Although the anti-VEGF treatments have constituted a milestone for anti-angiogenic purposes, another aspect of this framework has to be considered in that VEGF inhibitors often fail to produce enduring clinical responses in a great number of patients (Saltz et al., 2007). The anti-angiogenic therapy results in transitory improvements, in the form of tumor standstill or constriction, in some cases increasing survival. In spite of this, tumors begin to grow again, though after a transient period of clinical benefit, generally measured in months (Miller et al., 2005).

Regarding the resistance to the VEGF-targeted therapy, two mechanisms through which this endurance is highlighted can be defined: (i) tumors completely fail to respond from the outset of treatment (intrinsic resistance) or (ii) they respond initially, and then continue growing while still receiving the treatment (acquired resistance; Bergers and Hanahan, 2008).

A critical occurrence leading to the success or failure of anti-angiogenic therapy is the need for a proper timing window for tumor vascular normalization: long term anti-angiogenic therapy sometimes leads to tumor hypoxia (Winkler et al., 2004), and hypoxia triggers VEGF production, genetic instability in tumor ECs and vascular permeability (Taylor et al., 2010). The decrease in blood flow further reduces oxygen, nutrient and drug delivery, enforcing stress on the tumor (Figure 1A,a). In preclinical studies, VEGF-targeted therapy suppresses the growth of new tumor vessels, but is less effective against the established tumor vessels (Sitohy et al., 2012). Nagy and Dvorak (2012) postulated that “early” and “late” tumor vessels might differ in the susceptibility to anti-VEGF therapies. “Early” vessels would predominate initially; the “late” ones become proportionately greater, however, as tumors grow. While the former are responsive, the latter (though formed from the “early” vessels) lose their dependence to the growth factor and become resistant to anti-VEGF-A/VEGFR therapy. Furthermore, the enormous heterogeneity of the tumor vasculature has to be considered, and different types of evolving resistant surrogate tumor blood vessels, which vary between them in anti-VEGF therapy sensitivity, have been described (Sitohy et al., 2012). These aberrant new vessels may be VEGF-independent and therefore capable of mediating tumor vascularisation despite VEGF-inhibition. For example, brain tumors become more infiltrating after VEGF pathway inhibition, which may facilitate vessel co-option (Keunen et al., 2011).

Microenvironment adaptation to a cancer stress condition plays a key role in determining whether tumors respond to VEGF-targeted therapies (Rak et al., 2002). This event is driven in part by the molecular promotion of the translation of pro-survival genes, such as BCL2 (Sherrill et al., 2004), X-linked inhibitor of apoptosis protein (XIAP; Gu et al., 2009), and stress response genes (Somers et al., 2015). It has been shown that under the hypoxic condition, which inhibits the global protein synthesis, the eIF4E homolog 4EHP can promote the translation of certain mRNAs involved in the adaptation to hypoxia, such as EGF receptor and PDGF receptor-α (Uniacke et al., 2012). Preclinical data has shown that some adaptive mechanisms include a decreased propensity for certain cancer cells to die under stress conditions, sometimes following genetic aberrations such as loss of p53 function (Yu et al., 2002), by adapting their metabolism (McIntyre et al., 2012) or by autophagy (Xu et al., 2013). In addition to being constituted by transformed cells, tumors are characterized by the presence of infiltrating different stromal cells, which are often the cause of therapy resistance, including the resistance to anti-angiogenic therapies (McMillin et al., 2013). Among these, the immature myeloid cells (Chung et al., 2013), fibroblasts (Crawford et al., 2009) and endothelial progenitor cells (Shaked et al., 2006) infiltrate the tumor and mediate the resistance by incorporating themselves into vessels or by releasing pro-angiogenic growth factors, such as BV8 (Shojaei et al., 2007) or PDGF-C (Crawford et al., 2009). At present, there is conflicting evidence to whether the anti-angiogenic therapy could increase the tumor aggressiveness and cause flare phenomena. In some pre-clinical studies the anti-VEGF treatment promoted this serious downside, in terms of invasion and metastasis (Ebos et al., 2009). Also, evidence in patients showed that anti-VEGF therapy can promote tumor aggression. A study on metastatic renal cancer cell (mRCC) patient demonstrated a significant increase in tumor grade in the primary tumor after treatment with sunitinib and pazopanib (Sharpe et al., 2013), while data from the AVANT trial for colorectal cancer with adjuvant bevacizumab has highlighted that this treatment caused a higher incidence of relapses and deaths (de Gramont et al., 2012). Moreover, anti-VEGF therapy can promote invasion and the undergoing epithelial-to-mesenchymal transition (Lu et al., 2012). In our and other researchers’ opinion, a possible mechanism could be that the anti-angiogenic treatment damages the vessels, causing cancer cell extravasation; in preclinical models, TKIs may promote metastasis by damaging the vasculature integrity (Chung et al., 2012). On the other hand, flare-up phenomena have been described in patients with mRCC after the withdrawal of the anti-angiogenic therapy (Powles et al., 2013), and the analysis of the NSABP-C08 trial of adjuvant bevacizumab in colorectal cancer patients did not evidence the noxious effect of bevacizumab (Allegra et al., 2013).

Clinical practices have revealed a large number of adverse complications associated with anti-VEGF treatments. Among these, hypertension, proteinuria, hemorrhage, endocrine dysfunction, thrombosis, gastrointestinal perforation, fistula formation, cardiac toxicity, and reversible posterior leukoencephalopathy (Chen and Cleck, 2009). The need for hypertension treatment has been seen in approximately 25% of patients (Burger et al., 2011). The occurrence of reversible posterior leukoencephalopathy is correlated with uncontrolled hypertension and the permanent cessation of VEGF inhibitor therapy is greatly needed. VEGF inhibitors could also increase the risk of thromboembolism by about 5%, and anticoagulant treatments were active in reducing this side effect. Proteinuria might reflect the hypertension, and VEGF inhibitor treatment is usually stopped when it reaches 3 g protein loss in 24 h (Burger et al., 2011). Although the treatment with bevacizumab (Avastin, which neutralizes specifically the VEGF-A isoforms) and paclitaxel plus carboplatin (both chemotherapeutics) in the treatment of patients with non-small-cell lung cancer had significant survival benefits, febrile neutropenia and pulmonary hemorrhage were associated with the addition of anti-VEGF (Sandler et al., 2006). Moreover, bevacizumab is avoided in patients with ovarian cancer with substantial pelvic disease or with previous bowel surgery for the risk of bowel perforation or fistula formation (Simpkins et al., 2007). Owing to the fact that TKIs inhibit a lot of off-target kinases, they are associated with malaise, fatigue, hypothyroidism, diarrhea, and cardiac failure (Schmidinger and Bellmunt, 2010).

Measurements of vessel caliber by Magnetic Resonance Imaging (MRI) is a technique for in vivo monitoring of microvascular development during the treatment of cancer patients (Dennie et al., 1998). Some of MRI-based studies, have shown that treatment with anti-angiogenic drugs leads to an improvement in tumor microcirculation, which is less permeable and has an increased pericyte coverage (Goel et al., 2011). By using this technique, it is possible to obtain images of the structure of the tumor microvasculature (Vessel Architectural Imaging, VAI) whose evaluation provides an efficient parameter for monitoring disease progression and response to treatment in cancer patients. Emblem et al. (2013) conducted a retrospective analysis of 30 patients affected by glioblastoma and studied the structural heterogeneity of the tumor microcirculation by VAI, demonstrating that, during the anti-angiogenic therapy, tumor blood vessels of subjects who responded to treatment were similar to those of normal tissues (Sikov et al., 2015). Recent studies have shown a significant increase in the number of patients with breast cancer who are HER2-negative or triple negative and do not respond to conventional anti-angiogenic therapies (Earl et al., 2015), and that an improvement of vascular conditions causes an increase in tumor oxygenation and a better response to anti-angiogenic agents (Heist et al., 2015).

Tolaney et al. (2015) have demonstrated that high baseline microvascular density (MVD) in breast tumors is considered a positive response to vascular normalization index induced by bevacizumab. MVD is calculated by evaluating specific parameters including pericytes coverage and the number of α-SMA. In the case of high MVD, the effect of anti-angiogenic drugs would be to remove some vessels and increase the functions of others, by inducing their normalization. In the case of low MVD, the anti-angiogenic drug reduces them further and prevents their normalization. This makes some tumors insensitive to anti-angiogenic therapy. Consequently, knowing the baseline MVD is a key factor in predicting the success of treatment with anti-angiogenic drugs (Jeong et al., 2015).

A characteristic of tumor microenvironment is low oxygen tension, caused by an imbalance in oxygen delivery and consumption. At low pO₂ levels, cells become radioresistant and, as a vicious circle, the irradiation itself, which induces direct vessel damage, stimulates hypoxia with consequent recruitment of immunosuppressive myeloid cells, contributing to tumor resistance (Russell and Brown, 2013). Conversely, it has also been also demonstrated, in a xenograft tumor model, that VEGF is released at the onset of angiogenesis, independent of HIF (Hendriksen et al., 2009). These conflicting results are ascribable to the different origin of the tumors and to the different areas within the same tumor, characterized by chaotic and complex tumor vascular architecture that determines a better or worse oxygen distribution (Hida et al., 2016).

The lactate, produced by tumor glycolytic metabolism, predicts the response to irradiation of human carcinomas (Sattler et al., 2010), an increased risk of metastases (Brizel et al., 2001) and is able to induce angiogenesis (Hirschhaeuser et al., 2011). These findings indicate that the anaerobic metabolism of cancer cells is strongly related to the increased aggressiveness of a tumor and the possibility of measuring its amount is an important predictor.

HIF-1 activity can be silenced through inhibition of epidermal growth factor receptor (EGFR) or topoisomerase-1 or by anthracyclines (Semenza, 2010). Moreover, HIF-1 activity can be inhibited by new drugs which reduces HIF-1 mRNA amounts (Koh et al., 2008) or which provokes HIF-1 degradation (Kim et al., 2006). The treatment with these drugs during radiotherapy amplifies the irradiation effects prompting tumor vasculature destruction and reduction in growth (Harada et al., 2009). Unfortunately, these exciting results did not apply to all types of tumor because several cancers show little or no hypoxia and do not express HIF activation (Moeller and Dewhirst, 2006; Meijer et al., 2012).

Tumor hypoxia represents an important aspect of the tumor microenvironment. Clinical studies using needle-sensors (Eppendorf)^® have demonstrated that hypoxia varied on a tumor-to-tumor basis and represents a universal therapy resistance mechanism (Koch and Evans, 2015). For this reason, several methods have been developed to assess tumor hypoxia and to predict treatment outcome by evaluating the oxygenation status during therapy.

Several studies have been conducted to determine the presence of HIF inside tumor cells. Recent advances in imaging of hypoxia by positron emission tomography (PET) demonstrated that it is possible to select patients for specific therapies, improving the anti-hypoxia-direct radiotherapy (Baumann et al., 2016). Activation of HIF transcription factor can be also evaluated by using genetically encoded fluorescent sensors with different switching and their combination allows the distinction of hypoxic and re-oxygenated cells in glioma cell lines, focusing on regions devoid of blood vessels (Erapaneedi et al., 2016). A potential tracer, used as a biomarker in the context of anti-angiogenic therapy, is [¹⁸F]-FMISO: a low [¹⁸F]-FMISO-PET signal is correlated to decreased hypoxia and it is a predictor of vascular normalization (Hernandez-Agudo et al., 2016).

The dynamic contrast enhanced magnetic resonance imaging (DCE-MRI) technique has been used to evaluate the effect of bortezomib, by using multiple endogenous and exogenous markers to evaluate hypoxia (Sun et al., 2014). By DCE-MRI it has been demonstrated that tumor blood flow is significantly reduced after bortezomib administration and the results of this study are very important to monitoring the effects of treatment with an anti-tumoral drug.

It has been recently reported that a hypoxia visualization bio-imaging probe, protein transduction domain [PTD]-oxygen dependent degradation domain [ODD]-HaloTag (POH), was able to detect HIF-1 active (Takata et al., 2015).

HIF activity has also been monitored in a preclinical glioma model. After treatment with different drugs, imaging biomarkers through luciferase expression have been used to document the tumor response (Lo Dico et al., 2015).