Vascular co‐option and vasculogenic mimicry mediate resistance to antiangiogenic strategies

The concept that all the tumors need the formation of new vessels to grow inspired the hypothesis that inhibition of angiogenesis would have led to “cure” cancer. The expectancy that this type of therapy would have avoided the insurgence of resistance was based on the concept that targeting normal vessels, instead of the cancer cells which easily develop new mutations, would have allowed evasion of drug caused selection is, however, more complex as it was made apparent by the discovery of nonangiogenic tumors. At the same time an increasing number of trials with antiangiogenic drugs were coming out as not as successful as expected, mostly because of the appearance of unexpected resistance.


| INTRODUCTION
In the last decades, a large body of work has led to propose that the formation of new vessels, also known as angiogenesis, is necessary for a neoplasm to growth beyond the diameter of a few millimeters. Following a first period in which data from all over the world seemed to support this hypothesis, "inducing angiogenesis" has been considered to be a hallmark of cancer. 1 The idea that angiogenesis is necessary to cancer growth is the brainchild of Judah Folkman who went as far as to propose that, by inhibiting angiogenesis with appropriate drugs, it would be possible to treat any type of tumors or, at least, maintaining or inducing dormancy and preventing the formation of metastatic lesions. 2 The production of drugs which can actually block effectively angiogenesis in vitro and in preclinical animal model has led to a rush of large clinical trials which, in the beginning, had wide resonance. 3 As data started to come through it was, however, quite evident that this was not going to be the cases as antiangiogenic treatment turned out to provide, at the best, some very modest positive effects. [4][5][6][7] Contrary to expectation it became clear that resistance to antiangiogenic treatment did occur, the question was: which are the mechanisms? Resistance to antiangiogenic treatments had been first predicted in 1997 8 when an unexpected mechanism of cancer growth had been discovered: growth of nonangiogenic tumors which exploited the preexisting normal vessels 9 by co-opting them, 10 rather than induce, and rely upon, newly formed one. 8 Angiogenesis is defined as "the process of new blood-vessel growth" 11 and an angiogenic tumor is one in which the cancer cells induces angiogenesis in order to grow. 12 A nonangiogenic tumor is defined as one not inducing the formation of new nonneoplastic blood vessels. 9 Vascular co-option was firstly defined as a mechanism by which nonangiogenic cancer cells obtain a blood supply by hijacking the existing vasculature. 13,14 However, as shown in an example ( Figure 1), co-option is a physiological phenomenon as, for example, sub population of plasma cells co-opt vessels in normal bone marrow. 15 We are, therefore, proposing a more comprehensive definition of vascular cooption, including normal and neoplastic cells, which is "the establishment of a stable anatomical and/or functional interaction between the coopting cell and the co-opted blood vessels" (Figure 2). Vascular mimicry also occurs in tumors not inducing the formation of new nonneoplastic blood vessels, that is, nonangiogenic. However, these nonangiogenic cancer cells instead of co-opting preexisting vessels, form themselves channels that provide blood flow to the neoplastic mass. 16 This patterns of tumor vascularization are commonly ignored but can mediate cancer progression and establishment of metastasis.

| EVIDENCES FROM CLINICAL EXPERIENCE
As said above, for many years the data seemed to support Folkman's hypothesis. A first group of studies supporting it reported an association between the microvascular density (MVD) present in tumors and their aggressiveness, 17 an hypothesis successively challenged. 18 The discovery of the vascular endothelial growth factor (VEGF) family of proteins and their receptors, 19,20 their crucial involvement in inducing angiogenesis both in neoplastic and nonneoplastic conditions, plus their presence in a variety of tumors [21][22][23][24] provided more support. In 1996, the identification of angiostatin and its antiangiogenic efficacy in murine models, further seemed to confirm the fact that this approach could treat patients with both early and advanced malignancies. 25,26 At the time the idea developed that, because the targeted vessels were not neoplastic by themselves, as they were normal new vessels growing inside the neoplastic mass, this type of cancer treatment could be immune from resistance. 27 Contrary to expectations the clinical trials results have been disheartening. Representative is what has been observed in high-grade glioma: overall survival failed to improve with antiangiogenic drugs. 28 However, in these patients, bevacizumab treatment can relieve symptoms by reducing the severity of intracranial oedema. 29 In advanced breast cancer, improvement in disease-free but crucially not overall survival has been seen while the results in early breast tumors are inconclusive. 30 Antiangiogenic treatment has instead improved the results in patients with metastatic colorectal cancer, 31 still the benefits achieved are in the range of months rather than years. No benefit has been instead achieved for patients with early colorectal cancer. 32 Disappointing the results in small cell lung cancer 33 while positive results have been reported in non-small cell lung cancer, although the gained progression free and overall survival is once again limited. 34

| MECHANISMS OF RESISTANCE
It becomes evident that resistance to antiangiogenic treatment is widespread. It can be intrinsic, when it is observed at the beginning of the treatment, or acquired, that is, that it affects the relapsing disease after an initial response to therapy. 35,36 Following the first description of a putative mechanism, that is, the nonangiogenic cancer growth and progression by exploiting the preexisting vessels, 8 a number of different mechanisms of resistance started to be discovered and has been reviewed elsewhere. 37 This article is focused on the nonangiogenic tumor growth as a resistance mechanism.

| NONANGIOGENIC TUMORS AND VASCULAR CO-OPTION
Nonangiogenic tumors grow in the absence of angiogenesis by two main mechanisms. One way is by cancer cells infiltrating and occupying the normal tissues to exploit preexisting vessels, which is known as vascular co-option or vessel co-option. The second is one, less frequent, in which no new vessels are formed but the cancer cells themselves forms channels able to provide blood flow, the so called vasculogenic mimicry. 9 Nonangiogenic tumors can be very aggressive and induce metastatic spreading but as they lack new vessels, it was immediately evident that antiangiogenic treatment could not have worked in these patients as the target of the treatment itself is not present. 8 F I G U R E 1 An example of physiological vascular co-option: normal bone marrow. Immunostain with anti-CD79a antibody (dark brown). Nuclear counterstain with hematoxylin. Marrow blood vessels are coopted by a subpopulation of plasma cells organize in a single cell line F I G U R E 2 Legend on next page.
In 1999, Holash et al reported that tumor cells migrate to host organ blood vessels in sites of metastases, or in organs with an extensive vascularization likes the brain, and initiate nonangiogenic tumor growth instead of classic angiogenesis. These vessels then regress owing to apoptosis of the constituent endothelial cell, apparently mediated by angiopoietin-2 (Ang-2). Lastly, at the periphery of the growing tumor mass angiogenesis occurs by cooperative interaction of VEGF and Ang-2. Tumor cells often have immediate access to blood vessels, such as when they metastasize to or are implanted within a vascularized tissue, co-opt and often grow as cuffs around adjacent existing vessels. 13 A host defense mechanism is activated, in which the co-opted vessels trigger an apoptotic cascade, probably by autocrine induction of Ang-2, followed by vessel regression resulting in tumor death. However, successful tumors overcome this vessel regression by initiating neoangiogenesis. 38 The ability of cancer cells to co-opt the vessels initiates the process through few of them will intravasate, starting the metastatic process. 39 Maniotis et al 16  Vasculogenic mimicry can serve as a marker for tumor metastasis, a poor prognosis, worse survival, and a highest risk of cancer recurrence.

| CLINICAL STUDIES
Only a limited number of retrospective clinical studies looking at the correlation between the presence of nonangiogenic tumors and the effect of antiangiogenic drugs has been so far published (Table 1).
T A B L E 1 Clinical studies demonstrating that vascular co-option is associated with resistance to antiangiogenic treatment summary Hypothetical primary resistance to surgical treatment F I G U R E 2 Different co-option modalities. A, Mechanisms of co-option: the lung. Anatomical and functional co-option. In the normal lung: the alveolar spaces are lined by the pneumocytes which separate the air space from the vessels. In nonangiogenic cancer growth the tumor cells are filling the alveoli. Initially, the pneumocytes are still in place so the co-option is functional as the cancer cells are not in direct contact with the vessels. On the right, a deeper portion of the tumor: the pneumocytes have been detached from the abluminal surface of the vessels and have disappeared. The cancer cells now are in direct contact with the vessels and the co-option is not only functional but also anatomical. B, Liver: Liver sinusoid are vascular spaces in which a space (space of Disse) is present between endothelium and underlying liver cells. In this organ, the neoplastic cells exploit the liver sinusoids not by anatomically linking to the vascular structure but by taking the place of the hepatocytes (replacement). C, In the brain neoplastic both glioblastoma or metastatic cells can co-opt the vessels directly (blue neoplastic cells). The glioblastoma cells can also co-opt by merging and forming hybrid cells with the pericytes (yellow hybrid cell) alone, the only differences being that in the angiogenic tumors the percentage of viable tissue was higher (between 5% and 30%).
Untreated nonangiogenic metastases were comparable to the treated one, while untreated angiogenic had a more variable amount of necrosis. This study support the hypothesis that nonangiogenic tumors are not only more resistant to bevacizumab but also to the neoadjuvant chemotherapy used (the drugs used are not specified by the authors). 42 The lymph node is also involved in nonangiogenic growth by vascular co-option. 50 Jeong et al 51 confirmed this finding and described also how in node metastases from colorectal cancer patients treated or not treated with bevacizumab, no differences in the intratumor vascularization where found.
Antiangiogenic treatment of primary central nervous systems tumors has been used for several years but, again, the results have not lived up the expectations. 29 Post mortem histological examination of patients died after receiving treatment with cediranib, an inhibitor of VEGF receptor 2 (VEGFR2) tyrosine kinases, 43 or bevacizumab regimen 44 showed that the glioma cells were growing around preexisting vessels in a nonangiogenic fashion. In two case reports, vascular cooption and progression in absence of angiogenesis in human brain tumor samples, surgical and autoptic, has been illustrated. 29,45 Breast cancer patients treated with Bevacizumab showed, by dynamic contrast-enhanced magnetic resonance (DCE-MRI) imaging, three response patterns. In the first one, a decrease in K-trans values over the extent of the tumor was demonstrated, that is, decreased vascular permeability and/or vascular surface area, while the second was characterized by extensive necrosis. These two patterns indicate a response to antiangiogenic treatment. In the third one, instead, no changes were seen and the authors conclude that, in these lesions, the vessels were independent of VEGF. Whether they were preexisting normal vessels or not, it has not been investigated in this study 46 and therefore we can only say that nonangiogenic growth may be, but not necessarily, responsible for the resistance to These reports suggest that lung metastases of RCC can escape antiangiogenic treatments by switching phenotype and progress in an angiogenesis-independent fashion exploiting preexisting vessels. 48 A second case report form a patient with glioblastoma showed that, on biopsies taken after antiangiogenic treatment, a pattern of infiltration around the normal brain vessels is present. 45 Nonangiogenic growth and spread in the lung has also been linked to resistance to surgical treatment. 49 In nonangiogenic carcinomas growing in the lung, neoplastic cells can spread by moving from one alveolar cavity to the other trough the septa pores. 49 and C26 lines are growing in the lung in a nonangiogenic fashion, contrary to their angiogenic growth in the subcutaneous tissue. Otherwise, the lung metastases of the RENCA renal carcinoma are prevalently angiogenic; however, the residual secondaries after sunitinib treatment, presented a switch toward the nonangiogenic growth pattern. 53 Comparable presented in this study as far as the vascularization of the resistant metastases is concern but the hypothesis that vascular co-option could be involved it is discussed. 54 A third study show that sunitinib accelerated metastatic tumor growth and decreased overall survival in mice receiving short-term therapy in various metastasis assays. A faster metastatic process was identified in mice treated with sunitinib in advance of intravenous implantation of tumor cells, raising the possibility of a "metastatic conditioning" in multiple organs. As in the two studies discussed above, these observations of metastatic acceleration were in contrast to the demonstrable antitumor benefits obtained when the tumor cell lines were grown orthotopically as primary tumors and subjected to Sunitinib treatment. The authors, however, do not investigate where vascular co-option is occurring, 55 therefore, we can only say that nonangiogenic growth may be, but not necessarily, responsible for the resistance to treatment.
In an orthotopic hepatocellular carcinoma mouse model the implant is initially sensitive to sorafenib, another TKI blocking the An obvious one would be the vascular pattern of the tumors, however this at present require surgical resection to allow histological examination, but in most of the cases extensive surgery is not needed.
Radiological imaging will hopefully provide with methodology alternative to the histopathological characterization. An example is the work of Cheng et al 79 proposing that contrast-enhanced multidetector CT scan imaging can be used to distinguish vessel co-opting tumors from angiogenic tumors in patients with CRC liver metastases.
Another emerging approach is to combine antiangiogenic compounds with blockage of vascular co-option. This is likely to be the most successful because tumors can actually change their vascular status during progression in both ways 9,80 : an angiogenic tumor treated with antiangiogenic compounds can "escape" by turning nonangiogenic but a nonangiogenic tumor could "escape" anti-co-option drugs acquiring an angiogenic phenotype.
The combined approach is now being explored at preclinical level.
The use of bevacizumab with the antibody OS2966 against beta1integrin, one of the keys molecules in co-option, has demonstrated that the two have a synergic effect. 81  This model further indicates that vascular co-option is a mechanism of escape and suggest that "double therapy," antiangiogenic and anti-co-option, is predicted to give the best results. 83 10 | CE N'EST QU'UN DÉBUT (ON THE WALLS OF PARIS, MAY 1968) As always, things in biology are frequently more complicated that initially tough. Deemed as one of the six original Hallmark of Cancer, 84 the induction of angiogenesis it is actually not such an Hallmark. However, with all his limits and failures, the effort to treat tumors just by antiangiogenic therapy has nevertheless led to study with more attention the role of cancer in blood vessels. Once we will learn more about this relationship, both in angiogenic and nonangiogenic growth, hopefully we will be able to better exploit the antiangiogenic protocols so F I G U R E 3 Post co-option events and possible therapeutic strategies. Having co-opted a vessel, the cancer cell can behave in different way. It can grow around the vessels (1) or instead, can acquire migratory property (pericyte mimicry) and, rather than grow, migrate along the abluminal surface (2) and eventually produce distant metastatic lesions. A third possibility is that a neoplastic cell, having co-opted the vessels, will sit quiescent in a "perivascular niche" as a dormant cancer cell (3). As already well know some cells instead will immediately cross the blood vessels wall, enter the blood stream, and produce haematogenous metastases (4). Targeting the molecular apparatus that leads to co-option is an obvious approach to block both local growth (1) and perivascular spread (2). Requiring even more work but holding more promises, is the possibility of preventing the formation of perivascular niches, disrupts the niche already formed or even just to prevent the existing dormant cells from "weak up" in patients with high risk of relapsing with metastatic disease. This approach is very likely to be more effective than trying to treat established metastatic lesions far designed alongside new treatments which are emerging by the study of vascular co-option. Two are the main areas which promises to provide more targets for treatments. The first is the biology of cooption as this is an active process requiring specific pathways while the second is the biology of the cancer cell when in nonangiogenic mode, as it differs in several aspects form the angiogenic ones.
Another potentially very important issue in cancer treatment, is the plasticity cancer cells can have: one example is how some neoplastic cell lines produce an angiogenic tumor if injected subcutaneous but a nonangiogenic one if eventually lodging into the lung after inoculation in the tail vein. 53 This fact also suggest that there are not necessarily committed angiogenic or nonangiogenic tumors, but rather a tumor can have an angiogenic or nonangiogenic behavior as also suggested by clinical studies. 80 Such a plasticity can therefore probably explain why some neoplastic cells, able to grow in some microenvironment, remain apparently "dormant" in a quiescent status after co-opting a vessel. 85 This perivascular microenvironment in which this happens has been called "the perivascular niche." Should be confirmed that indeed these perivascular quiescent cells represent dormant disease, the knowledge of their biology could bring, if not to the ability to eliminate them, at least of preventing them from originating metastatic relapses. 86 The recent advances in cancer biology due to the discovery of nonangiogenic growth have therefore the potential to lead to further steps toward a more effective cancer treatment.

CONFLICT OF INTEREST
The authors declare that they have no conflicts of interest.

ETHICAL STATEMENT
Not applicable.

DATA AVAILABILITY STATEMENT
Data sharing is not applicable to this article as no new data were created or analyzed in this study.