statistical analyses. Two-tailed unpaired Student's t-test was used for comparing two groups of data. One-ANOVA was used to compare multiple groups of data. Pearson's correlation was used for relevance analysis. Kaplan– Meier analysis was used for survival analysis. Data from biological triplicate experiments were presented with error bar as mean ± SD. Statistical significance was considered at P < 0.05.

## RESULTS

#### YY1 Was Associated With Angiogenesis of HCC

In our previous research, we confirmed that YY1 correlates closely to HCC metastasis and recurrence (28). We analyzed 26 HCC cases by IHC analysis, angiogenesis was showed CD31 staining positive. The expression level of YY1 and angiogenesis were higher in high-degree of malignancy tissues than in lowdegree of malignancy (**Figures 1A,B**). Pearson's correlation and linear regression analysis showed that the expression levels of YY1 and CD31 were positively correlated (**Figure 1C**). The MVD was calculated by IHC staining with anti-CD31. The result showed that YY1 was positively correlated with MVD in HCC (**Figure 1D**).

### YY1 Indicated Tumor Malignancy in HCC

To explore the clinicopathologically relevant feature of YY1, the LIHC dataset from TCGA was analyzed. The expression level of YY1 in HCC was higher than that in normal tissues (**Figure 2A**), which suggests that YY1 may promote the malignant progression of HCC. Further analysis of these data showed that YY1 expression was positively correlated with clinical stages and pathological grades, except stage IV, which may due to few patients in the IV groups (**Figures 2B,C**). Meanwhile, diseasefree survival and overall survival analysis demonstrated that the high expression of YY1 in HCC indicates a poor clinical prognosis (**Figures 2D,E**). These results suggested that YY1 promotes the malignant progression of HCC.

### YY1 Binds to VEGFA Promoter to Enhance VEGFA Expression in HCC Cells

VEGFA is a prominent factor involved in the acquisition of endothelial cell-dependent angiogenesis. In order to elucidate the underlying mechanism that YY1 induces angiogenesis, we detected the effects of YY1 on VEGFA expression. To explore the regulation of YY1 to VEGFA, we analyzed the H3K4me3, H3K27ac, DNase, and YY1 ChIP-seq data of Cistrome Data Browser database. The results showed that YY1 binds to VEGFA promoter (**Figure 3A**). ChIP-seq was used to further analyze the DNA-binding motif of YY1 on the VEGFA promoter (**Figure 3B**). ChIP-PCR analysis was carried out on HepG2 cells by using specific antibodies against YY1, showing the occupancy of YY1 on the VEGFA promoter, which validated the ChIPseq results (**Figure 3C** and **Figure S1**). In addition, the effect of YY1 on the promoter activities of VEGFA were detected by dual-luciferase reporter system. YY1 increased VEGFA promoter activities, whereas YY1 knockdown decreased VEGFA promoter activities (**Figure 3D**). The protein expression in cells (**Figure 3E** and **Figure S2**) and secreted VEGFA (**Figure 3F**) were consistent with the mRNA expression (**Figure 3G**), and the results showed that VEGFRA expression levels increased after YY1 overexpression and decreased after YY1 silence. In addition, we confirmed the correction between YY1 and VEGFA in HCC tissues. IHC staining showed that high YY1 expression levels exhibited extremely strong stain of VEGFA in HCC tissues (**Figure 3H**). Correlation analysis showed that YY1 was associated with VEGFA expression in HCC tissues of TAGA database (R = 0.56, P = 0) (http://gepia.cancer-pku. cn) (**Figure 3I**). In summary, YY1 binds VEGFA promoter to upregulate its transcription activities, protein expression, and secretion in HCC.

### YY1 Stimulated HCC Cell Culture Media Accelerated Endothelial Cells Neovascularization

Angiogenesis was measured by in vitro tube formation assay. To examine the effects of YY1 in HCC cells on HUVEC or HAECs tube formation, we detected the morphologies of vessel-like structure of co-cultured GFP-labeled HepG2 cells and RFP-labeled HUVECs or HAECs in three-dimensional culture. The showed that YY1 and VEGFA could significantly induce the formation of vessel-like structures more than that in normal condition and siYY1 could reduce vessel-like structure compared with siCtrl (**Figure 4A**). Then, the tube formation, migration and invasion were detected in HUVECs or HAECs that cultured with condition medium from supernatants of HepG2 cells transfected with YY1 and YY1 siRNA or treated with VEGFA for 48 h. The results showed that the conditioned medium from YY1-overexpression treatment significantly enhanced HUVEC or HAECs tube formation and knockdown YY1 downregulated

conditioned media. \**P* < 0.05, \*\**P* < 0.01.

tube formation (**Figure 4B**). Conditioned medium from YY1 overexpression or VEGFA treatment promoted HUVEC or HAECs migration and invasion. However, YY1 knockdown inhibited the migration and invasion of HUVECs (**Figures 4C,D** and **Figure S3**). Phosphorylation level of VEGFR2 in HUVECs or HAECs were increased after conditioned medium from HepG2 cell with YY1 overexpression or VEGFA treatment. Conversely, phosphorylation level of VEGFR2 decreased after YY1 knockdown (**Figure 4E** and **Figure S4**). This result was validated by immunofluorescence staining (**Figure 4F**).

### Bevacizumab Blocked the Promotive Effect of YY1 on Tube Formation Through VEGFA

Bevacizumab is an anti-VEGFA monoclonal antibody (30). In HCC cells, the effects of YY1 upregulation on bevacizumab resistance through the VEGFA transcriptional activation were detected. YY1-overexpression or control HCC cells were treated with 250µg/mL bevacizumab for 48 h and supernatant of culture medium were collected. Tube formation assays were performed by treating the HUVECs or HAECs with the indicated cell supernatants. Ectopic expression of YY1 significantly increased the tube formation by HUVECs or HAECs. Bevacizumab

YY1, VEGFA, and pVEGFR2 expression levels in tumor tissue of the Ctrl, YY1, siYY1, and YY1 + bevacizumab groups. Scale bar = 20µm. \**P* < 0.05, \*\**P* < 0.01.

blocked the promotive effect of conditioned medium with YY1 overexpressing on tube formation (**Figure 5A**). In addition, we treated migration and invasion in HUVECs or HAECs treated with the same conditioned medium and the results showed that bevacizumab blocked the promotive effect of conditioned medium with YY1-overexpressing on migration and invasion (**Figures 5B,C**). WB analysis confirmed that bevacizumab inhibited the phosphorylation level of VEGFR2, and bevacizumab also blocked the effect of YY1 on the phosphorylation of VEGFR2 (**Figure 5D** and **Figure S5**). These results showed that bevacizumab blocked the promotive effect of YY1 on tube formation and YY1 overexpression increased bevacizumab resistance by inducing VEGFA transcription.

### YY1 Enhanced Tumor Vascularization in HCC Xenograft Model by Promoting VEGFA Expression

To assess the effect of YY1 on tumor angiogenesis in vivo, nude mice were subcutaneously implanted HepG2 cells. Tumorbearing nude mice in bevacizumab group were treated with 2 mg/kg bevacizumab twice per week by intraperitoneal injection. Compared with the control, YY1 overexpression promoted tumor growth (**Figures 6A,B**), tumor weight (**Figure 6C**) and MVD, which indicated by CD31-positive cells (**Figure 6D**). The opposite results were obtained after silencing YY1 expression. Bevacizumab treatment (2 mg/kg) abrogated the promotive effect of YY1 on the tumor volume and in vivo angiogenesis. Next, the protein expression of YY1, VEGFA, and pVEGFR2 in xenograft tumors were analyzed by immunohistochemistry. As shown in **Figure 6E**, the expression levels of VEGFA and phosphorylation level of VEGFR2 were increased in YY1 overexpressed group and decreased in YY1 silenced group that that in control group. Bevacizumab blocked the upregulated effect of YY1 on the phosphorylation of VEGFR2 in vivo. These results suggested that YY1 contributed to endothelial cell-dependent angiogenesis in vivo through promote VEGFA expression.

### DISCUSSION

Angiogenesis is associated with tumor metastasis, malignancy, and poor clinical prognosis of patients (31). Considering the association of aggressive tumors and angiogenesis, developing targeted therapies according angiogenesis formation and induction mechanism is important.

YY1 promotes epithelial–mesenchymal transition in HCC (28); however, the relationship between YY1 and endotheliumdependent angiogenesis has rarely reported. Analysis of the clinical stage and pathological grade in LIHC cases of TCGA database showed that YY1 expression is a risk factor that determines the survival of HCC patients. Kaplan–Meier analysis revealed that the disease-free survival and overall survival time in YY1-positive HCC patients were shorter than those in YY1 negative patients. The results showed that YY1 expressed highly in tumor tissues than normal tissues and upregulated in HCCs with a high degree of malignancy. YY1 plays an important role in poor prognosis. The results of CD31-positive endothelial celldependent microvessel density showed that YY1 expression was positively correlated with MVD.

The growth and maintenance of angiogenesis were modulated by various growth factor pathways (32). VEGFA is one of most critical growth factors that regulates angiogenesis (33). We also detected the correlation of YY1 and VEGFA in HCC in vitro. YY1 was positively related to VEGFA, which are crucial to tumor angiogenesis, promote endothelial cell proliferation, and increase vascular permeability. YY1 may promote angiogenesis formation by promoting VEGFA expression in HCC. We further found that YY1 interacts with the promoter of VEGFA and enhances its transcriptional activity in HCC cells. YY1 overexpression increased VEGFA transcriptional activity, whereas YY1 knockdown decreased VEGFA expression and secretion. HUVECs co-cultured with conditioned HCC cells or cultured with conditioned medium from HCC cells, and secreted VEGFA from HCC cells promoted tube formation, migration and invasion of HUVECs in vitro. These results showed that the exogenous overexpression of YY1 in HCC cells increased the secretion of VEGFA and continued to activate the VEGFR signaling pathway in endothelial cells. After VEGFA treatment, more receptors were induced to combine with VEGFA, leading to the activation of the VEGFA/VEGFR pathways. Secretion of VEGFA stimulated by YY1 promoted phosphorylation level of VEGFR2 in HUVECs, which activated VEGFR2 associated angiogenesis signaling pathway (**Figure 7**).

VEGFA plays a critical role in angiogenesis, and its expression is upregulated in HCC cells. Blocked of VEGFA signaling inhibits tumor growth and angiogenesis (34). Bevacizumab, the first and most commonly used anti-angiogenic drug, prevents the activation of VEGFR signaling by specifically targeting VEGFA to Ferrara et al. (16), Kerr (35), and Ramezani et al. (36) Although bevacizumab is a molecular-targeted therapy and served as the first-line treatment option for metastatic colorectal cancer, breast cancer, renal cell carcinoma, and advanced non-small cell

lung cancer (32), its resistance limits its therapeutic efficacy in the clinical treatment. Our results showed that bevacizumab blocked the promotive effect of YY1 on angiogenesis and YY1 overexpression increased bevacizumab resistance by inducing VEGFA transcription. In vivo, YY1 promoted tumor growth, and angiogenesis formation also relied on VEGFA. This finding indicates that YY1 promotes angiogenesis formation depending on the transcription activation of YY1 on VEGFA. Therefore, YY1 can be used as a potential target of angiogenesis.

In conclusion, our data indicated that YY1 promotes endothelial cell-dependent tumor angiogenesis by promoting VEGFA transcription of HCC in vitro and in vivo. This work also provides a potential antitumor therapy for inhibiting angiogenesis by targeting YY1 in HCC.

#### DATA AVAILABILITY STATEMENT

The datasets generated for this study will not be made publicly available. There is no omics dataset which requires submission to public databases. The datasets for this study were from TCGA public databases and Cistrome Data Browser database.

### ETHICS STATEMENT

The animal study was reviewed and approved by Laboratory Animal Ethics Committee of Nankai University.

#### REFERENCES


### AUTHOR CONTRIBUTIONS

TS and JM conceived and designed the projects. JM and ZL wrote the manuscript. WY, ZL, RQ, and HA performed the experiments. CY, TS, SCh, YW, YZ, YL, and SCa provided technical and material support. JM, ZL, and XW performed the data analysis.

### FUNDING

This work was supported by grants from National Natural Science Funds of China (Grant Nos. 81572838, 81872374, 81703581, 81871972, 81902441), Tianjin Science and Technology Project (Grant No. 18PTSYJC00060), Chinese National Major Scientific and Technological Special Project for Significant New Drugs Development (Grant Nos. 2018ZX09736- 005, SQ2018ZX090201), The National Key Research and Development Program of China (Grant No. 2018YFA0507203), Postdoctoral support scheme for innovative talents (Grant No. BX20180150), Project funded by China Postdoctoral Science Foundation (Grant No. 2018M640228), and The Fundamental Research Funds for the Central Universities, Nankai University.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fonc. 2019.01187/full#supplementary-material


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Yang, Li, Qin, Wang, An, Wang, Zhu, Liu, Cai, Chen, Sun, Meng and Yang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Role of Matrix Metalloproteinases in Angiogenesis and Cancer

Saray Quintero-Fabián1†, Rodrigo Arreola2†, Enrique Becerril-Villanueva<sup>3</sup> , Julio César Torres-Romero<sup>4</sup> , Victor Arana-Argáez <sup>5</sup> , Julio Lara-Riegos <sup>4</sup> , Mario Alberto Ramírez-Camacho<sup>6</sup> and María Elizbeth Alvarez-Sánchez <sup>7</sup> \*

<sup>1</sup> Multidisciplinary Research Laboratory, Military School of Graduate of Health, Mexico City, Mexico, <sup>2</sup> Psychiatric Genetics Department, National Institute of Psychiatry "Ramón de la Fuente", Clinical Research Branch, Mexico City, Mexico, <sup>3</sup> Psychoimmunology Laboratory, National Institute of Psychiatry "Ramón de la Fuente", Mexico City, Mexico, <sup>4</sup> Biochemistry and Molecular Genetics Laboratory, Facultad de Química de la Universidad Autónoma de Yucatán, Merida, Mexico, <sup>5</sup> Pharmacology Laboratory, Facultad de Química de la Universidad Autónoma de Yucatán, Mérida, Mexico, <sup>6</sup> Centro de Información de Medicamentos, Facultad de Química de la Universidad Autónoma de Yucatán, Mérida, Mexico, <sup>7</sup> Genomic Sciences Graduate Program, Universidad Autónoma de la Ciudad de Mexico, Mexico City, Mexico

#### Edited by:

Erika Ruiz-Garcia, National Institute of Cancerology (INCan), Mexico

#### Reviewed by:

Linda C. Meade-Tollin, University of Arizona, United States Ronca Roberto, University of Brescia, Italy

#### \*Correspondence:

María Elizbeth Alvarez-Sánchez maria.alvarez@uacm.edu.mx

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology

Received: 24 June 2019 Accepted: 20 November 2019 Published: 06 December 2019

#### Citation:

Quintero-Fabián S, Arreola R, Becerril-Villanueva E, Torres-Romero JC, Arana-Argáez V, Lara-Riegos J, Ramírez-Camacho MA and Alvarez-Sánchez ME (2019) Role of Matrix Metalloproteinases in Angiogenesis and Cancer. Front. Oncol. 9:1370. doi: 10.3389/fonc.2019.01370 During angiogenesis, new vessels emerge from existing endothelial lined vessels to promote the degradation of the vascular basement membrane and remodel the extracellular matrix (ECM), followed by endothelial cell migration, and proliferation and the new generation of matrix components. Matrix metalloproteinases (MMPs) participate in the disruption, tumor neovascularization, and subsequent metastasis while tissue inhibitors of metalloproteinases (TIMPs) downregulate the activity of these MMPs. Then, the angiogenic response can be directly or indirectly mediated by MMPs through the modulation of the balance between pro- and anti-angiogenic factors. This review analyzes recent knowledge on MMPs and their participation in angiogenesis.

Keywords: angiogenesis and cancer, immune system, metalloproteinases, MMP, MT-MMP

## INTRODUCTION

## Epithelial-Mesenchymal Transition (EMT) in Metastasis and Migration

Currently, cancer research is focused on understanding the functional mechanisms underlying cell transformation and tumor progression that can be used to develop new markers and therapies (1). Cancer metastasis, the final step of tumor progression and the leading cause of cancer morbidity and mortality, involves the spread of cancer cells from the primary tumor to nearby tissues and distant organs; it is mediated by complex molecular changes of in cell cycle regulation (2, 3). The molecular changes that regulate the cell morphology and functions of epithelial cells, that is epithelial-mesenchymal transition (EMT), include the destruction of intercellular relationships and cell-matrix adhesive characteristics, extracellular matrix (ECM) breakdown, and cleavage of basement membrane components by matrix metalloproteinase (MMP) activity modulation. For example, when epithelial cells lose their polarity through EMT, cell-cell tight junctions and adhesive connections are lost, resulting in infiltration and an enhanced migration ability of these cells (4, 5). Therefore, EMT enables malignant cells to become motile and invasive, which constitutes a fundamental requisite for cancer metastasis (6).

On the other hand, angiogenesis, in which MMP participation is well-recognized, was found to be involved in cancer metastasis over 45 years ago. Interest in angiogenesis related to cancer arose in 1968 when it was highlighted that tumors secrete a diffusible substance that stimulates angiogenesis (7). It is now recognized that angiogenesis plays a crucial role in the establishment of cancer and is the rate-determining step in tumor progression (7, 8). Numerous studies have demonstrated the key participation of MMPs along with EMT to promote angiogenesis, infiltration by cancer cells, and metastasis (9– 13). MMPs are a family of zinc-binding metalloproteinases that participate in the degradation of ECM components, including the basement membrane and the tumor surface, resulting in tumor cell migration into the near tissue. Furthermore, MMPs promote tumor growth and spread through the capillary endothelium and neovascularization (14).

Given the relevance of MMPs in diseases such as cancer, this work presents the most representative studies on the subject. We emphasize the role of cytokines and growth factors inducing EMT in various types of cancer together with the role of MMPs. We also analyzed the carcinogenic and angiogenic processes, and with the participation of MMPs, cytokines, and immune system cells in these processes along with the regulation, activation, and signaling pathways of MMPs in cancer cells.

### BIOCHEMICAL PROPERTIES OF MATRIX METALLOPROTEINASES

MMPs, also known as matrixins, are members of the metzincin protease superfamily of zinc-endopeptidases.They display a specific proteolytic activity against a broad range of substrates located on the ECM. Other members of the superfamily include A Disintegrin and metalloproteinases (ADAMs), and ADAMs with thrombospondin motifs (ADAMTSs), which contain a conserved methionine (Met or M) residue adjacent to the active site (15, 16). The first MMP (collagenase/MMP-1) was identified more than five decades ago (17). Since then, a total of 28 members have been named MMPs and given a distinctive numbering, and have been identified in vertebrates. In humans, there are 23 paralogs of MMPs (including a duplicated MMP-23 gene that encodes two identical forms of MMP-23), out of which at least 14 can be found expressed in the vascular endothelium (18, 19). The typical structure of MMPs consists of an N-terminal zymogenic propeptide domain (∼80 amino acids), a metaldependent catalytic domain (∼170 amino acids), a linker region (∼15–65 amino acids), and a C-terminal hemopexin-like domain (∼200 amino acids) (**Figure 1**) (19, 21). MMP classification is traditionally centered on the substrate specificities observed and the common structural domain architecture. The MMP family can be divided into at least six subfamilies: (1) collagenases; (2) gelatinases, (3) stromelysins, (4) matrilysins, (5) MMP membrane-type (MT)-MMPs, and (6) other MMPs. However, since they present a wide range of substrates and different functions, many of these are similar but have a different biological function that has yet to be clarified. We used a classification related to their evolutionary origin to locate MMPs that have not been properly classified (21, 22) (**Figure 2**).

All MMPs are produced as proenzymes and require a proteolytic cleavage under physiological conditions to promote the release of the propeptide domain (zymogen activation) and generate mature MMPs (22). This means that the activity of MMPs is regulated by a post-translational proteolytic cleavage and endogenous inhibitors (15, 21). Nevertheless, efforts to define the substrate recognition profile by MMPs have resulted in substrate selectivity conferred by key subsite interactions (P3, P1′ , P2′ , and P3′ ) with a motif sequence specificity "P-X-X-|-L-X-X," even though combined frequencies of subsites have been observed. It is known that subsite P3 maintains a high preference for Pro; still, many MMPs favor small residues (Ala/Val/Ile/Leu) and less frequent aliphatic residues. While subsite P1′ maintains hydrophobic residues with preferences for Leu/Ile/Val/Met, subsite P2′ maintains preference for Ile/Val, Glu/Asp, and Lys/Arg/His depending on the MMP. Finally, subsites P3′ and P2′ are inconsistent in all MMPs with any preference for Gly and Ala (26, 27). Therefore, the ability to recognize a wide variety of substrates selected by profile signatures by MMPs involves the peptide hydrolysis of latent protein targets, located on the ECM and the surface of the cell membrane.

Moreover, the MMP catalytic domain of the metzincin clan of metalloendopeptidases shares a general zinc-binding signature as core of the catalytic reactivity; the signature conserved sequence is the H-E-X-X-H-X-X-G-X-X-H/D region. Additionally, the conserved M residue of the superfamily is located on the methionine containing turn (Met-turn) which is part of the catalytic region and likely has structuralstability functions; nevertheless, the strict conservation of this residue remains unclear (28, 29) (**Figure 1**). All MMPs differ in expression, localization, substrate profile specificities, and structural organization. For further details about the structure and function of MMPs see (14, 15, 30).

### CANCER AND ANGIOGENESIS

Angiogenesis is a process by which new blood vessels or capillaries grow from the preexisting vasculature, and it is necessary for diffusion of nutrients and delivery of oxygen for tissue metabolism or cells involved in wound healing, myeloid and stromal cells. New blood vessels require the dismantling of endothelial lined vessels via the "sprouting" of endothelial cells (ECs), expanding the vascular tree (31). Moreover, the neo-vessel networks play more complex roles in diverse tissues such as the endometrium during the menstrual cycle, implantation, and endothelial cell migration out of the existing blood vessels (32). Given the complexity of a process as angiogenesis, the vascular endothelial growth factor, VEGF (VEGF-A), plays a remarkable role in signaling through the VEGF receptor-2 (FLK1) which induces angiogenesis in both health and disease processes. VEGF activity is enhanced by VEGF co-receptors, such as NRP1 and NRP2. In contrast, the loss of VEGF results in the interruption of vascular development. Placental growth factor (PlGF) is

FIGURE 1 | Structure and architectures of MMPs. The selected Protein Data Bank (PDBs) structures are comprehensive (when possible) full-length peptides found in the available coordinates files, all structures were overlapped at similar positions. For every structure, the propeptide domain and triple-helical collagen peptide appear in yellow, while the catalytic domains (right) appear in black, and hemopexin domains (left) in white. (A) MMP-1 family (collagenases and stromelysins) is represented by the structure of the MMP-1 from Human (PDB: 4AUO) in complex with triple-helical collagen peptide. Family members: MMP-1, MMP-8, MMP-13, MMP-3, MMP-10, MMP-12, MMP-20, and MMP-27. (B) Gelatinases family is represented by the full-length structure of the inactive MMP-2 with propeptide from Human (PDB: 1CK7). The additional fibronectin type II domains appear in white and are located under the catalytic domain (black). Family members: MMP-2 and MMP-9. (C) MT-MMPs transmembrane type I family. Represented by two structures mixed in two models of MMP-14 (MT1-MMP) from Human (PDBs: 2MQS and 3MA2). Models were built by the superposition of the homologous structure of MMP-1 (PDB: 4AUO). 2MQS structure is a complex of the hemopexin domains with triple-helical collagen peptide; 3MA2 structure is a complex of the catalytic domain with TIMP-1 inhibitor. The models show the hypothetical MMP-14 with hemopexin and catalytic domains in complex with TIMP-1 and triple-helical collagen peptide. The structure of helical membranal fragment is unknown (542–562) and the structure of the cytoplasmatic tail of the C-terminal fragment (563–582) is available in a complex with the FERN domain from Radixin (PDB: 3X23, structure not represented). Family members: MMP-14 (MT1-MMP), MMP-15 (MT2-MMP), MMP-16 (MT3-MMP), and MMP-24 (MT5-MMP). (D) Matrilysin family (shortest MMPs). Represented by the full-length structure of the inactive MMP-7 with propeptide from Human (PDB: 2MZE). This family lacks hemopexin domains. Family members: MMP-7 and MMP-26. (E) Global MMPs architecture by families. Families (a–d) are represented from (A,D). (e) is the MMP stromelysins type 3 family (structures available but not complete); the architecture is similar to that of MMP-1 family. Family members: MMP-11 (stromelysin 3), MMP-21, MMP-28 and MMP-19 (evolutionary close to MMP-11 and MMP-7). (f) is the MT-MMP GPI (Glycosylphosphatidylinisotol) anchored family (structures not available), the architecture is similar to that of MMP-1 family and closely related to stromelysin type 3 family, but it is attached to the membrane by the GPI. Family members: MMP-17 (MT4-MMP), MMP-25 (MT6-MMP). The (g) family is represented by the MMP-23 (structures not available) and shares the catalytic domain with other families; the architecture is different on the N-terminal of the catalytic domain, containing a type II helical membrane fragment. On the C-terminal are an ShKT (Stichodactyla toxin) domain (with potential channel-modulatory activity) and an Ig-like (Immunoglobulin) C2-type domain that mediates protein-protein interactions. Cyt: cytoplasmatic domain, PD: Propeptide domain, TD: transmembrane helix, FD: Fibronectin type-II domains, CAT: zinc-dependent metalloproteinase domain, Ig: Ig-like C2-type domain and ShKT type domain. All figures were made with VMD (Visual Molecular Dynamics) (20).

a cytokine VEGF homolog that stimulates angiogenesis and various types of cells, such as myeloids and stromals cancers, in addition to activating tumor cells, while their inhibition improves cancer treatment (33).

Collagenases (MMP-1, −8, and −13) are proteins associated with angiogenesis, and their loss leads to the irreversible rupture of the matrix. Type IV collagen participates in cell endothelial migration in the interstitial stromal spaces. It is known that the tissue inhibitors of metalloproteinases (TIMP-1, TIMP-2, TIMP-3, and TIMP-4) regulate them, playing a key role in angiogenesis regulation by inhibiting neovascularization (34).

In adults, angiogenesis is initiated only under inflammation or hypoxic conditions (35). In the early proliferative stage, vascular repair must predominate to control bleeding by vasoconstriction and coagulation. During menstruation, the endometrium is expelled if the ovule is not fertilized. Women who suffer from

include matrilysins, the GPI-anchored MMPs, and other metalloproteinases as MMP-11 (a stromelysin) and MMP-21 (an MMP with a specific function in embryogenesis). (3) Evolutionary group 3 (Figure 1C) includes the MT-MMP trans-membrane type I family (MT1-MMP, MT2-MMP, MT3-MMP, and MT5-MMP). All three groups share a basic architecture with PD-CAT-HD domains array with a few additions or deletions, as matrilysins. The shortest MMPs without HD domain (group 3) contain a transmembrane type I helix and cytoplasmic domains after the HD domain (Figure 1). (4) We added evolutionary group 4 that includes the MT-MMP transmembrane type II family (Figure 1E) with MMP-23A and MMP-23B proteins. MMP-23A gene is considered a pseudogene produced by duplication of the MMP-23B gene. Sources: "GeneCards: the human gene database" (24) and Uniprot databases (25).

endometriosis show aggressive angiogenesis in the peritoneal cavity (36).

On the other hand, several studies have established the importance of transmembrane receptors and ligands participating in cell differentiation. Their role in endothelial sprouting during angiogenesis has recently been studied. ECs express several Notch receptors (such Notch1 and Notch4), as well as the Notch1 protein and Notch ligand delta-like 4 (DLL4), which are important signals for vascular development (37). In most of the healthy population, resting ECs showed long half-lives through VEGF activation, Notch signaling, and angiopoietin-1 (ANG-1) and fibroblast growth factors (FGFs) expression (33). Recent knowledge concerning the complexity of angiogenesis indubitably shows the role of the participants in this event and allows for finding applications in anti-angiogenic therapy.

As previously mentioned, angiogenesis is a normal development and part of the healing process; however, it is key to tumor branching and arborization under pathological conditions such as cancer. The formation of new vascular networks promotes the growth, maintenance, and spread of cancer (38). During angiogenesis in cancer, alterations have been described at the level of lymphangiogenesis and vasculogenesis, both processes are highly involved in the propagation of cancer cells and an unfavorable prognosis (39).

The accelerated growth of the tumor leads to hypoxic tumor microenvironment, interstitial hypertension, and acidosis. To reverse these adverse physicochemical changes, VEGF-C and VEGF-D are synthesized by the activation of VEGFR-3/2, triggering a rise in diameter and density of the peritumoral lymphatic vessels, favoring the propagation of tumor cells toward sentinel lymph nodes (40, 41). It has been shown that the inhibition of these factors by the use of antibodies decreases lymphogenesis and metastasis in nearby ganglia (42–44). Then, angiogenesis maintains a constant and permanent supply of nutrients for cancer cells that leads to tumor growth. This aberrant revascularization begins after the loss of regulation of inhibitory factors (e.g., thrombospondin-1) and angiogenic promoters (VEGF) (45, 46). Hypoxia-inducible factor (HIF) is one of the first growth factors to initiate the abnormal process of vascular growth and responds to the low oxygen tension in the tumor mass. Subsequently, a wave of growth factors such as EGF, basic and acidic FGF, estrogen, prostaglandin E1 and E2, IL-8, TGF, TNF, neuropilins, and VEGF promotes the formation of a vascular network that ensures the exchange of oxygen and nutrients with the tumor (5, 31, 47, 48). This vascularization process is regulated primarily by VEGF-A/VEGF-1,2 and DLL4 signaling. The activation of ECs also triggers a branching process toward the central region of the tumor (49, 50). This new supply and drainage network that supports the tumor allows the latter to maintain a favorable microenvironment for its growth and dissemination. At present, the tumor niche is considered an independent organ able to maintain itself (51). Additionally, integrin receptors are overexpressed in tumor ECs and play a key role connecting the cell cytoskeleton to the extracellular matrix protein ligands such as arginine-glycine-aspartic acid (RGD). This binding interaction between integrin and protein ligands is an important mechanism during the angiogenesis of tumor endothelial cells (52).

#### MMPs in Cancer Angiogenesis

It is well known that MMPs have been implicated in angiogenesis regulation as well as in the anomalous relationship between cancer and the related processes of angiogenesis, vasculogenesis and lymphangiogenesis. MMPs also have a role in the immune system action in cancer development and progression (**Tables 1**, **2**). The pro- and anti-angiogenic effects of MMPs participate in crucial steps as the ability to degrade ECM or cleave several substrates. Specifically, MMP-2 and MMP-9 give rise to the modulation of the dynamic remodeling of ECM (editing aggrecan, collagens, elastin, fibronectin, laminins, and glycosaminoglycans, and latent signaling proteins), activating and deactivating by proteolytic cleavages releasing biological activities that induce cellular regulation (108, 109). MMP activation can be induced by several angiogenic factors, such as VEGF, basic fibroblast growth factors (bFGF), TGF-α and β, and angiogenin. Specifically, MMP-1 activity promotes the expression of the vascular endothelial growth factor receptor 2 (VEGFR2) and EC proliferation, stimulating serine/threonineprotein kinase MARK2 (PAR-1) and activating the transcription factor NF-κB, suggesting the existence of a mechanism by which MMP-1 stimulates vascular remodeling and angiogenesis (110). Similarly, MMP-7 modulates the VEGF pathway in human umbilical vein endothelial cells (HUVECs), degrading soluble VEGFR-1 and in turn promoting angiogenesis (111). TNF-α, IL-8 and other factors with a known pro-angiogenic capacity, stimulate the production of MMP-2,−8, and−9 in ECs and regulate the angiogenesis process (63, 112).

Angiogenesis studies using MMP-8 and MMP-2 knock-out mice, show an in vitro reduction of cell proliferation and neocapillary network growth, as well as a decrease in HUVEC migration and poor in vivo angiogenesis. Interestingly, ischemiainduced neovascularization is also affected by a reduction in ECs and invasive, proliferative, or mobilizing activities of endothelial progenitor cells (EPCs) derived from bone marrow (113, 114). Additionally, it is known that ECs secrete MMP-2 and−9-containing vesicles stimulated by VEGF and FGF-2 and thus regulate the proteolytic activity critical for the angiogenesis-related invasive and morphogenic processes (115). Furthermore, MMP-9 also generates the angiogenic and tumoral repressor, tumstatin by proteolysis of the non collagenous domain (NC1) from the collagen alpha-3(IV) chain. The antiangiogenic properties of tumstatin inhibit EC proliferation and induce apoptosis by interacting with alphaVbeta3 integrin (116).

Among the most studied MMPs participating in angiogenesis is the MMP-14 (MT1-MMP). It significantly contributes to angiogenesis regulation by cleaving ECM molecules as a matrixdegrading enzyme (**Figure 3**). This MMP also acts as a key effector in the production of pro-angiogenic factors such as VEGF. In addition, MT1-MMP interacts with cell surface molecules, such as CD44 and sphingosine 1-phosphate receptor 1 (S1P1), to induce EC migration, and plays a critical role in the proteolytic degradation of anti-angiogenic factors as decorin. Furthermore, evidence shows that MT1-MMP is able to degrade pro-TGF-β and endoglin (TGF-β receptor), suggesting a pivotal role in vessel maturation and angiogenesis, respectively (117) (**Figure 3A**). In addition, MT1-MMP appears to be an essential molecule that determines ECM adhesion and human endothelial cell tube formation through the modulation of MMP-2 expression (**Figure 3B**). This suggests an important role in regulating angiogenesis-related functions in human ECs (118).

### Soluble MMPs in Cancer Angiogenesis

Soluble MMP expression and its effects on cancer stabilization/proliferation are intimately linked via vascular angiogenesis mechanisms that are now well recognized. In this regard, MMP-1 expression has been reported to contribute to the progression of Head and neck squamous cell carcinomas (HNSCC) and the suggest metastatic phenotype of human breast and colorectal cancers, among others (119–121). Interestingly, MMP-1/protease-activated receptor-1 (PAR1) signaling axis has been implicated in tumor angiogenesis and intravasation of carcinoma cells by inducing vascular permeability (122), as well as, hypoxia-regulated MMP-1 expression in metastatic bladder cancer cells, which could be associated to a reactive oxygen species (ROS)-related regulation of the spheroid metastatic phenotype and cell spread (123). The increased expression of MMP-1 in human chondrosarcoma is an important prognostic factor and its function in the spread of tumor cells has been evaluated by silencing assays in which cancer metastasis is impaired but local tumor growth and angiogenesis are enhanced (124). These findings strongly support a role for MMP-1 in the diverse proliferative outcomes of human cancer through angiogenic processes.

Many studies have been published describing the relationship between MMP-2 expression and tumor angiogenesis. One of the earliest reports indicates that IL-8, an angiogenic factor, induces MMP-2 expression and activity in melanoma cells, enhancing their invasion (125). A relationship between MMP-2 expression and stromal support, angiogenesis, invasiveness, and tumor

#### TABLE 1 | Immune system proteins associated to MMPs in angiogenesis and cancer.






growth was demonstrated using an MMP-2-specific inhibitor in a mouse model of bladder cancer (126). Furthermore, an elevated expression of MMP-2 was correlated with VEGF expression in gastric cancer (127) which suggests that this MMP plays a critical role in the progression of cancer through ECM degradation, tumor neovascularization and metastasis.

On the other hand, MMP-9 promotes endothelial cell migration and triggers the angiogenic switch by releasing VEGF during carcinogenesis (128). Decreased expression of VEGF and MMP-9 in medulloblastoma cells that overexpress osteonectin, also referred to as Secreted Protein Acidic and Rich in Cysteine (SPARC), leads to decreased angiogenesis and tumor growth, indicating the pro-angiogenic role of MMP-9 in cancer tissues (129). In contrast, the direct proteolytic cleavage of osteopontin (OPN) by MMP-9 contributes to cancer metastasis, most likely associated with angiogenesis via the regulation of VEGF and angiostatin secretion (130, 131). This model suggests that cancer growth is accompanied by increased vascular permeability, due in part to the expression of MMP-9, leading to the regulation of angiogenic factors, and eventually, neovascularization in cancer tissue.

Studies have revealed that, both MMP-2 and MMP-9 can degrade type IV collagen and are frequently elevated in human cancer. Additionally, a cooperative effect of MMP-2 and MMP-9 was demonstrated in an in vivo experimental model establishing the angiogenic phenotype and invasiveness of tumor keratinocytes (132). The mechanism whereby MMP-2 and MMP-9 activity induces cancer angiogenesis involves the cleavage of latent TGF-β in a CD44-dependent manner, which can promote tumor growth and invasion (133). Together, these results confirm TABLE 2 | Proteins associated with MMPs in angiogenesis on cancer.


the activation of pro-MMP-2. MT1-MMP dimer forms a complex with one TIMP-2 inhibitor, the interaction is not a symmetric array. TIMP-2 binds to a single MT1-MMP monomer by the catalytic domain mediated by the N-terminal. The C-terminal of TIMP-2 binds to the hemopexin domain of pro-MMP-2, thus allowing the prodomain of MMP-2 to access the catalytic domain of the second monomer of MT1-MMP.

the contribution of MMP-2 and MMP-9 to cancer angiogenesis through the degradation of ECM components and the activation of pro-angiogenic factors VEGF and TGF-β in diverse cancer tissues (134). The above findings may explain the central role of the metalloproteinases MMP-2 and MMP-9 in tumor angiogenesis through the induction of pro-angiogenic factors.

Many other MMPs have also been implicated in the incipient establishment of cancer angiogenesis (**Tables 1, 2**). For example, MMP-3 and MMP-7 interact in vivo with osteopontin at tumor sites and may be related to the angiogenic process during tumor development (135). The interaction of diverse MMPs with another class of proteases also contributes to tumor angiogenesis. For example, the overexpression of the serine protease matriptase in human carcinoma cells regulates MMP-3 activity, promoting proliferation and angiogenesis of tumor tissues by degradation of surrounding ECM (136). MMP-13 has also been implicated in cancer angiogenesis promotion through tube formation and neocapillary network development mediated by stimulation of ERK-FAK signaling pathway stimulation. It also stimulates VEGF-A secretion, which contributes to the angiogenic process (137).

It has been widely accepted that MMPs likely play antagonistic roles in regulating cancer angiogenesis. MMP-7 and MMP-9 may be involved in the blockage of cancer angiogenesis by cleaving plasminogen and generating angiostatin molecules (138). Additionally, cross-talk between MMP-7 and MMP-9 leads to the cleavage of insulin-like growth factor-binding protein 2 (IGFBP-2), an angiogenic activator in major aggressive cancers via the transcriptional regulation of the VEGF gene, showing adverse effects in cancer angiogenesis in some tissues (139, 140). It has also been described that MMP-19 is essential to the development of nasopharyngeal carcinoma due to its tumor suppressive and anti-angiogenic functions which can reduce secreted MMP-2 and VEGF (141).

### Membrane-Type Metalloproteinases (MT-MMPs) in Cancer Angiogenesis

Membrane type 1 matrix metalloproteinase (MT1-MMP) is considered a key mediator of cancer progression and metastasis. The overexpression of MT1-MMP in malignant breast cells significantly enhances VEGF production via the Akt and mTOR signaling pathways activated by the MT1-MMP–VEGFR-2–Src complex, which promotes tumor growth and angiogenesis (142, 143). Apparently, a similar mechanism could be involved in glioblastoma angiogenesis (144, 145). Therefore, it is worth noting that this tumor phenotype appears to be associated with the dependence of Akt-mediated signaling pathway, which is stimulated by several angiogenic factors.

Quintero-Fabián et al. Metalloproteinases in Angiogenesis and Cancer

On the other hand, the proteolytic cleavage of semaphorin 4D into its soluble form by MT1-MMP provides a novel molecular mechanism to control tumor-induced angiogenesis in HNSCC (146). In addition, cross-talk between MT1-MMP, MMP-2, and laminin-5γ2 chain fragments contributes to the vasculogenic mimicry of melanoma cells (147). In contrast, the colorectal cancer cells that shed MT1-MMP–mediated endoglin fragments exhibit an anti-angiogenic effect (148). Other studies have demonstrated that MT1-MMP and membrane type 2 matrix metalloproteinase (MT2-MMP) work cooperatively as pro-invasive factors that directly lead to Snail1-triggered cell participation in cancer angiogenesis and metastasis (149). In addition, MT2-MMP is a potential EMT mediator in carcinomas that can degrade adherents and tight junction proteins (150). It has been reported that both MT1-MMP and membrane type 3 matrix metalloproteinase (MT3-MMP) modulate pro-MMP-2 activation, whose angiogenic role in cancer was mentioned above, through inhibition by TIMP-2 and TIMP-3 (151). These data show that a cooperative effect of MT-MMPs during cancer angiogenesis is required together with the angiogenic factors.

Moreover, MT4-MMP expression correlates with EGFR activation, which triggers an angiogenic switch through its catalytic activity and induces the dissemination of cancer cells by disturbing the vessel integrity of the primary breast tumor and promoting hematogenous but not lymphatic metastasis (152–154). Finally, it has been shown that a high MT6-MMP expression in cancer cells is associated with tumor growth; however, further experiments are necessary to determine the exact role of this MT-MMP in the angiogenic process (155).

#### MMPs and the Immune System in Cancer

It is known that uncontrolled angiogenesis, anomalous ECM turnover, decreased growth, and cell migration, as well as inflammatory response, are the result of an imbalance between MMPs activity and their inhibitors, which may be associated with different diseases. Several specific signals are responsible for coordinating the formation, growth, remodeling, and stabilization of blood vessels. It is recognized that excessive growth-promoting signal cues lead to pathological angiogenesis and cancer (15, 156).

In the tumoral microenvironment, there is a complex and dynamically interacting areas involving stromal cells (fibroblasts, myofibroblasts, neuroendocrine cells and immune cells), blood vessels, lymphatic network, and ECM (157, 158), resulting in a tremendous heterogeneity observed in cancer cells. This condition is because tumor cells express and modulate a broad group of signaling pathways including immune modulatory pathways of cytokines and chemokines, which participate in the progression and establishment of cancer cells (**Tables 1**, **2**) (159– 169). In this regard, both cytokines and chemokines induce the expression and activity of MMPs, which in turn allow for the activation of pro-inflammatory signaling pathways as well as the activating receptors, for example, expressed on the surface of T cells and NK cells. Of these molecules act as a powerful mechanism to regulate the immune response. We have analyzed and highlighted several studies showing the involvement of MMPs and their interactions with immune system proteins in angiogenesis and cancer processes as shown in **Tables 1**, **2**.

The presence of secreted extra-cellular vesicles (exosomes) has recently gained importance within the tumor microenvironment (170–172). Exosomes are specific bearers of multiple modulating molecules, such as the antigens for cluster of differentiation (CD), cytokines and chemokines, growth factors (EGF, FGF, PIGS), adhesion proteins (L1/CD171), nucleotides (non-coding RNA, miRNA), and metalloproteinases. Exosomes activate and modify the activity of diverse proteins such as immune proteins and receptor ligands into the circulation by proteolytic cleavage, playing a role as effectors and regulators to promote crosstalk between cancer and stromal cells (53, 157, 173–176). In addition, in vitro studies revealed that a type of TGF-β mediated exosome derived from lung cancer cells increase the expression of MMP-2 (104), while immunosuppressive exosome secretion from lymphoblastoid cells induces apoptosis in CD4<sup>+</sup> T cells (177). Therefore, exosomes could function as conversion markers of malignant cells.

Since they are diverse, MMPs influence multiple cellular processes, such as the inflammatory process regulating barrier function and the activity of inflammatory cytokines and chemokines. Chemoattractant proteins such as MCP-1, MCP-2, MCP-3, and MCP-4 are targets for MMP activity, as result the modified MCPs changing their activity from agonist to antagonist and causing inflammation. Inflammation produces immune tolerance and leads to specific micro-environment conditions, exploited by tumors to evade immune cells and enhance progression, angiogenesis and metastasis (78, 171). In the inflammation process, FGF2 expression can facilitate induction of FGF-dependent angiogenesis by mononuclear phagocytes, Tlymphocytes, and mast cells. The mechanism mediated through FGF2 release induces pro-inflammatory molecules, such as IFNα, IL-2, IL1-β, and nitric oxide. During the first phases of the EC angiogenesis process, FGFs (1, 2, and 4) upregulate urokinasetype plasminogen activator (uPA) in vitro and transform plasmin into plasminogen, an activator of MMPs, triggering ECM degradation and the secretion of exosomes containing MMP-2, MMP-9, TIMP-1, and TIMP-2 (156, 178).

Several studies, both clinical and experimental, have shown that elevated MMP (including MT1-MMPs) levels are associated with the modulation of tumor progression. In brain tumors, growth factors and cytokines modulate the activity of several MMPs. Additionally, it has been observed that MMP-2 positive tumor cells in patients are correlated with low mean survival (54, 179, 180).

Furthermore, in the context of immune cells, there are tumor-associated cells that contribute to the synthesis and upregulation of MMPs (70, 181–183). Importantly, tumorassociated macrophages (TAMs) secrete membrane-bound or soluble proteases, such as MMP-2, MMP-9, and MMP-12, which are involved in ECM degradation and promote the infiltration of tumor-associated blood vessels (184). Macrophages are known to promote cancer initiation and tumor development in an inflammatory environment (185). Bone marrow-derived myeloid cells are also involved in the process, through active regulation of blood vessel formation and maintenance in tumors (186).

Furthermore, accumulated evidence shows that primary tumors can recruit immune cells, such as MMP-9 positive neutrophils, B cells, and M2 polarized macrophages to produce tumor-associated immune cells, which are known to contribute to neovascularization by supplying MMP-9 and other MMPs (59–61, 90, 183, 187). Although most of the published studies consider that M2 macrophages produce high amounts of IL-10, IL-1β, VEGF and MMP, additional subsets have been described with different proportions of cytokines related to cancer microenvironment. A subset of high M1 and low M2 infiltration macrophages are associated with improved patient survival in non-small-cell lung cancer (188), while the activation of M2 macrophages is correlated with a negative prognosis in cancer progression (189). It is therefore important to reorient the associated functions of the M2 macrophage subset to stop and kill cancer cells.

Finally, the molecular role of MMPs in the immune system and cancer is to modulate a series of latent signaling proteins located in ECM, including cytokines and growth factors such as quiescent TGF-β forming a complex with TGF-β-binding protein-1 in ECM. Thus, TGF-β modulates MMP expression, resulting in a bidirectional regulatory loop enhancing TGF-β signaling and promoting cancer progression (133, 190–194). Another mechanism observed is MMP-9 activity, which truncates IL-8 (1–19, 21, 22, 26–52, 63, 108– 135) into more active chains, altering the function of the receptor and improving its biological activity, resulting in greater chemotaxis for neutrophils than the intact form of theof cytokine (195, 196).

Accordingly, immunomodulatory mechanisms of MMPs, cytokines, receptors, and growth drivers are involved in the development and progression of several types of cancer.

#### THERAPEUTIC PERSPECTIVE OF MMPS

MMPs and their inhibitors TIMP, control a wide variety of physiological processes. They constitute promising pharmaceutical targets for inhibition and other metastatic processes.

Currently, monoclonal antibodies are possible candidates to inhibit the activity of MMPs (MMP−14,−12,−9, and−2). However, studies have only managed to identify antibodies against MMP-9 activity, which has biological functions and not for the MMP−14,−12, and−2 (197). In prostate cancer, MMP-9 may amplify local angiogenesis by cleaving membranebound VEGF. Therefore, VEGF is a candidate to be blocked and controlled and to prevent the activation of the androgen receptor (AR)/phosphatidylinositol 4-phosphate 5-kinase type-1 alpha (PIP5K1α)/AKT/MMP-9/VEGF signaling axis required for cell survival and invasion of metastatic tumors (198). Similarly, the effect of phytochemicals on MMP-2, MMP-9, and their tissue inhibitors (TIMPs) has been tested in breast cancer, with no alterations observed in vitro (199). TIMP-3 has been another important target of study regarding the inhibition of cancer cell migration, invasion, and metastasis in vitro and in vivo by natural products (200).

Recently, the effect of MMP-2 gene silencing in normal and MCF-7 cells exposed to the irradiation has been studied. It is known that this MMP leads to the degradation of basement membranes; however, the differential response to DNA damage silencing the MMP-2 gene in normal and MCF-7 cells may be attributable to ROS generation (201). In addition, thrombospondin-2 (THBS2) is a target gene of microRNA-93- 5p (miR-93-5p) and THBS2 is closely associated with ECM and MMP-2 and 9. This MicroRNA is involved in the progression of malignant tumors and is highly expressed in cervical cancer tissues and cells. Thus, the THBS2/MMP signaling pathway is relevant for more studies in clinical trials (202). Moreover, the combined therapy for glioma treatment using temozolomidemarimast (a specific alkylating agent and an MMP inhibitor, respectively) results in tumor cell progression and invasiveness. An alternative treatment proposed in an in vitro study uses a combination of temozolomide and compounds 1 and 2 of N-Oisopropyl sulfonamido-based hydroxamates (MMP-2 inhibitors) to inhibit cancer cell invasiveness and viability (203).

All these studies represent advances in cancer drug development and cancer therapy, with a focus on the control of MMPs and the proteins with which they form complex networks of multifunctional interactions to modulate the signaling pathways that deviate during the development of metastatic cancer. Importantly, the emerging combined clinical therapy mitigates the side effects of existing treatments and raises the anticancer efficacy of chemotherapeutic drugs.

#### CONCLUSION REMARKS

In this work, we have highlighted the role that MMPs play in the cancer and its interaction with growth factors, inhibitor proteins, and the EMT process. The activity of MMPs is involved in the degradation, remodeling, and exchange of ECM, which, under normal physiological conditions, contributes to homeostasis as part of an extensive network of extracellular tissue modulation. In cancer, homeostasis is modified, leading to localized abnormal physiological conditions that modify this extensive network of extracellular tissue modulation.

The increase in MMP activities, as an abnormal process, is a way of producing/inducing an erroneous metabolic cascade. Erroneous metabolic cascades are signals that trigger the emergence of complex abnormal cell pathways, which give rise to tumor/cancerous phenotype cells. In this regard, the transformation into tumor/cancerous phenotypes suggests an exacerbated adaptive survival process. MMPs are not the only elements of this extensive network of extracellular tissue modulation; others such as TIMP proteins, which modulate MMPs, ADAMs and ADAMTSs (15), play a role in this anomalous process as numerous regulatory branches of the network do, namely interleukins.

Although the information on the role of MMPs in cancer is very broad and the way these proteins are expressed is well known, we observed a significant lack of data at the fine level of the relationship between MMPs and the continuity of both the normal and altered signals that positively modulate the carcinogenic process. MMPs produce modulatory elements that remain unclear and, considering that the ECM is a complex array of proteins, fibers, and carbohydrates in different tissues, there may be several variants that generate the loss of homeostasis, causing the diverse cancerous processes observed.

In the angiogenesis process, MMPs are well-known key factors involved in ECM degradation that induce angiogenesis initiation in both physiological and pathological processes. However, the experimental evidence thus far demonstrates that MMPs also play a decisive role in the activation of pro-angiogenic and, in some cases, anti-angiogenic factors in cancer tissues. Thus, MMPs can be considered angiomodulators, which could control new vessel formation necessary for cancer growth, progression, and spread. Therefore, we speculate that MMPs participate in cancer angiogenesis in a cell context-dependent manner.

Most of the experimental data regarding MMP participation in cancer development, vascular endothelium processes, other epithelia (such as periodontal), and inflammatory processes allow us to assume that MMPs are proteins that carry out a type of external cellular regulation/signaling on the ECM. These proteins possess a different regulatory action mechanism that is complementary to other mechanisms such as ligand-receptor signaling pathways. The MMP mechanism is based on editing macromolecules by proteolysis, mainly anchored to the ECM.

It is evident that the proportion of MMPs and other macromolecules (cytokines, grown factors, fibers such as integrins, polysaccharides, and others) in the ECMs of different tissues in normal conditions are metabolically, microenvironmentally, and epigenetically balanced for their functions in each type of tissue. There are tissue-specific proportions, although ECMs have a high degree of heterogeneity.

Inspecting our concept of MMP participation in ECM in the literature, we found an excellent review of ECM with similar concepts (204). The ECM is a highly dynamic system in constant remodeling and is undoubtedly an extension of communication/modulation/signaling among cells located outside the plasma membrane, where MMPs participate in protein editing by providing post-translational modifications.

On the other hand, EMT is a biological process aimed producing mesenchymal phenotype cells from epithelial cells. Its inverse process, mesenchymal-epithelial transition (MET), is carried out with the participation of ECM elements. EMT and MET lead to normal tissue regeneration and fibrosis [EMT type 2, according to Kalluri (205)]. Regarding the participation of MMPs in EMT type 3 (abnormal type), it is evident that their participation is associated with errors in the communication,

#### REFERENCES


modulation, and signaling of this process, which is induced and directed by tumor cells.

Evidence suggests that tumor cells induce the uncontrolled upregulation of MMPs, producing a large number of stimulating factors that disrupt EMT and immunological processes that prevent tumor cell elimination and migration. The upregulation progresses to generating anomalous tissue-specific type signals.

It is known that tumor cells have extensive heterogeneity in their metabolism and phenotype relative to normal tissue across cancer types. Furthermore, these abnormal signals coming from the tumor cells are tissue-specific, leading to the adaptation to the microenvironment where they developed. Finally, an interesting observation of the MMP family is the large, robust specificity profile, which suggests that its role is controlled in a tissuespecific manner; that is, MMP types are expressed accordingly to the regulatory proteins needed for the tissue.

However, more research efforts are needed to determine when abnormal signals begin, what the determinants are, and how microenvironmental tissue-specific conditions can lead a cell to change its metabolism and phenotype. In addition, the question remains: When does the high expression problem of MMPs become a problem metabolically? Although MMPs do not seem to be the cause of the appearance of tumor cells, they induce tumor development because they are targets to regulate development, contributing to increased invasiveness and growth of metastatic tumors.

#### AUTHOR CONTRIBUTIONS

MA-S conceived the idea and scripted the basis of the manuscript. RA and MA-S had equal contributions and a role in the design, analysis, and writing of the article. RA contributed in entirety to the design of the **Figure 1**. SQ-F conceived the idea. SQ-F and RA designed of the **Figure 2**, **Table 2**, and drafted the manuscript. RA conceived and designed the **Figure 3B** and participated with JT-R, VA-A, JL-R, MR-C, and MA-S in the designed the **Figure 3A**. EB-V participated in its construction. All authors reviewed the literature, critically reviewed the manuscript, and approved the final version.

#### FUNDING

This present study was supported by the Autonomous University of Mexico City (UACM) and by National Institute of Psychiatry Ramón de la Fuente Muñiz (INPRFM).


of the p21WAF1 expression. Cell Signal. (2013) 25:2025–38. doi: 10.1016/j.cellsig.2013.06.004


of Angiogenesis. New York, NY: Springer (2017). p. 335–56. doi: 10.1007/978-1-4614-5857-9\_18


as-sociated with tumor recurrence of meningioma. Int J Med Sci. (2013) 10:276–85. doi: 10.7150/ijms.5185


vascular intravasation and metastatic dissemination. Cancer Res. (2013) 73:4196–211. doi: 10.1158/0008-5472.CAN-12-4495


transition and metastasis of gastric cancer via JAK2/STAT3 signaling pathway. Oncotarget. (2017) 8:20741–50. doi: 10.18632/oncotarget. 15119


in angiogenesis. Cytokine Growth Factor Rev. (2005) 16:159–78. doi: 10.1016/j.cytogfr.2005.01.004


heart regeneration. Sci Rep. (2018) 8:1–14. doi: 10.1038/s41598-018- 25490-w


signal pathway. Eur Rev Med Pharmacol Sci. (2019) 23:5113–21. doi: 10.26355/eurrev\_201906\_18175


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Quintero-Fabián, Arreola, Becerril-Villanueva, Torres-Romero, Arana-Argáez, Lara-Riegos, Ramírez-Camacho and Alvarez-Sánchez. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Contribution of Angiogenesis to Inflammation and Cancer

Dolores Aguilar-Cazares <sup>1</sup> , Rodolfo Chavez-Dominguez 1,2, Angeles Carlos-Reyes <sup>1</sup> , César Lopez-Camarillo<sup>3</sup> , Olga N. Hernadez de la Cruz <sup>3</sup> and Jose S. Lopez-Gonzalez <sup>1</sup> \*

<sup>1</sup> Departamento de Enfermedades Cronico-Degenerativas, Instituto Nacional de Enfermedades Respiratorias "Ismael Cosio Villegas", Mexico City, Mexico, <sup>2</sup> Posgrado en Ciencias Biologicas, Universidad Nacional Autonoma de Mexico, Mexico City, Mexico, <sup>3</sup> Posgrado en Ciencias Genomicas, Universidad Autonoma de la Ciudad de México, Mexico City, Mexico

During carcinogenesis, advanced tumors are surrounded by both stromal and immune cells, which support tumor development. In addition, inflammation and angiogenesis are processes that play important roles in the development of cancer, from the initiation of carcinogenesis, tumor in situ and advanced stages of cancer. During acute inflammation, vascular hyperpermeability allows inflammatory mediators and immune response cells, including leukocytes and monocytes/macrophages, to infiltrate the site of damage. As a factor that regulates vascular permeability, vascular endothelial growth factor (VEGF) also plays a vital role as a multifunctional molecule and growth factor. Furthermore, stromal and immune cells secrete soluble factors that activate endothelial cells and favor their transmigration to eliminate the aggressive agent. In this review, we present a comprehensive view of both the relationship between chronic inflammation and angiogenesis during carcinogenesis and the participation of endothelial cells in the inflammatory process. In addition, the regulatory mechanisms that contribute to the endothelium returning to its basal permeability state after acute inflammation are discussed. Moreover, the manner in which immune cells participate in pathological angiogenesis release pro-angiogenic factors that contribute to early tumor vascularization, even before the angiogenic switch occurs, is also examined. Also, we discuss the role of hypoxia as a mechanism that drives the acquisition of tumor hallmarks that make certain cancers more aggressive. Finally, some combinations of therapies that inhibit the angiogenesis process and that may be a successful strategy for cancer patients are indicated.

Keywords: inflammation, angiogenesis, carcinogenesis, cancer, vascular hyperpermeability, vasculogenic mimicry, metastasis

#### INTRODUCTION

According to Hanahan and Weinberg, cancer cells demonstrate 10 common properties, including the ability to evade growth suppressors, and avoid cell death, sustained cellular proliferation, replicative immortality, genomic instability, energetic cellular deregulation, the ability to suppress immune destruction, and induce angiogenesis, the ability to invade surrounding tissues and promote metastasis, and the ability to promote tumorrelated inflammation (1). Tumors are comprised of both abnormal and normal cells, and this heterogeneous composition serves to maintain many complex, dynamic, and shifting interactions among the tumor, immune, and stroma cells in the microenvironment (2).

#### Edited by:

Simona Pisanti, University of Salerno, Italy

#### Reviewed by:

Ronca Roberto, University of Brescia, Italy Miguel Ángel Medina, University of Málaga, Spain

\*Correspondence:

Jose S. Lopez-Gonzalez slopezgonzalez@yahoo.com

#### Specialty section:

This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology

Received: 19 August 2019 Accepted: 26 November 2019 Published: 12 December 2019

#### Citation:

Aguilar-Cazares D, Chavez-Dominguez R, Carlos-Reyes A, Lopez-Camarillo C, Hernadez de la Cruz ON and Lopez-Gonzalez JS (2019) Contribution of Angiogenesis to Inflammation and Cancer. Front. Oncol. 9:1399. doi: 10.3389/fonc.2019.01399

**195**

The association between inflammation and cancer is widely recognized. In 1863, Rudolf Virchow (3, 4) reported that some tumors were infiltrated by inflammatory cells, leading to the hypothesis that inflammation is associated with cancer. In 2012, the International Agency for Research on Cancer established that infection with some pathogens, including Helicobacter pylori, human papillomavirus (VPH) variant 16, and Schistosoma haematobium, is associated with cancer. This and other observations support the notion that persistent infection and inflammation are concomitant with the process of carcinogenesis. In addition, chronic sterile inflammation induced by some non-infectious agents, such as asbestos, UV light, and silica crystals, may leads to cancer development (5–8).

Tumors have a high metabolic rate and require a constant supply of nutrients, along with the exclusion of waste material. These processes are successfully achieved through the induction of angiogenesis. To preserve physiological homeostasis, angiogenesis is rigorously linked with the inflammatory processes. However, in deregulated inflammatory processes that lead to chronic inflammation, pathological angiogenesis can be initiated (9). An intimate connection between immune cells and the endothelium occurs during inflammation. In addition, several studies have indicated that immune/endothelium cell interactions are maintained and encourage tumor development.

In this review article, we focus on mechanisms during acute inflammation that lead to vascular hyperpermeability. In addition, the development of pathological angiogenesis during chronic inflammation is discussed, highlighting the preservation of this process during carcinogenesis. Furthermore, how vascular hyperpermeability, angiogenesis, and inflammation work together in the development of cancer is examined. Therapeutic advances for the normalization of tumor vasculature are indicated. Finally, our particular vision in terms of the roles that angiogenesis and inflammation play in tumor development is presented.

#### ACUTE INFLAMMATION/VASCULAR HYPERPERMEABILITY

Inflammation is defined as the physiological response to infectious or non-infectious agents. The process of inflammation is activated in order to remove both damaged tissue cells and the source of injury (10). The overall goal of the inflammatory process is the reparation of damaged tissue in order to restore the typical tissue architecture, thus maintaining cellular/tissue homeostasis. During the inflammatory process, cells damaged by infectious agents, or cellular stress, release endogenous molecules known as alarmins or danger-associated molecular patterns (DAMPs) that translocate to the cell membrane. These DAMPs are then sensed by a wide variety of cells that express distinct pattern recognition receptors (PRRs), including toll-like receptors (TLRs) and nucleotidebinding oligomerization domain (NOD)-like-receptors (NLR) (11–15). In particular, leukocytes (M1 macrophages, monocytes, neutrophils, mast cells, eosinophils, and other cells) link DAMPs to induce the activation of the inflammasome and the NFκB signaling pathway. Subsequently, these cell types release several pro-inflammatory cytokines, such as VEGF, IL-1α, IL-1β, and TNF-α, along with the chemokines IL-8, MIP-1α, and RANTES (16). Other inflammatory mediators, including bradykinin, histamine, thrombin, and fibrinogen, and endotoxins such as lipopolysaccaride (LPS) are also released. The target cells for these cytokines and chemokines, particularly those of VEGF/VEGFR, are endothelial cells, which then induce vasodilatation (edema) and increase vascular permeability (17, 18). In addition, the expression of several adhesion molecules, such as E-selectin, P-selectin, ICAM-1, ICAM-2, and VCAM-1, is initiated (19, 20). The activation of the endothelium as mediated by these factors is important for the passage (transmigration) of inflammatory cells from the blood to the site of damage (21). Studies have indicated that the increased permeability that occurs during the inflammation process is localized to the microvasculature, primarily in the post-capillary vein (22, 23).

### Vascular Hyperpermeability and the Role of VEGF

The endothelial barrier consists of the joining of endothelial cells by diverse lateral cell-cell junctions. These tight-junctions (TJs) involve specific molecules, namely, claudins and occludins, which form a zipper-like structure between cells that controls the paracellular passage of ions and solutes. TJs are found primarily in the blood-brain barrier (22, 24).

Adherens junctions (AJs) are another important union and are formed by cadherins and catenins molecules. AJs serve to maintain the cell-cell adhesive contact. The vascular/endothelial (VE) cadherin mediates homotypic adhesion with the adjacent cell in a calcium-dependent manner. The intracellular domain of the VE-cadherin is anchored to the cytoskeleton by means of various catenins (α, β, γ, and p120 catenins) that comprise the AJ (25, 26). In addition, intracellular catenins also transmit signals for cell-cell communication (26).

Endothelial cells are also tethered to the extracellular matrix (ECM) by focal adhesions mediated by a family of actin-like proteins, including focal adhesion kinase, talin, and paxillin (22, 25).

The hyperpermeability of the endothelium is mediated by cytokines and chemokines during the inflammatory process and is carried out by two transport mechanisms that facilitate the arrival of immune cells to the damaged area (27, 28). In this process, caveolin-dependent vesicles or vacuoles (vesiculovacuolar organelles or VVOs) form transendothelial channels in specialized regions of the plasma membrane (27–29). Through the sequential fusion of the VVOs, transcellular transport is allowed and used to deliver the contents of the VVOs to the extravascular space (25, 27, 28). This transport mechanism serves to carry proteins of 50–100 nm from the luminal area to the abluminal area of the endothelium (27). While the precise activation mechanism is not fully known, it has been reported that exposure of the endothelium to various factors, such as histamine and VEGF-A, results in the activation of VVO transport (25). Ultrastructural studies have suggested that VVOs form grape-like structures with interconnecting vesicles and vacuoles throughout cells (28, 29). In addition, it has been suggested that G proteins and members of the Src tyrosine kinase family are important for the signaling cascade involved in this transport mechanism (24, 25).

Another mechanism of transport involves a paracellular process (22, 26, 28). During this type of event, cell-cell endothelial junctions are temporarily inhibited, with several inflammatory mediators released into the circulation, including histamine, thrombin, VEGF, and pro-inflammatory cytokines (25, 26). Various signaling pathways, including those involving Rho GTPases, MAP kinases, and protein kinases, are then activated by these factors, leading to the interruption of cellcell joints and the migration of phagocytic and other blood cells (25).

Although transendothelial transport occurs during inflammation in order to increase vascular permeability, paracellular transport is believed to be primarily involved in cell migration (22, 26).

Recent studies have indicated the importance of mural cells, including pericytes, smooth muscle cells, and macrophages, in the regulation of permeability (30–33). An excellent review of these aspects has been published by Goddard and Iruela-Arispe (34).

VEGF is the main soluble factor that modifies the endothelial barrier (35–37). This factor is secreted by neutrophils, platelets, macrophages, activated-T cells, dendritic cells, pericytes, and the endothelial cells themselves (38). VEGF was isolated in 1989 by Ferrera from the Genentech group (39). Several homodimeric glycoproteins comprise the VEGF family. In mammals, five members of the VEGF family have been identified, namely VEGF-A, VEGF-B, VEGF-C, VEGF-D, and placenta growth factor (PLGF) (36, 37, 40). As the prototypical VEGF, VEGF-A is considered the most potent stimulator of vasculogenesis and angiogenesis (38). In addition to increased vascular permeability, vasodilatation, and the recruitment of inflammatory cells, VEGF triggers the inhibition of apoptosis and increases cellular proliferation (38).

The biological activity of VEGF is mediated by the high affinity tyrosine kinase receptors VEGFR-1, VEGFR-2, and VEGFR-3. VEGFR-2 is expressed primarily in endothelial cells and its interaction with VEGF-A triggers increased vascular permeability. VEGFR-2 dimerization induces the autophosphorylation of tyrosine residues and the activation of specific signaling pathways, including the PI3K and p38 MAPK pathways (36, 37, 40). In addition, conformational changes induced by receptor dimerization lead to an increase in intracellular Ca2+, the activation of PLCγ and endothelial nitric oxide synthase (eNOS), with the latter resulting in increased production of nitric oxide (NO) (41, 42). In addition, Src kinase activation induces the phosphorylation of VE-cadherin and various catenins, preventing them from anchoring to the cytoskeleton (22, 25, 26).

Increased vascular permeability allows for platelets and immune cells such as neutrophils and monocytes/macrophages to reach the site of tissue damage (17). At the site of damage, platelets then participate in the coagulation process in order to prevent blood loss from damaged vessels (17). Subsequently, neutrophils arrive at the site of damage to eliminate the pathogen by means of reactive oxygen species (ROS) (43). Finally, the monocytes/macrophages arrive to phagocytose dead cells, cell debris, and various compounds of the ECM, including fibrin. In addition, neutrophils are removed by efferocytosis (44, 45). The resolution of the associated tissue damage and the return to a normal tissue structure with proper tissue-specific funcions are the goals of the vascular hyperpermeability associated with inflammation (22).

#### Resolution of Vascular Hyperpermeability

The resolution of inflammation is a highly orchestrated process involving numerous biochemical processes. In order for this resolution to be successful, inflammatory mediators must act on specific targets to initiate a series of events resulting in homeostasis (46, 47). In particular, the actions that must be accomplished are as follows: (i) turning off the recruitment of neutrophils/lymphocytes, (ii) normalization of the cytokine gradient and the apoptosis of neutrophils, (iii) activation of apoptosis signals for leukocytes and the silencing of pro-inflammatory signaling pathways, (iv) efferocytosis by macrophages and the reprogramming of macrophages from classically activated (M1) to alternatively activated (M2), (v) incorporation of myeloid cells into the local population or recirculation by blood or lymphatic routes, and (vi) tissue repair/return to homeostasis and basal permeability (46, 47). The particular molecules responsible for carrying out the above events include the cytokines produced by M2 macrophages and specialized lipids such as lipoxins, resolvins, protectins, and maresins (48). Proteins such as annexin-A1, adrenocorticotropic hormone, galectin-1, and adenosine are also involved (46). These molecules are synthesized by various cell types, including neutrophils, macrophages, and endothelial cells.

Although the mechanisms by which the hyperpermeability of the endothelium returns to the basal state have yet to be completely described, oxidized phospholipids are known to act as protectors of the endothelial barrier (49). At low concentrations, the oxidized 1-palmitoyl-2-arachidonic-snglycerol-3-phosphorylcholine (PAPC) (OxPAPC) inhibits TNF-α production in phagocytes by blocking the NF-κB pathway (49). In addition, OxPAPC is involved in the restoration of vascular permeability through the activation of the GTPases Cdc42 and Rac. This results in increased cortical actin, the stabilization of cell-cell junctions, and the inhibition of paracellular gap formation. Cdc42 and Rac also activate the Ras-associated protein-1 (Rap1) signaling pathway. Rap1 is an important regulator of various cell functions, including cellular polarization, and leads to increased VE-cadherin and β-catenin, as well as ZO-1 and ocluddin. Furthermore, OxPAPC interacts with the 78 kDa glucose-regulated protein GRP78, which is a multifunctional protein found in the endoplasmic reticulum and plasma membrane. This interaction then provides stability to the union of AJs with TJs (49–51).

### ANGIOGENESIS IN CHRONIC INFLAMMATION

The persistence of the harmful agent that induced the inflammation leads to the upregulation of the inflammatory response. As already mentioned, vascular hyperpermeability promotes the presence of inflammatory cells such as monocytes and macrophages. These cells release pro-inflammatory cytokines, including TNF-α, IL-1β, and IL-6 that increase the expression of adhesion molecules and chemokines for further recruitment of T-lymphocytes (52). In these immune cells, activation of signaling pathways such as, NF-κB, MAPK, and JAK-STAT increase cytokines production. The arrival of more immune cells exacerbates the inflammatory response inducing a chronic inflammation. In response to these factors, the endothelial cells promote angiogenesis. The endothelial cells proliferate and migrate to form new capillaries contributing to restoring nutrient levels and facilitating immune cell migration (53).

In this shifting microenvironment, the immune cells gradually modify their cytokine profile sustaining the inflammatory network. In particular, the presence of Th17 lymphocytes in the milieu contributes to the persistence of inflammation. IL-6, TGF-β, and IL-1β are necessary cytokines for Th17 lymphocytes development, these cells secrete IL-17, IL-21, and IL-22. Combination of IL-17 with other cytokines such as IL-6 and IL-8 contributes to the chronicity of inflammation (54, 55).

An example of pathological angiogenesis during chronic inflammation is diabetic retinopathy (56). Angiogenesis in the retina of patients with diabetes is initiated by ischemia produced by chronic inflammation. In addition, the hyperglycemic environment activates a series of events, culminating in increased vascular permeability, the accumulation of extravascular fluid, ischemia, and pathological angiogenesis (57). Some studies have shown high levels of pro-inflammatory cytokines, including VEGF, TNF-α, NO, and IL-6 in the vitreous humor of patients with diabetes mellitus (57).

Another example is prolonged peritoneal dialysis. In this pathology, adipocytes secrete pro-inflammatory cytokines, which culminates in pathological angiogenesis. The association of chronic inflammation and angiogenesis also occurs in inflammatory bowel disease where continuous ulceration and regeneration lead to the development of chronic inflammation and pathological angiogenesis (58).

Further investigation of the association between inflammation and angiogenesis, which can result in a number of pathological conditions, is required for a better understanding of the underlying molecular events in these process. In the future, selected molecules may be useful as therapeutic targets for the reprogramming of homeostasis.

#### ANGIOGENESIS AND INFLAMMATION IN CARCINOGENESIS

As discussed in the previous sections, increased vascular permeability during the inflammatory process is essential for the arrival of immune cells. The vast array of cytokines and chemokines that participate in the inflammatory process serve to activate and recruit immune cells, which also impacts the associated endothelial cells (59, 60).

Currently, the association of inflammation, angiogenesis, and cancer is well-known. Worldwide, 16% of cancers are caused by infections. In addition, 25% of all inflammatory processes are estimated to lead to tumor development (61). Unlike acute inflammation, during chronic processes of inflammation, inflammatory infiltrates consisting primarily of mononuclear cells that produce reactive oxygen and nitrogen species (RONS) are present. Most RONS have unpaired electrons; thus, they are considered free radicals. As such, while RONS are potent microbial agents, they can also cause cell damage when they are released. DNA is particularly sensitive to RONS, which can induce modified DNA structures such as 2'-deoxyribose. In particular, the modified DNA structure 7,8-hydroxy-2' deoxyguanosine (8-oxodG) induces a breakdown in the double and single strand of DNA (62). These alterations can affect cell cycle regulation and lead to an increase in mutation rate. In driver genes, the gain of mutations has been shown to trigger carcinogenesis (63). Although this event in itself is not enough to produce a tumor, the resulting microenvironment favors chronic inflammation which is another factor involved in cancer initiation. Pro-inflammatory cytokines, such as IL-1 and IL-6, and growth factors, such as TGF-β and VEGF, then can activate various signaling pathways, primarily those involving NF-κB and STAT3 (64). It has been demonstrated that both NF-κB and STAT3 stimulate various survival signals in cells, associated with triggering the carcinogenesis process (53, 65). Other soluble factors generated in the inflammatory process include VEGF-A, cyclooxygenase-2, and prostaglandins. The overall effect of all these molecules on the endothelium is crucial for the recruitment of cells, the production of inflammatory mediators, the increase in vascular permeability, and angiogenesis (66, 67).

In 1986, Dvorak made the analogy of the tumor and its associated tumor microenvironment (TME) to a wound that does not heal (68). In this study, the tumor vasculature captured radiolabeled fibrinogen several-fold faster than did control tissue, allowing for an increase in tumor microvascular permeability. This increase in vascular permeability was attributed to the vascular permeability factor, now known as VEGF-A (68). In addition, Dvorak demonstrated that in the phenomena studied, ECM molecules, including laminin, fibronectin, collagen, and proteoglycans, were involved.

Another similarity between tumors and wound healing is the presence of inflammatory infiltrates. During these processes, cells release a plethora of angiogenic factors, including fibroblast growth factor (FGF)2, CXCL8,WNT7b, ANGPT2, IL-1β, IL-6, IFN-γ, CXCL9/10, and MMP2/9 (32).

In 2005, Cao et al. (69) transfected HCT116 colon carcinoma cells and T4 breast cancer cells with a hypoxia-responsive promoter. The cells were then inoculated in a rodent model using dorsal skinfold window chambers; the presence of angiogenesis was demonstrated at a very early stage of tumor development and was hypoxia-independent.

Mizukami et al. reported that inoculation of HIF-1α knocked down by siRNA in colon carcinoma cell lines reduced tumor growth with no alteration of angiogenesis in a CD1 nude mouse model (70). In addition, this same group demonstrated that under HIF-1α-independent angiogenesis conditions, the RAS and NF-κB signaling pathways upregulated the production of VEGF, IL-8, COX-2, and prostaglandin E (70).

Thus, the process of angiogenesis may be occurring at a very early stage of tumor development and not necessarily at the point of the hypoxia-induced angiogenic switch. However, a deeper research in this issue is necessary to design therapeutic schemes in order to impact in cancer patient clinical outcome.

#### ANGIOGENESIS AND INFLAMMATION IN CANCER ESTABLISHMENT

It is widely known that hypoxia is another critical player in the tumor angiogenesis process. Several factors during cancer development contribute to the generation of hypoxia and the resulting VEGF release. The main molecular component of hypoxia-induced angiogenesis initiation is the hypoxia-inducible factor (HIF)-1α. In metazoan organisms, HIF-1α has been shown to play an essential role in oxygen homeostasis. Under normoxic conditions, HIF-1α is continuously synthesized and degraded. However, under hypoxic conditions, HIF-1α is stabilized and accumulates in the cytoplasm where it dimerizes with HIF-1β. The HIF-1α/HIF-1β complexes then control the expression of hundreds of genes, including VEGF (71). Hypoxia, in conjunction with angiogenesis, can also activate other cancerspecific biological pathways. Under hypoxic conditions, tumor cells present a metabolic shift from oxidative phosphorylation to aerobic glycolysis (72). In addition, hypoxia increases cellular proliferation and the avoidance of apoptosis, which contributes to the chemoresistance of tumors (71). Hypoxia also induces the fibroblasts surrounding the tumor, to acquire a cancer-associated fibroblast phenotype, which is associated with the release of bFGF, IL-6, PDGF, and TGF-β and favors a microenvironment conducive to the cellular evasion of the antitumor immune response (73). Hypoxia also induces the epithelial to mesenchymal transition (EMT), which encourages tumor cells motility, and MMP secretion, subsequently leading to an invasion phenotype (74). The angiogenic switch provides more advantages to the tumor than just angiogenesis, leading to the gradual acquisition of several tumor hallmarks, which allow the tumor to develop into more advanced stages (clinically advanced tumor).

### ANGIOGENESIS AND INFLAMMATION DURING CANCER METASTASIS

Metastasis is the leading cause of death from tumors. It can be described as the process by which tumor cells separate from the primary tumor, travel via the blood or lymph, and arrive at a distant site where they can establish a secondary tumor or metastasis.

In terms of a spatial-temporal context, after the angiogenic switch onset, the tumor establishes and grows. The resulting high rate of proliferation and mutagenesis then induces genetic heterogeneity in the tumor. Welch and Hurst proposed four characteristics of metastasis, namely, (i) Motility and invasion, (ii) microenvironment modulation, (iii) plasticity, and (iv) colonization. In addition to providing nutrients and oxygen, tumor angiogenesis also contributes to the metastatic cascade, which involves vasculogenic mimicry and co-option mechanisms (75).

Vasculogenic mimicry is the generation of structures such as channels and tubes, in conjunction with perfusion, and does not involve endothelial cells. The network formed by vasculogenic mimicry connects with blood vessels in order to supply blood and fluids to the tumor mass (76). Tumors that display vasculogenic mimicry are associated with greater aggressiveness and patients with these tumors typically have lower survival rates. In addition, vasculogenic mimicry is considered an evasion mechanism for antiangiogenic therapy (77, 78).

Within the motility/invasion phase of the metastatic cascade, a critical mechanism that allows tumor cells to acquire the necessary skills is the EMT. During embryogenesis, this mechanism is preponderant. However, in cancer progression, the EMT allows tumor cells to develop vasculogenic mimicry. As part of the EMT, VE-cadherin is expressed in tumor cells favoring both vasculogenic mimicry and metastasis. Moreover, inflammation associated with cancer contributes to both vasculogenic mimicry and the EMT (75, 78). Among the primary immune cells contributing to these mechanisms are tumor-associated macrophages (TAMs), which secrete MMPs for the remodeling of the ECM, favoring motility and tumor cells invasion. Furthermore, TAMs release an array of cytokines, including TGF-β, TNF-α, IL-1β, IL-6, and IL-8, which contribute to the activation of the EMT program (79, 80). Additional immune cells involved in these events include tumorassociated neutrophils (TANs) and myeloid-derived suppressor cells (MDSCs). This set of immune cells and the molecules they secrete then activate the PI3K and NF-κB signaling pathways for promoting the EMT and vasculogenic mimicry (81, 82). This perpetual tumor-associated inflammation and the ongoing redundancy of the factors released that gradually modulate the microenvironment undoubtedly impact the process of tumor progression.

A further important aspect is that tumor cells "appropriate" the pre-established vasculature during vessel co-option. This activity is not exclusive of angiogenesis but rather is considered as one mechanism of tumor cell invasion. Vessel co-option has been clinically associated with aggressive tumors, such as melanoma, glioblastoma, non-small cell lung carcinoma, and ovarian cancer (31, 83).

As observed throughout this review, the inflammatory response proceeds, or is intimately involved in the increase in vascular permeability and angiogenesis observed in both physiological and pathological processes. Indeed, the underlying inflammation in the tumor microenvironment promotes angiogenesis. Advantages conferred to the tumor by angiogenesis include an increase in cellular proliferation, metabolic reprogramming, invasion, and metastasis. Tumor angiogenesis also promotes the continuous arrival of immune cells at the site of the tumor. However, the changes in the tumor microenvironment at this step induce the immune cells to develop a phenotype that, instead of activating the antitumor immune response, favors tumor aggressiveness. As part of the tumor microenvironment, endothelial and immune cells, as well as tumor cells, continuously secrete VEGF (84). This growth factor has an immunosuppressive effect on some immune cells. Indeed, VEGF inhibits the maturation of dendritic cells (DC). In addition, it promotes the accumulation of MDSCs through the recruitment of monocytes/macrophages and, in addition with the IL-4 and IL-10 produced by tumor cells, induces the polarization to M2 macrophages (38).

In patients with colorectal cancer and advanced melanoma, a direct correlation between high concentrations of VEGF and Treg cells has been observed (85, 86). In a mouse model and in patients with colorectal cancer, a subpopulation of Treg cells expressing VEGFR-2 that expands with the exposition of VEGF has been reported (87). The tortuous blood vessels that the tumor develops as an outcome of angiogenesis, vasculogenic mimicry, and co-option all serve to prevent cytotoxic T-lymphocytes from reaching the tumor bed and exerting their antitumor action (88). In this case, the immune response acts promoting tumor growth.

### ANTIANGIOGENIC/IMMUNOTHERAPY COMBINATION

It has been reported that antiangiogenic therapy induces the "normalization" of the tortuous blood vessels that occur in pathogenic angiogenesis. Indeed, the combination of chemotherapy and antiangiogenic therapy appears promising and leads to increased survival of patients with cancer. This phenomenon is attributed to the normalization of blood vessels, which allows the chemotherapy drugs to reach the tumor bed (89). In addition, it has been demonstrated that radiotherapy in combination with antiangiogenic therapy leads to blood vessel normalization (90).

The first drug approved by the FDA for the antiangiogenic treatment of solid tumors was Bevacizumab, which is a humanized anti-VEGF monoclonal antibody. Bevacizumab, in combination with chemotherapy, has been shown to increase progression-free survival (PFS) and overall survival (OS) (91– 95). Aflibercept is a recombinant protein known as a "VEGF trap" that can bind all VEGF isoforms, along with PLGF, and inhibit their activities. Patients with metastatic colorectal cancer have been treated with Aflibercept, with resulting increases in PFS and OS. Ramucirumab is a monoclonal antibody against VEGFR-2 that has been tested as a second line of treatment in combination with other chemotherapeutic agents. Sorafenib and Sunitinib are tyrosine kinase inhibitors that block VEGFR-2. In particular, Sorafenib is an inhibitor of multiple kinases and shows antiproliferative, apoptotic, antiangiogenic and antifibrotic properties. Sorafenib has been approved for hepatocellular carcinoma treatment. Sunitinib is also a multiwide inhibitor approved for neuroendocrine pancreatic tumors and metastatic renal carcinoma (91, 96).

With respect to drugs that stimulate the immune system, several inhibitors of the various immunological checkpoints have been approved. The expression of the programmed deathligand 1 (PDL-1) has been reported in tumor cells, macrophages, DC, and MDSCs. These cells bind to the programmed cell death protein (PD-1) on T-lymphocytes and inhibit their effect or function.

The cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) receptor is another immune checkpoint regulated by hypoxia. In an oxygen-free environment, CD8+ T-lymphocytes increase the expression of CTLA-4, which binds to CD80 and CD86 present in antigen-presenting cells, resulting in the inhibition of CD8+ T-lymphocyte activity. PD1/PDL-1 and CTLA-4 represent therapeutic targets where their inhibitory activity can be blocked with the use of antibodies against these molecules. These kinds of immunotherapy favors the increase in Tlymphocytes to the tumor site and promotes their antitumor activity (88, 97–100).

Sustained angiogenesis and cancer-related inflammation share signaling pathways and molecules. New treatment strategies and the development of new drug combinations that inhibit angiogenesis and stimulate the antitumor response will undoubtedly lead to improved cancer treatments and patient survival in the near future.

### CONCLUDING REMARKS

This review was aimed to establish the relationship between inflammation and endothelial activation, which leads to increased vascular permeability and the initiation of angiogenesis. The relationship between these processes was reviewed for both non-tumor and tumor conditions.

In non-tumor conditions, soluble factors secreted by stromal and immune cells impact the endothelium and initiate its activation favoring the transmigration of cells to eliminate the harmful agent. The regulatory mechanisms of the oxidized phospholipids that contribute to the endothelium basal permeability state after acute inflammation were indicated. In addition, pathological angiogenesis during chronic inflammation were discussed.

The relationship between inflammation and angiogenesis in the advanced stages of cancer is supported by numerous studies. However, the few reports describing the association of these processes in the early stages of cancer are mentioned. It has been proposed that immune cells interact along with tumor development. Moreover, it has been suggested in the cancer immunoediting theory proposed by Dunn and Schreiber RD (101), that immune cells may interact with transformed cells for their elimination. When the eradication of the transformed cells does not occur, these cells gradually proliferate, increasing DNA mutations and the number of tumor cells. In this initial stage, more cells of the immune response arrive to the in situ tumor to eliminate only the susceptible tumor cells through their cytotoxic

by the emerging influx of leukocytes through vascular hyperpermeability. Sustained cellular damage may lead to carcinogenesis initiation. According to cancer immunoediting theory, immune cells recruitment might eliminate transformed cells (Elimination phase). However, in this complex microenvironment, some cytokines act as growth factors for transformed cells or in the endothelium increase vascular hyperpermeability and leukocyte transmigration. These immune cells destroy susceptible tumor clones (Equilibrium phase). Tumor development induces metabolic alterations leading to the angiogenic switch; while, immune cell infiltration now promotes tumor growth (Escape phase). At advanced cancer stages, tumor mass viability is maintained by sustained angiogenesis and vasculogenic mimicry. This complex and dynamic environment promotes phenotypic changes into aggressive tumors, which take advantage of the tortuous vascular branches generating metastatic foci. It should be noted that inside the endothelium circle, the three phases of the immunoediting cancer theory are indicated. The intensity of the color represents the gradual activation of the endothelium. Created with Biorender.com.

mechanisms. This premise was presented as the equilibrium phase of the immunoediting theory.

According to our point of view and based on this proposition, the angiogenesis process is required from the early development of an in situ tumor in order to favor the arrival of immune cells. For this purpose, the blood vessels adjacent to the incipient tumor increase their permeability to allow the transmigration of inflammatory cells to the tumor site. Therefore, it can be considered that tumor cell proliferation causes stress on the tumor-cell surroundings and the release of DAMPs. These molecules are then captured by receptors in both immune and endothelial cells which allows the endothelium activation and the arrival of inflammatory cells. The close relationship of these processes results in: (i) tumor cell proliferation, (ii) the release of DAMPs and pro-inflammatory cytokines, and (iii) endothelium activation and the recruitment of more inflammatory cells. This cyclic process gradually increases the region affected; thus, angiogenesis may contribute to the generation of a microenvironment that favors the presence of growth factors, secreted initially by the infiltrated cells and, tumor cell multiplication and genetic instability (see **Figure 1**).

Finally, owing to the high proliferation rate of tumors, hypoxia is induced; and the angiogenesis switch is turned on. The maintenance of this cyclic process further leads to cancer cells and the development of resistance mechanisms and evasion of the immune response. In this stage, the generation of a tumor microenvironment known as tumor-associated inflammation is induced. After this step, the established tumor can initiate the metastatic process, in which vasculogenic mimicry and co-option contribute to mechanisms of invasion and the migration of tumor cells. Tumor angiogenesis results in abnormal vasculature, with unstable, tortuous blood vessels uncovered by pericytes, which alter immune cell infiltration. Many patients with cancer are diagnosed at this advanced stage but due to the level of pathogenic angiogenesis, only the combination of antiangiogenic therapy and chemo/radiotherapy has been shown to increase the OS. A recent therapeutic option includes the combination of antiangiogenic therapy with inhibitors of various immunological checkpoints. This combination appears to "normalize" the abnormal blood vessels and favors the ability of T-lymphocytes to reach the tumor site and exert their antitumor activity (88).

In summary, sustained angiogenesis and cancer-related inflammation share important signaling pathways and molecules. These hallmarks ultimately serve to support tumor development.

#### REFERENCES


Therefore, improving the combination of therapies that inhibit pathological angiogenesis and stimulate the antitumor response may prove to be a successful strategy for the treatment of patients with cancer.

#### AUTHOR CONTRIBUTIONS

DA-C, RC-D, and JL-G organized the entire manuscript, wrote the draft, and revised the last version of the manuscript. AC-R, OH, and DA-C wrote the acute inflammation/vascular hyperpermeability. DA-C, CL-C, RC-D, and JL-G wrote the angiogenesis in chronic inflammation. AC-R, OH, and RC-D wrote the angiogenesis in the carcinogenesis process. **Figure 1** was designed and made by DA-C, RC-D, and JL-G.

#### ACKNOWLEDGMENTS

The authors acknowledge Instituto Nacional de Enfermedades Respiratorias Ismael Cosio Villegas, Universidad Autonoma de la Ciudad de Mexico, and Universidad Nacional Autonoma de Mexico.


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Aguilar-Cazares, Chavez-Dominguez, Carlos-Reyes, Lopez-Camarillo, Hernadez de la Cruz and Lopez-Gonzalez. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Investigation of the Prognostic Significance of Vasculogenic Mimicry and Its Inhibition by Sorafenib in Canine Mammary Gland Tumors

Maria Carolina Mangini Prado<sup>1</sup> , Sofia de Almeida Losant Macedo<sup>1</sup> , Giulia Gumiero Guiraldelli <sup>2</sup> , Patricia de Faria Lainetti <sup>1</sup> , Antonio Fernando Leis-Filho<sup>1</sup> , Priscila Emiko Kobayashi <sup>2</sup> , Renee Laufer-Amorim<sup>2</sup> and Carlos Eduardo Fonseca-Alves 1,3 \*

<sup>1</sup> Department of Veterinary Surgery and Anesthesiology, São Paulo State University—UNESP, Botucatu, Brazil, <sup>2</sup> Department of Veterinary Clinic, São Paulo State University—UNESP, Botucatu, Brazil, <sup>3</sup> Institute of Health Sciences, Universidade Paulista—UNIP, Bauru, Brazil

#### Edited by:

César López-Camarillo, Universidad Autónoma de la Ciudad de México, Mexico

#### Reviewed by:

Paul Dent, Virginia Commonwealth University, United States Sara Caceres, Complutense University of Madrid, Spain

#### \*Correspondence:

Carlos Eduardo Fonseca-Alves carlos.e.alves@unesp.br

#### Specialty section:

This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology

Received: 20 September 2019 Accepted: 03 December 2019 Published: 19 December 2019

#### Citation:

Prado MCM, Macedo SdAL, Guiraldelli GG, de Faria Lainetti P, Leis-Filho AF, Kobayashi PE, Laufer-Amorim R and Fonseca-Alves CE (2019) Investigation of the Prognostic Significance of Vasculogenic Mimicry and Its Inhibition by Sorafenib in Canine Mammary Gland Tumors. Front. Oncol. 9:1445. doi: 10.3389/fonc.2019.01445 Canine mammary gland tumor (CMT) is one of the most important tumors in intact female dogs, and due its similarity to human breast cancer (BC), it is considered a model in comparative oncology. A subset of mammary gland tumors can show aggressive behavior, and a recurrent histological finding is the presence of vasculogenic mimicry (VM). VM is a process in which highly aggressive cancer cells fuse, forming fluid-conducting channels without endothelial cells. Although, VM has been described in canine inflammatory carcinoma, no previous studies have investigated the prognostic and predictive significance of VM in CMT. Thus, this research aimed to investigate the prognostic significance of VM in vivo and the capacity of sorafenib to inhibit VM in vitro. VM was identified in situ in formalin-fixed paraffin-embedded CMT samples (n = 248) using CD31/PAS double staining. VM was identified in 33% of tumors (82/248). The presence of VM was more strongly related to tumor grade than to histological subtype. Patients with positive VM experienced shorter survival times than dogs without VM (P < 0.0001). Due to the importance of the VEGF-A/VEGFR-2 autocrine feed-forward loop in epithelial tumors, we investigated the association between VEGF-A and VEGFR-2 expression by neoplastic tumor cells and the associations of VEGF-A or VEGFR-2 expression with VM. Among the VM-positive samples, all (n = 82) showed high scores (3 or 4) for VEGF-A and VEGFR-2, indicating that VM was a common finding in tumors overexpressing VEGF-A and VEGFR-2. Thus, we cultured two CMT primary cell lines with VM abilities (CM9 and CM60) in vitro and evaluated the anti-tumoral effect of sorafenib. The CM9 cell line showed a half maximal inhibitory concentration (IC50) of 2.61µM, and the CM60 cell line showed an IC<sup>50</sup> of 1.34µM. We performed a VM assay in vitro and treated each cell line with an IC<sup>50</sup> dose of sorafenib, which was able to inhibit VM in vitro. Overall, our results indicated that VM was a prognostic factor for dogs bearing CMT and that sorafenib had an inhibitory effect on VM in CMT cancer cells in vitro.

Keywords: angiogenesis, dog, breast cancer, tubular assay, antiangiogenic drugs

## INTRODUCTION

Canine mammary gland tumor (CMT) is one of the most common tumors in intact female dogs and is a therapeutic challenge due to its metastatic rate (1). In a One Health perspective, CMT was considered a spontaneous model for studying human breast cancer (BC) (2). Thus, studies on CMT can benefit both humans and dogs. CMT and human BC share many clinical and pathological similarities, including hormonal regulation (2). To grow and metastasize, tumor cells require a sufficient supply of nutrients and oxygen (3). The process of forming new vessels from existing ones, known as neoangiogenesis, is induced by hypoxia and the production of proangiogenic factors (4). Among angiogenic factors, vascular endothelial growth factor-A (VEGF-A) overexpression and its receptor (VEGFR-2) play key roles (3, 4). VEGF-A/VEGFR-2 deregulation was previously demonstrated in canine CMT (2) and human BC (5).

Vasculogenic mimicry (VM) is defined as a process used by highly aggressive neoplastic cells to generate vascularlike structures without the presence of endothelial cells (6). VM has been extensively described in various tumors and participates in tumor spread and metastasis (6–9). Many signaling mechanisms are involved in the initiation of VM. Molecules that are involved in this process are being investigated with the aim of developing new strategies for therapeutic targets against cancer (6). Although, the mechanism of VM is not yet clear, studies have found that the ERK-1/PI3K/MMP-2 signaling pathway may be critical. In addition, VEGFR-2 can induce proliferation through activation of the canonical extracellular signal-regulated kinase (ERK) pathway. Therefore, VEGFR-2 expression by tumor cells may be associated with VM formation (10). Common anti-angiogenic drugs primarily target endothelial cells by inducing apoptosis in these cells and reducing the proliferation of aggressive tumors (6). Among antiangiogenic therapies, sorafenib is a tyrosine kinase inhibitor widely used in human medicine (11–13) that was recently used in veterinary medicine (14).

In humans with liver cancer, sorafenib has been shown to effectively inhibit angiogenesis and induce apoptosis, with good antitumor effects (15). Lee et al. (16) used sorafenib to inhibit the development of human BC cell lines and showed effective induction of apoptosis and autophagy, indicating the potential of sorafenib in human patients with BC. However, to our knowledge, there are no previous studies investigating the antitumor effect of sorafenib on canine mammary cancer cell lines. Several researchers have investigated models to study MV in vitro (3, 9, 17). Due to the importance of VM in the development of cancer metastasis and the relation of VM with patient prognosis, this research aimed to verify the role of VM in canine mammary tumors in vitro and evaluate the association between VEGF-A/VEGFR-2 expression in canine mammary carcinoma tumor samples. In addition, we evaluated the inhibitory effect of sorafenib on VM in canine mammary gland tumor cells in vitro.

### METHODS

#### Study Design

This study was performed in accordance with national and international guidelines for the use of animals in research. All procedures were approved by the institutional Ethics Committee for the Use of Animals (protocol number: CEUA 0091/2018). The experiment was designed in two steps. First, we selected cases of canine mammary gland tumor from the archives of the Veterinary Teaching Hospital of São Paulo State University (UNESP) between 2008 and June 2019. These cases were used to evaluate the associations of vasculogenic mimicry with clinical pathological information. The study design is detailed in **Figure 1**.

### Patients

We retrospectively included 248 canine mammary gland tumor-bearing dogs treated with surgery, with or without chemotherapy. Our inclusion criteria were treatment with surgery with or without chemotherapy, clinical information available in patient records, the presence of a paraffin block in the veterinary archive for immunohistochemical evaluation and no chemotherapeutic treatment prior to surgery. Histological classification was performed according to Goldschimidt et al. (1), and tumor histological grading was performed according to Karayannopoulou et al. (18). For clinical evaluation, the patients underwent a complete blood count, abdominal ultrasound and three-view thoracic radiographic examination. The clinical stage of disease was established according to the World Health Organization classification for CMT (stages I–IV), as modified by Sorenmo et al. (19). Patients with at least stage III and tumor histological grade II disease received adjuvant treatment, and patients with metastatic disease at diagnosis were treated with chemotherapy. Clinical follow-up was performed according to Dos Anjos et al. (2).

### CD31-Periodic Acid Schiff (PAS) Double Staining for VM

All procedures for CD31/PAS double staining were performed according to the protocol by (20). Briefly, tissue sections (N = 248) were stained using a rabbit polyclonal anti-CD31 primary antibody (PECAM-1, Thermo Fischer Scientific, Waltham, MA, EUA) for blood endothelial cell identification using a polymer system conjugated with peroxidase as the first staining step. Then, the sections were counterstained with 0.5% PAS and Schiff. The criteria for determining CD31- and/or PAS-positive VM and procedures for positive/negative control were those described by (20). VM was characterized by the formation of tubular or fracture-like structures by tumor cells containing red blood cells with positive CD31 and/or PAS expression (20).

### VEGF-A and VEGFR-2 Immunohistochemistry

Because we found VEGF deregulation by and previous publication (2), we performed immunohistochemistry to detect VEGF-A and VEGFR-2 expression in the 248 tumor samples

used to evaluate VM formation and prognosis. The procedures for VEGFR-2 immunohistochemical detection and evaluation and controls were previously described by our research group (2). VEGF-A immunostaining was performed using a mouse monoclonal antibody (clone VG1, Dako Cytomation, Carpinteria, CA, USA). Antigen retrieval was achieved by incubation in a citrate buffer pH 6.0 in a pressure cooker (Pascal, Dako, Carpinteria, CA, USA), and endogenous peroxidase activity was blocked with 8% hydrogen peroxide diluted in methanol for 10 min. Then, the samples were incubated with the primary antibody overnight, followed by incubation with a polymer system (Envision, Dako, Carpinteria, CA, USA) for 1 h. The samples were incubated with 3,30-diaminobenzidine (DAB; Dako, Carpinteria, CA, USA) for 5 min and counterstained with Harris haematoxylin for 1 min. The blood vessels in the tumor samples were used as an internal positive control. For the negative control, mouse (Negative Control Mouse, Dako, Carpinteria, CA, USA) immunoglobulin was used to stain a new CMT section. All antibodies were cross-reacted with canine tissue provided by the manufacturer. For the immunohistochemical analysis, the evaluators (MCMP and CEFA) were blinded to patient clinical data, histological type and grade.

### Primary Cell Culture and the Anti-tumoural Effect of Sorafenib

The establishment of canine mammary cell cultures followed the previous description published by our research group (21), and all procedures for the establishment, characterization and culture of CM9 and CM60 mammary primary cells were described previously (22). The anti-tumoural effect of sorafenib was determined by an assay based on the cleavage of an MTT salt into purple crystals by metabolically active cells. For this experiment, each cell line was seeded in a 96-well plate specific for cell culture containing DMEM F12 (Lonza, Basel, Switzerland) supplemented with 10% FBS (Lonza, Basel, Switzerland) and 1% penicillin and streptomycin. The cells were maintained for 24 h at 37◦C. After this initial period, the cells were cultured and incubated in medium without serum, and sorafenib was added to the medium at 2, 4, 6, 8, 10, 12, 14, or 16µM for 24 h. For MTT controls, we used untreated cells (basal control) and cells treated with the highest DMSO concentration (control for DMSO toxicity). In the same plate, each dose was tested in triplicate, and each replicate was performed in triplicate (3×3). After a 24-h incubation, 10 µL of MTT labeling reagent was added to each well, and the plate was incubated at 37◦C for 4 h. Then, the cultures were solubilized, and the spectrophotometric absorbance of the samples was detected using a microtiter plate reader at 570 nm.

## In vitro VM Assay and the Sorafenib Antitumor Effect

This experiment was based on two steps. First, we evaluated cell cultures to determine the time point with the most VM formation by the two CMT cell lines (CM9 and CM60). After determining this time point, we treated the cells with sorafenib (2.61µM for CM9 cells and 1.34µM for CM60 cells). All experiments were performed in triplicate with negative controls (cells treated with DMSO).

All experiments were performed with 80% confluent cell cultures. Three-dimensional (3D) cell cultures were prepared in a 24-well plate. In total, 200 µl of Matrigel (Matrigel <sup>R</sup> Growth Factor Reduced (GFR) Basement Membrane Matrix, <sup>∗</sup>LDEV-Free, Corning, New York, NY, USA) was added to each well and air-dried for 30 min at room temperature. Then, the cell cultures were trypsinized, and 50,000 cells were suspended in 500 µl of DMEM without fetal bovine serum and seeded in each well. The cells were incubated in a humidified atmosphere with 5% CO<sup>2</sup> at 37◦C. The cells were evaluated for VM with an inverted microscope at 1, 2 3, 4, 5, 6, and 7 h.

The cells showed the best tubular structure formation at 4 h. Thus, we performed 3D experiments in triplicate as described above. However, we seeded cells in a 24-well plate with sorafenib at the IC50 dose for each cell line. After 4 h, we evaluated VM formation with an inverted microscope, comparing treated cells with control cells. The control cells were seeded in triplicate and treated by adding the DMSO concentration of the IC50 dose for each cell line.

### Statistical Analysis

Clinicopathological data were evaluated in a descriptive way, with the data presented as percentages. We evaluated patient survival in the context of clinicopathological data, including the presence of VM and the expression of VEGF-A and VEGFR-2. Survival curves were generated using Kaplan-Meier analysis. Chi-square or Fisher exact tests were used to evaluate the correlations of VEGF-A and VEGFR-2 expression with clinicopathological parameters. Samples with scores of 1 or 2 were considered to have low VEGF-A or VEGFR-2 expression, and samples with scores of 3 or 4 were considered to have high expression. Statistical analysis was performed using GraphPad Prism v.8.1.0 (GraphPad Software Inc., La Jolla, CA, USA).

### RESULTS

### Clinical Information

Two hundred twenty-three patients (223/248) had malignant mammary gland tumors, and the remaining 25 patients had benign tumors (25/248). Regarding the malignant tumors, carcinoma in mixed tumor was the most common histological subtype (77/223), followed by complex carcinoma (49/223), tubulopapillary carcinoma (24/223), tubular carcinoma (21/223), solid carcinoma (16/223), comedocarcinoma (10/223), inflammatory carcinoma (9/223), malignant myoepithelioma (5/223), micropapillary invasive carcinoma (4/223), carcinosarcoma (3/223), adenosquamous carcinoma (2/223), anaplastic carcinoma (2/223), and mucinous carcinoma (1/223). Regarding the benign tumors, benign mixed tumor was the most common tumor subtype (14/25), followed by simple adenoma (6/25) and complex adenoma (5/25). Patients with inflammatory carcinoma or carcinosarcoma experienced shorter survival times than other patients (P < 0.0001). The complete clinical information can be found in **Table 1**.

Forty-two of the 248 canine mammary samples were not histologically graded since they were samples of benign tumor (25) or a special tumor subtype (17). Regarding the 206 graded tumor samples, grade I tumors were the most frequent (105/206), followed by grade II (60/206) and grade III (41/206) tumors. Unsurprisingly, the patients with grade III tumors experienced the shortest survival times (P < 0.001), followed by the patients with grade II tumors and the patients with grade I tumors. Regarding the lymph node status, in 14 out of 248 patients, lymph node histopathology was not performed. Thirty-two patients had lymph node metastasis at the time of diagnosis. Patients showing lymph node metastasis at diagnosis experienced shorter survival times than patients without lymph node metastasis (P < 0.0001).

#### CD31/PAS Double Staining

Among all CMT samples (N = 248), VM was identified in 33% of the tumor samples (82/248). The presence of VM had a stronger relation with tumor grade than with histological subtype (**Figure 2**). Thus, only tumors with a higher grade (II or III), independent of tumor subtype, presented VM-positive structures. Additionally, the patients that were positive for VM structures experienced shorter survival times than the negative patients (P < 0.0001) (**Figure 2**).

### VEGF-A and VEGFR-2 Immunostaining

Due to the evidence of several pathways involving tyrosine kinase binding, including the VEGF pathway, we evaluated VEGF-A and VEGFR-2 expression in a large number of CMT samples. VEGF-A and VEGFR-2 expression was identified in endothelial and neoplastic tumor cells. Among all tumor samples, 178 (74%) out of 248 were positive for VEGF-A. Sixth-five samples had a score of 1 (73/178), 43 scored a 2 (52/178), 41 scored a 3 (25/178), and 34 scored a 4 (28/178). Interestingly, the patients with relatively high VEGF-A scores experienced reduced survival (P < 0.0001). The VEGF-A score showed a positive correlation with VM. Regarding VEGFR-2 expression, 65 (26%) out of the 248

#### TABLE 1 | Clinical parameters of the 248 dogs used in this study.


\*Histological grading included only, 206 dogs.

CMT samples were negative. Among the positive VEGFR-2 samples (N = 183), 65 out of 183 had a score of 1, 43 scored a 2, 41 scored a 3, and 34 scored a 4. Patients with a VEGFR-2 score of 4 had the shortest survival times (P < 0.0001) (**Figure 2**). In addition, the samples with a VEGFR-2 score of 3 or 4 were also positive for VM. VEGF-A and VEGFR-2 immunohistochemical staining results are shown in **Figure 3**.

## Sorafenib IC<sup>50</sup> and VM in vitro

Since we identified VEGF-A and VEGFR-2 downregulation in CMT tumor samples, we investigated the antitumor effect of sorafenib on our CMT cells. Sorafenib has been shown to affect the viability of primary cell cultures of the CM9 cell line, showing an IC<sup>50</sup> of 2.61µM, and sorafenib has an IC<sup>50</sup> of 1.34µM for CM60 cells. Both cell lines also showed a VM ability in vitro after 4 h (**Figure 4**). The sorafenib IC<sup>50</sup> for each cell line was able to inhibit in vitro VM, and the treated CM9 and CM60 cell lines lacked the ability to form vascular-like structures in vitro (**Figure 4**).

### DISCUSSION

This paper describes the correlations of VM with prognostic factors in female dogs harboring mammary gland tumors. Interestingly, the dogs with tumors that were positive for VM formation exhibited reduced survival times, indicating that VM is an independent prognostic factor in CMT. This feature can be evaluated in HE slides, bringing a new histological tool for determining patient prognosis in CMT. Previously, VM was demonstrated in female dogs with mammary gland tumors (3, 7). However, these previous studies investigated VM only in inflammatory mammary carcinomas (3, 7). Overall, VM was associated with a high tumor grade and an undifferentiated histological subtype. In other types of tumors, such as human hepatocellular carcinomas, VM has been associated with an advanced tumor grade, invasion, metastasis and a short survival time, indicating VM occurs in relatively aggressive tumors (23). In human breast cancer (BC), VM was previously associated with poor patient outcomes and trastuzumab resistance in HER-2-positive tumors (24). Thus, new studies evaluating VM might provide a new therapeutic perspective. Since dogs are considered a model for human BC studies (2), dogs and humans can benefit from comparative oncology initiatives.

We identified by immunohistochemistry a strong correlation between VEGF-A and VEGFR-2 in our tumor samples. In addition, our linear regression analysis demonstrated a dependency between VEGF-A and VEGFR-2 expression. Thus, these findings are evidence that VEGFR-2 expression is dependent on VEGF-A expression by neoplastic cells. VEGFR-2 deregulation induces VM by autophagy (25), cancer stem cell activation (17) and hypoxia (11). Our results indicated that both VEGF-A and VEGFR-2 had associations with VM formation and patient overall survival. Since VEGFR-2 activation leads to the induction of vascular formation (17), VEGFR-2 expression occurs in normal endothelial cells during physiological vasculogenesis. However, cancer cells can also express VEGFR-2 to promote intratumoural vessel formation.

In human glioma patients, VEGFR-2 was implicated as a key protein for VM and associated with a poor prognosis. In dogs, VEGFR expression has been investigated in CMT (2, 26, 27). However, no previous studies associated VM with VEGFR-2 expression. As previously demonstrated in human gliomas (17), we believed that the VEGF-A/VEGFR-2 autocrine feed-forward loop could be involved in VM formation in CMT. Thus, we investigated the ability of a VEGFR-2 tyrosine kinase inhibitor (sorafenib) to prevent VM in vitro.

To investigate cell viability after sorafenib treatment, we determined IC<sup>50</sup> values using an MTT assay. Interestingly, the IC<sup>50</sup> values for our CMT cell lines were lower than those previously reported in the literature for different human cancer cells (28–31). This result reinforces the use of sorafenib in dogs with CMT as a preclinical model for human BC and as a therapeutic option for dogs with relatively aggressive CMT. Sorafenib toxicity and pharmacokinetics were previously investigated in dogs, demonstrating that sorafenib is safe in dogs with cancer (14). Our cancer cell VM structures were evaluated after a 4-h assay, and sorafenib inhibited structure formation. Thus, future clinical trials in dogs can elucidate whether sorafenib is effective in dogs with tumors showing VM.

Several clinical trials have been performed to evaluate sorafenib efficiency in prolonging patient survival; however, the results are controversial (32–36). Overall, the combination of sorafenib with chemotherapy or endocrine therapy has produced clinical improvements in patients. One important limitation of these previous studies is the inclusion criteria limiting the study population to only patients with advanced disease, with no predictive marker selecting which patients will benefit from the therapy (37). In this scenario, our study proposes that breast cancer-affected patients with histological evidence of VM can benefit from sorafenib treatment. However, prior to using VM as a marker favoring sorafenib treatment, a clinical study in CMT-affected dogs is necessary to provide stronger evidence for sorafenib use in clinical practice.

Since dogs with spontaneous canine mammary gland tumor can be an important model of human breast cancer, it is important to perform clinical studies in owned dogs, but it would not be ethical to use sorafenib in these dogs without prior evidence that VM can be inhibited by sorafenib. Thus, our study is the first preclinical study to show evidence that sorafenib can target cells with a VM ability.

### CONCLUSION

Our results strongly suggest that VM is a prognostic factor in female dogs with mammary gland tumors and

FIGURE 4 | Evaluation of in vitro vasculogenic mimicry by two canine mammary gland tumor cell lines (CM9 and CM60). It was possible to observe tubular-like structures in both cell lines after 4 h. After 6 h, both cell lines started to show tubule disruption, and a group of cells had formed at 8 h. The cells treated with sorafenib showed no tubular-like structure formation at 4 h. Additionally, at 6 h, the sorafenib-treated cells had not formed linked tubular structures.

is related to a shortened survival time. VM formation can be induced by VEGFR deregulation, opening a new perspective for treatment with specific inhibitors. We found that sorafenib inhibited VM in vitro and had an antitumoral effect, supporting its use in future clinical trials involving dogs.

### DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the article/supplementary material.

### ETHICS STATEMENT

The animal study was reviewed and approved by Institutional Ethics Committee for the Use of Animals (protocol number: CEUA 0208/2016). Written informed consent was obtained from the owners for the participation of their animals in this study.

## AUTHOR CONTRIBUTIONS

MP, PF, AL-F, and CF-A conducted all in vitro experiments. SM, GG, and CF-A retrieved the paraffin blocks from the pathology service, performed immunohistochemistry experiments, and conducted the survival analysis. MP, RL-A, and CF-A contributed to the experimental design, intellectual input, and data interpretation. MP and CF-A wrote the manuscript. All authors revised the final version of this manuscript.

### FUNDING

The two co-authors received scholarships from the São Paulo State Research Foundation — FAPESP, Grant Nos. #2018/17109- 9, #2015/25400-7, and #2019/26484-0. This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior — Brasil (CAPES) — Finance Code 001. We would also like to thank the research grant from the National Council for Scientific and Technological Development (CNPq) (#422139/2018-1).

## REFERENCES


for patients with HER2-negative locally advanced or metastatic breast cancer. J Clin Oncol. (2012) 30:1484–91. doi: 10.1200/JCO.2011. 36.7771


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Prado, Macedo, Guiraldeli, de Faria Lainetti, Leis-Filho, Kobayashi, Laufer-Amorim and Fonseca-Alves. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Inhibition of Dimethylarginine Dimethylaminohydrolase (DDAH) Enzymes as an Emerging Therapeutic Strategy to Target Angiogenesis and Vasculogenic Mimicry in Cancer

#### Julie-Ann Hulin<sup>1</sup> \*, Ekaterina A. Gubareva<sup>2</sup> , Natalia Jarzebska3,4, Roman N. Rodionov <sup>3</sup> , Arduino A. Mangoni <sup>1</sup> and Sara Tommasi <sup>1</sup>

<sup>1</sup> Clinical Pharmacology, College of Medicine and Public Health, Flinders University, Adelaide, SA, Australia, <sup>2</sup> N.N. Petrov National Medical Research Center of Oncology, Saint Petersburg, Russia, <sup>3</sup> Division of Angiology, Department of Internal Medicine III, University Center for Vascular Medicine, University Hospital Carl Gustav Carus, Technische Universität Dresden, Dresden, Germany, <sup>4</sup> Department of Anesthesiology and Intensive Care Medicine, University Hospital Carl Gustav Carus, Technische Universität Dresden, Dresden, Germany

The small free radical gas nitric oxide (NO) plays a key role in various physiological and pathological processes through enhancement of endothelial cell survival and proliferation. In particular, NO has emerged as a molecule of interest in carcinogenesis and tumor progression due to its crucial role in various cancer-related events including cell invasion, metastasis, and angiogenesis. The dimethylarginine dimethylaminohydrolase (DDAH) family of enzymes metabolize the endogenous nitric oxide synthase (NOS) inhibitors, asymmetric dimethylarginine (ADMA) and monomethyl arginine (L-NMMA), and are thus key for maintaining homeostatic control of NO. Dysregulation of the DDAH/ADMA/NO pathway resulting in increased local NO availability often promotes tumor growth, angiogenesis, and vasculogenic mimicry. Recent literature has demonstrated increased DDAH expression in tumors of different origins and has also suggested a potential ADMA-independent role for DDAH enzymes in addition to their well-studied ADMA-mediated influence on NO. Inhibition of DDAH expression and/or activity in cell culture models and in vivo studies has indicated the potential therapeutic benefit of this pathway through inhibition of both angiogenesis and vasculogenic mimicry, and strategies for manipulating DDAH function in cancer are currently being actively pursued by several research groups. This review will thus provide a timely discussion on the expression, regulation, and function of DDAH enzymes in regard to angiogenesis and vasculogenic mimicry, and will offer insight into the therapeutic potential of DDAH inhibition in cancer based on preclinical studies.

Keywords: DDAH, nitric oxide, ADMA, angiogenesis, vasculogenic mimicry, cancer

#### Edited by:

Laurence A. Marchat, National Polytechnic Institute, Mexico

#### Reviewed by:

Yingjie Chen, University of Minnesota Twin Cities, United States Gro Vatne Røsland, Haukeland University Hospital, Norway

> \*Correspondence: Julie-Ann Hulin julieann.hulin@flinders.edu.au

#### Specialty section:

This article was submitted to Cancer Molecular Targets and Therapeutics, a section of the journal Frontiers in Oncology

Received: 20 August 2019 Accepted: 05 December 2019 Published: 09 January 2020

#### Citation:

Hulin J-A, Gubareva EA, Jarzebska N, Rodionov RN, Mangoni AA and Tommasi S (2020) Inhibition of Dimethylarginine Dimethylaminohydrolase (DDAH) Enzymes as an Emerging Therapeutic Strategy to Target Angiogenesis and Vasculogenic Mimicry in Cancer. Front. Oncol. 9:1455. doi: 10.3389/fonc.2019.01455

### INTRODUCTION

Despite recent therapeutic advances, cancer remains among one of the leading causes of death worldwide, and the development of novel anti-tumor therapies is still a key priority. Clinical and experimental studies have documented the critical importance of an adequate blood supply for local solid tumor growth and distant metastasis (1–4). Furthermore, the ability of tumor cells to induce new blood vessel growth is a determining factor in both tumor size and spread. The process of angiogenesis, involving the formation, sprouting, extension, and remodeling of pre-existing blood vessels, is a well-accepted paradigm for the development of these intra-tumoral vascular networks (5, 6). Anti-angiogenic treatments for solid tumors have received much attention, yet studies have consistently revealed variable benefits among cancers of different origins. Positive results are often modest and not beneficial when long-term survival is considered (7–10).

Vasculogenic mimicry (VM) describes an alternative mechanism by which particularly aggressive tumors can acquire a micro-circulation: this process involves the formation of vessellike networks lined by the tumor cells, effectively mimicking a true vascular endothelium (11–14). Not only does this process occur de novo, without the need for endothelial cells and independently of angiogenesis (15), but the tumor-lined vessels are also able to fuse to the conventional vascular network (16). There is evidence for VM networks in a number of cancers including those of the breast (17), prostate (18), brain (19), and ovaries (20, 21). The presence of these networks is generally predictive of poor survival and increased metastatic potential due to entrance of the tumor cells into the vasculature (17, 22–25). Intriguingly, the use of anti-angiogenic treatments may actually be a driving factor in the development of VM (26, 27), which may be at least partly induced by the resulting hypoxia (28). The presence of VM in cancers therefore represents a highly clinically relevant challenge both from a prognostic and a therapeutic point of view.

The signaling molecule nitric oxide (NO), a small short-lived free radical gas, has a fundamental role in diverse physiological processes across different tissues. Perhaps the most well-studied and established of these is its role in maintaining physiological homeostasis of the cardiovascular system. Research published simultaneously in 1987 by Ignarro et al. and Palmer et al. first identified NO as the endothelium-derived relaxing factor (29, 30). It is now clear that NO is not only a powerful vasodilator, central to the control of vascular tone, and blood pressure (31, 32), but is also critical for inhibition of platelet aggregation and promoting anti-inflammatory effects (33, 34). Importantly, NO is known to participate in vascular permeability and angiogenesis mediated by vascular endothelial growth factor (VEGF) (35). Due to the essential and diverse roles of NO, it is not surprising that altered NO concentrations result in significant pathophysiological conditions. These include numerous cardiovascular disorders, as well as neurodegenerative disorders, inflammatory arthritis, septic shock, schizophrenia, and various cancers, as previously reviewed (36).

The importance of NO in a range of cellular processes is further highlighted by its tight regulation at multiple levels, which is critical for both its spatial and dosage control. Endogenous NO is the product of a two-step redox reaction requiring molecular oxygen and a series of cofactors including flavin mono- and di-nucleotide, calmodulin, nicotinamide adenine dinucleotide phosphate, and tetrahydrobiopterin (37, 38). This biochemical synthesis of endogenous NO is governed by the family of nitric oxide synthase (NOS) enzymes through the stereospecific conversion of the natural amino acid Larginine to L-citrulline and NO. The three distinct mammalian isoforms of NOS are NOS1 (also known as neuronal or nNOS), NOS2 (inducible or iNOS), and NOS3 (endothelial or eNOS), each exhibiting a unique expression pattern and named for their location of initial isolation; nNOS is predominantly expressed by resident cells of the central and peripheral nervous system including both neuronal and non-neuronal cells (39, 40), iNOS is expressed in inflammatory cells and can also be found in many other cell types in response to immunologic or inflammatory agents such as cytokines and lipopolysaccharides (41), and eNOS is predominantly expressed in endothelial cells. There is thus a regulation of NO synthesis that exists at the level of NOS transcription, post-translational modifications and specific cellular expression, as well as metabolic regulation at the level of NOS substrate availability (42). The activity of all three NOS isoforms is also regulated by the competitive inhibitors asymmetric dimethylarginine (ADMA) and monomethyl arginine (L-NMMA), which are ubiquitous endogenous metabolites of protein degradation that compete with the NOS substrate, L-arginine, for binding to the NOS active site (43–48). The two members of the dimethylarginine dimethylaminohydrolase (DDAH) family of enzymes, DDAH1 and DDAH2, are responsible for the degradation of the NOS inhibitors ADMA and L-NMMA (49) and are therefore key components in maintaining homeostatic control of NO.

There is a growing body of literature which demonstrates NO as a molecule of interest in carcinogenesis and tumor growth progression (50–52). In particular, dysregulation of the DDAH/ADMA/NO pathway, resulting in increased local NO availability, is often associated with promotion of tumor angiogenesis, growth, invasion, and metastasis. Increased expression of DDAH enzymes in tumors of different origins has been reported by numerous research groups in recent years, and inhibition of DDAH expression and/or activity in cell culture models and in vivo studies has indicated the potential therapeutic benefit of targeting this pathway (53–56). Additionally, whilst ADMA-mediated regulation of angiogenesis is highly relevant for tumor growth, DDAH enzymes may have dual ADMA-dependent and -independent effects on cancer progression. In this review we revisit the relevance of NO in cancer and provide an update in relation to cancer angiogenesis and VM. We also summarize a pioneering body of evidence for the potentially important expression, regulation, and function of DDAH enzymes in cancer initiation and/or progression. Finally, we discuss and offer insight into the therapeutic potential of DDAH inhibition as a cancer anti-angiogenic agent based on preclinical studies.

#### NITRIC OXIDE AS A CELLULAR MODULATOR OF ANGIOGENESIS

Nitric oxide (NO) is an endogenously and ubiquitously produced free radical gas that is readily able to permeate cell membranes due to its small size and high lipophilicity. The half-life of NO has been estimated to be within the range of 0.1–2s, thus allowing for rapid termination of NO signaling cascades following removal of the initial stimulus (57). Despite its short half-life, NO has a unique ability, as a result of its physicochemical properties, to diffuse over long distances (several 100µ) within milliseconds. In addition, in contrast to conventional biosignaling molecules which act solely by binding to specific receptor molecules, NO manifests many of its biological actions via a wide range of chemical reactions. The precise reaction is dependent upon local NO concentration as well as composition of the extracellular and intracellular environment (58, 59). NO thus acts as a pleiotropic messenger, directly influencing a number of biological processes and pathophysiological conditions (36, 60).

The first physiological role identified for NO was its ability to bind and activate soluble guanylyl cyclase (sGC) in the cGMP signaling cascade (61); to date this remains the only known receptor for NO. Here, NO targets the heme component of sGC which allows for further coupling with cGMP-dependent protein kinase G, phosphodiesterases, and cyclic nucleotide gated channels (62, 63). In addition to inducing immune and inflammatory responses, this binding of NO to sGC mediates relaxation of smooth muscle and blood vessels, with a consequent increase in blood flow (64), prevents leukocyte adhesion and inhibits platelet aggregation thus maintaining vascular homeostasis and preventing atherosclerosis (65). Importantly, a number of studies indicate that NO is vital in promoting angiogenesis (66, 67). Angiogenesis is stimulated by NO production and attenuated when NO bioactivity is reduced, however the exact mechanisms underpinning these processes are complex.

NO is considered an "endothelial survival" factor as it inhibits apoptosis (68, 69) and enhances endothelial cell proliferation (70, 71), migration (67, 72), and podokinesis (73). These events are in part due to NO-mediated (primarily via eNOS and iNOS) increase in vascular endothelial growth factor (VEGF) or fibroblast growth factor expression (71, 74), and suppression of angiostatin production (75). There is a bidirectional interaction between VEGF and NO; VEGF can also promote NO synthesis via PI3 K/AKT-mediated phosphorylation of eNOS (76, 77). NO has also been identified as a regulator of isoforms of the antiangiogenic matricellular protein thrombospondin (TSP) through phosphorylation of extracellular signal-regulated kinase (ERK). Specifically, NO represses transcription of TSP2 (78), and triphasically regulates TSP1 protein expression dosedependently (79). Furthermore, NO facilitates angiogenesis through stimulating the expression of matrix metalloproteinase (MMP). This is thought to be mediated by a cross talk between eNOS/iNOS and MMP via the VEGF receptor/cyclic adenosine monophosphate/protein kinase A/AKT/ERK signaling pathway. Consequently, ERKs upregulate the expression of membrane MMPs, thus favoring endothelial cell migration and vascular tube formation (80–82).

### THE DUAL ROLE OF NITRIC OXIDE IN CANCER

As synthesis of NO is generally a tightly regulated process, aberrant and dysregulated NO production is implicated in numerous pathophysiological conditions. It has been increasingly recognized that altered NO synthesis is associated with cancer initiation and progression, particularly cancer-driven angiogenesis, vasculogenic mimicry, and resulting metastasis. The dichotomous role of NO in cancer has been the subject of several reviews which highlight that NO can exhibit both oncogenic and tumor suppressing behavior depending on cancer type, location and stage, as well as local NO concentration and duration of exposure (50, 52, 83–87).

Modulation of NO concentration appears beneficial in mediating tumor regression and treatment for cancers characterized by reduced NO signaling, and this has been the focus of several research groups in recent years. An increase in NO concentration via the use of glyceryl trinitrate (GTN) reduced hypoxia-induced metastatic potential of an in vitro and in vivo model of murine melanoma (88) and exerted pro-apoptotic effects in colon cancer cell lines (89). Treatment with GTN has also shown potential for the treatment of prostate and small cell lung cancer by increasing sensitivity to chemotherapeutic agents (90–93). Similarly, the NO donor sodium nitroprusside has been demonstrated to suppress cell invasion in in vitro models of prostate and bladder cancer (94) and cell migration of gastric epithelial cells (95). Furthermore, it has shown protective effects due to apoptosis and growth inhibition in models of cervical cancer, pancreatic cancer, lymphoma, and glioma (96–99).

In contrast, other studies have demonstrated that excessive NO production is associated with poor prognosis and increased invasiveness of tumors of the breast (100–105) and with survival, proliferation and dedifferentiation of prostate cancer cells (106, 107). In head and neck cancer, excessive NO correlates with cancer risk and metastatic potential (52, 108, 109), and in colorectal cancer increased NO leads to enhanced angiogenesis and invasiveness (110, 111). Elevated NO concentrations have also been correlated with endometrial, cervical and gastric cancers, and tumors of the central nervous system (112–118). For these conditions, however, there is currently no targeted approach for intervention of NO production available for clinical use.

#### DDAH ENZYMES AS MODULATORS OF NO SYNTHESIS

Together the NOS enzymes share 50–60% homology (119) and are all inhibited by asymmetrically methylated arginines (43–48). Methylarginines are endogenous metabolites of protein degradation and consist of monomethyl arginine (NMMA), asymmetric dimethylarginine (ADMA), and symmetric dimethylarginine (SDMA). They are continuously produced as the combination of two cellular processes: post-translational N-methylation of arginine residues incorporated into proteins, catalyzed by a family of protein methyltransferase (PRMT) enzymes (1–9) (120), and their subsequent release into the cytosol following proteolysis (121). Free methylarginines can then accumulate in the cytoplasm or cross cellular membranes where they are able to exert their biological function of inhibiting NOS enzymes in neighboring cells. Transport of methylarginines across cell membranes is typically controlled through transporters of the cationic amino acid (CAT) family, particularly CAT1, CAT2A, and CAT2B (122, 123). Both ADMA and NMMA inhibit all NOS isoforms, however plasma ADMA concentrations are considerably higher than those of NMMA (46, 124) and as such the relative contribution of NMMA to NOS inhibition has often been underestimated. Whilst ADMA and NMMA both compete with L-arginine for binding to the NOS active site (43–48), SDMA is not a direct inhibitor of NO synthesis. It can, however, reduce the availability of the NOS substrate L-arginine, by competing for transport by the CAT transporters (125).

Different routes of elimination have been identified for all three methylarginines. Two pathways for the metabolism of ADMA and SDMA are: (1) the transamination to asymmetric dimethylguanidinovaleric acid (ADGV) for ADMA and to symmetric dimethylguanidinovaleric acid (SDGV) for SDMA, mediated by alanine-glyoxylate aminotransferase 2 (AGXT2) (126–128), and (2) N-alpha-acetylation, although the enzyme responsible for catalyzing this reaction is still currently unknown (129–131). Conversion to γ-(dimethylguanidino) butyric acid has previously been proposed as a catabolic route for ADMA and SDMA (131), but the significance of this metabolic pathway has not received any further investigation. NMMA concentrations can also be regulated by the enzyme peptidylarginine deiminase 4 (PAD4), which catalyzes the deamination of NMMA residues still incorporated into proteins into L-citrulline (132). Renal excretion is responsible for the elimination of the majority of SDMA, but accounts for only a small percentage of ADMA clearance (<10% in some species) (131, 133–135). Most importantly, ADMA and NMMA are primarily metabolized by DDAH enzymes into Lcitrulline and dimethylamine or monomethylamine, respectively (134, 136). The DDAH/ADMA/NO pathway is summarized in **Figure 1**.

Two DDAH isoforms have been identified in mammals (DDAH1 and DDAH2) and it is estimated that collectively more than 70% of ADMA is metabolized by these enzymes (137). Indeed, global heterozygous deletion of DDAH1 in mice increased plasma, brain, and lung ADMA concentrations by 20% (138). The DDAH isoforms are highly conserved at the amino acid level [62% in humans (49, 139)], particularly with residues important for substrate binding and hydrolysis. DDAH isoforms are also highly conserved across species, with high homology between the human, mouse, rat, and bovine gene sequences (DDAH1: 92%, DDAH2: 95%). While researchers are in agreement with DDAH1 being the key enzyme responsible for ADMA and NMMA metabolism (94, 140), there is conflicting evidence surrounding the metabolic activity of DDAH2.

Several lines of evidence suggest that under normal conditions DDAH1 is the isoform responsible for ADMA metabolism (141). Firstly, the tissues from DDAH1 KO mice do not display any DDAH activity (140). Secondly, silencing of DDAH1 in cultured vascular endothelial cells results in ADMA accumulation and a decrease in NO production, while silencing of DDAH2 has no effect (140). Consistent with this finding, overexpression of DDAH1 in cultured endothelial vascular cells decreases ADMA content and overexpression of DDAH2 does not (142). Purified recombinant DDAH2 was originally reported to metabolize NMMA (49) but following studies have failed to reproduce the metabolic activity of DDAH2 in vitro (143). Fluctuations in ADMA concentrations are observed in response to overexpression and/or knockout of the DDAH2 gene (144–146), but whether DDAH2 affects ADMA concentration via direct metabolism or by indirect regulation of its metabolism still remains unclear. The difficulties in recapitulating DDAH2 activity in vitro may suggest the requirement for additional cofactors or protein-protein interactions, or a missing step in the pathway of ADMA metabolism that is not functional in the cell lysates often used to assess recombinant DDAH2 protein function. Regardless, based on current available knowledge the DDAH1 enzyme appears to be key for metabolism of ADMA/NMMA and thus more relevant in regard to the treatment of cancer through the ADMA/NO pathway.

#### Implications for Angiogenesis

The DDAH enzymes play a key role in homeostasis of the cardiovascular system, and specifically in modulation of angiogenesis and neovascularization. Whilst it appears that the majority of DDAH function is attributed to degradation of ADMA and thus modulation of NO synthesis, ADMAindependent functions of DDAH have also been identified.

ADMA plays a key inhibitory role in the formation of new blood vessels; examples include inhibition of proliferation of bovine retinal capillary endothelial cells (147) and coronary artery endothelial cells (148). Furthermore, in vitro and in vivo studies show that ADMA modulates all the key aspects of VEGFinduced angiogenesis: activation, proliferation, differentiation, and migration of endothelial cells. Fiedler and colleagues demonstrated that increased ADMA concentrations inhibit the VEGF-induced capacity of human umbilical vein endothelial cells (HUVECs) to form tubes on Matrigel by disrupting chemotaxis, migration, protrusion formation, focal adhesion turnover and reducing cell polarity and gap junction intercellular communication (74). In the same study, ADMA was also reported to interfere with activation of Rho GTPases via RhoA activation and Rac1 and Cdc42 inhibition. By inhibiting NO synthesis, ADMA reduced VEGF-mediated phosphorylation of VASP and Rac1 activation in human endothelial cells (74). This is consistent with what has been previously observed in pulmonary endothelial cells (149). Moreover, it appears that ADMA can interfere with the activation of endothelial progenitor cells (EPCs) (150). ADMA supplementation has also been reported to accelerate high glucose-induced EPC senescence, whilst the

NOS) are endogenously inhibited by asymmetrically methylated arginines (ADMA or R-Me<sup>2</sup> and L-NMMA or R-Me). These endogenous inhibitors of the NO synthesis are generated and released in the cytosol as the product of 2 biomolecular processes: the post-translational methylation of arginine residues incorporated into proteins catalyzed by one members of the protein arginine methyltransferase (PRMT) family of enzymes and the release of said methylated residues into the cytosol by proteolysis. Methylated arginine can act as NOS inhibitors solely in their free form. The enzyme responsible for the metabolism of more than 70% of circulating and intracellular ADMA and L-NMMA is dimethylarginine dimethylaminohydrolase (DDAH), which converts ADMA and L-NMMA into L-citrulline and dimethylamine (DMA or Me2-NH2) or monomethylamine (MMA or Me-NH2).

opposite effect was observed with the overexpression of DDAH2 (151). This is in line with association studies showing an inverse correlation between the number of EPCs in blood and plasma ADMA levels in coronary artery disease (150), peripheral arterial disease (152), and after renal transplantation (153). Additionally, increased plasma ADMA concentrations are linked to higher cardiovascular risk and numerous vascular diseases, many of which are associated with low NO output and endothelial dysfunction (154–158).

The generation of heterozygous DDAH1 knockout mice by Leiper and colleagues first demonstrated that DDAH1+/<sup>−</sup> mice exhibited accumulation of ADMA and reduced NO concentrations, leading to vascular pathophysiology such as endothelial dysfunction, structural alterations in the pulmonary vasculature and decreased heart rate and cardiac output (138). Importantly, angiogenesis was significantly reduced in these mice, as assessed by quantification of microvessels sprouting from aortic rings (149) and hemoglobin content in plugs (74). Over-expression of DDAH1 reversed the antiangiogenic effects associated with increased ADMA (74). The more recent generation of global DDAH1 deficient mice further confirmed the importance of DDAH1, but not DDAH2, for ADMA metabolism and in cardiovascular physiology (140). DDAH1−/<sup>−</sup> mice exhibit impaired endothelial cell proliferation and decreased neovascularization (142). The generation of an endothelium-specific DDAH1−/<sup>−</sup> mouse using Tie-2 driven Cre expression demonstrated that intracellular ADMA concentrations are crucial in determining the endothelial cell response. Whilst the angiogenic response was significantly impaired both in vivo and ex vivo, plasma ADMA concentrations, vasoreactivity ex vivo and hemodynamics in vivo remained unaffected (159). Together, these studies further highlight the essential role of DDAH1 in ADMA and NMMA metabolism.

The expression and activity of both DDAH1 and DDAH2 appear to be critical for wound healing and angiogenesis. Overexpression of DDAH1 in endothelial cells resulted in enhanced tube formation when grown on Matrigel and an increase in VEGF mRNA expression; blocking DDAH activity reversed these effects (160). DDAH1 overexpressing mice exhibited enhanced neovascularization after hind limb ischemia (161, 162) and improved endothelial cell regeneration with reduced neointima formation following vascular injury (163). Conversely, DDAH1 knockout mice had reduced endothelial repair and angiogenesis, and impaired endothelial cell proliferation compared with WT mice in a model of carotid artery wire injury. Interestingly, VEGF-expression was reduced in this DDAH1 global KO mouse model via a mechanism that was independent from the NO/cGMP/PKG pathway, and regulated by the Ras/PI3K/Akt pathway (142). In fact, experiments performed in DDAH−/<sup>−</sup> mice (142), siRNA-mediated DDAH1 knockdown and DDAH1 overexpressing HUVEC cells (164) have demonstrated that DDAH1 regulates HUVEC cell cycle progression via Ras/Akt activation and modulation of cyclinD1, cyclinE, CDC2, and CDC25C concentrations. Moreover, DDAH1 was reported to regulate angiogenesis by increasing NO concentrations, which induces caspase-3 activation in human fetal pulmonary microvascular endothelial cells (165). Increased angiogenesis is also observed following transfection of endothelial cell lines with DDAH2 (166), a process that is partially mediated by the upregulation of VEGF expression through a Sp1-dependent and NO-independent mechanism (167). Furthermore, comparative studies performed in DDAH1+/−, DDAH2+/−, and DDAH2−/<sup>−</sup> mice have demonstrated the important role of DDAH2 in pathogenic retinal ischemia and ischemia-induced angiogenesis and the protective potential of DDAH2 inhibition against aberrant neovascularization (146). It seems that this is achieved through reduced ADMA metabolism and improved vascular regeneration in a VEGF-independent fashion. Another ADMAindependent mechanism by which DDAH2 appears to regulate angiogenesis involves the regulation of VEGF and kinasedomain insert containing receptor (KDR) expression within the silent information regulator 1 (SIRT1) pathways in EPCs (168).

Taken together, these studies demonstrate the key role of the DDAH/ADMA pathway in the regulation of neovascularization and endothelial cell proliferation, differentiation, and motility in vivo and in vitro. Impairment of the DDAH/ADMA/NO pathway and subsequent endothelial dysfunction have been extensively studied in relation to cardiovascular and renal disorders. The importance of the DDAH enzymes in cancer angiogenesis, neovascularization, and vasculogenic mimicry has only recently begun to be unraveled.

## EXPRESSION AND REGULATION OF THE DDAH ENZYMES

#### DDAH Expression

Whilst synthesis of ADMA occurs in all cells, expression of DDAH isoforms is variable. The two DDAH isoforms (DDAH1 and DDAH2) display distinct but overlapping tissue distribution, and additionally show some overlap with the constitutively expressed NOS isoforms. DDAH2 is expressed in the heart, vascular endothelium, kidney, placenta, and adipose tissue (169, 170). Sites of DDAH1 expression are considerably wider, but it is predominantly found within the brain, liver, and kidney (140, 171–176), the organs which represent the major sites of ADMA metabolism (141, 177, 178), as well as in the heart, lung, skeletal muscle, nervous system, spinal dorsal horn, and trophoblasts (138, 179–181). It is also important to mention that the expression pattern of DDAH1 and DDAH2 does not necessarily reflect the tissue activity of the enzymes. This issue is further complicated by the fact that the currently available DDAH activity assays do not distinguish between DDAH1 and DDAH2 isoforms. Therefore, even if two tissues display the same level of DDAH activity, it is unclear what amount of activity can be attributed to each DDAH isoform. This could be of particular importance given the additional ADMA-independent effects of both enzymes, as discussed later.

#### Expression of DDAH Is Altered in Cancer

Identification of genes which are differentially expressed in cancer relative to normal tissue can be highly beneficial in terms of developing new diagnostic, prognostic, and targeted therapeutic treatments for cancer development and progression. The recent advances in genome-wide transcriptomic and proteomic techniques have allowed for profiling of different cancers at various disease stages with this aim in mind. Interrogation of publicly available data generated by The Cancer Genome Atlas (TCGA) Research Network (http://cancergenome. nih.gov) and the Genotype-Tissue Expression (GTEx) project identified altered expression of DDAH1 and DDAH2 in various cancer tissues. The online web-tool Gene Expression Profiling Interactive Analysis (GEPIA; http://gepia.cancer-pku.cn/) (182) was used to analyse RNA-seq expression data sourced from these databases and to generate expression profiles of DDAH mRNA expression in comparable normal and tumor tissues for each cancer type. In pancreatic adenocarcinoma and thymoma DDAH1 and DDAH2 mRNA is significantly increased, whilst both DDAH1 and DDAH2 expression is decreased in lung squamous cell carcinoma (**Figure 2A**). Interestingly, with the exception of these three cancers, the expression of DDAH1 and DDAH2 does not change in the same direction. Instead, the expression of either isoform is altered independently of the other. There is also no evidence for an inverse correlation of DDAH1 and DDAH2 expression (e.g., an increase in DDAH1 expression paired with a decrease in DDAH2 expression, or vice versa) in any cancer type for which data is available in the TCGA database.

An increase in DDAH2 mRNA expression is further observed in glioblastoma, brain lower grade glioma and liver cancer samples, whilst a decrease is observed in cervical cancer

generated by the online web-tool Gene Expression Profiling Interactive Analysis (GEPIA; http://gepia.cancer-pku.cn/) (182).

samples relative to normal tissue (**Figure 2B**). The range of cancer types that display altered DDAH1 expression is significantly broader than that for DDAH2. In the majority of cases where there is a change in DDAH1 expression in tumor samples, it is significantly increased: these include breast cancer, colorectal cancer, lymphoid neoplasm diffuse large B-cell lymphoma, esophageal cancer, ovarian cancer, prostate cancer, rectal adenocarcinoma, stomach cancer, thyroid cancer, and uterine corpus endometrial carcinoma (**Figure 2C**). A decrease in DDAH1 expression is only found in chromophobe renal cell carcinoma (a rare form of kidney cancer), melanoma and in testicular germ cell tumors (**Figure 2D**). The relative expression of DDAH1 and DDAH2 mRNA in various cancers, and the sample number for each analysis, is shown in **Figure 2**.

In addition to RNA-seq data obtained through mining of TCGA datasets, a number of research groups have also identified altered DDAH mRNA and protein expression in various cancer cell lines and cancer tissues (**Table 1**). Studies to date have demonstrated an increase in DDAH1 protein expression in human glioma, meningioma, prostate cancer, and hepatocellular carcinoma, primarily by means of large-scale proteomic analysis. An upregulation of DDAH1 protein has also been observed in cohorts of melanoma and breast cancer cell lines, relative to normal melanocyte, and mammary epithelial cells, respectively (183, 187). Aside from the identification that DDAH1 expression is significantly altered in these cancers, only a handful of these studies undertook further analysis into the specific role and function of DDAH1 within each cancer context.

In addition to protein expression, Buijs et al. (188) further assessed DDAH1 catalytic activity in hepatocellular carcinoma (HCC) tissue relative to paired non-tumorous liver tissue. In tissue homogenates, mass spectrometry analysis of arginine and ADMA concentrations revealed a 74% increase in the arginine:ADMA ratio, which is indicative of increased ADMA metabolism and thus increased NO production. Furthermore, increased NO concentration was predicted in both tissue homogenates and serum from preoperative HCC patients, as measured by NO metabolites (nitrate and nitrite) using a colorimetric Griess assay. An increase in expression of the angiogenesis stimulating factor, VEGF, was also observed in HCC tissue samples. It is important to note that immunofluorescence analysis of tumor tissue samples confirmed expression of DDAH1 localized to hepatocytes, and absent from neighboring endothelial cells of vascular structures (188). We have also recently published evidence for a novel role of DDAH1 in breast cancer, particularly in the more aggressive and invasive triple negative breast cancer (TNBC) molecular subtype (187). In this study we demonstrated high expression of functional DDAH1 enzyme in TNBC cells relative to normal mammary epithelial cells. This was determined by both western blot analysis and mass spectrometry assessment of Lcitrulline formation with 200µM ADMA substrate. Inhibition of DDAH1 protein expression in these cells resulted in reduced L-citrulline formation, increased intracellular ADMA concentration and a reduced arginine:ADMA ratio; all consistent with decreased ADMA metabolism and consequently decreased NO production (187).

In 2011, a proteomics and pathway analysis study by Ummanni et al. identified DDAH1 overexpression in histologically characterized prostate cancer tissue, and highlighted its potential as a novel biomarker for prostate cancer development and/or progression (185). Intriguingly, whilst western blotting validated dysregulation of DDAH1 protein in tumor tissue, no significant change in DDAH was observed at the mRNA level. This is somewhat consistent with data in breast cancer cell lines, where a much greater change in DDAH1 expression was observed at the protein level compared to the transcript level (187). It is possible that this phenomenon is in part due to post-transcriptional regulation of DDAH1, likely mediated by multiple microRNA regulators in the unusually long DDAH1 3′ UTR (2,971 bp). In a recent follow-up study, tissue microarray analysis further confirmed higher DDAH1 expression in prostate cancer compared to benign prostatic hyperplasia and normal prostate tissues; the expression of which correlates well with the aggressiveness of prostate cancer and suggests its role in disease progression (54). In hormonedependent (PC3) and hormone-independent (LNCaP) prostate cancer cell lines, both of which express DDAH1, generation of L-citrulline from the enzyme-substrate ADMA was observed in a colourimetric assay. In alignment with findings in breast cancer cell lines (187), specific knockdown of DDAH1 protein in PC3 and LNCaP cell lines not only resulted in reduced L-citrulline formation, but also significantly increased intracellular ADMA concentration and decreased NO metabolite concentration.

In contrast to these studies, DDAH1 protein downregulation was frequently detected in gastric cancer tissues, where its low expression was associated with more lymph node metastasis and poorer clinical outcome (193). Knockdown and overexpression of DDAH1 in gastric cancer cell lines recapitulated these findings: cells overexpressing DDAH1 migrated more slowly and were less invasive in vitro, and displayed decreased metastatic potential in vivo, possibly through inhibition of epithelial-mesenchymal transition (EMT) pathways (193). The authors also reported reduced β-catenin expression following DDAH1 overexpression, and suggested that DDAH1 mediates β-catenin degradation via the Wnt signaling pathway, thus inhibiting EMT. The exact mechanism by which DDAH1 modulates β-catenin expression is currently undefined; there was no assessment of DDAH1 catalytic activity and subsequent NO production in this study. To the best of our knowledge, this represents the only study to date that identifies DDAH1 as a tumor suppressor. It is possible that the tumor suppressor role of DDAH1 in gastric cancer is independent of its role in the ADMA/NO pathway.

DDAH2 protein expression has been less extensively studied in cancer, but an upregulation has been reported in prostate cancer cell lines as well as the malignant stroma (but not tumor cells) of non-small-cell lung cancer tissue (166, 184). In the LNCaP prostate cancer cell line DDAH2 was more strongly expressed when compared to benign prostate hypertrophy cells, and was also accompanied by increased eNOS, iNOS, and VEGF expression (184). It is likely that a combination of these factors, and not specifically DDAH2 expression, is responsible for the increased NO production that was observed in these cells. Interestingly, the NOS inhibitor N <sup>G</sup>-nitro-L-arginine methyl



↑, increased expression; ↓, decreased expression; ↔, no change in expression; ND, not determined; WB, western blot; IHC, immunohistochemistry; TMA, tissue microarray analysis; qRT-PCR, quantitative realtime-PCR; IF, immunofluorescence; ISH, in situ hybridization.\*Upregulated in 78% of cell lines investigated.

ester (L-NAME), which is not degraded by DDAH, significantly increased DDAH2 expression and elevated NO production (184). A more recent study in 2016 identified increased expression of DDAH2 in the stroma fibroblasts of lung adenocarcinomas, where tumors with high stromal DDAH2 expression had a poorer prognosis (166). Almost all cases of minimally invasive adenocarcinoma and invasive adenocarcinoma were positive for DDAH2, while only half of pre-invasive lesions (atypical adenomatous hyperplasia and adenocarcinoma in situ) were positive. In contrast, in normal lung tissue only the vascular endothelium showed staining for DDAH2 (166).

#### DDAH Regulation

Regulation of both DDAH1 and DDAH2 expression and activity is mediated via various mechanisms at different levels.

#### Post-translational Modulators of DDAH Activity

DDAH exists as a holoenzyme bound to a single inhibitory zinc ion. Removal of the zinc by either phosphate or imidazole results in increased DDAH enzymatic activity, thus demonstrating the regulatory role that the zinc binding site plays (194). The crystal structure of DDAH1, purified from bovine brain, shows zinc bound to the active site cysteine (Cys273); 95% of total DDAH1 purified protein exists as the zinc-bound form. These data suggest DDAH1 exists predominantly in its inhibited conformation (195). NO itself is a reversible inhibitor of DDAH activity through S-nitrosylation of the active site cysteine residue (Cys273 in bovine DDAH1, Cys274 in human DDAH1, Cys249 in human DDAH2), which involves covalent attachment of nitrogen monoxide to the thiol chain of the specific cysteine residues. Typically, this is associated with increased expression of iNOS and thus increased NO synthesis, and does not occur under basal conditions (196). It has been demonstrated in vitro via incubation of purified bovine DDAH or recombinant bacterial DDAH with a NO donor (DEA NONOate; 2-(N,Ndimethylamino)-diazenolate-2-oxide) (197, 198). This represents a feedback loop whereby subsequent accumulation of the DDAH substrates, ADMA and L-NMMA, in turn reversibly inhibit the NOS enzymes. Intriguingly, NO-induced DDAH inhibition is significantly more potent in the absence of zinc (DDAH apoenzyme), which suggests zinc binding is protective of DDAH S-nitrosylation (198). Phosphorylation of rat DDAH1 at Ser33 and Ser56 has been reported (199), however the impact of this on DDAH1 activity is currently unknown. There is currently no further evidence to suggest additional posttranslational modification of DDAH enzymes.

There are a significant number of endogenous compounds, vitamins, and therapeutics identified to date that act as DDAH activators or inhibitors without altering gene expression. Many of these factors modulate DDAH activity via oxidative effects, such as via attenuation of low-density lipoprotein-induced endothelial dysfunction or by induction of reactive oxygen species. Key examples include 17β-estradiol (200), insulin (201), vitamin E (202), and the antioxidant Probucol (203) as DDAH activators. In contrast, the cytokine TNF-α (204), glucose (201, 205), s-nitrosohomocysteine (206), and erythropoietin (207) are significant inhibitors of DDAH activity. With the exception of s-nitrosohomocysteine, the exact mechanisms by which these compounds function to modulate DDAH activity is as yet undefined, however literature suggests that ultimately it is Snitrosylation of DDAH and/or a modulation of zinc availability or binding capacity to the DDAH active site which are likely contributors. For example, induction of DDAH enzymatic activity may require a zinc-binding protein to act as a zinc receptor, thus abolishing the zinc-mediated inhibition of DDAH. On the other hand, zinc released from a redox sensitive zincbinding protein, under conditions of oxidative or nitrosative stress, may bind to and inactivate DDAH. A recent study by Bollenbach and colleagues has also identified a DDAH inhibitory role for some naturally occurring amino acid derivatives, namely NG-hydroxy-L-arginine, Nω,Nω-dimethyl-L-citrulline and connatin (208).

Taken together, the number and diversity of endogenous compounds, vitamins, and therapeutics which are capable of altering DDAH activity highlights the importance of quantifying DDAH activity in tissues of interest. As protein expression may not necessarily reflect enzyme activity, a comprehensive understanding of the importance of DDAH enzymes in any given tissue or disease state requires assessment of transcript abundance, protein expression, and additionally activity of DDAH enzymes.

#### Transcriptional and Post-transcriptional Regulation of DDAH Expression

The understanding of what regulates DDAH expression in cancer is very limited. The only study to specifically address regulation of DDAH1 in cancer was performed in breast cancer cell lines and identified the microRNA miR-193b as a direct negative regulator through the DDAH1 3′UTR (187). In MDA-MB-231 cells expressing endogenous DDAH1, ectopic expression of miR-193b reduced DDAH1 mRNA and protein expression and decreased the conversion of ADMA to citrulline. Conversely, inhibition of miR-193b in the MCF7 cell line, which was absent for DDAH1 expression, was sufficient to induce DDAH1 (187). Mir-193b has been previously reported as a tumor suppressor in breast cancer tissues (209, 210) and is frequently downregulated in other solid tumors such as melanoma (211), liver cancer (212), and prostate cancer (213), all of which are reported to exhibit increased DDAH1 expression (**Table 1**). It is therefore plausible that miR-193b is an important regulator of DDAH1 expression in multiple cancers.

The DDAH1 3′UTR is unusually long (2,971 bp) and is therefore likely regulated by multiple microRNAs. In addition to miR-193b, various studies have demonstrated direct regulation of DDAH1 by miR-21 (214–216); however all studies to date have been performed in human endothelial cells. miR-21 was one of the earliest defined oncomiRs, and its role in carcinogenesis has been thoroughly investigated (217), particularly in gastric cancer where it is often upregulated (218, 219). In alignment with this, downregulation of DDAH1 is reported in gastric cancer tissue and cell lines (193). In HUVECs, transmembrane glycoprotein neuropilin-1 increases DDAH1 expression, mediated by a post-transcriptional mechanism involving miR-219-5p (220). Although this regulation has not been assessed in cancer, miR-219-5p has been reported to have a tumor suppressive role in colon cancer (221, 222) and ovarian cancer (223), which may in part relate to regulation of DDAH1.

Further studies on regulation of DDAH1 have identified that DDAH1 protein is increased in a time- and dose-dependent manner in cultured rat smooth muscle cells stimulated with IL-1β (224), and that O subfamily of forkhead (FoxO)1 is pivotal in regulation of endothelial activation as a negative regulator of DDAH1 (225). Agonists of the nuclear receptor farnesoid X receptor (FXR) have been shown to induce hepatic DDAH1 transcription through a promoter FXR response element, resulting in decreased plasma ADMA (172). Another study has also reported an increase in DDAH1 following stimulation with an FXR agonist in the liver and kidney, which was also accompanied by decreased plasma ADMA (226). Activation of FXR with bile acids has been found to enhance tumor angiogenesis (227), however whether FXR alters DDAH1 expression in cancer cells has yet to be identified. Furthermore, metal-responsive factor 1 (MTF1), a pluripotent transcriptional regulator induced by various stress conditions such as hypoxia and oxidative stress, increases DDAH1 expression via a direct binding site in the DDAH1 promoter (228). Hypoxia, which is often observed in solid tumors, induced DDAH1 expression in liver cancer HepG2 cells (188), however the exact mechanism underlying this induction remains to be elucidated.

The promoters of both DDAH1 and DDAH2 contain sterol response elements (DDAH1 more so than DDAH2). In cultured endothelial cells, the sterol response element binding protein (SREBP) transcription factor member, SREBP-2, was found to bind the DDAH1 promoter and activate transcription (229); knockdown of SREBP-2 led to a decrease in DDAH1 mRNA expression. SREBPs are key transcription factors which play a central role in lipid metabolism, and elevated SREBP levels are common in various cancers (230, 231). It appears that regulation by SREBPs is isoform-specific, however, as SREBP-1c decreased both DDAH1 and DDAH2 expression (229). Finally, an increase in DDAH activity in human and murine endothelial cell lines has been demonstrated following treatment with estradiol (200). In following studies, an estrogen receptor (ER) binding site was identified within the DDAH2, but not the DDAH1, promoter (232), suggesting a mechanism for estradiol in transcriptional regulation of DDAH2. In HUVECs, estradiol increased DDAH2 protein expression, decreased ADMA concentrations, and increased NO production (233); these effects could be blocked by ER antagonists (233, 234). Although not yet known, this regulation of DDAH2 by estradiol and ER may play an important role in cancers driven by excessive ER signaling, such as those of the breast.

#### IMPACT OF DDAH EXPRESSION ON TUMOR ANGIOGENESIS AND VASCULOGENIC MIMICRY

A key aspect of cancer progression involves tumor angiogenesis. In addition to providing blood flow and nutrients to the tumor to support growth, angiogenesis is also implicated in tumor invasion and metastasis as the vasculature provides the tumor with access to distant organs. This is of particular concern when vasculogenic mimicry (VM), the process in which vascular-like structures are generated by cancer cells, is present. These vascular-like structures are not only able to fuse to the conventional vascular network (16), but they can remodel the vasculature such that it becomes "leaky" (235). Several studies including our own have demonstrated the functional role that increased DDAH expression has on both tumor angiogenesis and VM.

To the best of our knowledge, the only study to assess the role of DDAH2 on tumor angiogenesis was undertaken in lung adenocarcinoma. In surgically resected specimens, high expression of DDAH2 in stroma of invasive lung adenocarcinoma correlated with stronger eNOS expression in the vascular endothelium of the malignant tissue (166). In vitro assessment of recombinant DDAH2 expression in HUVECs demonstrated a significant increase in cell proliferation and capillary-like tube formation, in a model of angiogenesis (166). Whilst together these findings may be indicative of a model whereby DDAH2 promotes tumor angiogenesis, a more definitive assessment of the role of DDAH2 in vivo is clearly required.

To study the effect of DDAH1 on tumor growth and vascular development, Kostourou and colleagues generated a rat C6 glioma cell line over-expressing the rat DDAH1 isoform (236– 239). The increased DDAH1 expression resulted in increased NO synthesis, as indicated by increased cGMP production, combined with increased expression and secretion of VEGF. Whilst no change in cell proliferation was observed in in vitro assays, DDAH1 overexpressing cells grew approximately twofold faster than wildtype cells following subcutaneous injection into the flanks of nude mice (236). The use of non-invasive magnetic resonance imaging (MRI) for the assessment of blood vessel development in vivo demonstrated significantly increased vascularity in these tumors; this was further supported by increased tumor perfusion as assessed by Hoescht 33342 staining of functionally perfused vessels. It is thus plausible that the increased growth of DDAH1 overexpressing tumors is a direct result of increased blood vessel development. Further analysis of the tumor angiogenesis identified no difference between vascular maturation, vascular function and microvessel size between wildtype and DDAH1 overexpressing cells, suggesting a role for DDAH1 in the initial stages of vasculogenesis (237). Collectively, these studies were the first to demonstrate the importance of DDAH1 in regulation of tumor vessel development and clearly demonstrated that DDAH1 expression leads to more hypoxic tumors, higher blood volume, better tumor perfusion, and increased number of functional vessels (236–238).

It has been further demonstrated that xenografts derived from cells over-expressing an active site DDAH1 mutant (incapable of metabolizing ADMA) display an intermediate phenotype between tumors overexpressing wildtype DDAH1 and control tumors in terms of growth rate, endothelial content (vessel area), and hypoxia (239). However, VEGF production by inactive DDAH1-expressing cells is not significantly altered compared to wildtype cells (239). Thus, it appears that whilst DDAH1 metabolic activity is essential for the change in VEGF production (236, 239), cell growth and tumor vascularity are not entirely dependent upon ADMA metabolism and VEGF production. One hypothesis put forth by Boult et al. is that metabolically inactive DDAH may still be able to bind and hold ADMA, thus sequestering it away from NOS and relieving NOS inhibition (239). Further support for this hypothesis is provided by an elegant study in which DDAH1 was overexpressed under control of a pTet-Off regulatable element in rat C6 glioma cells deficient in NO production. Xenografts derived from cells with DDAH1 overexpression, lacking the ability to produce NO, were not significantly different in terms of size, vessel density, vessel function, or vessel maturation when compared to cells absent for DDAH1 expression and NO function (240). Together these studies suggest that, at least in C6 gliomas, the effect of DDAH1 on tumor growth and angiogenesis is purely NO-dependent.

In prostate cancer cell lines, exogenous expression of human DDAH1 increases cell proliferation, migration and invasion, and induces expression of multiple NO-regulated genes such as VEGF, HIF-1α, and iNOS. In alignment with the studies in rat C6 glioma cells, inhibition of NOS by L-NAME or 1400 W is sufficient to reverse the induction of all three proangiogenic genes. Furthermore, overexpression of an active site mutant human DDAH1 does not significantly alter cell behavior or VEGF expression, providing additional evidence that hydrolytic activity of DDAH1 is required for mediation of prostate cancer growth. Similarly, in vivo assessment of mouse xenografts has demonstrated significantly increased tumor size, invasion into muscular regions, mitotic figures, necrosis, proangiogenic factor expression, and tumor microvessel number in wildtype DDAH1-overexpressing tumors compared to mutated DDAH1-overexpressing and control tumors (54).

In our own studies assessing VM in triple negative breast cancer cell lines, specific knockdown of endogenous DDAH1 significantly attenuated cell migration, but not proliferation. Formation of vessel-like networks in an in vitro assay of VM, and VEGF expression, were also significantly reduced (187). Interestingly, expression of a miR-193b mimic, a direct negative regulator of DDAH1, completely abolished vascular channel formation (187); this is perhaps suggestive of miR-193b regulating a network of genes involved in VM. In contrast, exogenous expression of DDAH1 in a DDAH1-null breast cancer cell line was not sufficient to induce VM (187), indicating that DDAH1 is required but not sufficient for VM in breast cancer. The extent to which DDAH1 can modulate breast cancer VM via ADMA-dependent or -independent processes is yet to be established.

#### Pharmacological Inhibition of DDAH1 Activity in Cancer

There are currently no synthetic compounds which act as specific DDAH1 or DDAH2 activators, nor are there any selective DDAH2 inhibitors. With the exception of a few compounds specifically targeting bacterial DDAH (241–243), all other synthetic DDAH inhibitors have been synthesized to selectively target DDAH1. Despite enhanced DDAH2 expression being linked to a handful of cancers such as lung (166) and prostate (184), the lack of a robust and reproducible in vitro DDAH2 activity assay represents a significant limitation for the development and pharmacokinetic characterization of DDAH2 activity modulators. As a consequence, studies investigating the effects of DDAH pharmacological inhibition focus solely on DDAH1. Over the last two decades various different classes of DDAH1 inhibitors have been synthesized; these exhibit different structures, features and mechanisms of action, and have been previously extensively reviewed (244). Whilst some of these molecules have structural similarity with the DDAH substrates (methylated arginines) (183, 245–249), others bear a very different chemical structure (56, 250–252). A comprehensive discussion on all DDAH inhibitors synthesized to date and

their impact on endothelial cells falls outside the scope of this review, however, here we summarize a small body of evidence that identifies the therapeutic potential for pharmacological inhibition of DDAH1 in cancer.

The first study to show some potential for DDAH1 inhibition by a small molecule in cancer was published by (183). The research group demonstrated that DDAH1 is overexpressed in melanoma cell lines compared to normal human epidermal melanocytes and that cellular inhibition of DDAH1 by N 5 -(1 imino-2-chloroethyl)-L-ornithine (Cl-NIO) resulted in reduced nitric oxide production in the A375 melanoma cell line. The reduction in NO synthesis was measured by quantifying 3 nitrotyrosine levels and total nitrate and nitrite in the cell culture supernatant and it was independent of changes in DDAH1 or iNOS expression (183). Unfortunately, this study did not assess the effects of DDAH1 inhibition by Cl-NIO on specific tumor parameters, such as tumor cell viability and proliferation in vitro and/or in vivo growth of xenograft tumors derived from A375 cells, or assess the impact on angiogenesis.

More recently, the potential therapeutic benefit of inhibiting DDAH1 was demonstrated for breast cancer (55). DDAH1 activity was inhibited in triple negative breast cancer cell lines by the potent DDAH1 inhibitors, arginine analogs ZST316 and ZST152 (244, 249), as identified by increased intracellular ADMA concentrations and decreased intracellular L-citrulline concentrations (55). In an in vitro Matrigel tube formation model of VM, both ZST316 and ZST152 significantly inhibited the number of vessel-like networks formed at concentrations above 1µM (55). Importantly, the endogenous NOS inhibitor L-NMMA, which is widely used as a tool to decrease NO availability, also significantly reduced tube formation in these assays. By contrast, no inhibition was observed when cells were treated with SDMA, which is neither a substrate for DDAH1 nor an inhibitor of NOS. Cell viability and proliferation were not affected by doses of up to 100µM of ZST316 or ZST152, however, a decrease in cell migratory potential was observed, which may be in part responsible for the reduced tube formation in the model of VM (55). Although, these results are somewhat preliminary and need further confirmation with in vivo studies, they suggest a promising role for DDAH1 inhibition as a novel treatment strategy in triple negative breast cancer.

The most recent and comprehensive study which describes a role for DDAH1 pharmacological inhibition in cancer demonstrates the ability of the compound 3-amino-6-tert-butyl-N-(1,3-thiazol-2-yl)-4-(trifluoromethyl)thieno[2,3-b]pyridine-2-carboxamide (DD1E5) to irreversibly inhibit DDAH1 activity in prostate cancer cells (56). Treatment with DD1E5 inhibited proliferation, migration and invasion of prostate cancer cell lines LNCaP and PC3, but was also able to attenuate proliferation of cells stably overexpressing DDAH1; this was accompanied by decreased DDAH1 enzymatic activity, increased ADMA concentration and decreased NO synthesis. Additionally, modulation of the angiogenic pathway was observed in prostate cancer cells following treatment with DD1E5: the pro-angiogenic factors VEGF, iNOS, c-Myc, and HIF-1α were all downregulated, indicating that DDAH1 inhibition attenuates the angiogenic potential of DDAH1+ cells (56). The release of pro-angiogenic signals bFGF and IL8 was also decreased following DDAH1 inhibition, and this translated into a decrease in endothelial cell tube formation when cells were cultured in conditioned media from the treated prostate cancer cells. Most importantly, in vivo analysis demonstrated that DD1E5 inhibited the growth of xenograft tumors derived from DDAH1 overexpressing PC3 cells, reduced the expression of VEGF, NOS, and HIF-1α in xenograft tumors, and resulted in poorer vascularization as assessed by micro vessel density (56).

#### CONCLUSION

The DDAH enzymes are responsible for the metabolism of the endogenous NOS inhibitors, the asymmetrically methylated arginines ADMA and L-NMMA, and are thus critical factors in both maintaining and modulating precise NO production. In endothelial cells, the significance of the DDAH/ADMA/NO axis is well-documented: NO has a regulatory role which is required for endothelial cell activation, proliferation and migration, and which overall is necessary for effective angiogenesis. Studies have consistently demonstrated that dysregulation of this pathway and NO synthesis, as a consequence of DDAH modulation, results in impaired angiogenesis (142, 149, 164, 253).

The role of NO has been extensively studied in cancer, particularly tumor angiogenesis, yet the literature is not always entirely clear. It appears that NO can have both oncogenic and protective roles depending on cancer type, location and stage, as well as local NO concentration and exposure duration. Nonetheless, excessive NO production has been associated with poor prognosis, increased vasculature and increased invasiveness of multiple cancers such as breast (102, 104, 105), prostate (107), and colorectal (110, 111). Until recently, the majority of studies which have assessed the impact of altered NO production in cancer have focused solely on the role (both expression and regulation) of the three NOS enzymes. In contrast, limited studies have addressed the potential impact of DDAH expression and function in the oncology setting. As discussed here, DDAH expression (particularly DDAH1) is significantly altered in a number of different cancers. In the majority of these, DDAH expression is increased and is associated with increased NO concentrations, increased VEGF expression and increased cell aggressiveness. Furthermore, in vivo studies using DDAH1 overexpression models have demonstrated increased tumor growth and corresponding increased tumor vasculature. Whilst one of the roles of DDAH1 in tumor vessel development is likely facilitation of endothelial cell migration and invasion, as supported by DDAH1 overexpression conditioned media studies (236), initial reports in breast cancer cell lines suggest that DDAH1 is also a modulator of VM. Whether the function of DDAH1 in VM is entirely ADMA/NO-dependent remains to be determined. In contrast to DDAH1, the importance of DDAH2 in ADMA metabolism and thus tumor angiogenesis is still unclear. Collectively, these studies begin to further elucidate the complex tumor-promoting pathways in multiple cancers.

Importantly, the upregulation of DDAH1 expression and consequent increased enzymatic activity may suggest a novel role for DDAH1 in tumor progression, providing novel diagnostic, and therapeutic opportunities for DDAH1 as a possible molecular drug target. Intriguingly, DDAH1 autoantibodies have been detected in sera of prostate cancer patients and proposed as a new marker for a novel prostate cancer and benign hyperplasia diagnostic, improving on the traditional prostate specific antigen (PSA) test which often yields falsepositive results (254). The exact mechanism responsible for production of DDAH autoantibody markers is unknown but may relate to changes in DDAH1 expression levels. A handful of studies have assessed the impact of DDAH1 inhibition by small molecules in cancer with promising results for inhibition of tumor growth, vasculature density, and VM. Taken together, they demonstrate that pharmacological inhibition of DDAH1 represents a novel, alternative strategy for the treatment of cancers associated with elevated DDAH1 expression and activity. Studies on breast cancer, prostate cancer, glioma, and melanoma have identified that these cancers typically express high levels of DDAH1 and are dependent on DDAH/ADMA/NO signaling for cell survival, proliferation, migration, and/or angiogenesis; these cancers would be prime candidates for treatment by DDAH1 inhibition. It is currently unknown as to whether DDAH1 inhibitors act exclusively by blocking enzymatic activity or whether they may modulate alternative functions of DDAH1 (e.g., potential protein-protein interactions).

Although studies are limited, the data to date suggests a basis for the development of DDAH1 inhibitors to be used as combined anti-angiogenic and anti-VM agents in cancer. It is important to continue to unravel the mechanisms of DDAH1-mediated tumor angiogenesis and VM, and to further explore the potential of selectively inhibiting DDAH1 activity across different tumor types and stages. Pending the results of animal studies, the use of DDAH1 inhibitors, alone or in combination with traditional anti-angiogenic therapies such as anti-VEGF drugs, might represent a novel strategy to

#### REFERENCES


suppress both angiogenesis and VM, key factors in early cancer development and dissemination. Based on the current evidence, which highlights the lack of a clear direct cytotoxic effect of small molecule DDAH1 inhibitors, it would appear that the best therapeutic window is within the early stages of cancer development, typically driven by neovascularization, in order to timely prevent dissemination and metastasis. This, however, does not exclude their potential use in the later stages of the disease, particularly if combined with other pharmacological strategies. Given the importance of DDAH1 in maintaining homeostasis of the cardiovascular system, particularly in attenuating cardiovascular disease and heart failure, potential negative impacts of inhibition of DDAH1 must be considered. As such, further studies are required to determine whether DDAH1 inhibitors can be safely administered systemically or whether approaches for a targeted, local, delivery are preferable.

#### AUTHOR CONTRIBUTIONS

J-AH and ST wrote the first draft of the manuscript and prepared figures. EG, NJ, RR, and AM contributed to manuscript revision, read, and approved the final submitted version.

#### FUNDING

This study was partially supported by the Flinders Medical Centre Foundation and the Flinders University Faculty of Medicine and Health Sciences.

#### ACKNOWLEDGMENTS

The results of DDAH transcript expression in normal and tumor tissue shown here are in whole or part based upon data generated by the TCGA Research Network: https://www.cancer.gov/tcga.

receptor 2-negative metastatic breast cancer. J Clin Oncol. (2010) 28:3239–47. doi: 10.1200/JCO.2008.21.6457


and organ function in experimental septic shock. Biochem J. (2014) 460:309– 16. doi: 10.1042/BJ20131666


carboplatin in patients with lung adenocarcinoma. Clin Cancer Res. (2006) 12:6748–57. doi: 10.1158/1078-0432.CCR-06-1124


correlation with tumor angiogenesis. Am J Pathol. (2003) 162:793–801. doi: 10.1016/S0002-9440(10)63876-X


endogenous methylarginines. Nephrol Dial Transplant. (2016) 31:57–63. doi: 10.1093/ndt/gfv390


angiogenesis in lung adenocarcinoma. Virchows Arch. (2016) 468:179–90. doi: 10.1007/s00428-015-1863-z


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2020 Hulin, Gubareva, Jarzebska, Rodionov, Mangoni and Tommasi. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Regulation Networks Driving Vasculogenic Mimicry in Solid Tumors

Olga N. Hernández de la Cruz <sup>1</sup> , José Sullivan López-González <sup>2</sup> , Raúl García-Vázquez <sup>1</sup> , Yarely M. Salinas-Vera<sup>1</sup> , Marcos A. Muñiz-Lino<sup>3</sup> , Dolores Aguilar-Cazares <sup>2</sup> , César López-Camarillo<sup>1</sup> and Ángeles Carlos-Reyes <sup>2</sup> \*

<sup>1</sup> Posgrado en Ciencias Genómicas, Universidad Autónoma de la Ciudad de México, Mexico, Mexico, <sup>2</sup> Laboratorio de Cáncer de Pulmón, Instituto Nacional de Enfermedades Respiratorias "Ismael Cosío Villegas," Mexico, Mexico, <sup>3</sup> Laboratorio de Patología y Medicina Bucal, Universidad Autónoma Metropolitana Unidad Xochimilco, Mexico, Mexico

#### Edited by:

Francesco Fiorica, Azienda Ulss 9 Scaligera, Italy

#### Reviewed by:

Zexian Liu, Sun Yat-sen University Cancer Center (SYSUCC), China Francesco Grignani, University of Perugia, Italy

> \*Correspondence: Ángeles Carlos-Reyes reyes\_cardoso@yahoo.com

#### Specialty section:

This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology

Received: 09 August 2019 Accepted: 28 November 2019 Published: 14 January 2020

#### Citation:

Hernández de la Cruz ON, López-González JS, García-Vázquez R, Salinas-Vera YM, Muñiz-Lino MA, Aguilar-Cazares D, López-Camarillo C and Carlos-Reyes Á (2020) Regulation Networks Driving Vasculogenic Mimicry in Solid Tumors. Front. Oncol. 9:1419. doi: 10.3389/fonc.2019.01419 Vasculogenic mimicry (VM) is a mechanism whereby cancer cells form microvascular structures similar to three-dimensional channels to provide nutrients and oxygen to tumors. Unlike angiogenesis, VM is characterized by the development of new patterned three-dimensional vascular-like structures independent of endothelial cells. This phenomenon has been observed in many types of highly aggressive solid tumors. The presence of VM has also been associated with increased resistance to chemotherapy, low survival, and poor prognosis. MicroRNAs (miRNAs) and long non-coding RNAs (lncRNAs) are non-coding RNAs that regulate gene expression at the post-transcriptional level through different pathways. In recent years, these tiny RNAs have been shown to be expressed aberrantly in different human malignancies, thus contributing to the hallmarks of cancer. In this context, miRNAs and lncRNAs can be excellent biomarkers for diagnosis, prognosis, and the prediction of response to therapy. In this review, we discuss the role that the tumor microenvironment and the epithelial-mesenchymal transition have in VM. We include an overview of the mechanisms of VM with examples of diverse types of tumors. Finally, we describe the regulation networks of lncRNAs-miRNAs and their clinical impact with the VM. Knowing the key genes that regulate and promote the development of VM in tumors with invasive, aggressive, and therapy-resistant phenotypes will facilitate the discovery of novel biomarker therapeutics against cancer as well as tools in the diagnosis and prognosis of patients.

Keywords: cancer, vasculogenic mimicry, microRNAs, long non-coding RNAs, epithelial-mesenchymal transition, tumor microenvironment

### INTRODUCTION

Solid tumors can form new blood vessels through complex neovascularization mechanisms that include the following: (i) angiogenesis the development of new blood vessels by endothelial cells (ECs) from pre-existing vessels, (ii) vasculogenesis generated from EC precursors, (iii) intussusception the splitting of vessels through the insertion of tissue pillars, (iv) vessel co-option migration of tumor cells migrate along existing vasculature, (v) cancer stem cell (CSC) trans-differentiation whereby cancer cells trans-differentiate to ECs leading to the formation of blood vessels, and (vi) vasculogenic mimicry (VM) where the tumor cells mimic ECs and form blood vessel-like threedimensional channels (1–3).

In particular, VM can enhance cancer cell migration, invasion, and metastasis as well as increased resistance to therapies. VM has been documented in diverse solid tumors such as breast cancer, liver cancer, ovarian cancer, gastric cancer, prostate cancer, and melanoma (4, 5). During tumorigenesis, the tumor microenvironment plays a vital role in the formation of the tumor vasculature. Deficient blood-vessel perfusion, hypoxia due to low oxygen pressure, and low-nutrient availability in the microenvironment lead to angiogenesis, metastasis, and tumor cell survival (6, 7). Hypoxia is a master regulator of various transcription factors and signaling pathways during of the development of VM in solid tumors (8). On the other hand, microRNA (miRNAs) and long non-coding RNAs (lncRNAs) regulate the expression of genes and signaling pathways in diverse tumor types, which contributes to the cancer hallmarks like metastasis via VM formation (9). Here, we summarize the latest advances in VM regulation in solid tumors. We first overview the role of the tumor microenvironment and the epithelialmesenchymal transition (EMT) in VM from different types of tumors. We further described some mechanisms of VM. Finally, we detail the regulation by miRNAs and the regulation networks by lncRNAs-miRNAs as well as their clinical impact on VM.

#### ROLE OF TUMOR MICROENVIRONMENT AND THE EMT IN THE DEVELOPMENT OF VM

The tumor microenvironment comprises the vasculature (blood vessels), extracellular matrix, stromal cells, immune cells, and signaling molecules. Poor blood-vessel perfusion in the tumor microenvironment is due to acidic pH, low nutrient levels, and hypoxia due to low oxygen pressure (10, 11). These aspects cause an imbalance in the angiogenic and anti-angiogenics factors that favor the invasion and metastasis of tumor cells. Tumor cell adaptation to the hypoxic microenvironment favors sustained angiogenesis (12). In particular, acidic pH and hypoxia are important factors for remodeling EMC. The acidic pH in solid tumors is due to the production of lactic acid during the fermentative metabolism produced by the high expression of Na+/H+ exchangers (NHE1), isoforms of anion exchangers, Na/HCO<sup>−</sup> 3 co-transporters, H+/ATPases, carbonic anhydrase IX isoform, monocarboxylate transporters, and the vacuolar ATPase. The released proton (H+) acidifies the tumor microenvironment and diffuses toward the stroma increasing the tumor survival, proliferation, and angiogenesis (13).

A crucial event in the development, progression, and metastasis of malignant tumors is neovascularization. It supplies growth factors, nutrients, and oxygen that alter the vascularization of the tumor including sustained angiogenesis (3). A new mechanism of neovascularization is VM that leads to the formation of blood vessels by the tumor cells themselves independently of ECs. VM is characterized by the deregulation of genes such as vimentin, cadherins, and metalloproteases and can be detected by double PAS/CD31 staining (5, 14). Tumors that show VM are highly aggressive and metastatic invasive phenotypes that are resistant to therapies (15). VM is promoted by the hypoxic tumor microenvironment, acidic pH, low nutrient levels, and the EMT (16).

Several studies report that the hypoxic tumor microenvironment regulates different transcription factors mediated by HIF-1α. These factors induce VM development as well as the regulation of epithelial markers that favor the EMT in different solid tumors (**Figure 1** and see **Table 1**) (16–42).

We show some reports of the role of the tumor microenvironment and EMT in the VM. In melanoma, hypoxia activates MMP-2 and MMP-9 expression promoting invasion to adjacent tissue. There is also deficient blood perfusion due to HIF-1α high expression (43). In addition, an increase in HIF-1α causes high expression of VEGF that facilitates the formation of VM. In a different study, the authors found that melanoma cells showed an increase in Bcl-2 expression promoting the formation of three-dimensional tubular structures via the VE-cadherin up-regulation mediated by Bcl-2 (44).

Ovarian cancer includes high expression levels of human chorionic gonadotropin and HIF-1α, which contribute to cell proliferation and tumor growth. They also lead to VM via the luteinizing hormone receptor (45–47). Only 25% of ovarian cancer biopsies are positive for VM, which correlates with hypoxia and EMT and is due to the high expression levels of HIF-1α, vimentin, VE-cadherin, Twist1, and Slug (21). These factors decrease E-cadherin expression. A hypoxic environment in breast cancer can increase the levels of HIF-1α, VE-cadherin, MMP-9, Cdc42, EGFR, p-Akt, and p-mTOR to promote the development of VM via the HIF-1α/VE-cadherin/MMP-9, MMP-2 signaling pathway (48). In the colorectal cancer HCT-116 cell line, hypoxia-induced development of VM is due to an increase in the zinc finger E-box binding homeobox 1 (ZEB1) and HIF-1α as well as with high vimentin expression and loss of E-cadherin expression in EMT (25).

The relationship between Bcl-2/Twist1 and Bcl-2-Bmi1 promotes EMT and development of VM through loss expression of E-cadherin and increased vimentin expression in hepatic cancer cells. VM is also caused by the translocation of Twist1 to the nucleus via Bcl-2 due to hypoxia (49–51). In biopsies of hepatic cancer, high expression levels of Notch1 and Hes1 were associated with VM. In hepatic cancer HepG2 and MHCC97-H cell lines, the Notch1 expression was higher in HepG2 favoring invasiveness by inducing the EMT through an increase in vimentin and loss of E-cadherin expression. These events are mediated by stimulation of TGF-β1 (30).

Glioma biopsies have increased expression in HIF-1α, MIF, and CXCR4. This has been observed in hypoxic regions of the tumor and is associated with VM development. In U87 and U251 glioma cell lines, high expression of MIF and CXCR4 promoted EMT and VM formation. In in vivo assays, MIF induces VM through the CXCR4-AKT-signaling pathway (35). In other reports, the glioma cell line SHG-44 transfected with pEGFP-Cl-LRIG1 and overexpressing LRIG1 inhibits VM promoted by hypoxia as well as migration, invasion, and proliferation. In addition, LRIG1 expression repressed signaling

of the EGFR/PI3K/AKT and EMT through an increase in Ecadherin and low vimentin expressions (52).

In melanoma, LRIG1 shows the same effects as glioma, but these are mediated by blocking via EGFR/ERK signaling (34). SACC-83 and SACC-LM salivary adenoid cystic carcinoma cell lines (SACC cells) have VM due to growth factors such as VEGFA that promote the development of VM mediated by hypoxia favoring migration and invasion as well as the EMT. Furthermore, the self-renewal capacity of SAAC-LM cells was due to the acquisition of the stem cell phenotype through VEGFA over-expression as well as an increase in the expression of N-cadherin, vimentin, CD44, and ALDH1 and loss of E-cadherin. These authors reported that only 26.3% of biopsies showed channel formation typical of VM. This phenomenon is related to HIF-1α and VEGFA expression (53). High expression levels of signal transducer and activator of transcription-3 (STAT3), p-STAT3, and HIF-1α in gastric cancer tissues for positive VM were associated with metastasis, degree of differentiation, and prognosis (54). On the other hand, esophageal squamous cancer cell lines Eca 109 and TE13 increased HIF-1α expression in hypoxic microenvironments. This promoted the VE-cadherin expression and led to VM development through the regulation of EphA2 and laminin subunit 5 gamma-2 (LN5γ2) expressions (55).

The EMT promotes the VM induced by hypoxia through the regulation of different transcriptional factors that promote the most aggressive, invasive, and metastatic phenotypes. These phenotypes are frequently therapy resistant with high recurrence.

#### MECHANISMS OF VM IN HUMAN CANCERS

Many studies have reported the participation of several transcription factors impacting diverse signaling pathways including EphA2, VE-cadherin, VEGFR2 (Flk-1), Rho, and integrins. These pathways regulate the VM development (**Figure 2**). In this review, we summarize some mechanisms related to the development of VM in solid tumors. Some of

Biorender.com.

TABLE 1 | Molecular factors that promote the EMT and VM in solid tumors.


↑ High expression, ↓ low expression. EMT, epithelial-mesenchymal transition; VM, vascular mimicry; NSCLC, non-small cell lung cancer.

these are driven by different receptors like the Eph receptor tyrosine kinases and the ephrin ligands that have been extensively implicated in carcinogenesis (56).

EphA1 and EphA2 are the most well-characterized molecules in solid tumors. They are implicated in VM formation and angiogenesis. High invasive MUM-2B melanoma cell line develops vascular pattern networks that are PAS positive. Such networks are associated with high expression and phosphorylation of EphA2 (57). Recent studies in vivo and in vitro in prostate cancer showed that VM development was associated with high expression of EphA2 and PI3K. PI3K/AKT regulates the activity of EphA2 through phosphorylation at Ser897 position (58). In addition, Yeo and coworkers showed that fetal bovine serum at 1–5% induced VM formation in prostate cancer PC-3 cells as well as phosphorylation of EphA2. This also increased the expression of VE-cadherin and Twist and activated AKT signaling. These changes were accompanied by an increase in MMP-2 and LN5γ2 protein levels (59). More experiments are needed to analyze whether cellular stress caused by reduced concentrations of fetal calf serum induced the development of VM.

Liang and coworkers found that Rictor, a component of mTOR2 signaling highly expressed in melanoma, was related to the presence of VM. These were associated with poor patient survival. The authors showed that VM was regulated through an increased activity of AKT-MMP-2/9 signaling and phosphorylation of AKT at Ser473 and Thr308 mediated by Rictor (60). In breast cancer, overexpression of B and C isoforms

of WT1 promoted VM development due to the EphA2/βcatenin/vimentin signaling pathway (61). In contrast, highly aggressive gallbladder carcinomas develop VM mediated by high levels of MMP-2 and MT1-MMP the overexpression of EphA2, FAK/PI3K, and LN-5γ2 signaling pathways, and via Paxillin-P signaling in vitro and in vivo (62).

Another mechanism for promotion of VM is mediated by VE-cadherin, one of the major endothelial adhesion molecule controlling intercellular junctions for blood vessel formation. This occurs via vascular endothelial growth factor receptor (VEGFR) function (63). Delgado-Bellido and coworkers found that malignant melanoma cells show constitutive expression of phosphorylated VE-cadherin at the Y658 position and forms a complex with p120-catenin and the transcriptional repressor kaiso. The high expression of nuclear phosphorylated VE-cadherin activated kaiso-dependent genes CCDN1 and WNT11 that increase the VM formation (64). Other study in melanoma showed that the c-Myc proto-oncogene was highly expressed in metastatic melanoma and induced VM through Snail activation inducing EMT by TGF-β/Snail/E-cadherin signaling pathways. c-Myc increases Bax expression causing a decrease in the Bcl2/Bax ratio through bilinearly patterned programmed cell necrosis. Necrosis forms empty spaces similar to blood vessels that serve as a support for VM formation under severe hypoxia (65). On the other hand, HER2-positive tumors showed high VM formation via VE-cadherin regulation (66).

Integrins are cellular adhesion receptors for cell attachment to the extracellular matrix. They transmit signals between cells and microenvironment. Integrins have multiple functions in cancer from initiation through metastasis, and they have been implicated in VM in various cancer types (67). In glioblastoma, Liu and coworkers analyzed tumors with VM and found a positive correlation between high levels of insulin-like growth factor–binding protein 2 (IGFBP2) and CD144/MMP-2 expression. They further found that overexpression of IGFBP2 increased tubule network formation through activation of CD144 and MMP-2, which was mediated by the binding of IGFBP2 to integrin α5/β1 and activation of the FAK/ERK/SPI pathway (68). Another report in glioblastoma multiforme showed that the presence of tumor-associated macrophages with M2 phenotype 2 infiltrating the VM-positive tissue area was associated with high levels of cyclooxygenase-2 (COX-2). Co-cultures of the U87 cell line with M2 macrophages activated by interleukin-4 promoted VM development through the prostaglandin E2/EP1/PKC pathway with high COX-2 and α-SMA expression and low VE-cadherin expression (69). In contrast, breast cancer cell lines that overexpress COX-2 vascular channels were also reported; this event was inhibited by COX inhibition (celecoxib) or siRNAs and was restored upon addition of exogenous prostaglandin E2 (70).

VEGF signaling via tyrosine kinase receptor VEGF receptor 2 (Flk-1) has a critical role in tumor angiogenesis and promotion of VM in cancer (71, 72). Blood vessels detected in the VM of glioblastoma cells are integrated by mural cell-lined vasculature. In glioblastoma, U87 and GDSC cell lines promoted VM formation. This formation was mediated by high expression of Flk-1 and VE-cadherin. Suppression of Flk-1 activity with SU1498 inhibitors in turn inhibits VM formation as well as FAK and ERK1/2 signaling pathways (73).

RhoC (Ras homolog gene family member C) regulates cytoskeletal organization and affects the motility of cancer cells favoring invasion and metastasis as well as progression and VM formation (74). In hepatocellular carcinoma (HCC), vascular channels were formed, and these vessels were associated with RhoC FAK/Paxillin signaling regulation as well as with high expression of RhoC/ROCK2, VE-cadherin, and MMP2 mediated by ERK/MMPs signaling (75).

### CSCs AND VM

CSCs are a subgroup of tumor cells that are multipotent with the capacity for self-renewal and differentiation as well as phenotypic and functional features of stem cells. In recent years, several reports have shown the role of CSC in the development of VM; they can form vascular-like structures that mimic embryonic vascular network patterns that are pivotal in tumor progression (76). For instance, in colorectal cancer, the Wnt/β-catenin pathway can induce VM formation through high Wnt3 and low β-catenin expressions. This pathway can also increase the expression of VEGFR2 and VE-cadherin. Thus, the Wnt/β-catenin activation favors the acquisition of endothelial phenotypes (77).

The glioblastomas undergo trans-differentiation into CD133 positive vascular ECs that promote the formation of glioma stem-like cells that can initiate the development of VM through VEGFR2 and VE-cadherin (72). In HCC, the Frizzled-2 gene (FZD2) induces proliferation, migration, and invasion due to its high expression. HCC also has decreased E-cadherin expression and increased N-cadherin, Snail, and Slug expressions. This promotes the EMT and VM formation. FZD2 also regulates the transcription factors Nanog and SOX2 in pluripotent cells favoring stemness. Moreover, the enrichment analysis of DEGs showed that FZD2 has a close relationship to the Hippo pathway mediated by the activity of YAP and TAZ (29). The tumorigenic growth of melanoma is shaped by stem-like cells. This cancer can grow to form spheroid cells that generate laminin networks similar to VM. These laminin networks have high expression of VE-cadherin, VEGFR-1, and nestin stem cell-associated biomarkers (78). These reports demonstrate that subpopulations of CSCs can transdifferentiate, and they contribute to development of VM in solid tumors.

#### REGULATION OF VM BY miRNAs IN SOLID TUMORS

miRNAs are post-transcriptional regulators of gene expression; they participate in degradation and/or inhibition of translation of their target genes (79). The alteration of the expression of multiple miRNAs has been reported in diverse types of solid tumors, and they act as oncogenes or tumor suppressors (80). Dysregulation of both groups correlates with diverse biological processes such as proliferation, invasion, migration, and VM in human cancer (see **Table 2**) (81–102). An important signaling pathway is VE-cadherin; it is one of the first factors identified as a regulator of VM. Liu and coworkers recently demonstrated the low expression of miR-27b in ovarian cancer cells. Restoring miR-27b expression in ovarian Hey1B and ES2 cancer cell lines significantly decreased intracellular VE-cadherin expression. This inhibits invasion, metastasis, and VM caused by the direct binding of miR-27b to the 3′ -UTR region of VE-cadherin (89). In HCC cell lines, the miR-27a-3p and miR-17 were downregulated and associated with high Bcl-2 expression. The VEcadherin, MMP-2, Twist1, HIF-1α, and VEGFA also lead to VM. In hepatic tumors, these genes were associated with poor prognosis of patients. Moreover, miR-27a-3p functions as a tumor suppressor for invasion and metastasis and is mediated by downregulation of VE-cadherin and EMT (96, 97). The miRNAs also regulate the EMT and facilitate the development of capillarylike structures in the tumors. They adopt invasive and metastatic properties. There was a decrease in miR-186 expression levels in P69 and M12 prostate cancer cell lines and tissues of patients with metastatic prostate cancer. The restoration of miR-186 suppresses cell motility, invasion, colony formation, and threedimensional culture growth, and inhibits the EMT through the negative modulation of Twist1 transcription factor (100). miR-124 in cervical cancer induced the suppression of the EMT process and decreased migration, invasion, and VM due to its specific interaction with 3′UTR of the angiomotin like 1 (AmotL1) protein that regulates cell migration related to angiomotin (87, 102).

Severalreports have shown that Eph2A expression is regulated by different miRNAs. In glioma and glioblastomas cell lines, overexpression of miR-26b and miR-141 inhibited VM formation through their specific binding with the Eph2A 3′ -UTR region. Also, both miRNAs regulate cell proliferation, migration, and invasion (92, 95). Ovarian cancer has a decrease in the expression

#### TABLE 2 | Modulation of VM by microRNAs (miRNAs).


levels of miR-200a that induce VM development. The restoration of miR-200a inhibits VM through modulation of the expression of EphA2. Also, low miR-200a expression has been associated with tumor grade and metastasis (88).

VEGF expression has been detected in hypoxic environments, and its secretion by tumor cells plays a crucial role in the tumor angiogenesis and VM formation (71, 72). Salinas-Vera et al. reported that miR-765 decreased VEGF expression after hypoxic conditions in ovarian cancer. Restoration in SKOV3 cells resulted in a significant inhibition of VM suggesting that miR-765 coordinates VM formation through modulation of the VEGFA/AKT1/SRC-α signaling pathway (90). The same group reported that miR-204 reduced the expression and phosphorylation of 13 proteins involved in the PI3K/AKT, RAF1/MAPK, VEGF, and FAK/SRC signaling pathways in MDA-MB-231 breast cancer cell line. These pathways impact VM development; its restoration repressed the VM formation and regulated the PI3K/AKT signaling pathway through its specific interaction with PI3K and SRC (86).

Multicellular spheroids derived from MCF-7, MCF10AT, and MCF10DCIS cell lines reproduce the architecture and tumor physiology observed in vivo. These models show high expression of osteopontin (OPN) oncoprotein via downregulation of miR-299-5p leading to vascular structures similar to VM (83). However, more studies are required in other type of tumors to correlate whether multicellular spheroids are linked with VM.

In gliomas, the expression decrease of miR-584-3p has been related with VM formation. The restoration of this miRNA expression inhibits VM formation in vitro through direct binding with the 3′ -UTR region of ROCK1 (94). In this same tumor, miR-Let-7f overexpression suppresses VM by repression of periostin (POSTN) that can induce migration of the cells. Overexpression of miR-9 a tissue-specific miRNA in the central nervous system increases apoptosis, suppresses tumor volume, and decreases cell proliferation and migration as well as VM formation in vivo and in vitro through negative regulation of the oncoprotein Stathmin (STMN1) (91, 93).

These studies show the essential role of the miRNAs in regulating the VM, in addition to other signaling pathways related to hallmarks cancer as cell proliferation, invasion, migration, and sustained angiogenesis in which new microcirculation process are implicated in therapy resistance and tumor recurrence.

#### lncRNAs-miRNAs-mRNAs REGULATION NETWORKS OF THE VM AND THEIR CLINICAL RELATIONSHIP

lncRNAs are a heterogeneous group of RNA molecules longer than 200 nucleotides. They have a dynamic role in the transcriptional and translational regulation of key genes in several diseases including of cancer. Their aberrant lncRNAs expression in the tumorigenesis contributes to metastasis, progression, and patient survival, as well as with VM development (103). The lncRNAs act as a competitive endogenous RNA. They change the expression transcriptional by competing for specific miRNAs binding sites altering their interaction. The regulation of lncRNAs in the miRNAs forms a complex regulatory network of lncRNAs-miRNAs-mRNAs

(104, 105). This network promotes the acquisition of cellular phenotypes such as migration, invasion, angiogenesis, and VM (**Figure 3**). lcRNAs are involved in a wide range of cellular processes regulating gene expression through multiple molecular mechanisms such as mRNA degradation and regulation of protein activity; scaffolds in the assembly of complexes, guides, decoys, or riborepressors; riboactivators; translational inhibition; chromatin remodeling; and miRNAs sponging (106). LncRNAs decrease miRNA target concentration within the cell through of their specific binding, which inhibits their function. Interestingly, hypoxia induces the expression of many lncRNAs and, similarly to miRNAs, lncRNAs are differentially expressed in diverse tumors leading to cancer hallmarks like metastasis via VM formation (107) (**Figure 4**).

Forinstance, Guo et al. compared the differential expression of lncRNAs in metastatic tissue, primary gastric cancer, and normal gastric tissue. High expression of the lncRNA olfactory receptor [family 3, subfamily A, member 4 (OR3A4)] is associated with gastric cancer metastasis, but in gastric cancer, this is related to clinicopathological features. This lncRNAs promotes cell proliferation, migration, and invasion in gastric cancer. The in vitro assays of OR3A4 cells induced VM and tubule formation in HUVECs cells. The high expression in cell lines induced angiogenesis in chicken embryos mediated by VEGF-C and MMP-9. This promotes the activation of target genes PDLIM2, MACC1, NTN4, and GNB2L1. Furthermore, high expression of OR3A4 has been observed in different types of cancer like esophageal, gastric, colon, gallbladder, pancreatic, hepatic, and

several gastric cancer cell lines (SNU-16, AGS, SNU-1, KATOIII, MKN45, NCI-N87, and SGC7901) and one immortalized gastric mucosa cell line (GES-1) (108). In another report, gastric cancer was detected via the overexpression of the lncRNA metastasisassociated lung adenocarcinoma transcript 1 (MALAT-1). MALAT-1 is associated with poor prognosis, endothelial vessels formation, and VM. Its overexpression increased migration, invasion, vascular permeability, and tumorigenicity in a nude mice model. Also, MALAT-1 regulates angiogenesis and VM through modulation of VE-cadherin/β-catenin complex, ERK/MMP, and FAK/Paxillin signaling pathways (109).

In cervical cancer, high expression of the lncRNA Ras suppressor protein 1 pseudogene 2 (RSU1P2) was associated with VM formation. Its overexpression reduces apoptosis, cell cycle progression, and the EMT in nude mice-induced tumorigenesis. This increases the cell viability, proliferation, migration, invasion, and VM. The lncRNA and miR-let-7a compete for binding sites for IGF1R, N-myc, and EphA4, which inhibits their suppressive effect. Interestingly, high expression of N-myc induced the overexpression of RSU1P2 and decreased the expression of let-7a forming a positive feedback loop (110).

In glioma tissues and glioma cell lines, the lncRNA LINC00339 was correlated with VM formation by its up-regulation. Knockdown of LINC00339 decreases cell proliferation, migration, and invasion, and led to the development of VM. There is also increased survival via reduction of tumor growth through interactions with miR-539- 5p. This system in turn increased the expression of TWIST1 that binds to promoters of MMP-2 and MMP-14 stimulating its transcription (111). Furthermore, the lncRNAHOXA cluster antisense RNA 2 (HOXA-AS2) was overexpressed in tissues and cell lines of glioma correlating with cell viability, migration, invasion, and VM formation via negative regulation of miR-373 and EGFR over-expression. EGFR enhances the expression levels of VE-cadherin as well as the activity of MMP-2 and MMP-9 metalloproteases in U87 and U251 cell lines. This favors VM development through activation of the PI3K/AKT signaling pathway (112).

Another report in gliomas found high expression of upstream transcription factor 1 and aldehyde dehydrogenase-1. These promoted cell proliferation, migration, invasion, and VM development. These molecules were related to histopathological grading. This kind of tumor has high expression of lncRNAs SNHG16 and linc00667 that induce the VM regulating of upstream transcription factor 1 and aldehyde dehydrogenase-1 targets, which are regulated by miR-212-3p and miR-429 and have low expression in gliomas. The inhibition of SNHG16 and linc00667 caused overexpression of miR-212-3p and miR-429 and the inhibition of VM (113).

In lung cancer, the estrogen receptor β interacts with different estrogen response elements located in the lncRNA MALAT-1 promoter to increase the expression. Upregulation of MALAT-1 decreases the expression of miR-145-5p and increases overexpression of the NEDD oncogene (a target of miR-145- 5p and linked to non-small cell lung cancer metastasis). This promotes VM formation and cell invasion in vitro and in vivo (114).

Li et al. reported the high expression of the lncRNA small nucleolar RNA host gene 7 (SNHG7) in colorectal cancer. SNHG7 is involved in cancer progression with poor prognosis and increased cell proliferation (in vitro and in vivo), cell cycle progression, migration, invasion, and VM formation. It blocks apoptosis in cell lines. They showed that miR-34a is a direct target of SNHG7. It regulates the expression of the GalNAc transferase 7 (GALNT7). Thus, SNHG7 could increase the expression level of GALNT7 oncogene by sponging miR-34a. They also demonstrated that SNHG7, miR-34a, and GALNT7 can increase the activity of the PI3K/AKT/mTOR pathway in colorectal cancer cell lines with different metastatic degrees (115). In triple-negative breast cancer, the high lncRNAs expression TP73 antisense RNA 1 (TP73-AS1) was associated with VM. Besides, TP73-AS1 was overexpressed in MDA-MB-231 cells and binds specifically to miR-490-3p. Furthermore, miR-490- 3p is regulated by TP73-AS1 inducing an increase in the TWIST1 (target of miR-490-3p) expression and promoting the development VM (116).

LncRNAs are regulated by various RNA-binding proteins. Li et al. demonstrated that the zinc-finger RAN-binding domain-containing protein 2 (ZRANB2) is one RNA-binding protein that is overexpressed in tissues and cell lines of glioma. Due to protein-RNA interactions, ZRANB2 stabilizes the SNHG20 lncRNA, promotes the degradation of Forkhead box K1 (FOXK1), and increases the transcription of MMP-1, MMP-9, and VE-cadherin; this stimulates proliferation, migration, invasion, and VM development in this type of cancer (117).

In HCC, high expression of n339260 lncRNA was related to the expression of stem cell markers (c-myc, sox2, and Nanog) as well as with high expression of VE-cadherin, VM formation, metastasis, poor prognosis, and low survival of the patients. Therefore, this lncRNA can induce VM through a CSC phenotype (118).

Another lncRNA related to the development of VM in lung cancer is LINC00312. Its high expression was observed in metastatic lung adenocarcinoma patients and is associated with poor survival. Overexpression of LINC00312 in mice increase the number of metastatic tumor nodules by increasing migration, invasion, and stimulation of VM. LINC00312 mediates the aforementioned effects through direct binding to the transcription factor Y-Box Binding Protein 1 (YBX1), which induces the expression of angiogenic genes such as VE-cadherin, TGF-β, VEGF-A, and VEGF-C (119).

In osteosarcoma, Ren et al. reported that lncRNAs and mRNAs are differentially expressed and are associated with VM in the extremely aggressive 143B osteosarcoma cell line. They found that lncRNA n340532 that is silenced in 143B cells by siRNA reduces the VM formation in vitro. Nude mice were injected with n340532-knockdown in 143B cells and develop smaller tumors with fewer metastatic nodules and VM channels vs. control mice (120).

These few studies show that lncRNAs-miRNAs-mRNAs play critical roles in the regulation of VM development. They have clinical implications in several types of highly aggressive cancer because they are involved with tumor progression, poor survival and prognosis, resistance to therapy, and tumor recurrence. Thus, some miRNAs and lncRNAs have been proposed as prognostic and diagnostic biomarkers in solid tumors, although the mechanisms of interaction between lncRNAs-miRNAsmRNAs have not been completely elucidated yet. A study of these interactions can better explain these regulatory networks and can explain how the cell coordinates complex events during VM. The data can also better explain the clinicopathological relationships in the development of tumors.

### CONCLUSION AND PERSPECTIVES

This review provides the most recent evidence on the impact of VM as an alternative way of generating blood vessels in tumors. The tumor microenvironment exerts a clonal selection pressure on the tumor cells to adapt to the microenvironment with low oxygen pressure and acidic/hypoxic environments. These changes promote the formation of VM in solid tumors where HIF-1α is the protagonist modulating different molecules such as VE-cadherin, EphA2, LN5γ2, MMPs, VEGF, STAT3, Bcl-2, and signaling pathways as TGF-β1, EGFR/PI3K/AKT, and RhoA/ROC. With all these antecedents, HIF-1α is considered a master regulator that promotes MV.

During EMT, the structure of the cytoskeleton of the tumor cell undergoes changes that contribute to the plasticity of the tumor; it mimics the characteristics of ECs, which help migration, invasion, and metastasis of cells. In this context, it is necessary to understand the mechanism of the EMT process and the relationship that exists with the VM because these events provide properties to the tumor cells that make the anti-angiogenic therapies inefficient. In addition, we can identify therapeutic targets that contribute to the treatment of the most frequent solid tumors that usually respond to the start of therapy and subsequently relapse and do not Hernández de la Cruz et al. miRNAs-lncRNAs Regulation During VM

respond to treatment. Examples include small cell lung cancer, pancreatic, osteosarcomas, melanomas, and sarcomas. Different miRNAs and lncRNAs may inhibit VM; they could be critical to the design of new therapeutic strategies. Nevertheless, the contribution of lncRNAs in VM remains largely unknown, few miRNAs and lncRNAs have been functionally studied in detail, and many important questions remain to be addressed. With the development of new and powerful genomic technologies, particularly next-generation sequencing, several lncRNAs can be identified in the future. These would be associated with different cancer types for their use in clinical practice.

### AUTHOR CONTRIBUTIONS

ÁC-R, OH, and JL-G organized the entire manuscript, wrote the draft, and revised the last version of manuscript. ÁC-R, RG-V, and CL-C wrote the Mechanisms of VM in Human Cancers and

#### REFERENCES


CSCs and VM sections. YS-V, DA-C, and MM-L wrote the Role of Tumor Microenvironment and the EMT in the Development of VM section. MM-L, JL-G, and DA-C wrote the section on the role of EMT in VM and wrote the section on the role of miRNAs and signaling pathway in vasculogenic mimicry in solid tumors and modulation of EMT-VM by miRNAs. YS-V, RG-V, CL-C, OH, JL-G, and CL-C wrote the section on the MiRNAs and lncRNAs and the regulation networks by lncRNAs-miRNAs and their clinical relation to the VM. **Figures 1**–**4** and **Tables 1**, **2** were designed and made by ÁC-R, MM-L, YS-V, and OH.

#### ACKNOWLEDGMENTS

The authors thank Instituto Nacional de Enfermedades Respiratorias Ismael Cosio Villegas, Universidad Autónoma de la Ciudad de México, and Universidad Autónoma Metropolitana Unidad Xochimilco.

emerging translational perspective? Semin Cancer Biol. (2017) 43:5–16. doi: 10.1016/j.semcancer.2016.12.001


metastatic potential of human triple negative breast cancer cells. Biochem Biophys Res Commun. (2017) 494:256–62. doi: 10.1016/j.bbrc.2017.10.043


formation via the MiR-373/EGFR axis. Cell Physiol Biochem. (2018) 45:131– 47. doi: 10.1159/000486253


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2020 Hernández de la Cruz, López-González, García-Vázquez, Salinas-Vera, Muñiz-Lino, Aguilar-Cazares, López-Camarillo and Carlos-Reyes. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Endothelial Phosphatase VE-PTP Participates in Vasculogenic Mimicry by Preventing Autophagic Degradation of VE-Cadherin

Daniel Delgado-Bellido, Concepción Bueno-Galera, Laura López-Jiménez, Angel Garcia-Diaz and F. Javier Oliver\*

Instituto de Parasitología y Biomedicina López Neyra, CSIC, CIBERONC, Granada, Spain

Aberrant extra-vascular expression of VE-cadherin has been observed in metastasis associated with Vasculogenic Mimicry (VM); we have recently shown that in VM prone cells VE-cadherin is mainly in the form of phospho-VE-cadherin in Y658 allowing increased plasticity that potentiates VM development in malignant cells. In the current study, we present results to show that human malignant melanoma cells VM+, express the VE-cadherin phosphatase VE-PTP. VE-PTP forms a complex with VE-Cadherin and p120-catenin and the presence of this complex act as a safeguard to prevent VE-Cadherin protein degradation by autophagy. Indeed, VE-PTP silencing results in complete degradation of VE-cadherin with the features of autophagy. In summary, this study shows that VE-PTP is involved in VM formation and disruption of VE-PTP/VE-Cadherin/p120 complex results in enhanced autophagy in aggressive VM<sup>+</sup> cells. Thus, we identify VE-PTP as a key player in VM development by regulating VE-cadherin protein degradation through autophagy.

Keywords: vasculogenic mimicry (VM), VE-PTP, vascular endothelial receptor protein tyrosine phosphatase, VE-cadherin, melanoma, autophagy

#### INTRODUCTION

The term vasculogenic mimicry (VM) describes the formation of perfusion pathways in tumors by highly invasive, genetically deregulated tumor cells: vasculogenic because they distribute plasma and may contain red blood cells and mimicry because the pathways are not blood vessels and merely mimic vascular function. While VM formation is a marker of highly invasive tumor phenotype, mechanisms by which these structures may contribute to adverse outcome are not well-understood. It has been proposed that VM formation may facilitate tumor perfusion and the physical connection between VM and blood vessels may also facilitate hematogeneous dissemination of tumor cells. There is a strong association between the histological detection of VM patterns in primary uveal and cutaneous melanomas and subsequent death from metastasis (1, 2), consistent with the in vitro observations that these patterns are generated exclusively by highly invasive tumor cells (3). ECs express various members of the cadherin superfamily, in particular, vascular endothelial (VE-) cadherin (VEC), which is the primary adhesion receptor of endothelial adherent junctions. Aberrant extra-vascular expression of VE-cadherin has been observed in specific cancer types associated with VM (4). VE-PTP (vascular endothelial protein tyrosine phosphatase) is an endothelial receptor-type phosphatase whose name was coined for

#### Edited by:

Erika Ruiz-Garcia, National Institute of Cancerology (INCan), Mexico

#### Reviewed by:

Frederique Gaits-Iacovoni, Institut National de la Santé et de la Recherche Médicale (INSERM), France Tao Sun, Nankai University, China

> \*Correspondence: F. Javier Oliver joliver@ipb.csic.es

#### Specialty section:

This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology

> Received: 21 June 2019 Accepted: 07 January 2020 Published: 24 January 2020

#### Citation:

Delgado-Bellido D, Bueno-Galera C, López-Jiménez L, Garcia-Diaz A and Oliver FJ (2020) Endothelial Phosphatase VE-PTP Participates in Vasculogenic Mimicry by Preventing Autophagic Degradation of VE-Cadherin. Front. Oncol. 10:18. doi: 10.3389/fonc.2020.00018

**248**

its prevalence to bind to VE-cadherin (5). VE-PTP poise endothelial barrier through helping homotypic VE-cadherin to keep at minimum basal endothelial permeability (6). Knockdown of VE-PTP increases endothelial permeability and leukocyte extravasation (7). VE-PTP also counterbalances the effects of permeability-increasing mediators such as VEGF, which increase endothelial permeability and leukocyte trafficking, by dephosphorylating VE-cadherin at Tyr658 and Tyr685, leading to stabilization of VE-cadherin junctions (8, 9).

p120-catenin was initially described as an Src kinase substrate, and then as a component of the cadherin-catenin complex. p120-catenin promotes cadherin stability, lowering the complex's susceptibility to endocytosis, ubiquitination, and proteasomal destruction (10). Phosphatases such as SHP-1, SHP-2, DEP1, and RPTPµ act upon p120-catenin. The RPTPµ tyrosine phosphatase binds p120 in a manner independent of p120's central Armadillo domain (11).

While studies have focused on the connection between VE-PTP and VE-cadherin in ECs. No reports have determined the role of VE-PTP in VM. Recent reports show that phospho-VE cadherin is highly expressed in VM+ cells and facilitates their pseudo-endothelial behavior by favoring p120/kaiso-dependent gene regulation (12). In the current study, we elucidated a mechanism linking VE-PTP expression with the induction of VM in metastatic melanoma cells: VE-PTP is present in the VE-Cadherin/p120 complex and the absence of VEPTP in this complex leads to autophagy. These results place VE-PTP as a dynamic component of VM transformation of melanoma cells owing to its ability to retain/safeguard VE-cadherin from being degraded by autophagy in aggressive cells.

## RESULTS AND DISCUSSION VE-PTP Expression Is Essential for VE-Cadherin Stability and to Form VM

Aberrant extra-vascular expression of VE-cadherin has been observed in specific cancer types associated with VM, and it has previously been shown that most of the VE-cadherin present in VM+ melanoma cells is phosphorylated form in Y658 (12). The current study is focused on the role of the phosphatase VE-PTP, its interaction with non-endothelial VE-cadherin and its consequences in VM development. Total VE-cadherin and VE-PTP expression were measured in different melanoma cell lines from either cutaneous (C8161, C81-61) or uveal (MUM 2B, MUM 2C) origin as shown in **Figure 1A** (protein) and **Figure 1B** (mRNA). Recently, our group reported that human malignant melanoma cells have a constitutively high expression of pVEcadherin at position Y658, pVE-cadherin Y658 is a target of focal adhesion kinase (FAK) and forms a complex with p120 catenin and the transcriptional repressor Kaiso in the nucleus (12). We have also shown that FAK inhibition enabled Kaiso to suppress the expression of its target genes and enhanced Kaiso recruitment to KBS-containing promoters (CCND1 and WNT 11). Silencing of VE-PTP induced a significant reduction of CCND1 and WNT 11 (Kaiso-dependent genes) (**Figure 1C**) and disrupted VM formation quantified by Wimasis program (**Figures 1D,E**) suggesting that VE-PTP was also involved in the intracellular dynamic of VE-cadherin resulting in the regulation of Kaiso-dependent genes. To evaluate the correlation between the levels of VE-PTP and VE-cadherin we performed a western blot after siVE-PTP in MUM 2B (**Figure 1F**) and found an almost complete vanishing of VE-cadherin and pVE-cadherin suggesting that VE-PTP was involved in VEcadherin stability and was needed for pVE-cadherin to reach the nucleus, as indirectly suggested the results obtained in **Figure 1C**. Interestingly, a completely different situation was found in primary endothelial cells HUVEC where siVE-PTP leads to an accumulation of pVEC in both cytosolic and nuclear compartments (**Figure 1G**). These results suggested that VE-PTP in malignant melanoma cells was protecting pVEC from degradation.

#### VE-Cadherin and VE-PTP Form a Complex With p120 Catenin in Melanoma Cells

The VE-cadherin-catenin complex provides the backbone of the adherent junction in the endothelium. Nonetheless, in nonendothelial cells, the proteins interacting with VE-cadherin have not been identified. Using a coIP or co-immunofluorescence (**Figure S1A**) approach to analyse the VE-cadherin and VE-PTP interacting proteins showed that VE-Cadherin forms a fragile complex with VE-PTP in MUM2B (**Figures 2A,C**) compared with the stronger complex in HUVECs (used as a positive control on VE-PTP/VE-Cadherin complex in normal endothelial cells) (13) (**Figure 2B**). Surprisingly, the presence of p120-catenin in complex with VE-PTP appears in MUM2B cells as compared to HUVEC cells (**Figures 2B,C**) suggesting that VE-PTP might be involved in the control of p120-catenin phosphorylation status in melanoma cells. To analyse the possible impact of p120 on VEcadherin stability in MUM2B, cytosol-nucleus subfractionation assay after silencing p120, found that p120 protect the stability of VE-cadherin (**Figure 2D**). Finally, we performed a coIP of p120 after siVE-PTP in MUM 2B cells, and we observed that binding of VE-cadherin to p120 was lost and resulted in increased global tyrosine phosphorylation of p120, suggesting that VE-PTP safeguard of VE-Cadherin/p120 binding in VM<sup>+</sup> cells by balance between phosphorylation and dephosphorylation of p120 through VE-PTP activity (**Figure 2E**) and p120-catenin is likely to be a substrate for VE-PTP. These results are compatible with the increased phospho-p120 (as result of VE-PTP inactivation) being responsible for complex dissociation to initiate VE-cadherin proteolysis. In fact, it has been described that only isoform 1A of p120 can be substrate of RPTPmu, correlating with the results in **Figure 2C** (increased binding of VE-PTP to isoform 1A of p120) and 2E (see the band with the arrow corresponding to pY-p120 in the blot for global p-Try (4G10) in the immunoprecipitation of p120).

## VE-Cadherin/VE-PTP Complex Dissociation-Enhanced Autophagy

To elucidate the mechanism leading to VE-Cadherin degradation after the inhibition of VE-PTP, we treated MUM2B cells with proteasome inhibitor lactacystin or MG-132 (**Figure S1B**)

cells; images were acquired using an Olympus CKX41 microscope (bars 500µm, the formation of tube-like structures was then quantified by Wimasis program. Each treatment was performed in triplicate, and the experiment was independently repeated at least three times. Results represented as fold enrichment over input. Asterisks denote statistically significant differences in an unpaired t-test (p < 0.05, \*\*p < 0.01, \*\*\*p < 0.001), and error bars denote SD, (F) to confirm the implications of VE-PTP in VM, we performed a siVE-PTP, and we show almost total decrease of VE-cadherin and Y658 expression in MUM 2B, (G) siVE-PTP in HUVEC in cytosol-nucleus subfractionation experiments showing the phosphorylation of Y658 VE-Cadherin.

and did not prevent VE-Cadherin degradation after VE-PTP disabling (**Figure 3A**) suggesting a proteasome-indenpendent pathway. Macroautophagy (referred to only as "autophagy") is a homeostatic "self-eating" pathway that has been conserved among eukaryotic cells. This is a lysosomal-associated process in which intracellular components, small portions of cytosol or chaperone-associated cargo, are engulfed in double-membrane vesicles, called autophagosomes, to be degraded with lysosomal hydrolyses (14). Recently, different studies reported that the formation of VM was promoted by bevacizumab-induced autophagy in GSCs, which was associated with tumor resistance to antiangiogenic therapy through the high expression of VEGFR-2 (15). In our study, inhibition of the fusion of autophagosomes and lysosomes with chloroquine suppressed the degradation of VE-cadherin after siVE-PTP (**Figure 3A** and **Figure S1C**). Even more, the levels of the mTOR substrate p-p70 (as a readout of mTOR activity and autophagy status) decreased in MUM 2B knockout for VE-cadherin (**Figure 3B**) suggesting that the complex VE-cadherin/VE-PTP might be restraining autophagy. Electron microscopy experiments (**Figure 3C**) were performed in MUM 2B and C8161 cell deficient for siVE-PTP or MUM 2B knockout for VE-cadherin; these results showed an enhanced autophagic morphology after the VE-PTP silencing or in VE-cadherin knockout cells, suggesting that the absence of either protein may have implications in the dynamic of protein turnover involoving the activation of autophagy. To confirm the implication of autophagy on VEcadherin degradation, we quantified autophagosomes formation under the same conditions described above after transfection of LC3-GFP and observed that the number of autophagosomes

(LC3-GFP punctuated) increased following silencing of VE-PTP in both MUM 2B and C8161 VM<sup>+</sup> cells (**Figure 3D** and **Figure S2C**). By analyzing the cBioPortal database, a platform of 48333 tumor samples, we found that high mRNA levels of PTPRB (the gene encoding for VE-PTP) were inversely associated with the expression of two essential genes involved in autophagy, LAMP1 and ATG7 in uveal melanoma (**Figure S2D**), suggesting that in patients this interaction might be relevant to determine autophagic features of the tumor.

While the role of VE-cadherin as a determinant of the pseudo-endothelial behavior of malignant melanoma cells have been widely described, no studies have addressed so far the implications of VE-PTP in VM development. Previous results have reported that VE-cadherin in VM-prone tumor cells is mostly as pVE-cadherin (Y658) and in several intracellular locations (including the nucleus) conferring the cells with the necessary plasticity to undergo pseudoendothelial differentiation (12). To get further information on the cause of these phosphorylated VE-cadherin population, we focalized in the phosphatase VE-PTP that keeps VE-cadherin unphosphorylated in endothelial cells. Despite the massive amounts of VE-cadherin, VE-PTP levels in melanoma cells were sharply diminished, suggesting that a majority of the VE-cadherin population is not in complex with VE-PTP (as compared with endothelial cells **Figures 2A,B**), then tolerating the accumulation of pVE-cadherin while endothelial cells tight junctions require a stable and abundant VE-PTP/VE-cadherin complex to keep vascular permeability strictly under control. In aggressive melanoma cells, that not presence (VE-PTP null cells) of unbound VE-Cadherin to VE-PTP initiates the proteolysis of VE-cadherin trough autophagy (increase p-p120) and finally decrease VM capacity or the reverse situation VE-PTP positive cells, increase the capacity to form VM (**Figure 4**). The question remains how a relatively small amount of VE-PTP protects from proteolysis and what signal emerges from the complex dissociation to activate autophagy. Contrary to endothelial cells,

p120-catenin is also firmly attached to VE-PTP in MUM 2B cells (**Figures 2B,C**), and the loss of this complex (after siVE-PTP silencing) leads to p120-catenin increased phosphorylation and VE-cadherin degradation by disunity of p120. Globally these results shed light to a new mechanism to control VM through the balance between VE-PTP/VE-cadherin and the phosphorylation status of p120 in aggressive melanoma VM<sup>+</sup> cells.

### MATERIALS AND METHODS

#### Reagents and Antibodies

The following reagents were used: Chloroquine 100µM during 3 h, Lactacystin 30µM during 30 min and MG-132 3µM during 3 h. Corning Matrigel Basement Membrane Matrix for in vitro angiogenesis experiments. Antibodies used were: Y658 VEC rabbit (1:1000 WB, 1:100 IF, Thermofisher), VEC C-ter mouse (1:500 WB, 1:50 IF, 2µg IP, clone F-8, sc-9989), anti-phosphotyrosine p-Tyr mouse (1:1000 WB, clone 4G10, Millipore), α-tubulin mouse (1:10000 WB, clone B-5-1-2, Sigma-Aldrich), p120 catenin mouse (1:1000 WB, 1:100 IF, 2µg IP, BD Biosciences), lamin B1 rabbit (1:1000 WB, Abcam) and VE-PTP mouse (1:1000 WB,1:100 IF, 2µg IP, Clone 12/RPTPb, BD Biosciences).

#### Cell Lines

Human melanoma cells MUM 2B, MUM 2C, C8161, and C81-61 were grown in RPMI medium supplemented with 10% fetal bovine serum, 2 mM of L-glutamine, and 1% penicillin/streptomycin (PAA laboratories). Human umbilical vein endothelial cells (HUVEC) were grown in endothelial cells growth medium-2 (EGM-2) (Lonza). All cells were cultured at 37◦C and 5% CO2 in incubator cells.

#### In vitro Angiogenesis Assay

The effect of siVE-PTP on the formation of tube-like structures in Matrigel (BD Biosciences) was determined according to the manufacturer's instructions. Briefly, 96-well plates were coated with 50µl of BD MatrigelTM Basement Membrane Matrix and allowed to solidify at 37◦C in 5% CO2 for 30 min. Cells were treated Scb, siVE-PTP transfected for 48h as described previously. After 48 h, respectively, of incubation, images were acquired using an Olympus CKX41 microscope (10X lens). The formation of tube-like structures quantified by Wimasis program. Each treatment was performed in triplicate, and the experiments independently repeated at least three times.

#### Quantitative RT-PCR

Total RNA was isolated by RNeasy Mini Kit (Qiagen) according to the manufacturer's recommendations. About 1µg of RNA from each sample was treated with DNase I, RNasinRibonuclease inhibitors (Invitrogen) and reversetranscribed using iScriptcDNA synthesis kit (Biorad) following the manufacturer's protocols. cDNA was amplified using the iTaq Universal SYBR green supermix (Biorad). Each reaction was performed in triplicate using CFX96 Real-time PCR detection systems. Primer sequences for the targets and the annealing temperature (60◦C): 36B4: Forward 5′ -CAGATT GGCTACCCAACTGTT-3′ , Reverse 5′ -GGCCAGGACTCGT TTGTACC-3, CCND1: Forward 5′ -CCGTCCATGCGGA AGATC-3′ , Reverse 5′ -GAAGACCTCCTCCTCGCACT-3′ ; WNT11: Forward 5′ -GCTTGTGCTTTGCCTTCAC-3′ , Reverse 5 ′ -TGGCCCTGAAAGGTCAAGTCTGTA-3′ , VE-PTP: Forward 5 ′ -TGCTAAGTGGAAAATGGAGGCT-3′ , Reverse 5′ -GCCC ACGACCACTTTCTCAT-3′ .

#### Gene Editing

MUM2B knockout (ko) cells for the VE-Cad gene were generated using the CRISPR-Cas9 technology. Five different sgRNAs were designed using the Zhang Lab Optimized CRISPR design tool and cloned into the pL-CRISPR.EFS.GFP which purchased from the Addgene public repository (#57818). sgRNA guides were validated in HEK293T Cells using the GeneArt Genomic Cleavage Detection Kit (Invitrogen, Carlsbad, USA) according to the manufacturer's instructions. Lentiviral particles for the best two sgRNAs in terms of allelic disruption (GGCAGGCGCCCGATGTGGCG and GATGATGCTCCTCGCCACATC).

#### Transfection of Small Interfering siRNA

Cultured cells were transiently transfected with an irrelevant siRNA (5′ -CCUACAUCCCGAUCGAUGAUG-3′ ) 50 nM. siVE-PTP: 5′ - GACAGUAUGAGGUGGAAGU−3 ′ , 50 nM, sip120: 5 ′ - GGATCACAGTCACCTTCTA−3 ′ , 50 nM, were transfected for 48 h using JetPrime (Polyplus transfection) according to the recommendations.

#### Immunobloting, Immunoprecipitation, Subfractionation Cytosol-Nucleus

For simple coimmunoprecipitation, cells were lysed in lysis buffer (50 mM Tris/HCl ph 8, 120 mM NaCl, 0,1 % NP-40, 1 mM EDTA, 10 mM NaF, 1 mM Na3VO4 and supplemented with a protease inhibitor cocktail (1 tablets to 10 ml of lysis buffer, Roche) for 30 min at 4◦C. Lysates were cleared by 13.000 rpm centrifugation for 10 min at 4◦C and incubated overnight at 4◦C with respective antibodies. Consequently, the next day, IP lysates were incubated for 2 h at 4◦C with 50µl of-of DynabeadsTM Protein G for Immunoprecipitation (ThermoFischer). Dynabeads were washed three times with low salt 120 mM lysis buffer and two times with high salt 300 mM lysis buffer. All lysates separated by dodecyl sulfate-polyacrylamide (7,5%, Biorad) gel electrophoresis and transferred to PVDF membrane (Pall laboratory) by semiwet blotting.

Accordingly with the article of Rockstroh et al. (16), for subfractionation cytosol-nucleus, cells were lysed in lysis buffer (250 mM sucrose, 50 mM Tris-HCl ph 7,4, 5 mM MgCl2, 1 mM Na3VO4, 0,25 % NP-40 and supplemented with a protease inhibitor cocktail (1 tablet to 10 ml of lysis buffer, Roche) for 10 min at 4◦C, lysates centrifuged at 500 g for 5 min, supernatant was considerate cytosolic fraction, pellet was resuspended in buffer 2 (1M sucrose, 50 mM Tris-HCl ph 7,4, 5 mM MgCl2) and centrifuged at 3,000 g for 5 min, supernatant was discarded. Pellet was resuspended in nuclear buffer (20 mM Tris-HCl ph 7,4, 0,4 M NaCl, 15% glycerol, 1,5% Triton X-100) for 45 min at 4◦C in agitation. This lysate centrifuged at 5,000 g for 5 min; the supernatant was considered a nuclear fraction.

#### Immunofluorescence

Immunostaining was performed on cells plated onto coverslips and grown for 24 h prior to experimental treatment. The cultured media was removed and wash two times with PBS 1X and the cells were fixed (Paraformaldehyde 3%, 5% sucrose) for 15 min at room temperature. Permeabilization was performed using 0,25% Triton-100 in PBS for 10 min. Before start with the antibodies incubation, cells were blocked with BSA 2% for 1 h. Respective primary antibodies were incubated for 45 min and secondary antibodies Alexa Fluor 488 anti-mouse (1:500, green) or Alexa Fluor 594 anti-rabbit (1:250, red) were incubated for 20 min. Nuclear counter staining with 4′ ,6′ -diamidino-2-phenylindole dihydrochloride (DAPI) was performed after removal of secondary antibody. Immunofluorescence images were obtained in the linear range of detection to avoid signal saturation using a fluorescent microscoper confocal microscopy (Leica SP5, 63X lens).

#### Electron Microscopy

The MUM 2B extracted were washed with PBS, prefixed for 30 min in a fixation solution (0.1 M cacodylate buffer pH 7.4 and osmium tetraoxide) for 60 min at 4◦C. After this treatment, tissues were washed with MilliQ water, and the samples were stained with uranyl acetate. The ultrathin sections were cut with a diamond knife in an ultramicrotome (Reichert Ultracut S). The samples were analyzed in a TEM Zeiss 902 with 80 KV of voltage acceleration (CIC-UGR).

### Autophagy Assay

GFP-LC3-expressing cells have been used to demonstrate the induction of autophagy. The GFP-LC3 expression vector was kindly supplied by Dr T Yoshimori (National Institute for Basic Biology, Okazaki, Japan). MUM 2B and C8161 were transiently transfected (0,5µgr) with this vector together with jetPrime (Polyplus transfection, Illkirch, France) according to the manufacturer's protocol. The assay was performed on cells grown in six-well plates. To determine LC3 localization, GFP-LC3-transfected cells were observed under a Zeiss (Zeiss Axio Imager A1) fluorescence microscope (20X lens). To determine LC3-II translocation, performed western blot of LC3-I and its proteolytic (phosphatidylethanolamine) derivative LC3-II (18 and 16 kDa, respectively) using a monoclonal antibody against LC3 (NanoTools,1:1000 WB, 1:100 IF, clone 5F10, Ref 03231- 100/LC3-5F10).

#### DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the article/**Supplementary Material**.

### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

### FUNDING

This manuscript has been released as a Pre-Print at BiorXiv (https://www.biorxiv.org/content/10.1101/634584v1). This work was supported by the grants from the Spanish Ministry of Economy and Competitiveness SAF2015-70520-R, and the Spanish Ministry of Science and Technology RTI2018-098968- B-I00, CIBERONC ISCIII CB16/12/00421, and Fundación Domingo Martínez to FO.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fonc. 2020.00018/full#supplementary-material

Figure S1 | (A) Co-immunofluorescence of VE-PTP (green) and VE-Cadherin (red), DAPI (nuclear stain, blue) in MUM 2B cells. Bars 15µm. (B) Inhibition of proteasome through MG-132 (3µM during 3 h) with or without siVE-PTP conditions not prevent the VE-cadherin degradation. (C) Co-immunofluorescence of Y658 VE-Cadherin (red) and LC3 I/II (green) with or without siVE-PTP (Cloroquine treatment: 20µM during 3 h) conditions in MUM 2B cells. Bars 15µm.

Figure S2 | (A) Quantification of autophagosomes (LC3-GFP punctuated cells) in MUM 2B cells and C8161 cells after LC3-GFP transfection (0.5µgr). (B) cBioPortal database, a platform of 48,333 tumors samples, we found that high mRNA levels of PTPRB, associated with the expression of two essential autophagy genes, ATG7 and LAMP1 in uveal melanoma samples.

## REFERENCES


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2020 Delgado-Bellido, Bueno-Galera, López-Jiménez, Garcia-Diaz and Oliver. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Resistance Mechanisms to Anti-angiogenic Therapies in Cancer

Yolla Haibe1†, Malek Kreidieh1†, Hiba El Hajj 1,2, Ibrahim Khalifeh<sup>3</sup> , Deborah Mukherji <sup>1</sup> , Sally Temraz <sup>1</sup> and Ali Shamseddine<sup>1</sup> \*

<sup>1</sup> Division of Hematology/Oncology, Department of Internal Medicine, American University of Beirut-Medical Center, Beirut, Lebanon, <sup>2</sup> Department of Experimental Pathology, Immunology and Microbiology, American University of Beirut-Medical Center, Beirut, Lebanon, <sup>3</sup> Department of Pathology and Laboratory Medicine, American University of Beirut Medical Center, Beirut, Lebanon

Tumor growth and metastasis rely on tumor vascular network for the adequate supply of oxygen and nutrients. Tumor angiogenesis relies on a highly complex program of growth factor signaling, endothelial cell (EC) proliferation, extracellular matrix (ECM) remodeling, and stromal cell interactions. Numerous pro-angiogenic drivers have been identified, the most important of which is the vascular endothelial growth factor (VEGF). The importance of pro-angiogenic inducers in tumor growth, invasion and extravasation make them an excellent therapeutic target in several types of cancers. Hence, the number of anti-angiogenic agents developed for cancer treatment has risen over the past decade, with at least eighty drugs being investigated in preclinical studies and phase I-III clinical trials. To date, the most common approaches to the inhibition of the VEGF axis include the blockade of VEGF receptors (VEGFRs) or ligands by neutralizing antibodies, as well as the inhibition of receptor tyrosine kinase (RTK) enzymes. Despite promising preclinical results, anti-angiogenic monotherapies led only to mild clinical benefits. The minimal benefits could be secondary to primary or acquired resistance, through the activation of alternative mechanisms that sustain tumor vascularization and growth. Mechanisms of resistance are categorized into VEGF-dependent alterations, non-VEGF pathways and stromal cell interactions. Thus, complementary approaches such as the combination of these inhibitors with agents targeting alternative mechanisms of blood vessel formation are urgently needed. This review provides an updated overview on the pathophysiology of angiogenesis during tumor growth. It also sheds light on the different pro-angiogenic and anti-angiogenic agents that have been developed to date. Finally, it highlights the preclinical evidence for mechanisms of angiogenic resistance and suggests novel therapeutic approaches that might be exploited with the ultimate aim of overcoming resistance and improving clinical outcomes for patients with cancer.

Keywords: VEGF, VEGF-R, bevacizumab, colorectal cancer, angiogenesis, resistance mechanisms

#### INTRODUCTION

#### Angiogenesis

Angiogenesis is the process of formation of new blood vessels from pre-existing vessels. It is a highly regulated process that involves migration, growth, and differentiation of endothelial cells (ECs). This regulated mechanism is crucial in embryonic development, wound healing, and reproduction (1). Nonetheless, alterations in any of its regulatory pathways may lead to metabolic

#### Edited by:

Erika Ruiz-Garcia, National Institute of Cancerology (INCan), Mexico

#### Reviewed by:

Luca Tamagnone, Institute for Cancer Research and Treatment (IRCC), Italy Lasse Dahl Ejby Jensen, Linköping University, Sweden

> \*Correspondence: Ali Shamseddine as04@aub.edu.lb

†These authors have contributed equally to this work and share first authorship

#### Specialty section:

This article was submitted to Cancer Molecular Targets and Therapeutics, a section of the journal Frontiers in Oncology

Received: 20 June 2019 Accepted: 10 February 2020 Published: 27 February 2020

#### Citation:

Haibe Y, Kreidieh M, El Hajj H, Khalifeh I, Mukherji D, Temraz S and Shamseddine A (2020) Resistance Mechanisms to Anti-angiogenic Therapies in Cancer. Front. Oncol. 10:221. doi: 10.3389/fonc.2020.00221

**256**

diseases, cardiovascular disorders, diabetic retinopathy, psoriasis, systemic lupus erythematosus, and importantly tumor growth and metastasis (2–5).

In the avascular phase, tumor growth is usually restricted in size due to a balance between pro-angiogenic and antiangiogenic factors that control vascular homeostasis (6). Beyond a few millimeters in size, solid tumors build, and increase their own blood supply to provide adequate oxygen and nutrients **(Figure 1)**. This process, referred to as the angiogenic switch, from an avascular state to an angiogenic phase, is crucial for tumors to grow and continue unrestricted proliferation (7). Hence, unlike normal physiological processes favoring negative regulation of angiogenesis, tumors favor its upregulation.

Multiple non-mutually exclusive mechanisms have been described as major players in tumor neovascularization. These include sprouting angiogenesis, non-sprouting angiogenesis, vasculogenesis, vasculogenic mimicry, and intussusception. Sprouting angiogenesis, however, remains the most well-studied mechanism used by tumor cells to produce their vasculature (8). Due to the importance of this latter process in tumor cell growth, invasion, and extravasation, different angiogenesis inhibitors (AIs) have been developed.

In this review, we will discuss the different driver molecules promoting angiogenesis in cancer. These include the angiogenic or angiostatic chemokines, the contribution of the endothelial progenitor cells (EPCs), the tumor vasculogenic mimicry, the markers for tumor-derived ECs, and pericytes. We will also provide an overview on the clinically tested anti-angiogenic drugs slowing down angiogenesis and leading to tumor starvation. Finally, the resistance mechanisms arising in cancer cells against these drugs and the potential therapeutic solutions will be discussed.

#### Angiogenesis: Pathophysiology During Tumor Growth

Unlike normal angiogenesis and neovascularization, tumor angiogenesis is an uncontrolled and disorganized process. It results in vessels with thin walls, incomplete basement membranes, and atypical pericytes (8). Since the needs of rapid tumor cell proliferation surpass the capacity of host vasculature, hypoxia and low supplies of nutrients characterize early stages of tumor development. Hypoxia triggers the expression of pro-angiogenic factors such as vascular endothelial growth factor (VEGF) and platelet-derived growth factor (PDGF) (9–11).

Matrix metalloproteinases (MMPs) secreted by tumor cells degrade the basement membrane as a first essential step to initiate angiogenesis (12). This alters cell-cell interactions and facilitates the migration of ECs through the created gap into the tumor mass, which in turn results in the proliferation and formation of new blood vessels, followed by vessel pruning and pericyte stabilization **(Figure 2)**.

#### Angiogenesis: Regulation

Angiogenesis is a tightly balanced mechanism regulated by both pro-angiogenic and anti-angiogenic factors (13). In malignant tumors, this balance is shifted toward a pro-angiogenic milieu to maintain sustainable angiogenic processes (14). Involved soluble growth factors include VEGF, PDGF, fibroblast growth factor (FGF)-2, angiopoietins (Angs), transforming growth factors (TGFs)- beta and alpha, and epidermal growth factors (EGF). Insoluble membrane-bound factors include integrins, ephrins, cadherins, MMPs, and hypoxia inducible factor-1 (HIF-1).

From these, VEGF was broadly studied and shown to significantly contribute to the induction and progression of angiogenesis (15). We will start by listing the different members of the VEGF family. In the following sections, a general overview on the role of the other angiogenic factors in normal and tumor angiogenesis will be described. In addition, direct and indirect angiogenesis inhibitory mechanisms will be discussed.

#### Vascular Endothelial Growth Factor Family

The VEGF family comprises seven members, VEGFs A to F and placenta growth factor (PGF) (16). These members are ligands

that interact with multiple receptors present on the vascular endothelium (17) (**Figure 3**).

#### Vascular Endothelial Growth Factor A

VEGF-A is the most potent angiogenic factor that is encoded by a gene located on the short arm of chromosome six (18). Its interaction with the transmembrane tyrosine kinase receptors, VEGF receptors (VEGFRs)-1 and 2, and their coreceptors, NRPs-1 and 2, present on vascular ECs results in the dimerization and phosphorylation of intracellular receptors (19). This further activates downstream signaling cascades involving phosphatidylinositol 3-kinase (PI3K)/protein kinase B (Akt), mitogen-activated protein kinase (MAPK), and extracellular regulated kinase (ERK) (20, 21).

VEGF-A expression is stimulated by hypoxia, growth factors, and cytokines such as IL-1, EGFs, PDGFs, and tumor necrosis factor (TNF)-α (16). It was noted in most solid tumors and some hematologic malignancies (20). VEGF-A is considered the backbone of angiogenesis during physiologic as well as pathologic processes. The deletion of one or both VEGF-A alleles in mouse pre-clinical models resulted in either vascular abnormalities or complete absence of vasculature leading to death (22). Interestingly, a striking positive correlation between the level of VEGF-A expression, tumor progression, and cancer patients' survival was observed (23, 24).

#### Vascular Endothelial Growth Factor B

VEGF-B is encoded by a gene located on chromosome eleven. It differs from VEGF-A by its promotor region (25, 26). It was found to be upregulated in many types of tumors including prostate, kidney, and colorectal cancers (CRCs) (27, 28). Since the VEGF-B promoter lacks the HIF-1 and AP-1 sites found in the VEGF-A promotor, stimuli such as hypoxia or cold do not induce VEGF-B expression (29, 30).

A study was conducted to explore the role of VEGF-B in cancer development. Results revealed that VEGF-B-deficient transgenic mice with pancreatic endocrine adenocarcinoma had

larger tumors compared to transgenic expression of VEGF-B but no difference in tumor vasculature (31). In addition, knockout studies have highlighted the role of VEGF-B in inflammatory angiogenesis and regeneration of coronary collaterals through arteriogenesis (32, 33).

#### Vascular Endothelial Growth Factors C, D, and E

The VEGF-C encoding gene is located on chromosome four (34– 36). Experiments performed on transgenic mice demonstrated the ability of VEGF-C to induce selective lymphangiogenesis without accompanying angiogenesis (37). Several studies showed a positive correlation between VEGF-C expression, lymphatic invasion, metastasis, and survival in cancer patients. For instance, while the 2-year survival rate of patients with uterine cervical cancers with high VEGF-C level in metastatic lymph nodes was 38%, that of patients with normal levels was 81% (38, 39).

VEGF-D is closely related to VEGF-C with which it shares 61% homology (40). Similar to VEGF-C, VEGF-D can bind and activate the VEGFRs 2 and 3 (41, 42). Depending on the activated receptor, separate downstream cascades are activated to induce the growth and proliferation of ECs in the vascular and lymphatic systems (43). As such, VEGF-D activity is crucial for hypoxia-induced vascular development (44) in melanoma, lung, breast, pancreatic, and esophageal cancer (43, 45–48).

VEGF-E is a potent angiogenic factor. Its isoform, VEGF-E nz-7, binds with high affinity to VEGFR-2 to stimulate efficient angiogenesis and increase vascular permeability (49).

#### Placental Growth Factor

PlGF is a member of the VEGF subfamily that binds to VEGFR-1 and its co-receptors, NRP-1 and 2. PlGF/VEGFR-1 signaling activates the downstream PI3K/Akt and p38 MAPK pathways independent of VEGFA signaling (50, 51). This stimulates the growth and migration of ECs, macrophages, and tumor cells (52, 53).

Upregulation of PlGF expression has been observed in tumors resistant to anti-VEGF therapy suggesting that PlGF might serve as a promising therapeutic target in this setting (54– 57). In addition, PlGF knockout (pgf−/−) mice were noted to have normal embryonic angiogenesis and impaired pathological angiogenesis following exposure of their tumors to ischemia (58). Thissuggests that by neutralizing PlGF, pathological angiogenesis can be inhibited without affecting normal blood vessels (59).

TABLE 1 | List of some FDA-approved anti-angiogenic agents.


#### CURRENTLY APPROVED ANTI-ANGIOGENIC THERAPIES

Since sprouting angiogenesis plays an essential role in tumor growth, invasion, progression, and metastasis, targeting this process may potentially halt the growth and spread of cancer (60). **Table 1** lists antiangiogenic agents approved for clinical use and their targets.

Angiogenesis inhibitors (AIs) are classified into direct and indirect agents. Direct endogenous inhibitors target vascular ECs and include endostatin, arrestin, and tumstatin. Unfortunately, phase II or III clinical trials did not result in significant effects on patients (14, 61). In the last decade, a number of molecules have been described, including semaphorins, netrins, slits, and others (62–64). Netrin-1, Netrin-4, and their receptors can have a repulsive or attractive signals in angiogenesis, partially via the regulation of VEGF signaling. There are still some contradictions reported on the positive and negative role of Netrin-1 in regulation of angiogenesis, and studies are still on going to identify its exact role in angiogenesis. Semaphorin-3A and Semaphorin-3E have negative effects on angiogenesis in central nervous system (CNS) and non-CNS tissues.

TABLE 2 | List of indirect angiogenesis inhibitors.


Indirect AIs target tumor cells or tumor associated stromal cells and include several types (14) **(Table 2)**. They prevent the expression of pro-angiogenic factors or block their activity.

Among the AIs, VEGF inhibitors were extensively studied and reached phase III clinical trials. They caused a modest increase in overall survival (OS) (65). Bevacizumab (BVZ), a humanized anti-VEGF monoclonal antibody, was the first drug to be approved by the Food and Drug Administration (FDA) for the treatment of metastatic colon, ovarian, renal, non-squamous cell lung cancer (NSCLC), and glioblastoma mutliforme (GBM) (66, 67). It failed to show clinical significance when used as monotherapy, except in GBM. In contrast, its clinical benefits were evident in association with other chemotherapeutic agents. For instance, since the tumor vasculature induced by VEGF is usually tortuous and dysfunctional, the use of BVZ was thought to normalize the blood vessel texture. It was also hypothesized that the combination of BVZ and chemotherapy increases the delivery of the chemotherapeutic agent to the cancer tissue by increasing its blood flow (68, 69). However, contrary evidence was reported by a decrease in cytotoxic drug delivery to tumors following treatment with AIs (70). Such inconsistency could be due to differences in blood vessel setups among various cancer types (71, 72). BVZ combined with chemotherapy was also studied in the adjuvant setting in colorectal cancer (CRC), but it failed to prove any clinical significance compared to chemotherapy alone in two phase III clinical trials (73–75).

Aflibercept is a soluble VEGF decoy receptor that consists of the extracellular domains of VEGFRs 1 and 2 and the Fc portion of human IgG1. It was FDA approved for the treatment of metastatic CRC in combination with 5-fluorouracil, leucovorin, and irinotecan in 2012 (76). Owing to its structure, Aflibercept can neutralize both, VEGF and PlGF (77). Compared

to treatment with BVZ, the use of Aflibercept in patientderived xenograft models resulted in higher tumor suppressive activity (78). Unfortunately, neutralizing both, PlGF and VEGF, had a minimal effect on tumor suppression in vivo (79). In a phase I clinical trial, relapsing GBM patients treated with BVZ monotherapy were compared to those treated with the combination of an anti-PlGF agent and BVZ. Similar results were obtained with no added benefit in the combination arm (80).

Unlike BVZ and Aflibercept, tyrosine kinase inhibitors, which are small molecules able to interact with the kinase domain on the VEGFRs, showed a remarkable clinical benefit when used as single agents, and with no added value when combined with chemotherapy. This was reported in the treatment of renal cell carcinoma (RCC), hepatocellular carcinoma (HCC), thyroid cancer, gastrointestinal stromal tumor (GIST), and pancreatic neuroendocrine tumor (PNET) (81).

### MECHANISMS OF RESISTANCE TO ANTI-ANGIOGENIC THERAPIES AND WAYS TO OVERCOME THEM

Although anti-angiogenesis therapies may prolong progressionfree survival (PFS), they have limited impact on overall survival (OS) and do not constitute a permanent cure in RCC, CRC, or breast cancer (73, 75, 82, 83). This limited clinical significance might be due to different innate and acquired molecular resistance mechanisms with no clear genetic explanations (65). Hypoxia plays an important role in tumor resistance to chemotherapeutic agents favoring more aggressive metastatic disease and hence worse prognosis. HIF-1 plays a critical role in resistance to anti-angiogenic therapy and is the main survival factor used by cancer cells to adapt to oxygen deprivation (84, 85). In this section, an overview on different mechanisms of resistance to anti-angiogenic therapies in the clinical and preclinical settings will be discussed **(Figure 4)** and the ways to overcome them will be provided **(Table 3)**. Some of these mechanisms are likely influenced by hypoxia. These include the production of alternative proangiogenic factors, the recruitment of BM-derived cells, the vasculogenic mimicry, as well as the increased tumor cell invasiveness and metastatic behavior.

### Hypoxia Caused by Anti-angiogenic Therapies

Treatment with anti-angiogenic agents results in vascular regression and intra-tumoral hypoxia. Several studies have made use of pimonidazole injections, to demonstrate an increase in hypoxic regions in primary tumors following anti-angiogenic treatment (86, 89, 115). Further analysis showed a concomitant increase in HIF-1a expression during treatment.

HIF-1a and hypoxia are known drivers of EMT, a process that promotes tumor metastasis. Upregulation of EMT-related genes, such as Twist and Snail, have been noted following anti-angiogenic treatment. This is in addition to the loss of TABLE 3 | List of mechanisms of resistance to anti-angiogenic therapies and ways to target them along with the outcomes associated with each approach.


the epithelial marker, E-cadherin, and the induction of the mesenchymal marker, vimentin (86, 116). Hypoxic environments also induce upregulation of VEGF expression through the upstream transcription factor HIF-1a (117). These factors cause tumors to acquire more angiogenic and invasive capacities, thus promoting metastasis (118).

#### Effect of Hypoxia on the Hepatocyte Growth Factor/Tyrosine Protein Kinase Met Pathway

The increase in tumor invasiveness and metastasis in response to AI-induced hypoxia from anti-angiogenic therapies can be explained by the over-expression of the tyrosine protein kinase, c-MET. For instance, in vitro studies revealed a direct positive effect of hypoxia on c-MET and phospho-c-Met expression (87). Other studies confirmed that this promotion of c-MET transcription that follows hypoxic conditions occurs via the direct regulation of HIF-1 (119).

The HGF/c-MET pathway is one of the most investigated signaling pathways in tumors resistant to anti-VEGF therapy. Binding of HGF to c-MET activates MAPK/ERK cascades, STAT3 pathway, PI3K/Akt axis, and/or NF-κB inhibitor-α kinase (IKK)- NF-κB complex (119–121). This usually promotes tumor growth and invasiveness.

VEGF exerts a negative feedback on c-MET activation in a GBM mouse model, resulting in the direct suppression of tumor invasion (122). For instance, compared to GBM patients who were not treated with BVZ, those treated with BVZ had more recurrence rates and their tumors had an upregulation in c-MET expression (123). This increased invasiveness of GBM after BVZ treatment was recently linked to inhibitory actions of VEGF and to the increase in c-Met and phospho-c-Met expression upon treatment (122).

MET activation in response to hypoxia can occur in endothelial cells, as well as in tumor cells or other cells of the tumor microenvironment. In fact, in one study (124) this had very diverse functional impacts.

#### Blocking c-MET to Overcome Resistance to Anti-vascular Endothelial Growth Factor Treatment

To overcome the c-MET protein overexpression that occurs with the neutralization of VEGF by BVZ, the addition of a c-MET inhibitor would be helpful. In the phase III METEOR trial, the administration of the inhibitor of tyrosine kinases including MET, Cabozantinib, after previous vascular endothelial growth factor receptor-targeted therapy in patients with advanced RCC resulted in improved survival (125).

#### Effect of Hypoxia on β1 Integrin Expression

It is thought that the hypoxic microenvironment generated during anti-angiogenic therapy induces HIF-1α expression, thus stimulating β1 integrin expression. β1 integrin is the member that is mostly implicated in cancer treatment resistance, especially that its expression has been upregulated in clinical specimens of BVZ-resistant GBM tumors (126–128). The expression levels of integrins are correlated with disease progression and poor survival of patients (129, 130). Upon interacting with c-MET, integrins ultimately enhance tumor cell invasiveness (113, 131, 132).

#### Blocking β1 Integrin to Overcome Resistance to Anti-vascular Endothelial Growth Factor Treatment

Several preclinical studies have demonstrated benefit from β1 integrin blockade in BVZ-resistant and non-resistant GBM tumors in xenograft models (113, 114).

#### Increased Tumor Invasiveness and Metastasis

Despite their overall inhibition of tumor growth, therapeutic AIs were associated with increased local invasiveness and distant metastasis. These phenomena seem to be major contributors to resistance against anti-angiogenesis therapies. They were first described by Ebos et al. and Paez-Ribes et al. in different preclinical models (115, 133).

Angiogenesis blockade enhances tumor invasiveness. For instance, RCC cells demonstrated an accelerated growth capacity and an invasive profile following treatment with BVZ (134). Similarly, GBM cells in mouse models developed enhanced invasiveness following VEGF inhibition (115).

Treatment with AIs also promotes tumor metastatic potential. Treatment with sunitinib has been shown to result in vascular changes that include decreased adherens junction protein expression, reduced basement membrane and pericyte coverage, and increased leakiness (89, 91, 135, 136). These phenotypic changes were observed in both, tumor vessels and normal organ vessels, so they tend to facilitate local intravasation and extravasation of tumor cells, resulting in metastatic colonization (136).

#### Factors Promoting or Affecting Tumor Invasiveness and Metastasis

Increased metastasis and enhanced invasiveness in response to anti-angiogenesis therapy are variable and depend on the treatment type, dose, and schedule. Singh et al. observed that sunitinib and anti-VEGF antibody monotherapy had different effects on mouse tumor models. While treatment with sunitinib enhanced the aggressiveness of tumor cells, using an anti-VEGF antibody did not (91). This was supported by Chung et al. who compared the efficacy of different RTK inhibitors and antibody therapies in murine models (135). While pretreatment with imatinib, sunitinib, or sorafenib enhanced lung metastasis following the injection of 66c14 cells, using an anti-VEGFR2 antibody inhibited the formation of lung nodules (135). Altogether, these results prove that the increased metastasis and enhanced invasiveness that result from use of AIs are largely dependent on treatment type.

Dosing and scheduling of administration of AIs can also induce resistance. Indeed, treatment with short-term and high-dosage sunitinib (120 mg/kg per day) before and after intravenous breast tumor cell inoculation into severe combined immune-deficient mice had the most deleterious effects (133). The high-dose of sunitinib increased tumor growth and enhanced metastasis to the liver and lung, resulting in reduced survival. Although similar results were observed using sorafenib, contradictory results were reported with sunitinib in different studies (115, 133). In fact, treatment with high-dose sunitinib before intravenous inoculation of tumor cells increased metastatic potential of lung cancer cells but not of RCC cells. In contrast, treatment with low-dosage sunitinib (30 and 60 mg/kg per day) did not stimulate metastasis (136).

It was documented that hypoxia and EMT also contribute to the increased invasiveness and metastasis of tumors, and c-Met, Twist, and HIF-1a are the key molecular players (11, 116). In contrast, semaphorin 3A (Sema3A), an endogenous antiangiogenic molecule, is frequently lost in tumors, resulting in increased invasiveness and metastasis (137).

#### Overcoming Resistance by Targeting Increased Tumor Invasiveness and Metastasis

Different inhibitors of c-Met were tested in preclinical studies and demonstrated promising effects. Crizotinib, a dual c-Met and ALK inhibitor, was effective in reverting sunitinibinduced invasion and metastasis in different models (86–88). Interestingly, this resulted in a reduction in the expression of EMT markers such as Vimentin, Snail, and N-cadherin downstream of c-Met (86, 87). By blocking c-Met and silencing Twist, the master regulator of EMT (138), metastasis was almost fully abrogated in both wild-type and pericyte-depleted tumors (86).

Sunitinib-treated transgenic mice tumors that were subjected to adenoviral Sema3A expression witnessed an impressive increase of 10 weeks in median survival and a reduction in metastasis and hypoxia (89). Normalization of the tumor vasculature was evident, and the expression of EMT markers, including c-Met, were reduced.

Rovida et al. investigated the use of conventional chemotherapeutics to counteract sunitinib-induced metastasis. Gemcitabine and topotecan, but not paclitaxel, cisplatin, and doxorubicin, were effective in reverting sunitinib-induced metastasis and in reducing primary tumor growth (90). Mechanistically, topotecan was shown to inhibit HIF-1a accumulation, thereby preventing hypoxia-driven invasiveness. Gemcitabine was moderately effective in combination with anti-VEGF antibody therapy in an established pancreatic ductal adenocarcinoma model but had no effect in a preventive setting (91).

#### Redundancy in Angiogenic Signaling Pathways

Initially, the primary focus in angiogenesis blockade was to target VEGF, which is the best known angio-stimulatory protein family responsible for EC activation and functional vessel formation and stabilization. Cancers that are highly dependent on the induction of angiogenesis by VEGF, were the best responders to anti-VEGF agents. These include CRC, RCC, and neuroendocrine tumors (139).

Cancers relying on angiogenic factors other than VEGF are less susceptible to anti-VEGF agents and include malignant melanoma, pancreatic cancer, breast cancer, and prostate cancer (98). The presence of several anti-VEGF resistant cancers suggests alternative angiogenic pathways. These involve Ang-1, EGF, FGF, granulocyte colony-stimulating factor (G-CSF), hepatocyte growth factor (HGF), insulinlike growth factor, PDGF, PGF, stromal cell-derived factor-1 (SDF-1), and TGF (140). Except for P1GF, which binds VEGF receptors, most angiogenic factors signal through specific transmembrane receptors, which are expressed on ECs (141). This variety of growth factors culminates in a plethora of pathways that tumor cells can exploit to induce angiogenesis.

Results from preclinical models and clinical trials suggest that inhibition of a specific growth factor can induce the expression of others (140, 141). In a study by Willett et al. in which rectal cancer patients were treated with BVZ, significantly increased plasma levels of PlGF were noted 12 days following the start of treatment (142). In a phase II study by Kopetz et al. in which metastatic CRC patients were treated with a combination of FOLFIRI and BVZ, the levels of several angiogenic factors including PlGF and HGF were found to increase before disease progression (54). Similarly, the levels of FGF2 and PlGF increased in GBM patients following treatment with cediranib, a pan-VEGF receptor tyrosine kinase inhibitor (71, 143). Similarly, treatment of transgenic mouse models of pancreatic tumors with an anti-VEGFR2 antibody for a prolonged period of time, associated with an increase in the expression of the proangiogenic growth factors, Ang-1, Ephrin-A1, Ephrin-A2, and FGF1, FGF2a, resulting in transient tumor growth delay and modest survival benefit (98, 144).

Redundancy in angiogenic signaling and potential in malignant tissues is nowadays more studied. In addition, the therapeutic effect of targeting a single angiogenic growth factor or its receptor became limited due to intrinsic resistance. This resistance arose either from redundancy in activated pathways or alternative growth factor signaling pathways. Thus, targeting multiple growth factors simultaneously or sequentially would be a successful approach to overcome such resistance. In the following subsection, we discuss potential angiogenic factors that might play a role in the escape from anti-VEGF treatment. We also shed light on results of studies evaluating the effects of targeting one or more of these factors on overcoming resistance to anti-VEGF therapies.

#### Angiopoietin

#### **Role of angiopoietin in the escape from anti-vascular endothelial growth factor treatment**

Ang-Tie signaling system is a vascular-specific RTK pathway that regulates vascular permeability and blood vessel development and remodeling through Ang-1 and Ang-2. Ang-1 binds to the Tie2 receptor on the M2 subpopulation of monocytes, HSCs, and ECs of blood and lymphatic vessels. This activates the Ang-Tie pathway and results in the maturation or stabilization of blood vessels (145). In contrast, Ang-2 blocks this pathway resulting in the remodeling or initiation of vascular sprouts following exposure to VEGF (146). Upregulation of Ang-2 expression was described in many types of cancers and presumable contributes to resistance against anti-VEGF therapy (147–151). For example, in CRC patients, elevated serum Ang-2 levels were associated with a poor response to BVZ treatment (152).

#### **Targeting angiopoietin to overcome resistance to anti-vascular endothelial growth factor treatment**

Blockade of both, VEGF and Ang2, in preclinical studies suppressed revascularization and tumor progression of cancers resistant to anti-VEGF therapy (92–95). However, results of ongoing clinical trials evaluating the efficacy of the humanized bi-specific monoclonal antibody against VEGF-A and Ang-2, vanucizumab, are still pending (153, 154).

#### Bombina Variegate Peptide 8 (Bv8)

#### **Role of bombina variegate peptide 8 in the escape from anti-vascular endothelial growth factor treatment**

Tumor-infiltrating T helper type 17 (Th17) cells produce interleukin-17 (IL-17), initiating a paracrine network to confer resistance to anti-VEGF therapy (38). IL-17 induces G-CSF secretion by tumor cells through nuclear factor κB (NF-κB) and ERK signaling (155). The increase in G-CSF induces the expression of Bv8, also known as prokineticin-2, in the bone marrow. Bv8 is a pro-angiogenic growth factor that was initially purified from the skin secretion of a yellow-bellied toad. It binds to the G-protein coupled prokineticin receptor (PROKR) and activates the downstream MAPK/ERK pathway (156, 157). As such, Bv8 promotes differentiation of myeloidderived **(**suppressor) stem (remove word stem) cells (MDSCs) and induces their mobilization to the peripheral blood and infiltration into the tumor microenvironment. This culminates in the promotion of angiogenesis and results in the escape from anti-VEGF therapy (158–161).

#### **Targeting bombina variegate peptide 8 to overcome resistance to anti- vascular endothelial growth factor treatment**

Treatment with the Bv8 antagonist, PKRA7, suppressed tumor formation in vivo by inhibiting angiogenesis in GBM and infiltration of MDSCs in pancreatic cancer (96). Neutralization of Bv8 and upstream G-CSF using monoclonal antibodies also resulted in tumor suppression (162). Results of ongoing clinical trials evaluating combination regimens using Bv8 inhibitors with or without other anti-angiogenic reagents are still pending.

#### Fibroblast Growth Factor (FGF)

#### **Role of fibroblast growth factor in the escape from anti-vascular endothelial growth factor treatment**

The FGF family consists of 22 members. Four of these are intracellular cofactors of voltage-gated sodium channels, while the remaining 18 members are secretory proteins that bind to RTK–FGF receptors (FGFRs) (163). FGFR is expressed on tumor cells and several types of stromal cells, including cancer-associated fibroblasts (CAFs), ECs, and tumor-infiltrating myeloid cells (164).

Binding of FGF to RTK–FGFR activates the downstream pathways such as MAPK/ERK, PI3K/Akt, STAT, and diacylglycerol (DAG)/protein kinase C (PKC) (165–168). One of the roles of this signaling pathway is cancer development and progression through the amelioration of angiogenesis (164, 169). Indeed, upregulation of FGF2 expression correlated with resistance to anti-VEGF agents in several tumors resistant, especially those exposed to hypoxic environments (54, 71, 98, 170).

#### **Targeting fibroblast growth factor to overcome resistance to anti- vascular endothelial growth factor treatment**

Simultaneous blockade of VEGF and FGF signaling pathways was very beneficial in many preclinical models of cancer (98, 171– 173). Combining the FGFR inhibitor, PD173074, with BVZ in xenografted mouse models with head and neck squamous cell carcinoma (HNSCC) completely abolished tumor growth (97). FGF blockade using the soluble FGF receptor, FGF-trap, was combined with an VEGFR2 inhibitor, and yielded comparable results in late stage pancreatic islet tumors (98). Unfortunately, in the clinical setting, patients with recurrence following anti-VEGF therapy did not benefit from the dual blockade of VEGFR and FGFR by dovitinib or nintedanib (99, 100).

#### Platelet-Derived Growth Factor

#### **Role of platelet-derived growth factor in the escape from anti-vascular endothelial growth factor treatment**

The PDGF family consists of four homodimers and one heterodimer. Binding of the PDGF dimers to tyrosine kinase PDGF receptor (PDGFR) results in the activation of downstream signal transduction pathways, such as PI3K and PLCγ (174). This plays an important role in mesenchymal cell growth and motility during embryonic development and tissue repair (175). When PDGF signaling is over-active in the tumor microenvironment, angiogenesis and tumor growth are promoted (176). Upregulation of PDGF-C expression was observed in vivo in CAFs infiltrating into tumors resistant to anti-VEGF therapy (101).

#### **Targeting platelet-derived growth factor to overcome resistance to anti- vascular endothelial growth factor treatment**

Sunitinib has many targets, including VEGFR and PDGFR. Following its FDA approval in 2006 for the treatment of metastatic RCC, it was assumed that combining PDGF and VEGF blockades might offer an additional therapeutic benefit (101). Several studies were initiated to evaluate the safety and efficacy of this combination (177). Unfortunately, combining BVZ with imatinib, which inhibits PDGF-R in addition to other tyrosine kinases such as Abl and Kit, was toxic and not effective treatment against RCC (102–104).

#### Transforming Growth Factor-β

#### **Role of transforming growth factor-**β **in the escape from anti-vascular endothelial growth factor treatment**

The TGF-β/Activin and bone morphogenetic protein (BMP) are the two main branches of the TGF-β superfamily. When TGFβ binds its type II receptors, it activates type I receptors and results in the phosphorylation of the receptor-regulated Smads (R-Smads) corresponding to each branch. R-Smads then complex with the common partner Smad4 (Co-Smad4) and work as transcription factors (178).

TGF-β signaling regulates cellular growth, differentiation, and apoptosis (179). Although signaling has tumor suppressive effects during the early stage, it switches toward malignant conversion and tumor progression at later stages (180, 181). It activates the production of extracellular matrix (ECM) by fibroblasts and stimulates tube formation by ECs, thus inducing angiogenesis (182–184).

Tumor tissues express higher levels of TGF-β and these levels can be correlated with patient survival (185–187). Upregulation of TGF-β expression was also observed in glioma models resistant to anti-VEGF therapy (188). This suggests a role of TGF-β in the acquired resistance to antiangiogenic therapy.

#### **Targeting transforming growth factor-**β **to overcome resistance to anti- vascular endothelial growth factor treatment**

Several preclinical studies revealed the anti-angiogenic benefits when inhibiting TGFβ in CRC, HCC, and GBM xenografts (189– 191). This offers the rationale to combine TGFβ inhibitors with anti-VEGF agents (192). In that sense, combining galunisertib, a small molecule inhibitor of TGFβRI, with sorafenib and ramucirumab in HCC is currently under evaluation (189, 193). Similarly, the combination of an anti-TGFβ monoclonal antibody, PF-03446962, with regorafenib in CRC is also under investigation (194).

#### Matrix Metalloproteinases

#### **Role of matrix metalloproteinases in the escape from anti-vascular endothelial growth factor treatment**

MMPs play an important role in angiogenesis and in different stages of cancer (195, 196). They are divided into six categories **(Table 4)** (197). MMP can promote or inhibit angiogenesis. For instance, the secreted MMP-9 plays an important role in the angiogenic switch process and in releasing VEGF from the ECM (1, 198). The membrane type MMP-1 induces degradation and remodeling of matrix during vascular injury and is responsible for invasion and migration of ECs and formation of capillaries (199–201). On the other hand, MMPs such as MMP-3, 7, 12, 13, and 20, inhibit angiogenesis through endostatin and angiostatin production. Endostatin that blocks the activation of pro-MMP-9 and inhibits capillary formation of Deryugina and Quigley (202).

#### **Targeting matrix metalloproteinases to overcome resistance to anti- vascular endothelial growth factor treatment**

Targeting MMPs released by bone marrow derived cells (BMDCs) prevents the release of sequestered growth factors in the ECM, and can help overcoming resistance to anti-angiogenic therapy (203). Despite the fact that doing so has proven some clinical efficacy in patients with advanced and refractory solid tumors in a phase I clinical trial (105), most MMP inhibitors failed to offer any clinical benefit (204). Few agents are still being developed and evaluated. Results from an ongoing phase II clinical trial evaluating one MMP inhibitor in patients with Kaposi's sarcoma are still pending (205).

#### Recruitment of Bone Marrow-Derived Cells

Long-term administration of AIs up-regulates HIF-1α and induces hypoxia in the tumor microenvironment by overpruning blood vessels (206). Hypoxic conditions due to antiangiogenic therapy result in the expansion and recruitment of myeloid cells and CAFs into the tumor environment. The presence of these BMDCs in the tumor microenvironment leads to a weakened antitumor response and an immunosuppressive tumor microenvironment (207). This promotes angiogenesis, tumor growth, EMT transition, and metastasis (208, 209). As a result, it has become evident that myeloid cells and CAFs play a major role in the induction of resistance to anti-angiogenic drugs.

#### Myeloid Cells

#### **Recruitment of myeloid cells**

Myeloid derived suppressor cells (MDSCs), also known as Gr1+ CD11b+ myeloid cells, consist of neutrophils, macrophages, and dendritic cells (DCs). An excessive production of MDSCs was described in cancer patients and tumor-bearing mice (210–213). This was linked to the immunosuppressive and tumor promoting capacities (214, 215). In a study by Shojaei et al., resistant tumors to anti-VEGF treatment had increased mobilization and infiltration of MDSCs into their microenvironments as compared with treatment-sensitive tumors (216).

Neutrophils are considered predictive biomarkers for patients treated with BVZ (217–222). Increased recruitment of neutrophils during anti-VEGF therapy promotes tumor progression and treatment resistance (216). This is mediated by the expression of the calcium-binding protein that regulates cell growth, survival, and motility, S100A4. As such, blocking granulocytes and S100A4 may be beneficial in diminishing anti-angiogenic therapy resistance (223).

TABLE 4 | Categories of Matrix Metalloproteinase-1 and their corresponding members.


metalloproteinases

Monocytes and macrophages are possibly implicated in resistance to anti-angiogenic therapy as well. Recruitment of these cells to the tumor microenvironment is mediated by different cytokines, including VEGF, chemokine C-C motif ligand 2 (CCL2), and macrophage colony stimulating factor (MCSF) (224, 225). Tumor associated macrophages actively participate in vascular sprouting by functioning as bridging cells between two different tip cells (226–228). They also secrete MMPs, promotingangiogenesis (198, 226, 229, 230). In addition, they can release pro-angiogenic growth factors including TGF-b, VEGF, EGF, and the chemokines, CCL2 and CXCL8 (226, 227, 231–233).

In different murine tumor models, anti-VEGF therapy reduced macrophage infiltration (217, 234–236). However, this was not the case with the tyrosine kinase with immunoglobulinlike and EGF-like domains 2 (TIE2)-expressing macrophages that constitute a specific subset of macrophages. These are usually recruited by HIF1a and tumor-secreted chemokines such as ANG2 in the setting of anti-angiogenic therapy (237– 240). They tend to associate with tumor vessels and release proangiogenic growth factors including VEGF (237, 241). As such, macrophages contribute to the resistance against anti-angiogenic therapy. Preclinical studies on models of mammary carcinoma and insulinoma evaluated the effect of inhibiting ANG2 on TIE2-expressing macrophage infiltration and angiogenesis. Although this approach did not block the recruitment of these macrophages, it hindered the upregulation of their TIE2 receptor. This reduced the production of pro-angiogenic growth factors and the association of TIE2 macrophages with blood vessels (242–244). As a result, MDSCs represent promising targets for therapy. Since G-CSF expression stimulated by tumor infiltrating T helper type 17 cells results in MDSC recruitment into the tumor microenvironment, inhibition of Th17 cell function might sensitize tumors to anti-VEGF therapies (155, 207).

#### **Targeting myeloid cells to overcome resistance to antivascular endothelial growth factor treatment**

Since SDF1 is the major BMDC recruiting factor, targeting its signaling pathway could potentially decrease BMDC infiltration and overcome resistance to anti-angiogenic therapy. In a transgenic mouse model of breast cancer, treatment with an SDF1 neutralizing antibody inhibited MDSC infiltration and angiogenesis (106). Since Bv8 leads to the recruitment of MDSCs into the tumor tissue after VEGF blockade, its inhibition can possibly improve the effect of anti-angiogenic therapy. A recent study showed that the combination of gemcitabine and an anti-Bv8 monoclonal antibody treatment in mice with adenocarcinoma inhibited tumor regrowth, angiogenesis, and metastasis (107). In addition, anti-Bv8 antibodies blocked MDSC recruitment and tumor angiogenesis in an RIP1-Tag2 insulinoma model of pancreatic cancer (245).

Blocking the recruitment of monocytes and macrophages can be another therapeutic opportunity to overcome resistance to anti-angiogenic therapy. In a phase I clinical trial, patients with solid tumors were treated with the human anti-CCL2 monoclonal antibody, carlumab, which targets the monocyte chemotactic protein-1 (MCP1). In addition to causing a drop in free CCL2 levels and a reduction in the level of tumorinfiltrating macrophages, this therapy resulted in a temporary antitumor activity (108). Treatment of RIP1-Tag2 pancreatic neuroendocrine tumors with combined ANG2 and VEGFR2 blockers decreased infiltration of TIE2 expressing monocytes and suppressed revascularization and tumor progression (92). Since macrophages express colony stimulating factor-1 receptor, its targeting is currently being evaluated by several phase I clinical trials (NCT01346358; NCT01004861; NCT01596751). This is supported by results from earlier studies showing a reduced macrophage infiltration into tumor tissue and clinical objective responses following treatment of diffuse-type giant cell tumor patients with the anti-colony-stimulating factor-1 receptor antibody, RG7155 (246).

Macrophage Migration Inhibitory Factor (MIF) suppresses the anti-inflammatory activity of macrophages. TAMs, mainly M2-polarized macrophages, stimulate angiogenesis thus promoting tumor cell migration and progression (247). VEGF increases MIF production in a VEGFR-dependent manner. Compared to tissue specimens of BVZ-sensitive GBM patients, BVZ-resistant ones had a decreased MIF expression and an increased TAM infiltration (248). As such, blocking the VEGF pathway using BVZ can deplete MIF expression. This explains the enhanced recruitment of TAM and M2 in BVZ-resistant GBM tumors. Data is lacking when it comes to evaluating the application of this target in the clinical setting.

#### Endothelial Progenitor Cells

#### **Recruitment of endothelial progenitor cells**

Anti-angiogenic therapy causes hypoxia which results in the activation of HIF1a in tumor cells (249). This causes tumor cells to secrete SDF1 and VEGF,main chemotactic factors for EPCs (209, 215, 250, 251). Upon stimulation of the C-X-C chemokine receptor-7 (CXCR7) by SDF1, EPCs secrete proangiogenic cytokines and promote angiogenesis (252, 253). For instance, in multiple myeloma, this occurs through regulating the trafficking of angiogenic mononuclear cells into areas of tumor growth (254). EPCs can also promote angiogenesis by differentiating into ECs and subsequently incorporating into newly forming blood vessels.

### Recruitment of Local Stromal Cells Pericytes

#### **Recruitment of pericytes**

Pericytes, also known as Rouget cells, are cells that interact with ECs. They regulate endothelial proliferation and differentiation and modulate vessel diameter and permeability, thus stabilizing the newly formed endothelial tubes (255, 256). In a study by Abramsson et al., paracrine co-signaling via PDGF-B and PDGFR-b played a major role in pericyte recruitment to ECs (257).

Several studies revealed enhanced pericyte recruitment to and coverage of the microvasculature in the tumor after treatment with AIs. Reduction in tumor vascularity following anti-VEGF therapy is accompanied by a tightly pericyte covered vessels (258). For instance, after treatment with sunitinib and the chemotherapy drug, temozolomide, a preclinical malignant glioma model revealed an increased number of vessels covered with pericytes (259). In addition, esophageal and ovarian cancer xenografts showed increased pericyte coverage around vessels following treatment with BVZ (260).

Tumor vessels that are heavily covered by pericytes have a reduced sensitivity for anti-angiogenic therapies (261) As such, the increase in pericyte infiltration was suggested to be a mechanism of resistance to anti-VEGF and anti-VEGFR therapies. By suppressing EC proliferation and by providing survival signals that contribute to the maintenance of ECs, pericytes mediate vascular maturation and stability hence allowing tumor cells to proliferate during the course of an antiangiogenic therapy (262–264). As a result of protecting ECs from anti-angiogenic agents, pericytes were implicated in clinical resistance to VEGFR inhibitors (249).

While there is a broad consensus on the fact that pericytecovered vessels are less sensitive to AI, several recent studies have highlighted that tumor vessels typically lack pericyte coverage due to their immaturity and rapid growth phase while normal quiescent vessels are well covered (265–267). This could identify a selective therapeutic window to target abnormal tumor blood vessels, rather than suggesting to target pericyte coverage.

In keeping with that, accumulating evidence supports the idea that—in addition to pruning non-covered vessels- cancer therapies should aim at promoting the establishment of a normal vasculature in tumors in order to favor wide distribution of standard chemotherapeutics and innovative drugs into the tumor mass and improve radiotherapy efficacy. This process is known as "vascular normalization" that many adopt as the future of anti-angiogenic therapy. By therapeutically improving, rather than reducing, the stability and function of tumor blood vessels, these may be exploited for delivery of therapeutics including endogenous anti-cancer immune cells. This would also improve perfusion, reduce hypoxia, and thereby reduce metastasis. Tumor vessel normalization for cancer therapy has been achieved by the application of molecules directly targeting endothelial cells, such as semaphorins (268, 269).

Although ANG1 is a growth factor that provides ECs with survival signals, its introduction in CRC tumor cells displays an anti-angiogenic therapy in one study (270). Although this approach was accompanied by a major increase in tumor microvessel pericyte coverage, it resulted in smaller tumors with less vasculature, suggesting a decreased sensitivity for angiogenesis (270). In a more recent study, tumorbearing mice were treated with antibodies against ANG2A, and a similar observation was noted (261). Combining the chemotherapeutic agent, topotecan, with pazopanib significantly inhibited tumor growth, despite an increase in the number of vessels that were infiltrated by pericytes (271). Similar results were observed in a preclinical malignant glioma model following treatment with the combination of temozolomide and sunitinib (272).

#### **Targeting pericytes to overcome resistance to anti- vascular endothelial growth factor treatment**

Targeting blood vessel maturation by inhibiting pericyte coverage of the tumor vasculature was suggested as a promising strategy, to break the resistance to anti-angiogenic therapies and improve their efficacy. ECs secrete PDGF-B that mediates migration and proliferation of pericytes expressing PDGFR-b (273). Since SDF1, and the heparin-binding EGF-like growth factor also play a major role in pericyte behavior (274), blocking the PDGF pathway alone might not be sufficient to prevent pericyte coverage of vasculature.

Although several studies showed that targeting pericytes and ECs leads to impaired tumor growth and improved efficacy to anti-angiogenic agents, data negating the potentiation of treatment outcome with dual blockade exists (275). For instance, in a study by Nisancioglu et al., treatment of lung cancer in pericyte-deficient PDGF-B (ret/ret) mice with the anti– VEGFA antibody, G6-31, did not have any additional anti-tumor benefit (276).

Other pathways like sphingosine-1-phosphate (S1P)/edg-1, TGF-b1/Alk5, or MMPs should be considered while trying to overcome resistance associated with pericyte coverage (277). As a result, anti-pericyte agents should always be combined with other therapies, including chemotherapeutic agents. For instance, in a preclinical study by Pietras et al., transgenic mouse models of cancer were treated with a combination of the two anti-PDGFR agents, imatinib and SU11248, cyclophosphamide, and/or an anti-VEGFR agent (109). Compared to monotherapies, combination therapies significantly improved anti-tumor responses. Of note, the combination of all three approaches resulted in complete responses. Also, treatment of neuroblastoma mouse xenograft models with a combination of metronomic topotecan and pazopanib resulted in a sustained anti-angiogenic effect. but induced resistance mediated by elevated glycolysis (109).

#### Cancer-Associated Fibroblasts

#### **Recruitment of cancer-associated fibroblasts**

CAFs are activated by growth factors released from tumor and inflammatory cells, including TGFb, PDGF, and FGF (169, 278, 279). CAFs also secrete several pro-angiogenic growth factors, including EGF, HGF, and FGF. For instance, VEGF-producing CAFs maintain tumor angiogenesis in VEGF-deficient tumor cells (280).

When CAFs were isolated from a mixture of EL4 tumors resistant to anti-VEGF agents and TIB6 tumors sensitive to anti-VEGF agents, they were able to promote tumor cell proliferation and growth even when VEGF was blocked. When CAFs were isolated from TIB6 tumors sensitive to anti-VEGF agents, no tumor growth was observed (215). This supports the role of CAFs in the acquired resistance to anti-angiogenic therapy. Further analysis revealed an upregulation in the expression of proangiogenic genes in CAFs derived from therapy-resistant tumors, and these included PDGF-C and Ang-like protein 2. As a result, it is assumed that a PDGF-C neutralizing antibody could be used in the treatment of tumors refractory to anti-VEGF agents (215).

CAFs can promote tumor growth and angiogenesis through the release of certain growth factors and proteases. For instance, CAFs secrete the chemokine SDF1 which directly stimulates tumor cells and recruits EPCs and other BMDCs into the tumor tissue (250, 251). They also produce proteases, including MMPs that stimulate the release of matrix-bound pro-angiogenic growth factors, thus promoting angiogenesis and resistance to anti-angiogenic agents (281–283).

#### **Targeting cancer-associated fibroblasts to overcome resistance to anti- vascular endothelial growth factor treatment**

Targeting CAFs might play a role in overcoming resistance to anti-angiogenic therapy. Treatment of nude mice human HCC xenografts with the anti-FGF2 monoclonal antibody, GAL-F2, inhibited tumor growth and angiogenesis by blocking the effect of the proangiogenic FGF in CAFs. Also, the addition of an anti-VEGF antibody or the tyrosine kinase inhibitor, sorafenib, led to an additive treatment effect (110). Similarly, treatment of patients with recurrent and persistent endometrial cancer with the dual VEGFR/FGFR inhibitor, brivanib, extended their progressionfree survival (PFS) by blocking the effect of the proangiogenic VEGF in CAFs. (111). Neutralization of PDGF-C suppressed CAF-mediated tumor progression.

### Adoption of Different Neovascularization Modalities

Besides acquiring resistance to angiogenesis inhibition through growth factor redundancy and recruitment of different cells, tumor cells may also escape the effect of AIs by adopting different neovascularization modalities (284–286). These include vascular co-option and vasculogenic mimicry.

#### Vessel Co-option

#### **Role of vessel co-option in the escape from anti-vascular endothelial growth factor treatment**

Vessel co-option refers to the process by which cancer cells incorporate into and grow along pre-existing vessels rather than inducing new vasculature (287). This strategy provides oxygen and nutrients for efficient tumor outgrowth. It was first described in brain tumors arising from well-vascularized brain parenchyma (288). For instance, vessel co-option was also observed in gliomas and other cancer types including lung cancers (289). It was shown to sustain the growth of cerebral metastases from melanomas, liver metastases from breast cancers and NSCLCs, and lung metastases from different primaries (290, 291). Interestingly, vessel co-option is independent of the classic angiogenic switch and doesn't require any angiogenic growth factors. As such, vessel co-opting tumors are usually not sensitive to anti-angiogenic agents. For example, patients with CRC and liver metastases demonstrated a poor response to BVZ therapy due to vessel co-option.

An interesting question is whether this process represents an intrinsic resistance mechanism to anti-angiogenic therapies or whether it occurs in response to treatment. According to results from several studies, an increase in vessel co-option tends to follow, rather than precede, the inhibition of angiogenesis (292). For instance, the use of an anti-VEGF antibody in GBM patients resulted in an increase in vessel co-option (293, 294). Similarly, the growth cerebral melanoma metastasis was sustained by vessel co-option following treatment with the anti-angiogenic agent, ZD6474 (290). Nevertheless, more data is needed to check whether this applies to different tumor types and to evaluate its impact in the clinical setting.

#### Vasculogenic Mimicry

#### **Role of vasculogenic mimicry in the escape from anti-vascular endothelial growth factor treatment**

Vasculogenic mimicry refers to the process in which vascular-like structures are formed by tumor cells, after they trans-differentiate and gain features of ECs such as the expression of the endothelial markers, VE-cadherin, TIE1, and ephrin A2 (295, 296). Since no new blood vessels are formed, this phenomenon is different from vasculogenesis and angiogenesis. Nevertheless, the fact that blood can still be transported through the vascular-like networks and tumors can still be well-oxygenated, vasculogenic mimicry strongly associated with poor patient survival. This process was described in different tumor types, including gliomas, malignant melanomas, sarcomas, and breast cancers (284, 297–300).

Since tumor cells trans-differentiate into endothelial-like cells as part of vasculogenic mimicry, it might be assumed that the process can be inhibited by anti-angiogenic agents. However, tumor cells that make use of this phenomenon were not found to develop sensitivity to anti-angiogenic therapies in early studies (301). Instead, they were shown to upregulate this process following treatment with BVZ or induction of hypoxia by several preclinical studies (302, 302, 303). As such, vasculogenic mimicry might serve as an escape mechanism from anti-angiogenic therapies. The idea of combining AIs with chemotherapeutic agents has been suggested but more data is needed to evaluate its impact in the clinical setting.

#### **Targeting vasculogenic mimicry to overcome resistance to anti- vascular endothelial growth factor treatment**

Following the emergence of vasculogenic mimicry as an alternative vascular-like network in tumors, researchers have realized the importance of combining angiogenesis inhibition with an anti-tumor cell strategy. This is particularly challenging because the transition of tumor cells into a more stem cell–like phenotype is linked to reduced responsiveness to chemotherapy and radiotherapy.

In an attempt to better understand the regulators of vasculogenic mimicry, several studies tried to recognize the molecular players of this process. Direct targeting of these molecules, including VEGF, is thought to serve as a promising therapeutic approach (302, 303). Other regulators of mimicry were also involved in the plasticity and stem cell-like phenotype of tumor cells. An example is the overexpression of the marker of brain development, NODAL (304–307). In addition, the overexpression of CD44 on vasculogenic tumor cells led to the initiation of the ongoing clinical study (NCT01358903). This trial evaluates the effect of an anti-CD44 agent on the process of vasculogenic mimicry during the treatment of solid tumors.

### CONCLUSION AND FUTURE OUTLOOK

The concept of targeting tumor angiogenesis is an important advancement in cancer therapy and has resulted in the development of therapeutic agents such as BVZ, sunitinib, and sorafenib. Benefits of using anti-angiogenesis therapies seem to be limited due to several reasons.

Despite the resulting stabilization of disease and increased PFS, treatment with anti-angiogenic agents may give rise to more resistant tumors with higher patient relapse rates. This lack of clinical benefit could be associated with preexisting resistance or with rapid adaptation to anti-angiogeneic agents. It is clear that multiple mechanisms of resistance against AIs exist, including upregulation of alternative angiogenic factors by tumor cells, involvement of stromal cells, and cooption/mimicry. The fact that the process of angiogenesis is complicated and involves a network of mechanisms suggests that the tumor microenvironment could mediate resistance to AIs (308). In addition, the vascular regression that is caused by AIs could elevate intra-tumoral hypoxia, which in turn, ameliorates resistance to radiotherapy, chemotherapy, and AIs. Also, the regression in tumor vasculature and the reduction in blood flow that result from AIs would impede the delivery of chemotherapeutic agents into tumors. All these complications of AI use would allow for tumor metastasis and would hence serve as practical limitations to drug development (309).

With the progress in several scientific and medical fields and with the growing surge in knowledge about angiogenesis and its resistance mechanisms, new pharmacological strategies ought to be developed in the near future. For instance, new ways of targeting tumor vessels should be designed. This could be made possible by developing novel therapeutics that can either optimize the function of tumor vessels to allow adequate tumor response to cancer therapy or directly target tumor vessels (310).

In addition, in the light of the wide gap between our improving knowledge in the mechanobiology of MSCs and our satisfactory understanding of their clinical implications, novel approaches should be suggested to fill the gap. This could be made possible by engineering MSCs to selectively deliver antiangiogenic molecules (309).

In addition, the use of combination strategies as a means to target multiple pathways involved in angiogenesis has been suggested to be a promising approach in overcoming resistance to AIs. To date, these either include a combination of multiple anti-angiogenic agents or a combination of anti-angiogenic drugs and other treatment regimens.

This process of selecting the most effective combination regimen is challenging because it requires extensive profiling of angiogenesis signaling pathways and involves a careful patient selection. Not only do combination regimens require regular dose adjustments to enhance efficacy and reduce toxicity, but also they require intermittent monitoring of treatment efficacy through biomarkers. Although combinations of different anti-angiogenic agents might increase treatment benefit, the presence of many alternative pathways can still result in acquired resistance. They can also induce excessive hypoxia that leads to additional resistance. Hence, the initiation of clinical trials to evaluate the efficacy and safety of such new combination strategies seems to be of utmost importance. In addition, the development of genetically engineered animal models whose tumor microenvironment can mimic that of humans could be of so much help in the development of reliable treatment approaches. This, in addition to clinical trials, would enable scientists and clinicians to make use of precision medicine for coming up with effective combinations of AIs and other therapies that would hopefully prevent the early acquisition of resistance or even impede its occurrence (141).

It is likely that the future therapy will make use of genomic, transcriptomic, and proteomic techniques as part of diagnostic profiling. Different therapeutic combinations can then be personalized and matched to current stages of tumor progression. Since tumors have rapid genetic drifts and might rapidly develop resistance to treatment, diagnostic profiling would have to be repeated during the course of treatment (141).

Nanotechnology enables researchers to develop novel nanotherapeutics, but this requires more knowledge about metabolic behaviors of tumor cells and possible physiological barriers or material properties that would improve or impede the efficiency of nano-therapeutics, respectively (311). It can therefore be foreseen that the future of AI-based therapies is heavily dependent on the efforts of basic scientists who can provide a clearer image regarding the response of cancer cells to the agents and on the ability of clinicians to make use of this knowledge to benefit patients (312).

These issues highlight the major challenges for future research. We look forward to the results of ongoing and future clinical trials discussed in this review paper in hopes that outcomes can be improved for all patients with cancers that are resistant to angiogenesis.

#### AUTHOR CONTRIBUTIONS

All authors made substantial contributions to study conception and design. YH, MK, HE, IK, DM, ST, and AS have been involved in drafting the manuscript and revising it critically for important intellectual content. All authors have provided final approval of the version to be published.

#### REFERENCES


by hypoxia-inducible factor 1. Mol Cell Biol. (1996) 16:4604–13. doi: 10.1128/MCB.16.9.4604


endothelial growth factor-A blockade. Cancer Res. (2010) 70:5109–15. doi: 10.1158/0008-5472.CAN-09-4245


**Conflict of Interest:** DM reports honoraria, travel support and institutional research funding from Roche, Pfizer, Novartis, and Amgen.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2020 Haibe, Kreidieh, El Hajj, Khalifeh, Mukherji, Temraz and Shamseddine. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# An Overview of Vasculogenic Mimicry in Breast Cancer

Marco A. Andonegui-Elguera<sup>1</sup> , Yair Alfaro-Mora<sup>2</sup> , Rodrigo Cáceres-Gutiérrez <sup>1</sup> , Claudia Haydee Sarai Caro-Sánchez <sup>3</sup> , Luis A. Herrera1,4 and José Díaz-Chávez <sup>1</sup> \*

<sup>1</sup> Unidad de Investigación Biomédica en Cáncer, Instituto Nacional de Cancerología-Instituto de Investigaciones Biomédicas, UNAM, Mexico City, Mexico, <sup>2</sup> Departamento de Genética y Biología Molecular, Centro de Investigación y de Estudios Avanzados del IPN, Mexico City, Mexico, <sup>3</sup> Departamento de Patología, Instituto Nacional de Cancerología (INCAN), Mexico City, Mexico, <sup>4</sup> Dirección General, Instituto Nacional de Medicina Genómica, Mexico City, Mexico

Vasculogenic mimicry (VM) is the formation of vascular channels lacking endothelial cells. These channels are lined by tumor cells with cancer stem cell features, positive for periodic acid-Schiff, and negative for CD31 staining. The term VM was introduced by Maniotis et al. (1), who reported this phenomenon in highly aggressive uveal melanomas; since then, VM has been associated with poor prognosis, tumor aggressiveness, metastasis, and drug resistance in several tumors, including breast cancer. It is proposed that VM and angiogenesis (the de novo formation of blood vessels from the established vasculature by endothelial cells, which is observed in several tumors) rely on some common mechanisms. Furthermore, it is also suggested that VM could constitute a means to circumvent anti-angiogenic treatment in cancer. Therefore, it is important to determinant the factors that dictate the onset of VM. In this review, we describe the current understanding of VM formation in breast cancer, including specific signaling pathways, and cancer stem cells. In addition, we discuss the clinical significance of VM in prognosis and new opportunities of VM as a target for breast cancer therapy.

Keywords: vasculogenic mimicry, breast cancer, angiogenesis, cancer stem cell, epithelial-mesenchymal transition, triple negative breast cancer

## BACKGROUND

Breast cancer is the most prevalent malignant tumor in women worldwide. Approximately 2.1 million cases were diagnosed in 2018, and it is the leading cause of cancer death in women (2). According to the WHO, breast cancer is classified histologically into invasive carcinoma, and other specific types, such as invasive lobular carcinoma, metaplastic carcinoma, carcinoma with medullary factor, among others (3, 4). However, chemotherapy of breast cancer is determined by another tumor classification. Up to 70% of invasive breast tumors show estrogen receptor alpha (ERalpha) or progesterone receptor (PR) expression. This group of patients is treated with ER-alpha inhibitors or aromatase inhibitors alone or in combination with standard chemotherapy (taxanes plus anthracyclines). About 20% of the patients have amplification or overexpression of the ERBB2 gene (HER2/neu). For these patients, treatment includes the use of antibodies directed against the ERBB2-encoded protein, which is a receptor of the EGFR family, and small molecules that inhibit the tyrosine kinase activity of the receptor. Finally, there is a group of tumors in which none of these markers is detected; these tumors are called triple-negative breast cancer (TNBC). They are a heterogeneous group of tumors with unfavorable prognosis, in which standard chemotherapy is used (5, 6). Recently, new therapies for breast cancer have been approved. For example, the use

#### Edited by:

Laurence A. Marchat, National Polytechnic Institute, Mexico

#### Reviewed by:

Francesco De Francesco, Azienda Ospedaliero Universitaria Ospedali Riuniti, Italy Francesco Grignani, University of Perugia, Italy

#### \*Correspondence:

José Díaz-Chávez jdiazchavez03@gmail.com

#### Specialty section:

This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology

Received: 21 September 2019 Accepted: 07 February 2020 Published: 27 February 2020

#### Citation:

Andonegui-Elguera MA, Alfaro-Mora Y, Cáceres-Gutiérrez R, Caro-Sánchez CHS, Herrera LA and Díaz-Chávez J (2020) An Overview of Vasculogenic Mimicry in Breast Cancer. Front. Oncol. 10:220. doi: 10.3389/fonc.2020.00220

**279**

of talazoparib or olaparib (poly (ADP-ribose) polymerase inhibitors(PARP) enzymes) in patients with mutations in BRCA1 and BRCA2 (7). CDK4 and CDK6 kinase inhibitors have been approved as a therapy for patients with estrogen receptor-positive and HER2-negative tumors (8). Patients with the same type of tumors, bearing mutations in the PIK3CA gene have been approved for PI3K kinase inhibitors (9). In the case of TNBCs, a high percentage have been shown to exhibit expression of PD-L1, a PD-1 ligand that inactivates the immune response. In this group, atezolizumab (a humanized anti-PD-L1 antibody) has been approved for use in combination with nab-paclitaxel (10, 11).

On the other hand, while a tumor is growing, hypoxic zones are formed due to the lack of blood vessels. Tumor vessel formation can occur through angiogenesis, i.e., the development of new blood vessels from pre-existing ones. When stimulated by the tumor, endothelial cells from normal vessels begin to migrate and proliferate, forming new vessels inside the tumor. Tumor angiogenesis is regulated by the VEGF (Vascular Endothelial Growth Factor) and the transcription factor HIF1alpha (Inducible Hypoxia Factor 1alpha). Discovery of some factors that regulate angiogenesis has led to the development of specific drugs that block this process, such as antibodies against VEGF (Bevacizumab) or molecules like sunitinib or sorafenib, which inhibit crucial kinases in angiogenesis (12, 13). In breast cancer, angiogenesis is considered a poor prognostic factor for survival (14). However, anti-angiogenic therapies in breast cancer have not demonstrated benefit in overall survival as adjuvant treatment or in metastatic disease (14, 15).

#### VASCULOGENIC MIMICRY AND RELATED SIGNAL PATHWAYS

In 1999, Maniotis et al. described the formation of tumor vessels lacking endothelial cells in uveal melanoma. These vessels were positively stained with periodic acid-Schiff (PAS), and they did not possess endothelial cell markers such as Factor VII-related antigen or CD31. They had a characteristic pattern and erythrocytes inside. In highly invasive cell lines grown in matrigel, structures similar to tumor vessels with cells positive for PAS and negative for CD31 were observed. Besides, these structures allowed the perfusion of a dye, showing that they were functional vessels. This phenomenon was termed vasculogenesis mimicry (VM) (1). Subsequently, VM was reported in other tumors, such as breast, ovary, prostate, and lung, among others (16–18). Positive PAS staining without CD31 detection (PAS+CD31–) is the most widely used marker for defining the presence of VM (**Figure 1**). The presence of erythrocytes in the vessels and their perfusion capacity suggest that they can irrigate tumors to avoid hypoxia and to transport nutrients. In addition, the presence of VM has been associated with the appearance of metastasis (19).

Since the discovery of VM, several factors regulating the formation of these vessels have been described. Like in the case of angiogenesis, hypoxia promotes VM. In cell lines derived from esophageal carcinoma, it was observed that inhibition of HIF1alpha inhibits the formation of VM and decreases the levels of proteins involved in the creation of these vessels, such as VE-cadherin, EPHA2 (ephrin A2) and Laminin 5gamma2 (20, 21). VE-cadherin is a relevant protein in VM. Under normal conditions, VE-cadherin is located in the plasma membrane of endothelial cells where it regulates intercellular unions. However, it has been observed that it is overexpressed in cells capable of performing VM. VE-cadherin is positively regulated by VEGF and by HIF1alpha (**Figure 2A**). VE-cadherin directs the location of EPHA2 to the intercellular junctions between cells that form the characteristic tubes of VM. EPHA2 is a kinase that activates two essential pathways in VM: PI3K (phosphoinositide 3-kinase) and ERK1/2 (extracellular signalregulated kinase 1/2) (through FAK kinase) which are associated with survival, proliferation, and migration (**Figure 2C**). PI3K also allows the activation of MMP14 (matrix metalloproteinase-14) which in turn activates MMP-2. This metalloproteinase cuts laminin 5gamma2, producing gamma2' and gamma2x fragments, which promote cell migration. Inhibition of the factors involved in VM signaling prevents the formation of vessels (16, 22, 23).

On the other hand, some microRNAs (miRNAs) are related to the regulation of vascular mimicry. MiRNAs are non-coding 19-to 24-base RNAs that control gene expression by binding to mRNAs, usually in the 3′untranslated region (3′ -UTR). MiRNAs can decrease transcription or prevent translation. In cancer, different microRNAs have been found to modify the regulation of oncogenes and tumor suppressor genes (24). MicroRNAs also regulate VM by interacting with specific genes; for example, miR-141 controls the expression of EPHA2. A decrease in miR-141 expression has been observed in high-grade gliomas. Besides, in glioma-derived cell lines, a decrease in miR-141 is associated with an increase in EPHA2 and an increase in VM (25).

### FACTORS INVOLVED IN VM IN BREAST CANCER

The presence of VM in breast cancer has been associated with poor prognosis in several clinical parameters (**Table 1**). Overexpression of factors regulating VM in breast tumors, such as HIF1alpha, VE-cadherin, and EPHA2 has also been reported (33–36). In a mouse breast cancer model, inhibition of angiogenesis promoted VM by expression of VE-Cadherin and other VM regulators in triple-negative tumors (37). In the MDA-MB-231 cell line (derived from a triple-negative tumor) which can form a pattern of tubular structures in matrigel, low expression of miR-204 was observed, while overexpression of miR-204 decreased the VM. This study also showed that PI3K-alpha and c-SRC are targets of miR-204. Therefore, it was proposed that miR-204 regulates critical pathways in VM, such as PI3K, MAPK, and SRC (38). Another factor associated with VM in breast cancer is osteopontin, a phosphoprotein related to tumor progression in different types of cancer. In a spheroid model of cell lines derived from breast tumors, an increase in osteopontin expression was observed in cells that formed vessels in matrigel. The expression of osteopontin was associated with a decrease in hsa-mir-299-5p, which targets osteopontin

FIGURE 1 | Vasculogenic mimicry in Triple Negative Breast Cancer. (A) CD31-PAS Double-staining (magnification 40x). (B) CD31-positive endothelial vessel (black arrow). (C) Tubular-type vasculogenic mimicry (VM) channel (black arrow), PAS-positive cuboidal tumor cells (red asterisks), PAS reaction in the luminal surface (blue arrow).

cells lining the lumen of the VM vessel is intended to represent PAS staining. Dotted line indicates an indirect interaction.

TABLE 1 | Vasculogenic mimicry and its association with prognosis in cancer.


(39). Besides, in a study that analyzed 200 breast cancer patient samples, an association of the presence of Osteopontin and VM was observed (27). On the other hand, in the MDA-MB-231 cell line, overexpression of WT-1 isoforms (Wilm's tumor 1) promoted VM, by increasing the expression of EPHA2 and VE-cadherin (40). The enzyme DDAH1 (dimethylarginine dimethylaminohydrolase-1) has also been associated with the formation of VM: inhibition of DDAH1 in MDA-MB-231 cells prevents the formation of VM. Interestingly, miR-193b decreases the levels of DDAH1 and, therefore, inhibits the formation of VM (41, 42). Other miRNAs regulate VM in breast cancer. In endothelial cells, cisplatin treatment was shown to promote the production of IL-6, which, through the STAT3 signal transducer, promotes cisplatin resistance and vessel formation by VM in MDA-MB-231 cells. The miR-125a targets IL-6 and STAT3. Decreased levels of miR-125a in endothelial cells were associated with increased production of IL-6 which promotes vessel formation by VM in breast cancer cells (43). On the other hand, the non-coding long RNA TP73-AS1 was shown to decrease the levels of miR-490-3p, which negatively regulates the TWIST1 gene. TWIST participates in the epithelial-mesenchymal transition and promotes the formation of VM. Therefore, the expression of TP73-AS1 stimulates the formation of VM through the overexpression of TWIST1 (44).

#### THE ROLE OF CSCS IN VM

In normal adult tissues, there are cells with the ability to proliferate, self-renew, and differentiate that allow tissue regeneration. These cells are known as stem cells. Similarly, it has been proposed that in malignant tumors, there is a cell subpopulation with the ability to self-renew and undergo less differentiation. In addition, it is hypothesized that these cells show mesenchymatous features, higher invasive capacity, and improved resistance to chemotherapeutic treatment. These cells have been called Cancer Stem Cells (CSCs). CSCs are characterized by specific markers, including CD44, CD133, CD166, ABC transporters, or metabolic enzymes such as Aldehyde dehydrogenase-1 (ALDH1) (45). It has recently been described that in different types of cancer, cells with stem characteristics actively participate in the formation of VM (46).

In human breast tumor xenografts transplanted in mice, it was demonstrated that a CD44+CD24– cell subpopulation presented CSC characteristics. CD44+CD24– cells obtained from mouse tumors were able to form tumors in other mice when as few as 1,000 cells were injected, while CD44+CD24+ cells did not form tumors even when injected more than 10,000 cells. In addition, tumors formed from CD44+CD24– cells presented cell heterogeneity, demonstrating that these cells were able to differentiate into a heterogeneous tumor (47). On the other hand, ALDH1 expression has been shown to be a marker of stem cells in normal tissue and breast tumors. In murine models, ALDH1+ cells derived from breast tumors were shown to have a superior ability to form tumors. Furthermore, the expression of ALDH1 is associated with lower overall survival and a higher probability of developing metastases in breast cancer patients (48, 49). In addition, the presence of ALDH1 is associated with the formation of VM. Both factors were shown to be associated with poorer overall and disease-free survival. Both the expression of ALDH1 and the presence of VM were most prevalent in triple-negative tumors (28). In an in vitro model using the HCC1937/p53 cell line (a triple-negative cell line with inducible p53 transfection) it was observed that ALDH1A3+ cells (one of the isoforms of the ALDH1 enzyme) could form tubular structures when they were grown in matrigel, while ALDH1A3- cells were not capable of creating such structures. The expression of ALDH1A3 coincided with the presence of Ki67, a proliferation marker, so it is inferred that cells that express the stem cell marker also have a greater proliferative capacity (50). On the other hand, in a study that included 134 samples of breast cancer patients, it was demonstrated that the CD133 marker was associated with VM in different breast cancer subtypes. The subtype that presented a more significant number of cases with VM and vessels with higher volume was the triple-negative. In addition, in the MDA-MB-231 cell line, a subpopulation characterized by the expression of the CD133 marker was described. This subpopulation was able to establish vessels in a matrix and expressed VE-cadherin and the metalloproteinases MMP-2 and MMP-9 (29). Therefore, as in other tumors, CSCs in breast tumors are actively involved in the formation of vessels of tumor origin. However, not all reports agree on the specific presence of CSC markers and the presence of VM. For example, Sun et al. found an association between VM formation and the presence of ALDH1 and CD44+CD24– phenotype, but not with the presence of CD133 (30). Therefore, it will be important to determine whether there is a single type of CSC in breast cancer or whether populations with stem cell characteristics are variable among tumors. Ginestier et al. demonstrated that only a fraction of the ALDH1-positive cells also possesses the CD44+CD24– phenotype. In addition, these cells had greater tumorigenic capacity compared to those with only one or none of these markers (49). Hence, stem cell markers used so far in breast cancer are not universal and may represent variants, sometimes synergistic, but with specific characteristics relevant to the treatment and progression of breast cancer.

### VM IN TRIPLE-NEGATIVE TUMORS

Triple-negative Breast Cancer (TNBC) includes a heterogeneous group of tumors characterized by the absence of expression of ER, PR, and that do not possess overexpression or HER2 amplification. Although these tumors have a high response to chemotherapy, they also have a poor prognosis for overall survival and relapse (51, 52). There is a higher proportion of triple-negative tumors with VM compared to tumors positive for ER, PR, and/or HER2. Accordingly, these tumors also have a greater number of vessels formed by VM (28–31). However, this association is controverted (26, 32). Indeed, Liu et al. found a correlation between the expression of HER2 and VM (32). Nonetheless, none of these studies grouped triple negative tumors, and the ER and PR, or HER2 mark were evaluated independently. Finally, in vitro analyses have shown differences in the ability to form vessels by VM and the mechanisms involved in this process between TNBC and no-TNBC cells. However, most of these studies use the MDA-MB-231 cell line as the TNBC tumor model and, only occasionally compare it to a different cell line. Despite the importance of the MDA-MB-231 line as a breast cancer study model, it is difficult to make a generalization regarding all TNBC tumors, due to their heterogeneity between patients and even within single tumors (40, 42, 44, 53). Although vessel formation by VM is more common in TNBC tumors, it is not exclusive to this type of tumor. However, due to the lack of specific therapies in this group, VM inhibition is a good candidate for therapeutic targeting.

### RELATIONSHIP BETWEEN CSCS, VM AND THE EPITHELIAL-MESENCHYMAL TRANSITION

As mentioned above, the presence of CSC markers is associated with the formation of VM. In addition, other factors related to morphological and cellular motility have a role in VM. During tumor progression, the epithelial-mesenchymal transition refers to the change of epithelial tissue, with very close cells interacting through intercellular unions to a mesenchymallike tissue, i.e., cells with greater invasive capacity, a large amount of intercellular material and without the apicobasal polarity characteristic of the epithelium. (54, 55). The epithelialmesenchymal transition is regulated by three families of transcription factors: SNAI (SNAI1/Snail and SNAI2/Slug), ZEB (Zinc finger E-box-binding homeobox; ZEB1 and ZEB2) and TWIST (TWIST1 and TWIST2). The activation of these gene families has been described. Nonetheless, the main mechanism by which these transcription factors promote TMS is through the repression of genes essential for the epithelial structure, such as CDH1, which encodes for E-cadherin, involved in adherens junctions (56). These factors bind epigenetic regulators and, together, regulate gene expression. For example, TWIST1 increases the expression of BMI1 (a repressor complex of the Polycomb family), and both are essential to repress the expression of CDH1 (57, 58). TGF-beta is one of the pathways that initiate the epithelial-mesenchymal transition. TGF-beta is a family of ligands that bind serine/threonine kinase receptors. In turn, these receptors phosphorylate and activate SMAD proteins. Finally, SMAD activation regulates the transcription of factors associated with EMT, such as SNAI1 or ZEB. On the other hand, it has been observed that the activation of the EMT program entails the cellular acquisition of CSC characteristics (**Figure 2B**) (54). In immortalized cells of breast epithelium, the overexpression of SNAIL increases the percentage of CD44+CD24– cells. Furthermore, CD44+CD24– cells show EMT-distinctive morphology. This phenomenon was also observed in cells transformed by the introduction of the HER2/neu oncogene (59). Therefore, EMT promotes the occurrence of CSCs in breast cancer.

Both EMT and CSC are related to VM. In TA2 mice (a mammary tumor model) MDA-MB-231 xenograft tumors, hypoxia-induced with the anti-angiogenic agent sunitinib is associated with VM and an increase in CD133+ cells. In addition, in matrigel cell cultures, activation of the HIF1alpha factor promotes TWIST1 transcription, which increases the percentage of CD133+ cells and vessel formation through VM. In this model, the inhibition of TWIST1 prevents the formation of VM and the emergence of cells with stem markers (31). On the other hand, the ZEB1 factor decrease in MDA-MB-231 cells inhibits VM and increases the expression of E-cadherin. In doing so, EMT is reversed while VM is inhibited (60). In breast cancer tumors, overexpression of TWIST is associated with a lower expression of epithelial factors such as E-cadherin (61, 62). Increased TWIST levels also correlate with more advanced stage tumor and are more common in TNBC tumors and HER2+ (63). Overexpression of TWIST and SLUG has also been observed in stromal tumor cells (64). However, the association of TWIST expression with disease-free survival and overall survival has not been consistently observed throughout these studies (63–65). Moreover, in a sample of 100 breast tumors, Nodal expression was associated with VM formation and VE-cadherin expression in a subgroup of tumors. Nodal is part of the TGF-beta family and participates in the development and regulation of differentiation (66). In vitro studies have demonstrated that Nodal expression is necessary for the formation of vessels by VM (67). Therefore, EMT, VM, and the presence of CSCs are interrelated and not isolated phenomena. Common features are the change toward an epithelium with invasiveness and migration capacity, less differentiation, and the ability to create tumor vessels.

### VM AS A THERAPEUTIC TARGET

As mentioned above, VM vessel formation is a process that includes proliferation, migration, invasiveness, and alterations in intercellular junctions. Accordingly, therapeutic inhibition of VM can target any of these processes. For example, it has been proposed that the use of a cytotoxic drug such as vincristine in combination with a specific inhibitor of the sarcoma family kinases (SFKs), which regulate signaling pathways involved in processes associated with VM, could have an additive effect on VM inhibition. In fact, an in vitro model using liposomes showed that both drugs can cause cell death and inhibition of vessel formation in MDA-MB-231 cells grown in matrigel. In addition, the use of these liposomes decreases the tumor volume of xenografts in nude mice (68). The authors also demonstrated that the use of liposomes for transporting compounds with different targets, such as epirubicin (a DNA intercalant) and celecoxib (a cyclooxygenase 2 inhibitor) are able to inhibit VM in breast cancer cells (69). On the other hand, it has also been proposed that the best strategy to inhibit vessel formation in tumors will be the simultaneous inhibition of angiogenesis and VM. New drugs, like acridine in complex with metals, such as gold, have shown the ability to promote apoptosis of cancer cells and inhibit the formation of vessels formed by endothelial cells (angiogenesis) or cancer cells (VM) (70).

The use of compounds obtained from natural extracts, such as brucine, has also been associated with VM inhibition. Brucine inhibits migration and invasiveness of MDA-MB-231 cells (71). Besides, brucine modifies the structure of actin and tubulin cytoskeleton and inhibits the formation of vessels by VM (72). Hinokitiol is also a natural compound with antitumor properties. In cells obtained from mammospheres, it was demonstrated that hinokitiol diminishes levels of the EGFR protein by increasing its proteasome-mediated degradation and, consequently, inhibits VM (73).

On the other hand, vessel formation by VM depends on the EGFR receptor in CSCs ALDH+ derived from breast tumors (74). Another compound that has demonstrated the ability to inhibit vessel formation by VM is 6'-bis (2.3-dimethoxybenzoyl) a,a-D-trehalose (DMBT) a derivative of brartemicin, a metabolite isolated from actinomycetes (75).

Currently, there are no specific therapies to inhibit VM. However, it is possible to propose that epithelial-mesenchymal transition, invasiveness and the presence of cancer stem cells may be useful targets to slow the formation of vessels by VM, in addition to having antitumor effect per se.

## CONCLUSION

VM is an alternative mechanism to angiogenesis that allows vessel formation without the involvement of endothelial cells. These vessels provide nutrients to the tumor and can serve as a means of spreading cancer cells. The absence of a therapeutic benefit of anti-angiogenic therapies in breast cancer may be due to the formation of vessels by VM. In addition, the formation of vessels with tumor cells may be a factor explaining the increased aggressiveness of tumor subtypes such as TNBC. However, VM also occurs in other breast cancer subtypes. The role of CSCs in VM in breast cancer will be better defined when specific stem markers are found to classify these cells. In addition, it will be important to use a greater variety of in vitro and in vivo models of breast cancer cells to determine the specific factors associated with the formation of VM in breast cancer. Finally, the discovery of particular factors involved in VM in breast cancer will make it possible to more precisely target therapies that inhibit the formation of vessels and may affect several processes that are important for tumor progression.

## AUTHOR CONTRIBUTIONS

JD-C conceived and designed the structure of the review and revised the manuscript. MA-E, YA-M, RC-G, and LH contributed in the manuscript writing and CC-S performed the CD31-PAS double staining and histopathology description. All authors have read and approved the final version of the manuscript.

## FUNDING

This study was supported by the Consejo Nacional de Ciencia y Tecnología (Grant number 261875) to JD-C.

### REFERENCES


by targeting dimethylarginine dimethylaminohydrolase 1. Sci Rep. (2017) 7:13996. doi: 10.1038/s41598-017-14454-1


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2020 Andonegui-Elguera, Alfaro-Mora, Cáceres-Gutiérrez, Caro-Sánchez, Herrera and Díaz-Chávez. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Cancer Stem Cells and Its Role in Angiogenesis and Vasculogenic Mimicry in Gastrointestinal Cancers

Erik Lizárraga-Verdugo<sup>1</sup> , Melisa Avendaño-Félix <sup>1</sup> , Mercedes Bermúdez <sup>1</sup> , Rosalio Ramos-Payán<sup>1</sup> , Carlos Pérez-Plasencia<sup>2</sup> and Maribel Aguilar-Medina<sup>1</sup> \*

<sup>1</sup> Facultad de Ciencias Químico Biológicas, Universidad Autónoma de Sinaloa, Culiacán, Mexico, <sup>2</sup> Laboratorio de Genómica, Instituto Nacional de Cancerología, Ciudad de México, Mexico

Cancer stem cells (CSCs) are able to promote initiation, survival and maintenance of tumor growth and have been involved in gastrointestinal cancers (GICs) such as esophageal, gastric and colorectal. It is well known that blood supply facilitates cancer progression, recurrence, and metastasis. In this regard, tumor-induced angiogenesis begins with expression of pro-angiogenic molecules such as vascular endothelial growth factor (VEGF), which in turn lead to neovascularization and thus to tumor growth. Another pattern of blood supply is called vasculogenic mimicry (VM). It is a reminiscent of the embryonic vascular network and is carried out by CSCs that have the capability of transdifferentiate and form vascular-tube structures in absence of endothelial cells. In this review, we discuss the role of CSCs in angiogenesis and VM, since these mechanisms represent a source of tumor nutrition, oxygenation, metabolic interchange and facilitate metastasis. Identification of CSCs mechanisms involved in angiogenesis and VM could help to address therapeutics for GICs.

Keywords: CSCs, esophageal, gastric, colorectal cancer, angiogenesis, vasculogenic mimicry

#### INTRODUCTION

Gastrointestinal cancers (GICs) are among the most common malignancies worldwide that mainly include gastric, esophageal and colorectal cancers (1). Treatments for GICs commonly are chemotherapy, radiotherapy, surgery and most recently anti-angiogenic therapy. However, the efficiency of these treatments depends on multiple factors such as cancer staging and resistance to treatment and relapse, which are related to Cancer Stem Cells (CSCs) (2).

In normal and tumoral tissues, vasculature supply the nutrients and oxygen required to maintain homeostasis. Blood vessel formation in the embryo occurs by vasculogenesis, a process that involve de novo production of endothelial cells (ECs) (3). On the other hand, the process through which new blood vessels are formed by sprouting and splitting from pre-existing ones is called angiogenesis (4), which is an important cancer hallmark.

Self-renewal of CSCs and initiation of tumor is accompanied by the promotion of angiogenesis, through the secretion of proangiogenic factors such as Vascular Endothelial Growth Factor (VEGF) (5). However, angiogenesis is not the unique source of nutrients and oxygen for tumors (6), given that CSCs are able to transdifferentiate into endothelial-like cells enhancing neovascularization (7). This process, called vasculogenic mimicry (VM), is present in different types of cancers and is responsible of providing a sufficient blood supply to tumor tissues (8). Interestingly, CD133 positive glioma cells express that express VEGF are able to increase vascular density (9) and higher recruitment of endothelial progenitor cells (EPCs) is observed in tumors enriched with CSCs (10).

#### Edited by:

Laurence A. Marchat, National Polytechnic Institute of Mexico, Mexico

#### Reviewed by:

Ali Syed Arbab, Augusta University, United States Danfang Zhang, Tianjin Medical University, China

\*Correspondence:

Maribel Aguilar-Medina maribelaguilar@uas.edu.mx

#### Specialty section:

This article was submitted to Molecular and Cellular Oncology, a section of the journal Frontiers in Oncology

> Received: 22 August 2019 Accepted: 10 March 2020 Published: 31 March 2020

#### Citation:

Lizárraga-Verdugo E, Avendaño-Félix M, Bermúdez M, Ramos-Payán R, Pérez-Plasencia C and Aguilar-Medina M (2020) Cancer Stem Cells and Its Role in Angiogenesis and Vasculogenic Mimicry in Gastrointestinal Cancers. Front. Oncol. 10:413. doi: 10.3389/fonc.2020.00413

**287**

The aim of this review is to compile recent knowledge of gastrointestinal CSCs and their participation in VM and angiogenesis in order to understand the underlying mechanisms that lead to the development of more effective therapies.

### GASTROINTESTINAL CSCs

Tumors are characterized by cell heterogeneity, according to CSCs theory, which hypothesizes that tumors are driven by a small cell subpopulation with stem cell properties, such as selfrenewal and differentiation capacity (11, 12). Also, CSCs promote tumor initiation, growth and proliferation, leading to aberrant growth and slow cycle cell replacement, making them resistant to therapies (13) and are able to move outside of the primary site and metastasize (14).

CSCs were first isolated (CD34+CD38−) from Acute Myeloid Leukemia (AML) patient samples in late 90s. This small population, was capable to transfer AML from human patients to NOD/SCID mice (15). Since then, surface markers have been used to identify and isolate CSCs in several types of cancers, for instance, CD24, CD44, CD90, CD133, and CD166 for Gastrointestinal CSC, and it was demonstrated that they are generally tissue specific (**Table 1**) (2).

Regarding to Esophageal Cancer Stem Cells (ECSCs), they were first isolated from Esophageal Squamous carcinoma cell line (ESCC) using colony morphology criteria (27). Nevertheless, isolation of ECSCs now is performed using CD44 and ALDH1 (19, 28).

CD44 was the first marker used to identify Gastric Cancer (GC) Stem Cells (GCSCs) (29). Moreover, the embryonic markers OCT-4, SOX2, NANOG and the surface maker CD133/Prom1 are highly expressed in GCSCs (30). Interestingly, CD44+/CD24<sup>+</sup> GCSCs subpopulation has shown stem cell properties in vivo and in vitro (16). Also, EpCAM+/CD44<sup>+</sup> phenotype present stem cell characteristics in GC tissues (18) Besides, isolated CD44+/CD54<sup>+</sup> GCSCs from tumors and peripheral blood, are able to generate tumors both in vitro and in vivo (17). However, other molecules, such as, CD90, CD71, ABCB1, ABCG2, CD133, ALDH1, and Lgr5 are also considered as potential markers to GCSCs isolation (31–35).

Finally, Colorectal Cancer (CRC) Stem Cells (CRCSCs) were first isolated by CD133 expression, showing tumorigenic capabilities in mice (25, 36). Nevertheless, molecules such as EpCAM+/CD44+/CD166+, ALDH+, EphB2+, LGR5+, and CD44v6<sup>+</sup> are commonly used to CRCSCs isolation from cell lines (23, 24, 37–39), despite these markers are shared with normal mesenchymal stem cells (MSCs). In this regard, it has been recently reported that Dclk1 discriminates between cancer and normal stem cells in the intestine (40).

#### CSCs in Vascular Niche

Vascular niches are key for maintaining the stem phenotype, such as, self-renewal, undifferentiated state and dormancy in normal stem cells (41). In cancer context, neo-vascularization plays an important role during carcinogenesis and metastasis. This process was first described by Scherer in glioblastoma, where the cancer cells growth is possible by the proximity of surrounded TABLE 1 | Surface markers of gastrointestinal cancers stem cells.


blood vessels, now called "cancer vascular niche" (42). Normal stem cells and CSCs primordially growth in vascular niches, due to a perivascular microenvironment. However, cancer vascular niche is rich in abnormal blood vessels, connected and organized with each other in a different pattern from normal vessels (43, 44). These abnormalities are induced by hypoxia, low pH and high interstitial hostile fluid pressure, making a selection of hostile cells that can escape from the tumor through aberrant blood vessels to metastasize (45). Angiogenesis within the tumor mass harbors a variety of host-derived cells, regulated by monocytes Tie-2 expression, fibroblasts, ECs, as well as, innate and adaptive immune cells (46, 47).

#### PROMOTION OF ANGIOGENESIS BY GASTROINTESTINAL CSCs

Angiogenesis can be divided in two types: sprouting and intussusceptive (48–50). In the first one, ECs proliferate and sprout toward an angiogenic stimulator (e.g., VEGF), forming flat structures called filopodia, producing proteolytic enzymes to enhance angiogenic process (51). On the other hand, intussusceptive angiogenesis is independent of ECs, where an existing vessel is divided into two new vessels only by cellular reorganization (52). Interestingly, neovascularization is an important process to support tumor growth and metastasis; usually, tumors reach a size of ∼2 mm in diameter when not fed by neovascularization (53). In this regard, CSCs are able to modify tumoral microenvironment by expressing angiogenic factors in order to enhance tumor neovascularization, contributing finally in their maintenance and proliferation (5).

#### Esophageal Cancer

Positive cells to Placental growth factor (PLGF), appear to be CSCs in esophageal cancer and have the capability to release PLGF, promoting cancer metastasis by the activation of MMP9 (54). Besides, CSCs that express PLGF are important due to the promotion (55) or inhibition of tumor angiogenesis depending on its interaction with VEGF (56).

#### Gastric Cancer

Bone marrow mesenchymal stem cells (BM-MSCs) are implicated in the promotion of tumor angiogenesis in gastric cancer (GC) since SGC-7901 cells in both, in vitro and in vivo models, increases VEGF release from tumor cells by the activation ERK1/2 and p38 MAPK pathways, resulting in angiogenesis promotion (57). Moreover, gastric cancer-derived MSCs (GC-MSCs) are also able to promote angiogenesis when interact with BGC-823 and MKN-28 GC cell lines, inducing overexpression of pro-angiogenic factors, such as, VEGF, MIP-2, TGF-β1, IL-6, and IL-8 favoring tube formation (58).

Recently, the Leucine-rich repeat and immunoglobulinlike domain-containing Nogo receptor-interacting protein 2 (LINGO2) a novel gastric cancer stem cell-related marker has been associated with cancer progression (59). In this regard, gastric tumor tissues overexpressing LINGO2 shows elevated expression of the angiogenic marker pVEGFR2 and a blood vessel marker CD34, meanwhile the silencing of LINGO2 in Human Umbilical Vein Endothelial Cells (HUVEC) cells results in inhibition of tube formation, suggesting the involvement of positive-LINGO2 CSCs in angiogenesis (59).

#### Colorectal Cancer

CRCSCs are able to initiate vascularization via pericytes by growth promotion (5, 60). Thus, lack of pericytes recruitment impacts negatively in tumor size owing to poor vascular structure (61). This is also correlated to worst prognosis, due to leaky vessels that produces elevated local pressure, enhancing progression and metastasis. Nevertheless, higher vascular density has been associated with recurrence, metastasis and patient mortality (5, 62).

Co-cultivation of CRCSCs and SW620 cells enhances its stemness properties. Also, transplantation of SW48 and MSCs support angiogenesis in vivo (63). Additionally, conditioned media (CM) from SW480 cells pre-treated with CRCSCs CM enhances HUVEC tube formation and higher levels of VEGFA expression (63). Besides, BM-MSCs are able to induce angiogenesis, when treated with IFN-γ and TNF-α, by VEGF expression via the HIF-1α signaling pathway (64), meanwhile, IL-8 allows tumor angiogenesis (65).

Participation of CRCSCs in tumor neovascularization has been demonstrated in tumor tissues by CD31/CD133/Lgr5 coexpression (10). Besides, CRC cell lines HCT116 and HT29 spheroid-derived cells are able to co-act with endothelial progenitor cells (EPCs) in order to promote migration and tube formation by secreting VEGF. Meanwhile, EPCs also increases tumorigenesis of CRC cells through angiogenesis (10).

### SIGNALING PATHWAYS OF CSCs IN ANGIOGENESIS

Little is known about cellular and molecular mechanistic features of CSCs roles in angiogenesis (**Figure 1**). For instance, Bone Morphogenic Protein 4 (BMP-4) plays a crucial role in angiogenesis by mediating vascular integrity. Besides, VEGF suppression is strongly regulated through BMP-9/ALK1. Conversely, TGFβ1/ALK5 pathway enhances angiogenesis by VEGF expression (66), being a critical signaling molecule for angiogenesis in CSCs (67). Moreover, VEGF-A/NRP-1 interaction promotes stemness properties in breast cancer (BC) cell lines by activation of Wnt/β-catenin pathway, since its inhibition relies in the attenuation of HUVEC-tube formation induced by co-culturing with extracts from Breast Cancer Stem Cells (BCSCs) (60). Moreover, glioblastoma stem-like cells (GSCs) produce VEGF-A, which is secreted in extracellular vesicles promoting permeability and angiogenesis in brain (68). Additionally, angiogenesis promotion can be stimulated by GSCderived exosomes (GSC-EXs) trough miR-21/VEGF/VEGFR2 axis (69).

Notch signaling pathway is also required for stem cell survival and vascular development and it is a crucial angiogenesis stimulator (70). Interestingly, inhibition of self-renewal capabilities and angiogenesis are orchestrated by Notch signaling repression in GSCs, as well as, reduction of vasculogenic markers, such as, CD105, CD31 and von Willebrand factor (vWF) (71).

#### VASCULOGENIC MIMICRY FORMATION BY CSCs IN GASTROINTESTINAL CANCERS

The generation of vascular channels (VC) without ECs or fibroblasts was first identified in aggressive and metastatic melanoma in 1999, and was termed vasculogenic mimicry (6). In this specific case, the relationship between aggressive melanoma cells that co-expressed Vimentin and epithelial (keratin 8,18) intermediate filaments was particularly interesting, since these cells, where able to be aligned along the external walls of microvascular channels conducing red blood cells, without ECs (72).

Channels formed by VM are composed of a basement membrane and tumor cells that facilitate microcirculation plasma and blood supply from host normal vessels (73). VM can be classified in classical patterns in matrix type (6) and the tubular type (74). Besides, it has been described that VM is composed by matrix proteins such as Laminin, Heparan sulfate proteoglycan, and Collagens IV and VI (75).

VC network may be an independent angiogenesis mechanism for blood source, since angiogenesis inhibitors induce extracellular matrix-rich tubular network formation in vitro and are not able to suppress VM in several types of cancers, showing that VM works as an alternative mechanism for blood cells supply (76). Besides, VM is associated with tumor size, short overall survival (OS), high tumor grade, clinical staging, invasion and metastasis (77–79).

Interestingly, tumor cells associated to VM structures acquire an undifferentiated phenotype as well as ECs characteristics (80). Nowadays, CSCs have been involved in VC formation in cancer (81–87). For instance, in salivary adenoid cystic carcinoma (ACC) specimens CD133 is positively associated with VM formation. Besides, CD133<sup>+</sup> ACC CSCs and xenograft tumors of nude mice injected with these cells show overexpression of VE-Cadherin and VM mediators (MMP-2, MMP-9) (86).

Furthermore, an holoclone CD133<sup>+</sup> isolated from MDA-MB-231 form VM and display MMP-2 and MMP-9 expression (87). In addition, VEGF-silenced cells, attenuate growth and promotes VM as adaptation mechanism associated to HIF-1α expression. Furthermore, enrichment of CD133+/CD271<sup>+</sup> Melanoma CSCs is found in the perivascular niche in vivo (81).

#### Esophageal Cancer

It has been shown that epithelial–mesenchymal transition (EMT) cells present stem phenotype, showing a remarkable relationship between EMT and CSCs (88). For instances, Ginseng extract showed a negative effect on EMT, as well as, VM in ESCC lines (89). Besides, recombinant Endostatin (rh-Endo) protein combined with radiotherapy downregulates EMT characteristics and VC formation in ESCC through inactivation of AKT/GSK-3β signaling pathway (90).

#### Gastric Cancer and Colorectal Cancer

In Gastric adenocarcinoma tissues, a positive relationship between CD133/Lgr5 expression and VC formations, microvessel density, tumor grade, lymph node metastasis and TNM staging has been shown (85). In the case of CRC, the upregulation of ZEB1 results in epithelial phenotype restoration, while, its silencing results in VM inhibition and VE-Cadherin and Flk-1 downregulation in HCT116 cell line (91).

#### SIGNALING PATHWAYS OF CSCs IN VM

CSCs and VM are involved in cell plasticity, which is the capability of an aberrant population to ECs transdifferentiation (**Figure 1**) (92). VEGF receptors regulate expression of specific marker for ECs, such as VE-Cadherin (93). In this regard, it has been described that primary and established sarcoma cell lines in contact with post-surgery fluids from Giant cell tumors of bone patients can enrich CD44/CD117 cell population and AKT/mTOR pathway activation. Moreover, it has been proved that prolonged stimulation results in transdifferentiation of tubule-like structures that express endothelial markers, such as, VE-Cadherin and CD31 (94). Additionally, CSCs switch on NFκB and STAT3 signal pathways via CCL5-CCR1/CCR3/CCR5, stimulating endothelial differentiation and tubule formation (95).

It has been demonstrated that DKK1 enhances VM formation via EMT by developing CSC characteristics in not small cells lung carcinoma (NSCLC) (96). Besides, the Wnt signaling receptor FZD2 drives EMT process, enhancing stem-like properties and VM capacity in HCT116 cells (97). Interestingly, inhibition of IL-8/CXCR2 signaling by Transgelin results in suppression of VM with increased IL-8 levels due to IL-8 uptake inhibition in breast cancer stem cells (BCSCs) (98).

In CRC, the poorly differentiated cell line HCT116 expresses endothelial markers and form tube-like structure in vitro after endothelial-conditioned medium co-culture. In addition, under hypoxic conditions cells exhibit higher levels of VEGFR2/VEGFA, as well as, CD31, CD34 and VE-Cadherin overexpression (99).

#### THERAPEUTICS STRATEGIES: NEW PERSPECTIVES

Little is known about the role of CSCs promoting angiogenesis and VM. It has been shown that abnormal blood vessels are capable to obstruct immune response to the tumor, as wells as, the transportation and distribution of oxygen and chemotherapeutics. This hostile tumor microenvironment can also lead to selection of cells resistant to radiotherapy and chemotherapy (43). Altogether might suggest that antiangiogenic drugs often induce tumor hypoxia, allowing CSCs to survive and propagate, thus driving tumor progression.

Nevertheless, some inhibitors of VM are potential molecules to use in therapy of different types of cancers, such as LCS1269 that is capable of overcoming multidrug resistance for DNAdamaging agents in melanoma by VM inhibition (100). In addition, Hinokitiol, a tropolone-associated natural compound, has an important effect over EGFR expression and VM in BCSCs through proteasome-mediated EGFR degradation (101).

Molecules and signal pathways involved in angiogenesis and VM supported by CSCs are novel targets of cancer therapeutics. Nevertheless, information of GICs therapeutics in this matter is limited. Has been described that anti-CD133 has a great potential in treating CRC (96). Besides, targeting signaling pathways is possible, for instance, BBI-608 drug targeting STAT3 could be used for advanced CRC resistant to standard therapeutics or in mixture with Paclitaxel for advanced GC (2, 97). Moreover, Ginsenoside Rg3, a derived from ginseng, represses growth cells and CSCs properties in CRC cells, as well as, inhibits angiogenesis-related genes, suppressing vascularization in xenograft tumors (98).

Several authors suggest that interfering on growth and survival of tumoral ECs can be enough to inhibiting angiogenesis and CSCs self-renewal (99). In this regard, VEGF secreted by cancer cells is a well-recognized therapeutic target and several angiogenic inhibitors have been developed with the capability of also suppress self-renewal of CSCs leading to reduced tumor growth. It has been shown that, Bevacizumab expands survival time by targeting the perivascular niche by the inhibition of VEGF (102). Additionally, bevacizumab reduces metastatic niche formation in rectal carcinoma patients (103) and combined with an anti hepatoma-derived growth factor antibody prevents tumor relapse and progression in NSCLC by impairing CSCs (104). Conversely, the administration of Bevacizumab combined to Sunitinib (VEGF inhibitor) induces tumor hypoxia in BC cell lines resulting in the augment of CSCs population (105).

#### CONCLUDING REMARKS

Recently, emerging evidence shows that tumors are heterogeneous, being constituted by multiple subpopulations such as CSCs that share self-renewal and differentiation characteristics with normal stem cells. Also, they are able to express specific surface markers that depend on the organ of origin. For instance, CD44, ALDH1, EpCAM, and Lrg5 are characteristics markers of gastrointestinal CSCs, in EC, GC, and CRC. Besides, vascular niches are important for maintaining tumor progression, since CSCs prefer a perivascular microenvironment, rich in blood vessels that often have an abnormal structure and is supported by hostile conditions such as, hypoxia, which in turn, enhances selection of more aggressive cells, able to invade and metastasize. In this regard, CSCs can be transdifferentiated into endothelial-like cells and pericytes, important lineages for maintenance of cancer vascular niche.

Some signaling pathways have been implicated in angiogenesis and VM. The most important molecules and pathways are VEGF/VEGFR2, Notch, BMP9/ALK1, PI3K/AKT/mTOR, NF-κB, and STAT3, that regulate different pivotal processes involved in angiogenesis promotion, such as permeability, endothelial and tubule-like transdifferentiation and promotion of endothelial markers expression, stem cell survival and vascular development.

Clinical relevance of angiogenesis in GICs is remarkable as poor pericyte coverage is correlated with worst prognosis due to leaky vessels that produce elevated local pressure and enhances progression and metastasis. Besides, a higher vascular density in the invasion front has been associated with recurrence, metastasis and patient mortality in CRC. Importantly, Dclk1 can discriminates between cancer and normal stem cells in the intestine.

CSCs are implicated in VM in different cancers, such as ACC, breast cancer and melanoma. In addition, there is a remarkable relationship between EMT and CSCs, due to EMT cells acquired stem phenotype. Importantly, GICs show that the use of drugs, certain proteins or radiotherapy that affect the EMT leads to inhibition of VM. Finally, clinical relevance of VM relies on its association with tumor size, short OS, high tumor grade, clinical staging, invasion and metastasis.

On this front, several drugs have been tested, for instance, Bevacizumab is able to expand survival time by targeting the perivascular niche by the inhibition of VEGF with effect on angiogenesis However, more studies are necessary in order to elucidate CSCs participation on VM and angiogenesis since this could help to address therapeutics for GICs.

#### AUTHOR CONTRIBUTIONS

EL-V, MA-F, MB, and MA-M conceived and designed the content of this review. EL-V, MA-F, MB, and RR-P wrote the paper. CP-P and MA-M contributed to the final version of the manuscript.

## FUNDING

Consejo Nacional de Ciencia y Tecnología CONACYT, Mexico (Grant 290311).

### REFERENCES


#### ACKNOWLEDGMENTS

The authors acknowledge CONACyT for MA-F (575985) and EL-V (304939) fellowships.


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The handling Editor declared a past co-authorship with several of the authors CP-P, RR-P.

Copyright © 2020 Lizárraga-Verdugo, Avendaño-Félix, Bermúdez, Ramos-Payán, Pérez-Plasencia and Aguilar-Medina. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.