Tumor vascularization plays a fundamental role in cancer progression and metastasis, allowing tumor cells to reach a continuous source of oxygen and nutrients and to evade host immunosurveillance. Sprouting angiogenesis, intussusception, vasculogenesis, vessel co-option, and vasculogenic mimicry (VM) are the main processes that contribute to tumor vascularization, generating an intricate net of vessels, some of which are composed of a mosaic of endothelial and tumor cells. While angiogenesis is commonly described as a complex mechanism including the remodeling of the extracellular matrix, the formation of the Tip-Stalk cells hierarchy, and the involvement of pericytes, VM is typically characterized by highly perfused vessels with significant deposition of matrix proteins, and it is often associated with highly invasive and metastatic tumors, frequently paired with a poor patient prognosis [1]. In previous years, periodic acid-Schiff (PAS)-CD31 was deemed to be the golden standard to distinguish between the two biological processes [2], but recently an alternative method of vascularization, the Epithelial-to-Endothelial Transition (EET), has been observed ^[3][4][3,4]. EET is the acquisition by epithelial cancer cells of endothelial markers, such as CD31, VE-Cadherin, Ephrin A2, and others [3]. Although not completely understood, it has been found that EET is induced by several microenvironmental factors and is established in highly plastic cancer cells, the so-called cancer stem cells (CSCs). One of the master inducers of all the above-described phenomena is hypoxia, which is undoubtedly one of the most typical features of the tumor microenvironment, determining the activation of the hypoxia-inducible factors (HIF) which were first discovered and elucidated by Semenza in 1992, who won the 2019 Nobel Prize in Physiology or Medicine along with Ratcliffe and Kaelin ^[5][6][7][5,6,7]. In particular, HIF proteins also enhance the stem features of cancer cells and contribute to their differentiation through the EET, by losing cell polarity, increasing the invasive ability, and upregulating Twist and Snail, causing the consequent downregulation of the tight junction proteins, E-cadherin and Occludin, together with the upregulation of angiogenesis-related molecules such as VE-cadherin, vitronectin and fibronectin [8]. During severe hypoxic conditions, generated by tumor expansion, the impaired electron transport chain reactions and, more generally, mitochondrial dysfunctions, induce high reactive oxygen species (ROS) production [9] which, consequently, affects cancer cells proliferation, migration, and metabolism. It is noteworthy that ROS may also be generated during tumor growth by the increased activity of peroxisomes, oxidases, cyclooxygenases, lipoxygenases, and thymidine phosphorylase [10]. Gastric cancer (GC), which is currently the fourth leading cause of cancer-related death and the sixth for incidence globally ^[11][12][11,12], is closely dependent on vascularization. Indeed, several studies have shown that Helicobacter pylori, which is related to more than half of GC cases, is able to penetrate normal, metaplastic, and neoplastic epithelia, triggering an immune-inflammatory response, and thus not only promoting gastric carcinogenesis but also stimulating the release of cytokines, matrix metalloproteinases, and angiogenic factors by gastric epithelial cells upon NF-kB activation [13]. It is well established that H. pylori-infected GC patients showed increased tumor vascularization compared to those who underwent H. pylori eradication [14]. Moreover, elevated gastrin secretion was also associated with NF-kB activation, as well, enhancing ROS generation and vasculogenesis [15]. When ROS levels were reduced by the use of N-Acetylcysteine, metastatic capability and chemoresistance resulted in inhibition, demonstrating that ROS play a central role in GC pathogenesis. Moreover, it was observed that inoperable GC patients subjected to chemotherapy demonstrated a ‘U-like’ association of mortality rate with vascular density, suggesting that very low and very high vascularization are both linked to poor outcomes [16].

Angiogenesis is a multistep process that involves the formation of new blood vessels starting from pre-existing ones. During cancer development, the tumor mass constantly grows until the oxygen concentration is too low and the catabolic products of the accelerated glycolysis affect the microenvironmental pH. In particular, hypoxia often triggers the early endothelial response by activating the production of pro-angiogenic factors such as the vascular endothelial growth factor (VEGF), angiopoietin-2 (ANG-2), or fibroblast growth factor-2 (FGF-2) [17]. These signals wake up the pericytes, detaching them from the vessel wall and allowing the endothelial cells (ECs) to degrade the basement membrane through the activation of matrix metalloproteinases (MMPs). After such events, ECs lose their junctions, the vessel dilates and several plasma proteins extravasate, creating a new temporary matrix layer for ECs migration [17]. The EC closest to the higher VEGF concentration gradient is then selected by the Notch-Dll4 axis as the “Tip” cell [18], which will lead to the building of the nascent vessel, while neighbor cells will be consequently selected as “Stalk” cells and they will divide to follow the Tip. Once the vessel is mature, all the junctions between ECs have been restored, pericytes have covered the vessel again and the basement membrane has been deposited, the blood will flow and the new vessel will be completely perfused. However, while in physiological conditions Dll4 selects the Tip cell only, activating the Notch pathway, in tumor-induced angiogenesis Dll4 expression was observed in the majority of tumor vessels [19]. Sprouting angiogenesis is a complex and long process, especially in tumors, where ECs often compete for the leading positions, generating a continuous change between Tip and Stalk status. Therefore, endothelial tip and stalk cells’ fate determination is not fixed by the initial conditions, but instead there is a constant dynamic phenotype switch [20]. Despite some interesting clinical trials aimed at treating and controlling angiogenesis in GC, their efficacy in improving patients’ Overall Survival (OS) remains limited. In 2011 and 2014, the trials AVAGAST and AVATAR ^[21][22][21,22] demonstrated that the use of Bevacizumab, a monoclonal antibody inhibiting VEGF-mediated angiogenesis by binding and inactivating the VEGF-A ligand, was able to modestly increase the Overall Response Rate (ORR) and the Progression-Free Survival (PFS) but with no significant improvement in the OS. Moreover, it clearly appeared that the efficacy was dissimilar depending on territorial differences, which might be due to different hospitalization conditions or genetic discrepancies [23]. Better results were achieved through the trials REGARD and RAINBOW ^[24][25][24,25], which tested Ramucirumab, a human monoclonal antibody binding to the extracellular region of VEGFR-2 and blocking the downstream effects of VEGF, improving PFS and OS, especially when combined with Paclitaxel. Such a promising result induced the American Society of Clinical Oncology to confirm Ramucirumab as the second-most-effective targeted drug after Trastuzumab [26]. Although tyrosine kinase inhibitors (TKi) are not widely used in the treatment of GC, heresearchers we report Sunitinib, one of the first TKi approved for use in imatinib-resistant GISTs. Indeed, it not only blocks VEGFRs, PDGFR-α, PDGFR-β and c-Kit [27] molecules involved in the vasculogenic process but also exerts a possible ROS-mediated cytotoxic effect, albeit one that is still not completely clear [28]. ItWe is also necessaryed to mention that such anti-angiogenic therapies often result in only a transitory improvement of the clinical picture. Resistance to such regimens is divided into two main categories: the first is defined as the evasive or adaptive resistance to angiogenesis inhibitors, which involves revascularization due to the upregulation of alternative pro-angiogenic signals, the protection of the tumor vasculature by recruiting pro-angiogenic inflammatory cells or by increasing protective pericyte coverage, the increased invasiveness of tumor cells into adjacent tissues to co-opt normal vasculature and the augmented metastatic relapse and tumor cell growth in lymph nodes and distant organs. The second one is called intrinsic resistance and includes those individuals who had never benefited from the anti-angiogenic treatments due to an innate indifference, probably due to the redundancy of several pro-angiogenic signals, an inflammatory cell-mediated vascular protection, and the invasive angiogenic-independent co-option of normal vessels [29].

Described for the first time in uveal melanoma by Maniotis et al., VM is a biological mechanism exploited by several cancer histotypes to generate PAS-positive vessels lined by tumor cells [30]. Although early evidence indicated that the inner wall of such vessels was composed of only tumor cells, Chang et al. demonstrated that the luminal surface might be lined by an intricate mosaic of tumor and ECs [31]. The intricate transformation that leads cancer cells to perform VM is, to date, not fully understood, but accumulating evidence points to a close correlation between CSCs and VM formation. Indeed, during VM, high plasticity cancer cells that do not express any typical endothelial markers behave like proper ECs to achieve new sources of oxygen and nutrients. To distinguish between the two histotypes, the golden standard is the immunohistochemical analysis of two endothelium-related proteins (CD31/CD34) coupled with the PAS reaction [2]. Currently, very little is known about the formation of VM in GC. A close association has been reported between GC patients with VM and the formation of hematogenous metastasis, probably due to the aggressive features developed by cancer cells and to the fact that, when lining the luminal surface of the vessels, cancer cells are directly exposed to the bloodstream, increasing their proficiency to detach and form distant metastases [32]. MMP-2 and MMP-9 have been positively correlated with VM, conferring on cancer cells the capability to remodel the extracellular matrix and degrade the vascular basement membrane [33], while EphA2 has been reported to be directly involved in the formation of tubular networks [34]. These phenomena were confirmed by Sun et al., who demonstrated that out of a collection of 84 specimens of gastrointestinal stromal tumors (GIST), 21 were found to be VM positive, with a high production of MMP-2 and MMP-9 [35]. Although GISTs are typically not aggressive tumors, patients with VM-enriched vessels experienced a worse prognosis with respect to the ones with low VM-dependent vascularization. VM is commonly associated with advanced-stage GCs and with patients’ poor prognosis, although the vast majority of the data are available only in Eastern countries [36]. Moreover, one of the main issues in identifying VM in GC samples is the heterogeneous mucous tissues of the gastrointestinal tract, which might lead to false PAS-positive results, thus overestimating the VM histological grade. It is today well-known that cells resistant to anti-angiogenic therapies, such as those reported above, are often more prone to generate VM networks in vivo and in vitro, and therefore research into new therapeutic strategies to target both phenomena at the same time is urgently needed.

Under particular microenvironmental conditions, i.e., hypoxia, high tumor interstitial fluid pressure, and altered extracellular matrix, VM-prone cells begin to express typical endothelial antigens through the so-called Epithelial-to-Endothelial transition. Hypoxia is the best-known and most potent VM inducer, activating HIF-1α and HIF-2α, which in turn bind to the hypoxia-response elements of target genes such as VEGF, VEGF receptors, EMT inducers, and stem-associated genes [37]. Under hypoxic conditions, Twist1 translocates into the nucleus and promotes VE-Cadherin expression, triggering the transition of epithelial cells to an endothelial-like phenotype ^[38][39][38,39]. Indeed, VE-Cadherin activates PI3K via EphA2 phosphorylation, promoting the matrix metalloproteinases secretion and thus extracellular matrix remodeling [40]. In particular, HIF-1α involvement was reported to promote the stabilization of Notch, a typical endothelial antigen, by binding to its intracellular domain and consequently promoting Nodal transcription [41]. It was recently reported that the onset of chemoresistance, and in particular to 5-Fluorouracil, might induce the EET in AGS cells via the upregulation of TYMP, EphA2, and VEGFR2 [42]. However, the EET is still a poorly understood phenomenon, with narrow limits that denote the difference with a proper VM process. Additionally, from the clinical point of view, the treatment of such a biological event is currently debated as all the above-described anti-angiogenic therapies might induce the onset of VM cells due to the generation of hypoxic regions. Even after discontinuing treatments, endothelial vessels might rebound and link to the neo-formed VM channels [39]. To date, the only therapeutic options evaluated in vivo are the use of Doxycycline, a tetracycline derivative, acting by the inhibition of the degradation of E-Cadherin preventing in this way both the EET and VM [43], anti-Notch4 antibodies which downregulate Nodal expression [41] and dual antiplatelet therapy, a gamma-secretase inhibitor, which was found to inhibit glioblastoma CSCs from differentiating into endothelial-like progenitor cells through blockade of Dll4-Notch signaling [44]. However, no therapies nor experimental drugs are currently reported for GC treatment.