Pei Wei
Yanxin Lu
Zhiyong Wang- Department of Immunology, Zunyi Medical University, Zhuhai, China
Targeted anti-angiogenic therapies against vascular endothelial growth factor (VEGF) are standard treatments for various advanced cancers. However, their clinical benefits are often limited by acquired resistance, with some patients even exhibiting paradoxical tumor aggressiveness and accelerated metastasis. Traditional views primarily attribute this to hypoxia-driven mechanisms within the tumor microenvironment. Based on an analysis of receptor tyrosine kinase (RTK) interactions and pathway rewiring within tumor cells, this paper proposes an integrated “brake-accelerator” model. We posit that in the presence of VEGF, VEGFR2–Mesenchymal-Epithelial Transition factor (MET) heterocomplexes and neuropilin (NRP) co-receptors restrict bypass pathway activity. Upon VEGF blockade, this “brake” is released, and bypass “accelerators” are passively or actively amplified, collectively driving invasive transformation and immune evasion. We identify “direct signal remodeling” in tumor cells as a potentially underappreciated contributor. This process acts as a potential upstream integration node that, alongside treatment-induced hypoxic stress and immune–stromal microenvironment remodeling, constitutes a “tripartite stress” framework. Through Darwinian clonal selection and induced phenotypic plasticity, this framework drives the evolution of tumors towards a more aggressive phenotype. This paper systematically reviews the molecular mechanisms supporting this model, including the regulatory role of VEGFR2-MET complexes, the signal hub function of NRP, and the networked characteristics of escape pathways. Finally, we discuss the implications of this conceptual model for future clinical practice, including the development of dynamic biomarkers based on intrinsic tumor cell characteristics and the design of more precise combination and adaptive treatment strategies to overcome resistance to anti-VEGF therapy and improve patient prognosis.
1 Introduction
Inhibition of tumor angiogenesis has been one of the most significant advances in oncology over the past two decades (1–3). Anti-angiogenic agents, exemplified by the humanized anti-VEGFA monoclonal antibody bevacizumab, have become a cornerstone of treatment for various solid tumors, including colorectal cancer (CRC) (4, 5), non-small cell lung cancer (NSCLC) (6, 7), and glioblastoma (GBM) (8, 9), significantly extending patients’ progression-free survival (PFS). However, the long-term efficacy of this strategy has not met initial expectations. The vast majority of patients eventually develop acquired resistance, and the improvement in overall survival (OS) has been modest (10, 11).
More concerning is the “paradox of aggression” observed in preclinical models and some retrospective clinical studies. In these cases, although tumors were initially controlled radiographically after anti-VEGF therapy, the recurrent disease exhibited enhanced local invasion and distant metastatic potential (12, 13). For instance, in patients with recurrent GBM, the non-enhancing, diffuse infiltrative recurrence pattern observed after bevacizumab treatment portends a worse prognosis (14–16). This phenomenon suggests that while anti-VEGF therapy inhibits tumor angiogenesis, it may inadvertently activate molecular programs that promote malignant progression.
To explain the resistance and paradox of aggression, several non-mutually exclusive models have been proposed. Model 1: The classic hypoxia-driven model. This model posits that anti-angiogenic therapy exacerbates tumor hypoxia, thereby upregulating hypoxia-inducible factors (HIFs). This in turn activates alternative pro-angiogenic factors [e.g., fibroblast growth factors (FGF), platelet-derived growth factors (PDGF)] and pro-invasion/metastasis pathways [e.g., mesenchymal-epithelial transition factor (MET), AXL receptor tyrosine kinase (AXL)] to compensate for the loss of VEGF signaling (17–19). Model 2: The metabolic/autophagy reprogramming model. This model focuses on metabolic shifts in tumor cells—such as switching to oxidative phosphorylation or enhancing autophagy—to adapt to limited nutrient and oxygen supply (19–21). Model 3: The immune microenvironment regulation model. This model posits that while VEGF blockade might improve T cell infiltration, it could also lead to the enrichment of pro-angiogenic and immunosuppressive myeloid-derived suppressor cells (MDSCs), thereby fostering resistance (22–24). Model 4: The direct signaling model. This model acknowledges that many tumor cells express VEGF receptors (VEGFRs), and thus anti-VEGF drugs may directly influence tumor cell behavior (25–27).
While these models elucidate partial truths, they predominantly regard tumor cells as passive adapters, failing to fully integrate the direct and active regulatory roles of anti-VEGF drugs on intrinsic tumor cell signaling. Building upon this, this paper proposes a more comprehensive “direct signal remodeling” framework. We argue that the direct effect of anti-VEGF drugs on tumor cells (i.e., Model 4) should not be viewed as merely one of several parallel mechanisms. Instead, we propose that this drug-induced signal network remodeling could be an early organizing node that helps orchestrate other adaptive changes, including treatment-induced hypoxic stress and immune microenvironment remodeling. The precise temporal hierarchy between these processes, however, remains to be fully elucidated, and we therefore present this as a working model rather than a proven sequence. This provides a more causally-linked, integrated framework.
We contend that the effects of anti-VEGF therapy extend far beyond indirect “angiocentricity”; in addition to their vascular effects, these agents can also profoundly and directly influence VEGFR-expressing tumor cells. This influence is dual-faceted: on one hand, it inhibits VEGF-dependent survival signals. On the other hand, and equally importantly, it may release a constitutive brake that VEGF signaling imposes on key pro-invasive pathways. A typical example is the disruption of the VEGFR2-MET complex, which liberates MET’s pro-invasive activity (28). This therapy-induced direct signal remodeling, coupled with the ensuing hypoxic stress and immune–stromal microenvironment changes, collectively forms a “tripartite stress” framework. Through Darwinian selection and induced phenotypic plasticity, this framework ultimately drives tumor evolution toward a more aggressive phenotype. Understanding this process is key to overcoming current therapeutic dilemmas.
2 Core principles of direct VEGF signaling
In this section, we outline four tumor cell–centric principles that explain why anti−VEGF therapy can paradoxically promote invasive escape. We progress from (1) tumor cells as direct and semi−enclosed VEGF targets, to (2) a prototypical brake–accelerator module (VEGFR2–MET), (3) networked escape via NRP−centered hubs and non−canonical VEGF signaling, and (4) evolutionary enforcement through clonal selection and induced plasticity, which together set the stage for the integrative framework in Section 3.
2.1 Tumor cells as direct and semi-enclosed VEGF targets
Traditionally, anti-VEGF therapy has been understood as targeting endothelial cells. However, abundant evidence indicates that various tumor cell types—including GBM, CRC, breast cancer, and renal cancer—themselves express functionally relevant levels of VEGF receptor family members such as VEGFR1, VEGFR2, and the co-receptors neuropilins (NRPs) (29–32). These receptors are not “silent” but constitute functionally active signal transduction systems. On one hand, tumor cells secrete VEGF, forming autocrine or paracrine loops that directly promote their own proliferation, survival, and migration (33, 34). More importantly, several studies have uncovered an intracrine VEGF signaling loop, in which newly synthesized VEGF can directly bind and activate VEGFRs on organelles like the endoplasmic reticulum or Golgi apparatus, without prior secretion, thereby maintaining cell survival and motility in certain tumor contexts (35–38). Collectively, these findings suggest that autocrine/intracrine VEGF signaling is observed across multiple solid tumor types, although its prevalence and functional importance appear to be context-dependent and require further clinical validation.
This “semi-enclosed” signaling system, particularly the intracrine loop, may provide a mechanism contributing to resistance against antibody drugs like bevacizumab, which only neutralize extracellular VEGF. More critically, its importance extends beyond explaining partial resistance. This suggests a core biological context wherein the VEGFR system in tumor cells can function as a continuously active, intrinsic regulatory hub. This persistent endogenous signaling implies that even when extracellular VEGF is blocked, intracellular signaling may remain active, at least transiently, potentially maintaining the suppression of certain bypass pathways. Because tumor cells are direct and complex targets of VEGF signaling, we must understand how their internal signaling network is remodeled upon therapeutic blockade. This leads directly to our core mechanism—the “brake-accelerator” effect.
2.2 A prototypical brake–accelerator module: the VEGFR2–MET complex
Against this backdrop of direct and semi-enclosed VEGF signaling in tumor cells, the effect of VEGF signaling on tumor cells is bidirectional and context-dependent. Beyond its classic role as a pro-survival “accelerator,” it can also function as a “brake” inhibiting invasion. A central feature of this proposed mechanism appears to be the dynamic interaction between receptor tyrosine kinases (RTKs). A landmark study in GBM found that VEGF promotes the formation of a heterocomplex between VEGFR2 and MET (the receptor for hepatocyte growth factor, HGF). Mechanistically, ligand-activated VEGFR2 recruits protein tyrosine phosphatases (such as PTP1B) to the complex, which dephosphorylates MET and suppresses its kinase activity (28). This suppression effectively halts downstream pro-invasive and epithelial-mesenchymal transition (EMT) signaling (28). Consequently, when anti-VEGF drugs block this pathway, the heterocomplex dissociates. This “releases the brake,” liberating suppressed MET signaling and powerfully driving tumor invasion (Figure 1). It should be noted that while research demonstrates this “suppression-release” mechanism in various models (28, 39), its universality and effect intensity across different cancer types are not yet confirmed with equivalent evidence. At present, the VEGFR2–MET heterocomplex is therefore best viewed as a prototypical “brake module” rather than the only form of VEGF-mediated restraint. Emerging data suggest that VEGF/VEGFR signaling can functionally interact with other pro-invasive RTKs—including AXL, fibroblast growth factor receptors (FGFRs), and the epidermal growth factor receptor (EGFR)—through diverse mechanisms such as reciprocal transcriptional regulation, shared adaptor proteins, and co-localization within membrane microdomains (40–43). Whether any of these interactions involve analogous cis-inhibitory complexes remains to be determined and represents an important area for future investigation. In the following sections, we use the VEGFR2–MET module as a mechanistically well-characterized example to illustrate how direct VEGF blockade can shift the balance from “brake” to “accelerator”, while acknowledging that additional, parallel restraints may exist within the broader RTK network.
Figure 1. The “brake-accelerator” model of direct signal remodeling involving Nrp in response to anti-VEGF therapy. This model illustrates how anti-VEGF therapy can paradoxically promote tumor aggressiveness by directly altering the interplay between RTKs and their co-receptors on the tumor cell surface. (Left) Before anti-VEGF therapy (brake on): In the presence of VEGF, its binding to VEGFR promotes the recruitment of Nrp, forming a VEGFR-Nrp-MET heterocomplex. This ternary complex acts as a “brake,” constraining MET-driven pro-invasive signaling pathways and thus maintaining a low invasive phenotype. (Right) After anti-VEGF therapy (accelerator on): Anti-VEGF agents prevent VEGF from binding to its receptor, leading to the dissociation of the ternary complex. However, Nrp remains associated with MET, acting as a signaling hub to stabilize MET or facilitate HGF binding. This distinct reconfiguration “releases the brake” on MET, which then becomes phosphorylated and activated, functioning as an “accelerator” to potently drive downstream signaling and result in a high invasive phenotype.
2.3 Networked escape and NRP-centered signal hubs
Tumor cells exhibit high signaling plasticity in response to VEGF pathway blockade. The escape process may be better conceptualized not as a linear compensation, but rather as a multi-node, networked remodeling. The activation of the MET pathway is the most studied example but represents only a fraction of the network (44, 45). This network encompasses at least three levels: 1) “Brake-release” mechanisms, such as the aforementioned MET signal liberation (28, 39, 44–49); 2) “Bypass activation,” involving the transcriptional upregulation of other RTKs like FGFR, EGFR, platelet-derived growth factor receptor (PDGFR) (18, 50–53); and 3) “Hub amplification,” which upregulates key signal integrators.
Within this escape network, NRP-1 and NRP-2 are critical signaling and network amplification hubs. As key VEGF co-receptors, they regulate stemness, migration, and invasion, while also integrating anti-VEGF pressure with other signals (54–56). Specifically, NRP1 is often associated with angiogenesis, stemness, and transforming growth factor-β (TGF-β)-mediated fibrosis/immunosuppression (57, 58), whereas NRP2 is more prominent in metastasis and lymphangiogenesis (59, 60). Both can crosstalk with the HGF/MET, Ephrin-Eph, and TGF-β networks and promote EMT and metabolic plasticity by linking to the phosphatidylinositol 3-kinase (PI3K)/AKT and mitogen-activated protein kinase (MAPK) axes via integrins/focal adhesion kinase (FAK). Although early NRP inhibitors showed activity, safety and druggability remain challenges (61, 62). Accumulating preclinical work and clinical cohort analyses in several solid tumors support a prognostic—and in some settings potentially predictive—role for NRP1/2 expression, but robust, prospective validation in large trials is still lacking. For instance, a systematic review and meta-analysis confirmed that high NRP1 expression was associated with worse overall survival in both hepatocellular and colorectal cancer cohorts (63). Furthermore, a correlative analysis of the landmark IMvigor210 trial in muscle-invasive bladder cancer revealed that high NRP1 expression not only predicted poor survival but was also associated with resistance to PD-L1 blockade, linking it to an immunosuppressive tumor microenvironment (64). Therefore, we propose that NRPs should be prioritized as key biomarkers and co-inhibitory targets for further validation in sequential combinations with MET/AXL or TGF-β pathway inhibitors.
Beyond canonical VEGFR2-dependent signaling, non-classical VEGF pathways further expand this networked escape. VEGFR1, often described as a regulatory or “decoy” receptor, can nonetheless transduce pro-migratory and pro-invasive signals in certain metastatic models and may modulate the availability of VEGF for VEGFR2 and NRP complexes (65–68). In parallel, NRPs can mediate VEGF-dependent but VEGFR2-independent signaling, and can also engage ligands such as TGF-β, HGF, and class 3 semaphorins to drive tumor growth, DNA damage repair, and therapy resistance even when VEGFR2 is pharmacologically inhibited (69–71). These non-canonical routes provide additional conduits for sustaining invasive and adaptive signaling under VEGF/VEGFR2 blockade and should be considered integral components of the “network remodeling” framework.
2.4 Evolutionary enforcement: clonal selection and induced plasticity
Anti-VEGF therapy acts as a powerful selector, driving the overall malignant evolution of tumors through two parallel and non-mutually exclusive mechanisms. The first mechanism is classic Darwinian clonal selection. In the early stages of treatment, tumor cell clones highly dependent on VEGF autocrine signaling are effectively eliminated, leading to tumor volume reduction. However, pre-existing minority cell clones within the tumor, independent of VEGF signaling or already having activated aggressive pathways such as MET/NRP, gain a significant survival advantage under this selective pressure and eventually proliferate and thrive, becoming the predominant component of recurrent tumors (28, 50, 72). The second mechanism is induced phenotypic plasticity. Therapeutic pressure itself can serve as a potent environmental signal, actively inducing molecular and phenotypic reprogramming in tumor cells. This includes: 1) EMT, which confers motility and invasive capabilities to cells (73, 74); 2) Acquisition of stemness, enabling cells to possess stronger self-renewal and drug resistance (56, 75); 3) Metabolic plasticity, whereby tumor cells adjust their energy metabolism patterns to adapt to resource-scarce environments (19–21). For instance, in GBM models, treatment with the anti-angiogenic drug sunitinib has been found to induce tumor cells to accumulate stem cell markers and exhibit a mesenchymal phenotype (75). These two mechanisms operate in parallel, jointly shaping tumors that are more aggressive and resistant after treatment, and they provide the evolutionary foundation for the “tripartite stress” framework integrating direct signal remodeling, hypoxia, and immune–stromal reprogramming discussed in the following section.
3 Integration and new framework: a “tripartite stress” framework
To more comprehensively understand the malignant evolution of tumors after anti−VEGF therapy, we propose a “tripartite stress model.” This framework explicitly integrates the tumor cell–intrinsic mechanisms detailed in Section 2 (direct signal remodeling and evolutionary enforcement) with the extrinsic pressures of the tumor microenvironment. We posit that three mutually coupled selective stresses act on VEGFR−expressing tumor cells and their microenvironment, and—via clonal selection and phenotypic plasticity—drive tumor evolution toward invasive, therapy−resistant states. Stress 1: Exogenous hypoxic stress. This is the most direct consequence of anti-angiogenic treatment. Vascular regression intensifies intratumoral hypoxia and activates HIF-1α and related transcriptional programs, creating an environment that facilitates activation of alternative pro-angiogenic and pro-invasive pathways (17–19). Stress 2: Endogenous signaling stress (direct signal rewiring). A key component of our working model is that anti-VEGF agents can act directly on tumor cells, abrogating pro-survival signaling (an “accelerator” effect) while simultaneously releasing invasion-suppressive constraints (a “brake” effect)—for example, via dissociation of VEGFR2–MET heterocomplexes—thereby promoting reconfiguration of the RTK network toward MET-dominated invasive phenotypes from within the cell (28, 39). Stress 3: Immune–stromal microenvironmental stress (immune and fibrotic/stromal modulation) (22–24, 76–78). VEGF is itself an immunosuppressive mediator, and its blockade can, in principle, enhance anti-tumor immunity (22, 23, 79, 80). Crucially, emerging evidence reveals that various immune subsets (including T cells and dendritic cells) directly express VEGFRs; thus, VEGF functions not only as an angiogen but as a direct immunosuppressive signal, and its blockade fundamentally rewires intracellular signaling within these immune populations (79, 80). However, prolonged therapeutic pressure and treatment-induced cell death may provoke sterile inflammation and recruit myeloid populations (e.g., MDSCs) that possess both immunosuppressive and pro-angiogenic functions (22–24). Concomitantly, anti−VEGF−induced perturbation of vascular and perivascular niches can activate cancer−associated fibroblasts (CAFs) and promote extracellular matrix (ECM) remodeling and stiffening, which further constrains perfusion and facilitates invasive programs (76–78). Cytokines produced by these cells (such as IL-6 and TGF-β) can in turn promote EMT, stemness, and invasion in tumor cells.
Critically, the precise in vivo temporal sequence between these stresses—specifically, whether direct signal rewiring (VEGFR2–MET dissociation) acts as an initiating event preceding hypoxia-driven HIF stabilization, or whether they occur concurrently—remains to be empirically defined. Establishing this causality requires dedicated experimental validation. Future studies should employ longitudinal sampling (e.g., serial biopsies) in murine models receiving anti-VEGF therapy, encompassing broad time windows from hours to weeks post-treatment. By combining time-resolved assessment of p-MET, p-VEGFR2, and HIF1-α levels with functional imaging of perfusion and hypoxia, researchers can map the temporal trajectory of “brake release” relative to hypoxic adaptation. Additionally, perturbation experiments manipulating the stability of the VEGFR2–MET complex prior to VEGF blockade would serve as a powerful tool to dissect the hierarchical contribution of direct signaling remodeling within this tripartite framework.
These three stresses are not independent but mutually reinforcing, forming positive-feedback loops. For example, hypoxia upregulates MET and NRP1/2, amplifying the consequences of signal rewiring; rewiring-driven EMT and acquisition of stem-like traits increase secretion of IL-6, CCL2 and promote CAFs activation and ECM remodeling/stiffening, which recruits myeloid cells and strengthens TGF-β/signal transducer and activator of transcription 3 (STAT3) signaling (81, 82); myeloid and TGF-β-driven immune/fibrotic programs activate AXL/FAK/PI3K pathways (83, 84), further promoting invasion and vascular co-option and thereby sustaining hypoxia and bypass angiogenesis. The relative timing and dominance of these stresses are likely context− and tumor−type−dependent, but together they form a multidimensional, positive−feedback cooperation that efficiently selects and sculpts tumor cell populations that escape therapy and display high invasive potential (Figure 2).
Figure 2. The “tripartite stress” framework for tumor evolution under anti-VEGF therapy. This framework illustrates how three interconnected stresses converge on the VEGFR-expressing (VEGFR+) tumor cell population following anti-VEGF treatment. These include: (1) endogenous signaling stress resulting from direct drug effects on tumor cells (i.e., the “brake-accelerator” effect); (2) exogenous hypoxic stress caused by vascular regression; and (3) a composite immune-stromal microenvironmental stress. This latter stress arises from the dynamic remodeling of not only immune cells and cytokines but also the stromal compartment, including key effectors like CAFs and the restructured ECM. The colored arrows connecting the three stress nodes represent a cyclic, mutually reinforcing network of crosstalk: for example, hypoxic stress exacerbates intracellular signaling pressure; altered signaling promotes stromal activation (recruiting CAFs); and a remodeled, stiff ECM further aggravates hypoxia by compressing residual vessels. Collectively, these forces drive clonal selection and phenotypic plasticity, ultimately promoting the emergence of aggressive and resistant tumor cells.
From this model we derive several falsifiable predictions: (1) concomitant inhibition of MET during VEGF blockade should markedly attenuate the invasion/EMT rebound observed with anti-VEGF monotherapy; (2) increases in hypoxia markers [HIF-1α, carbonic anhydrase IX (CA9)] should coincide with elevated p-MET and upregulation of NRP1/2 and with imaging-detectable infiltrative recurrence; (3) higher MDSC burden and elevated TGF-β/IL-6 levels should predict enhancement of EMT/stemness features and clinical resistance. At an operational level, the model supports using dynamic biomarkers (e.g., phosphorylated MET (p-MET), circulating tumor DNA (ctDNA), evidence of MET amplification, extracellular vesicle (EV)-associated VEGF, NRP1/2 expression, and MDSC/TGF-β/IL-6 levels) as triggerable readouts to stratify patients for addition or sequencing of MET/AXL/NRP or TGF-β pathway inhibitors. These candidates differ in their current levels of clinical validation and should be prioritized and categorized accordingly. It further argues for aligning anti-VEGF dosing with the “vascular normalization window” as suggested by preclinical and imaging studies and for optimizing the temporal integration of anti-angiogenic therapy with immunotherapy.
4 Counterexamples, heterogeneity, and boundary conditions
Any biological model has its applicability boundaries. The “direct signal remodeling” model proposed in this paper is no exception; the intensity and manifestation of its effects are constrained by various factors, which are crucial for the precise application of the model.
4.1 Cancer-specific signal remodeling and modes of aggressive evolution
In clear cell renal cell carcinoma (ccRCC), upregulation of AXL/MET and an enhanced mesenchymal phenotype are frequently reported after prolonged VEGF axis inhibition (85). Clinically, analysis of tumor tissues from patients with TKI-resistant ccRCC has confirmed that resistance to agents like sunitinib or sorafenib is actively driven by the downregulation of the tumor suppressor MCPIP1 and the concomitant activation of c-Met signaling, which directly promotes metastatic progression (86). The clinical benefit of multi-targeted tyrosine kinase inhibitors (TKIs) like cabozantinib, which directly inhibit MET/AXL, is consistent with this biological directionality (87, 88). This suggests that if the “MET-AXL brake” can be simultaneously applied after “releasing the VEGF brake,” the net effect of aggressive rebound might be reduced. Such observations provide a biological rationale for preventive or trigger-based combinations utilizing dynamic biomarkers (e.g., p-MET, AXL expression, and NRP spatial co-localization).
In CRC, compensatory FGF/FGFR and HGF/MET signaling, along with vessel co-option and infiltrative transformation, are observed after anti-VEGF therapy, particularly in liver metastases (89, 90). This phenomenon suggests that “non-angiogenesis-dependent growth” constitutes a critical boundary condition: even if VEGF axis inhibition is further intensified, it may be difficult to reverse the aggressive migratory advantage if FGFR/MET and associated adhesion-migration programs are not simultaneously inhibited. Correspondingly, FGFR activation markers and histological vessel co-option phenotypes should be incorporated into future stratification and monitoring strategies. Conceptually, this represents a vascular-pattern boundary condition within the tripartite-stress framework, where co-option-driven invasion can partially replace hypoxia-driven bypass angiogenesis.
In breast cancer (BC), particularly in triple-negative breast cancer (TNBC) and other Human Epidermal Growth Factor Receptor 2 (HER2)-negative subtypes, the association of AXL and NRP2 with invasion/metastasis and immunosuppression is especially prominent in several cohorts and preclinical models (91, 92). In a hypoxic-VEGF context, the interaction of Ephrin/AXL/NRP can act as an “amplifier” of the plasticity network, enhancing EMT, stemness, and immune evasion (93, 94). This implies that if VEGF is merely suppressed in a single dimension without constraining the NRP2/AXL hub, an enhanced phenotype due to “pathway diversion” might occur. In combination strategies, attention should be paid to the spatial localization and dynamic levels of NRP2 and AXL, serving as trigger signals for the addition of inhibitors.
In hepatocellular carcinoma (HCC), FGFR4/FGF19 and MET are often found to be activated in parallel after anti-angiogenic therapy, with emerging evidence linking this to reprogramming related to lipid and bile acid metabolism and immune evasion (95–97). This “growth factor-metabolism-immunity” linkage characteristic suggests that continuous inhibition of a single axis can easily be offset by parallel circuits. Therefore, the timed addition of inhibitors targeting FGFR4/FGF19 or MET needs to be accompanied by dynamic co-reading of metabolic and immune biomarkers (e.g., FGF19 levels, tumor-infiltrating CD8+ cells, and MDSC burden).
In NSCLC, in EGFR wild-type or kirsten rat sarcoma viral oncogene homolog (KRAS)-mutated populations, the combined benefit of VEGF inhibition and immunotherapy has been associated with baseline or treatment-induced HGF/MET activity (72, 98), which aligns with the “vascular-immune-bypass” tripartite stress model. At baseline, elevated HGF/MET activity may indicate a greater dependence on bypass circuitry and thus a higher rationale for upfront combination strategies. During therapy, treatment-induced rises in p-MET or ctDNA-MET amplification can serve as triggerable thresholds for the addition or sequencing of MET/NRP/AXL inhibitors to achieve feed-forward blockade of aggressive rebound. Strategically, during the early “vascular normalization window,” T cell entry can be optimized.
Counterexamples and boundary cases are also informative. In some tumor types or metastatic niches with low intrinsic dependence on VEGF autocrine signaling, or with pre-existing non-angiogenic growth programs, anti-VEGF therapy may predominantly yield vascular normalization or immune activation without a marked invasive rebound (99). Conversely, in lesions dominated by vessel co-option or alternative vascularization modes, the explanatory power of direct VEGF–RTK rewiring may be reduced unless coupled to stromal and vascular-pattern readouts (89, 100).
In summary, this cross-cancer analysis suggests that after VEGF axis inhibition, the “accelerator effect” of MET/AXL/FGFR and NRP hubs, along with histological vessel co-option, jointly shape aggressive adaptation. The predominant bypass pathways and ecological boundaries differ among cancer types, emphasizing the complexity and application specificity of the model in different clinical contexts (Table 1).
Table 1. Heterogeneity of resistance to anti-VEGF therapy and precision response strategies.
4.2 Model universality and clinical considerations
First, there is universal heterogeneity in the intrinsic tumor context. The expression levels, spatial localization, and dependence on VEGF signaling of VEGFR/MET/NRP vary significantly across different tumor types and even different molecular subtypes within the same tumor, which determines the intensity of the “direct signal remodeling” effect. For example, in GBM and ovarian cancer, the VEGFR2-MET complex-mediated “brake” mechanism has the strongest experimental evidence (28, 39), but its universality and functional intensity in other cancer types still require broader validation, which corroborates the cross-cancer heterogeneity evidence we reviewed in Section 4.1.
Second, differences in drug mechanisms are a key limiting condition for understanding clinical phenomena. When discussing the direct effects of anti-VEGF therapy, it is crucial to precisely distinguish specific drug types, as they play distinct roles in the “brake-accelerator” model. For instance, the humanized anti-VEGFA monoclonal antibody bevacizumab primarily neutralizes extracellular VEGF, leading to the “release of the brake” (101). Given their high structural and functional similarity and comparable clinical efficacy and safety reported for approved bevacizumab biosimilars, these agents are expected to produce largely similar pharmacodynamic effects, consistent with the same “brake-release” and downstream signal-remodeling logic. Small molecule TKIs, on the other hand, present a more complex picture. Some TKIs, such as apatinib, primarily target VEGFR. Because they can penetrate cell membranes, they may suppress intracellular VEGFR activity to some extent; their “brake-release” effect is similar to bevacizumab, but they may exert a stronger inhibitory effect on endogenous survival signals (102). More unique are multi-target TKIs, such as cabozantinib, which are themselves potent inhibitors of both VEGFR and MET (103, 104). The mechanism of action of these drugs is highly specific: while they inhibit VEGFR2, leading to the “release of the brake,” they simultaneously directly inhibit the liberated MET pathway. This is equivalent to an operation within a complex mechanical system that simultaneously releases brake A and applies brake B. This unique dual inhibitory effect means that their clinical manifestations and resistance mechanisms may be distinctly different from drugs that solely target VEGF. Therefore, the specific patterns of “direct signal remodeling” induced by different drugs vary significantly. This clinical concordance supports, rather than proves, our model and highlights the importance of biomarker-guided attribution, as cabozantinib’s benefit is likely multifactorial. It underscores our central thesis: a drug’s clinical behavior is dictated by its precise impact on the tumor cell’s internal signal network. Simply classifying all anti-angiogenic drugs as one category is insufficient; only by delving into the tumor cell level and precisely understanding the direct interaction patterns between specific drugs and intracellular signaling networks (especially key nodes like VEGFR-MET) can we truly predict and ultimately overcome therapeutic resistance.
Finally, the complexity of clinical outcomes also requires comprehensive consideration. Not all patients receiving anti-VEGF therapy will exhibit enhanced aggression. The “vascular normalization window” theory—which posits that at specific doses and times, anti-VEGF therapy can suppress rather than promote invasion by repairing distorted tumor vascular structures, thereby improving oxygen supply and drug delivery (105, 106)—does not contradict our model but rather defines its temporal and dose-dependent boundaries. It suggests that therapeutic pressure is a continuum: below a certain threshold, it may lead to beneficial normalization; above it, it triggers the “tripartite stress” and adaptive resistance cascade we describe. Understanding where a patient’s tumor lies on this continuum is therefore a key clinical challenge. Defining this patient-specific threshold will likely require longitudinal vascular, hypoxia, RTK-rewiring, and immune–stromal readouts (see Table 2). Furthermore, differences in the immune–stromal tumor microenvironment (e.g., stromal cell types, initial immune cell infiltration status) also greatly influence the ultimate therapeutic response (99, 107). These factors, alongside the dose- and time-dependent nature of the therapeutic pressure, collectively constitute the complex, context-dependent boundaries for understanding and applying the model presented in this paper.
Table 2. Comparison of anti-VEGF agents within the “brake-accelerator” framework.
5 Clinical implications and future directions
To bridge the gap from our proposed model to improved patient outcomes, this section outlines future strategies. Crucially, we aim to establish translational realism as a cornerstone by addressing not only the strategic rationale for novel biomarkers and combinations but also the practical hurdles inherent in their clinical implementation, including feasibility, standardization, and toxicity management. Based on the “direct signal remodeling” and “tripartite stress” models, we can propose the following specific directions for optimizing anti-VEGF therapeutic strategies.
5.1 Next-generation dynamic biomarkers
Future efficacy prediction and resistance monitoring should shift from single, indirect serum VEGF levels to multi-dimensional, dynamic biomarkers that better reflect the intrinsic state of tumor cells. Importantly, these candidates differ in their current levels of clinical validation and should be tiered and prospectively verified in biomarker−integrated trials (Tables 1, 2). The focus of detection should shift to the tumor tissue itself, including the use of immunohistochemistry or spatial transcriptomics to assess the levels of p-MET/total MET, and the expression and spatial co-localization of VEGFR/NRP on tumor cells before treatment. However, recognizing that spatial readouts are currently research-grade and require rigorous standardization before clinical deployment, and that repeated tissue biopsies are often limited by invasiveness and patient compliance, dynamic liquid biopsy emerges as a critical alternative. During treatment, monitoring changes in MET amplification in ctDNA in plasma, or detecting the levels of VEGF protein encapsulated in exosomes that can evade antibody neutralization, as suggested by preclinical and early translational studies (e.g., EV−bound VEGF that is bevacizumab−insensitive) (108). Mechanistically, these extracellular vesicles can be internalized by tumor cells via endocytosis, thereby delivering VEGF to intracellular compartments to sustain intracrine signaling, potentially bypassing surface blockade. This may provide sensitive early signals for warning of aggressive transformation and guiding treatment adjustments.
5.2 Precision combination and timed dosing strategies
Managing the balance between efficacy and toxicity is paramount when strictly translating these models into practice. The most direct inference from this model is that key escape pathways should be preventatively blocked concurrently with anti-VEGF therapy. Precision combination therapy based on biomarkers is a core strategy; for example, for patients with high baseline MET expression or MET activation during treatment, MET inhibitors (e.g., Cabozantinib, Crizotinib) may be considered early in combination. However, the combination of anti-VEGF agents with multi-target TKIs creates a risk of overlapping toxicities (e.g., exacerbated hypertension, proteinuria, or fatigue). Additionally, combination strategies targeting other key hubs like NRP1/2 are also worthy of exploration, although the druggability and safety profiles of clinical−grade NRP inhibitors remain to be established. Concurrently, innovative treatment timing and dosing strategies should be explored. Sustained high-intensity VEGF blockade may create the strongest selective pressure. Therefore, low-dose metronomic dosing based on the “vascular normalization” theory, or “adaptive therapy” aimed at controlling rather than eradicating tumors, by intermittent dosing to slow down selective pressure, may maintain efficacy while maximally delaying the emergence of aggressive resistance (105, 106). This rationale is supported by preclinical work indicating that lower, normalization−range anti−angiogenic dosing can sustain vascular function and reduce immunosuppressive rebound relative to prolonged high−dose blockade (99).
Building on this, the combination of anti-VEGF and immune checkpoint inhibitors (ICIs) has also shown clear mechanistic complementarity and significant clinical benefits. The short-term “vascular normalization window” can alleviate tumor hypoxia, reduce the accumulation of MDSCs and regulatory T cells (Tregs), improve dendritic cell maturation and CD8+ T cell infiltration, and mitigate immune exclusion caused by endothelial barrier dysfunction and adhesion molecule imbalance, thereby creating a more favorable environment for immune attack (107, 108). The magnitude of this synergy is dose−, timing−, and cancer−type−dependent, consistent with the normalization−to−stress continuum proposed in our model. Clinically, multiple key studies have confirmed the effectiveness of this combination strategy. For instance, the IMpower150 study demonstrated that in NSCLC, atezolizumab+bevacizumab+chemotherapy significantly improved PFS and OS compared to the control group, with patients in the liver metastasis subgroup benefiting particularly, providing strong external validation for “vascular-immune coupling” (109). In RCC, multiple Phase III studies (e.g., KEYNOTE-426 with Axitinib+Pembrolizumab, JAVELIN Renal 101 with Axitinib+Avelumab, CLEAR with Lenvatinib+Pembrolizumab, and CheckMate 9ER with Cabozantinib+Nivolumab) have all demonstrated significant efficacy advantages in first-line treatment (110–113). This evidence collectively supports a strategy where, when the VEGF axis is inhibited, synchronous immune activation is employed to counteract the combined pressure of bypass activation and immunosuppression. Based on the “Tripartite Stress” framework, we suggest considering lower doses or metronomic anti-VEGF therapy during the early vascular normalization window to optimize T cell infiltration. Simultaneously, based on dynamic biomarker thresholds (e.g., rising p-MET, complex dissociation, ctDNA-MET amplification), MET/NRP/AXL inhibitors can be layered or adjusted in sequence. Crucially, this dynamic regulation of dosing intensity and sequencing aims to achieve a “feed-forward blockade” of aggressive rebound while explicitly mitigating the cumulative toxicity observed in concurrent distinct-pathway inhibition.
5.3 Key considerations for future clinical trial design
To translate the above theories into clinical benefits, future clinical trial designs should be adjusted accordingly. It is recommended that early Phase I/II studies prospectively integrate dynamic biomarker monitoring. For example, trial protocols should specify collection of blood samples and (where possible) tumor tissue at multiple time points to systematically monitor changes in p-MET, EV-VEGF, and circulating tumor cell (CTC) EMT markers, and correlate them with imaging assessments and clinical outcomes. Furthermore, beyond biomarker monitoring for clinical decision-making, future trials should particularly emphasize the systematic, longitudinal collection of tumor tissue samples from patients after recurrence/progression. Only by performing multi-dimensional comparative analysis, such as single-cell sequencing and spatial transcriptomics (114, 115), on paired pre- and post-treatment samples, can we ultimately dissect the relative contributions of “clonal selection” and “phenotypic plasticity” at the clinical level, thereby providing the highest level of direct evidence for the model proposed in this paper. Additionally, adaptive trial designs based on biomarkers can be attempted (116, 117). While such designs are still largely prospective proposals in anti−VEGF settings, they provide a practical route to test the model’s trigger−based combination logic. For example, design a trial where all patients initially receive a standard anti-VEGF regimen. When a biomarker (e.g., p-MET) exceeds a preset threshold, patients are randomized into either the group continuing the original regimen or the group combining with a MET inhibitor. This design can more efficiently validate the value of combination strategies and provide high-level evidence for individualized treatment pathways.
6 Conclusion
Aggressive resistance emerging after anti-VEGF therapy is a central obstacle limiting its clinical application potential. The “direct signal remodeling” model, and its extended “tripartite stress” framework, presented in this paper, emphasize that academia and clinicians need to focus more on tumor cells themselves, beyond the traditional “angiocentric” view. As a working prototype rather than a universal mechanism, this framework encourages us to recognize that anti-VEGF drugs may not merely “starve” tumors but also actively and directly remodel the intrinsic signaling networks of tumor cells, resulting in complex dual effects of “accelerator” inhibition and “brake” release. This process, under the synergy of hypoxic and immune–stromal microenvironmental pressures, driven by clonal selection and phenotypic plasticity, can lead to the malignant evolution of tumors, with the relative timing and dominance of each stress likely varying across tumor types and treatment contexts. Future research should focus on utilizing multi-omics, spatially resolved, and dynamic monitoring biomarkers to precisely delineate this process, and accordingly develop mechanism-based precision combination therapies and innovative adaptive treatment strategies, ideally validated through longitudinal paired-sample studies and trigger−based adaptive trials. Only then can we truly wield this “double-edged sword” of anti-VEGF therapy, through biomarker−guided, dose− and timing−adapted, cancer−type−specific approaches, maximizing its therapeutic potential to bring more lasting and profound survival benefits to a wide range of cancer patients.
Data availability statement
The original contributions presented in the study are included in the article/supplementary material. Further inquiries can be directed to the corresponding author.
Author contributions
PW: Writing – original draft. JL: Writing – review & editing. XC: Writing – review & editing. YL: Writing – review & editing. QX: Writing – review & editing. ZW: Writing – review & editing.
Funding
The author(s) declared that financial support was received for this work and/or its publication. The present research was supported by the National Natural Science Foundation of China (32260157), the Science and Technology Support Program of Guizhou Province ZK[2022]609), and The Key Construction Discipline of Immunology and Pathogen biology in Zhuhai Campus of Zunyi Medical University (ZHGF2024-1).
Acknowledgments
The authors would like to express their gratitude to Figdraw (www.figdraw.com) developed by HOME for Researchers (www.home-for-researchers.com) for the assistance with the “brake-accelerator” model illustration (ID No. PYWIT94508).
Conflict of interest
The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Generative AI statement
The author(s) declared that generative AI was not used in the creation of this manuscript.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
References
1. Liu ZL, Chen HH, Zheng LL, Sun LP, and Shi L. Angiogenic signaling pathways and anti-angiogenic therapy for cancer. Signal Transduct Target Ther. (2023) 8:198. doi: 10.1038/s41392-023-01460-1
2. Guo Z, Jing X, Sun X, Sun S, Yang Y, and Cao Y. Tumor angiogenesis and anti-angiogenic therapy. Chin Med J. (2024) 137:2043–51. doi: 10.1097/CM9.0000000000003231
3. Garcia J, Hurwitz H, Sandler AB, Miles D, Coleman RL, Deurloo R, et al. Bevacizumab (Avastin®) in cancer treatment: A review of 15 years of clinical experience and future outlook. Cancer Treat Rev. (2020) 86:102017. doi: 10.1016/j.ctrv.2020.102017
4. Song Y, Mao Q, Zhou M, Liu CJ, Kong L, and Hu T. Effectiveness of bevacizumab in the treatment of metastatic colorectal cancer: a systematic review and meta-analysis. BMC Gastroenterol. (2024) 24:58. doi: 10.1186/s12876-024-03134-w
5. Chen H, Lei Y, Zhou J, Lv C, and Xuan Q. Advances in bevacizumab in colorectal cancer: a bibliometric analysis from 2004 to 2023. Front Oncol. (2025) 15:1552914. doi: 10.3389/fonc.2025.1552914
6. Di Costanzo F, Mazzoni F, Micol Mela M, Antonuzzo L, Checcacci D, Saggese M, et al. Bevacizumab in non-small cell lung cancer. Drugs. (2008) 68:737–46. doi: 10.2165/00003495-200868060-00002
7. Seiwerth F, Bitar L, Samaržija M, and Jakopović M. Long-term progression-free survival in non-small cell lung cancer patients: a spotlight on bevacizumab and its biosimilars. Expert Opin Biol Ther. (2024) 24:1017–24. doi: 10.1080/14712598.2024.2405562
8. Fu M, Zhou Z, Huang X, Chen Z, Zhang L, Zhang J, et al. Use of Bevacizumab in recurrent glioblastoma: a scoping review and evidence map. BMC Cancer. (2023) 23:544. doi: 10.1186/s12885-023-11043-6
9. Liu X, Chen Z, Yan P, Yang T, Zong D, Guo W, et al. Treatment mechanism and research progress of bevacizumab for glioblastoma. Am J Cancer Res. (2025) 15:1874–901. doi: 10.62347/RNUE7193
10. Haibe Y, Kreidieh M, El Hajj H, Khalifeh I, Mukherji D, Temraz S, et al. Resistance mechanisms to anti-angiogenic therapies in cancer. Front Oncol. (2020) 10:221. doi: 10.3389/fonc.2020.00221
11. Schiffmann LM, Bruns CJ, and Schmidt T. Resistance mechanisms of the metastatic tumor microenvironment to anti-angiogenic therapy. Front Oncol. (2022) 12:897927. doi: 10.3389/fonc.2022.897927
12. Pàez-Ribes M, Allen E, Hudock J, Takeda T, Okuyama H, Viñals F, et al. Antiangiogenic therapy elicits Malignant progression of tumors to increased local invasion and distant metastasis. Cancer Cell. (2009) 15:220–31. doi: 10.1016/j.ccr.2009.01.027
13. Ebos JM, Lee CR, Cruz-Munoz W, Bjarnason GA, Christensen JG, and Kerbel RS. Accelerated metastasis after short-term treatment with a potent inhibitor of tumor angiogenesis. Cancer Cell. (2009) 15:232–9. doi: 10.1016/j.ccr.2009.01.021
14. Lomax AJ, Yannakou CK, and Rosenthal MA. Spinal cord metastasis in a patient treated with bevacizumab for glioblastoma. Target Oncol. (2013) 8:153–5. doi: 10.1007/s11523-013-0270-0
15. Khattab MH, Marciscano AE, Lo SS, Lim M, Laterra JJ, Kleinberg LR, et al. Antiangiogenic therapies and extracranial metastasis in glioblastoma: A case report and review of the literature. Case Rep Oncol Med. (2015) 2015:431819. doi: 10.1155/2015/431819
16. Keunen O, Johansson M, Oudin A, Sanzey M, Rahim SA, Fack F, et al. Anti-VEGF treatment reduces blood supply and increases tumor cell invasion in glioblastoma. Proc Natl Acad Sci USA. (2011) 108:3749–54. doi: 10.1073/pnas.1014480108
17. Bergers G and Hanahan D. Modes of resistance to anti-angiogenic therapy. Nat Rev Cancer. (2008) 8:592–603. doi: 10.1038/nrc2442
18. Itatani Y, Kawada K, Yamamoto T, and Sakai Y. Resistance to anti-angiogenic therapy in cancer—alterations to anti-VEGF pathway. Int J Mol Sci. (2018) 19:1232. doi: 10.3390/ijms19041232
19. McIntyre A and Harris AL. Metabolic and hypoxic adaptation to anti-angiogenic therapy: a target for induced essentiality. EMBO Mol Med. (2015) 7:368–79. doi: 10.15252/emmm.201404271
20. Chandra A, Rick J, Yagnik G, and Aghi MK. Autophagy as a mechanism for anti−angiogenic therapy resistance. Semin Cancer Biol. (2020) 66:75–88. doi: 10.1016/j.semcancer.2019.08.031
21. Reyes−Castellanos G, Hadi NA, and Carrier A. Autophagy contributes to metabolic reprogramming and therapeutic resistance in pancreatic tumors. Cells. (2022) 11:426. doi: 10.3390/cells11030426
22. Bourhis M, Palle J, Galy−Fauroux I, and Terme M. Direct and indirect modulation of T Cells by VEGF-A counteracted by anti−angiogenic treatment. Front Immunol. (2021) 12:616837. doi: 10.3389/fimmu.2021.616837
23. Yang J, Yan J, and Liu B. Targeting VEGF/VEGFR to modulate antitumor immunity. Front Immunol. (2018) 9:978. doi: 10.3389/fimmu.2018.00978
24. Horikawa N, Abiko K, Matsumura N, Baba T, Hamanishi J, Yamaguchi K, et al. Anti−VEGF therapy resistance in ovarian cancer is caused by GM−CSF−induced myeloid−derived suppressor cell recruitment. Br J Cancer. (2020) 122:778–88. doi: 10.1038/s41416-019-0725-x
25. Epstein RJ. VEGF signaling inhibitors: More pro-apoptotic than anti-angiogenic. Cancer Metastasis Rev. (2007) 26:443–52. doi: 10.1007/s10555-007-9071-1
26. Ntellas P, Mavroeidis L, Gkoura S, Gazouli I, Amylidi AL, Papadaki A, et al. Old player-new tricks: Non angiogenic effects of the VEGF/VEGFR pathway in cancer. Cancers. (2020) 12:3145. doi: 10.3390/cancers12113145
27. Wang ZY, Li JQ, Guo JJ, and Wei P. Direct antitumor activity of bevacizumab: an overlooked mechanism? Front Pharmacol. (2024) 15:1394878. doi: 10.3389/fphar.2024.1394878
28. Lu KV, Chang JP, Parachoniak CA, Pandika MM, Aghi MK, Meyronet D, et al. VEGF inhibits tumor cell invasion and mesenchymal transition through a MET/VEGFR2 complex. Cancer Cell. (2012) 22:21–35. doi: 10.1016/j.ccr.2012.05.037
29. Szabó E, Schneider H, Seystahl K, Rushing EJ, Herting F, Weidner KM, et al. Autocrine VEGFR1 and VEGFR2 signaling promotes survival in human glioblastoma models in vitro and in vivo. Neuro Oncol. (2016) 18:1242–52. doi: 10.1093/neuonc/now043
30. Grandclement C, Pallandre JR, Valmary Degano S, Viel E, Bouard A, Balland J, et al. Neuropilin-2 expression promotes TGF-β1-mediated epithelial to mesenchymal transition in colorectal cancer cells. PLoS One. (2011) 6:e20444. doi: 10.1371/journal.pone.0020444
31. Ghosh S, Sullivan CA, Zerkowski MP, Molinaro AM, Rimm DL, Camp RL, et al. High levels of vascular endothelial growth factor and its receptors (VEGFR−1, VEGFR−2, neuropilin−1) are associated with worse outcome in breast cancer. Hum Pathol. (2008) 39:1835–43. doi: 10.1016/j.humpath.2008.06.004
32. Roskoski R Jr. Vascular endothelial growth factor (VEGF) and VEGF receptor inhibitors in the treatment of renal cell carcinomas. Pharmacol Res. (2017) 120:116–32. doi: 10.1016/j.phrs.2017.03.010
33. Goel HL and Mercurio AM. VEGF targets the tumour cell. Nat Rev Cancer. (2013) 13:871–82. doi: 10.1038/nrc3627
34. Shaw P, Dwivedi SKD, Bhattacharya R, Mukherjee P, and Rao G. VEGF signaling: Role in angiogenesis and beyond. Biochim Biophys Acta Rev Cancer. (2024) 1879:189079. doi: 10.1016/j.bbcan.2024.189079
35. Wiszniak S and Schwarz Q. Exploring the intracrine functions of VEGF-A. Biomolecules. (2021) 11:128. doi: 10.3390/biom11010128
36. Bhattacharya R, Fan F, Wang R, Ye X, Xia L, Boulbes D, et al. Intracrine VEGF signalling mediates colorectal cancer cell migration and invasion. Br J Cancer. (2017) 117:848–55. doi: 10.1038/bjc.2017.238
37. Lee SH, Jeong D, Han YS, and Baek MJ. Intracrine VEGF signaling mediates the activity of prosurvival pathways in human colorectal cancer cells. Cancer Res. (2016) 76:3014–24. doi: 10.1158/0008-5472.CAN-15-1605
38. Peng S, Zhang Y, Peng H, Ke Z, Xu L, Su T, et al. Intracellular autocrine VEGF signaling promotes EBDC cell proliferation, which can be inhibited by apatinib. Cancer Lett. (2016) 373:193–202. doi: 10.1016/j.canlet.2016.01.015
39. Morgan RD, Ferreras C, Peset I, Avizienyte E, Renehan AG, Edmondson RJ, et al. c-MET/VEGFR-2 co-localisation impacts on survival following bevacizumab therapy in epithelial ovarian cancer: an exploratory biomarker study of the phase 3 ICON7 trial. BMC Med. (2022) 20:59. doi: 10.1186/s12916-022-02270-y
40. Yadav M, Sharma A, Patne K, Tabasum S, Suryavanshi J, Rawat L, et al. AXL signaling in cancer: from molecular insights to targeted therapies. Signal Transduct Target Ther. (2025) 10:37. doi: 10.1038/s41392-024-02121-7
41. Boichuk S, Dunaev P, Galembikova A, and Valeeva E. Fibroblast growth factor 2 (FGF2) activates vascular endothelial growth factor (VEGF) signaling in gastrointestinal stromal tumors (GIST): An autocrine mechanism contributing to imatinib mesylate (IM) resistance. Cancers. (2024) 16:3103. doi: 10.3390/cancers16173103
42. Patel SA, Nilsson MB, Le X, Cascone T, Jain RK, and Heymach JV. Molecular mechanisms and future implications of VEGF/VEGFR in cancer therapy. Clin Cancer Res. (2023) 29:30–9. doi: 10.1158/1078-0432.CCR-22-1366
43. Huang L and Fu L. Mechanisms of resistance to EGFR tyrosine kinase inhibitors. Acta Pharm Sin B. (2015) 5:390–401. doi: 10.1016/j.apsb.2015.07.001
44. McCarty JH. Glioblastoma resistance to anti-VEGF therapy: Has the challenge been MET? Clin Cancer Res. (2013) 19:1631–3. doi: 10.1158/1078-0432.CCR-13-0051
45. Carvalho B, Lopes JM, Silva R, Peixoto J, Leitão D, Soares P, et al. The role of c-Met and VEGFR2 in glioblastoma resistance to bevacizumab. Sci Rep. (2021) 11:6067. doi: 10.1038/s41598-021-85385-1
46. Sennino B, Ishiguro-Oonuma T, Wei Y, Naylor RM, Williamson CW, Bhagwandin V, et al. Suppression of tumor invasion and metastasis by concurrent inhibition of c-Met and VEGF signaling in pancreatic neuroendocrine tumors. Cancer Discov. (2012) 2:270–87. doi: 10.1158/2159-8290.CD-11-0240
47. Ciamporcero E, Miles KM, Adelaiye R, Ramakrishnan S, Shen L, Ku S, et al. Combination strategy targeting VEGF and HGF/c-MET in human renal cell carcinoma models. Mol Cancer Ther. (2014) 14:101–10. doi: 10.1158/1535-7163.MCT-14-0094
48. Piao Y, Park SY, Henry V, Smith BD, Tiao N, Flynn DL, et al. Novel MET/TIE2/VEGFR2 inhibitor altiratinib inhibits tumor growth and invasiveness in bevacizumab-resistant glioblastoma mouse models. Neuro Oncol. (2016) 18:1230–41. doi: 10.1093/neuonc/now030
49. Tomida C, Yamagishi N, Nagano H, Uchida T, Ohno A, Hirasaka K, et al. VEGF pathway-targeting drugs induce evasive adaptation by activation of neuropilin-1/c-Met in colon cancer cells. Int J Oncol. (2018) 52:1350–62. doi: 10.3892/ijo.2018.4291
50. Ichikawa K, Miyano SW, Minoshima Y, Matsui J, and Funahashi Y. Activated FGF2 signaling pathway in tumor vasculature is essential for acquired resistance to anti-VEGF therapy. Sci Rep. (2020) 10:2939. doi: 10.1038/s41598-020-59853-z
51. Zahra FT, Sajib MS, and Mikelis CM. Role of bFGF in acquired resistance upon anti−VEGF therapy in cancer. Cancers. (2021) 13:1422. doi: 10.3390/cancers13061422
52. Liu T, Ma W, Xu H, Huang M, Zhang D, He Z, et al. PDGF−mediated mesenchymal transformation renders endothelial resistance to anti−VEGF treatment in glioblastoma. Nat Commun. (2018) 9:3439. doi: 10.1038/s41467-018-05982-z
53. Cascone T, Herynk MH, Xu L, Du Z, Kadara H, Nilsson MB, et al. Upregulated stromal EGFR and vascular remodeling in mouse xenograft models of angiogenesis inhibitor-resistant human lung adenocarcinoma. J Clin Invest. (2011) 121:1313–28. doi: 10.1172/JCI42405
54. Rizzolio S and Tamagnone L. Multifaceted role of neuropilins in cancer. Curr Med Chem. (2011) 18:3563–75. doi: 10.2174/092986711796642544
55. Prud’homme GJ and Glinka Y. Neuropilins are multifunctional coreceptors involved in tumor initiation, growth, metastasis and immunity. Oncotarget. (2012) 3:921–39. doi: 10.18632/oncotarget.626
56. Mercurio AM. VEGF/neuropilin signaling in cancer stem cells. Int J Mol Sci. (2019) 20:490. doi: 10.3390/ijms20030490
57. Chaudhary B, Khaled YS, Ammori BJ, and Elkord E. Neuropilin 1: Function and therapeutic potential in cancer. Cancer Immunol Immunother. (2014) 63:81–99. doi: 10.1007/s00262-013-1500-0
58. Varanasi SM, Gulani Y, Rachamala HK, Mukhopadhyay D, and Angom RS. Neuropilin-1: A multifaceted target for cancer therapy. Curr Oncol. (2025) 32:203. doi: 10.3390/curroncol32040203
59. Lungulescu C, Ghimpau V, Gheonea DI, Sur D, and Lungulescu CV. The role of neuropilin-2 in the epithelial to mesenchymal transition of colorectal cancer: A systematic review. Biomedicines. (2022) 10:172. doi: 10.3390/biomedicines10010172
60. Islam R, Mishra J, Bodas S, Bhattacharya S, Batra S, Dutta S, et al. Role of neuropilin-2-mediated signaling axis in cancer progression and therapy resistance. Cancer Metastasis Rev. (2022) 41:771–87. doi: 10.1007/s10555-022-10048-0
61. Zhao L, Chen HY, Lu L, Wang L, Zhang XK, and Guo XL. New insights into the role of co-receptor neuropilins in tumour angiogenesis and lymphangiogenesis and targeted therapy strategies. J Drug Target. (2021) 29:155–67. doi: 10.1080/1061186X.2020.1815210
62. Grandclément C and Borg C. Neuropilins: A new target for cancer therapy. Cancers. (2011) 3:1899–928. doi: 10.3390/cancers3021899
63. Fernández-Palanca P, Payo-Serafín T, Fondevila F, Méndez-Blanco C, San-Miguel B, Romero MR, et al. Neuropilin-1 as a potential biomarker of prognosis and invasive-related parameters in liver and colorectal cancer: A systematic review and meta-analysis of human studies. Cancers. (2022) 14:3455. doi: 10.3390/cancers14143455
64. Yu Y, Zeng H, Jin K, You R, Liu Z, Zhang H, et al. Immune inactivation by neuropilin-1 predicts clinical outcome and therapeutic benefit in muscle-invasive bladder cancer. Cancer Immunol Immunother. (2022) 71:2117–26. doi: 10.1007/s00262-022-03153-0
65. Ceci C, Atzori MG, Lacal PM, and Graziani G. Role of VEGFs/VEGFR-1 signaling and its inhibition in modulating tumor invasion: experimental evidence in different metastatic cancer models. Int J Mol Sci. (2020) 21:1388. doi: 10.3390/ijms21041388
66. Weddell JC, Chen S, and Imoukhuede PI. VEGFR1 promotes cell migration and proliferation through PLCγ and PI3K pathways. NPJ Syst Biol Appl. (2017) 4:1. doi: 10.1038/s41540-017-0037-9
67. Lee C, Kim MJ, Kumar A, Lee HW, Yang Y, and Kim Y. Vascular endothelial growth factor signaling in health and disease: from molecular mechanisms to therapeutic perspectives. Signal Transduct Target Ther. (2025) 10:170. doi: 10.1038/s41392-025-02249-0
68. Sharma S, Ehrlich M, Zhang M, Blobe GC, and Henis YI. NRP1 interacts with endoglin and VEGFR2 to modulate VEGF signaling and endothelial cell sprouting. Commun Biol. (2024) 7:112. doi: 10.1038/s42003-024-05798-2
69. Elaimy AL, Amante JJ, Zhu LJ, Wang M, Walmsley CS, FitzGerald TJ, et al. The VEGF receptor neuropilin 2 promotes homologous recombination by stimulating YAP/TAZ-mediated Rad51 expression. Proc Natl Acad Sci USA. (2019) 116:14174–80. doi: 10.1073/pnas.1821194116
70. Hu C, Zhu P, Xia Y, Hui K, Wang M, and Jiang X. Role of the NRP-1-mediated VEGFR2-independent pathway on radiation sensitivity of non-small cell lung cancer cells. J Cancer Res Clin Oncol. (2018) 144:1329–37. doi: 10.1007/s00432-018-2667-8
71. Jeon HM, Shin YJ, Lee J, Chang N, Woo DH, Lee WJ, et al. The semaphorin 3A/neuropilin-1 pathway promotes clonogenic growth of glioblastoma via activation of TGF-β signaling. JCI Insight. (2023) 8:e167049. doi: 10.1172/jci.insight.167049
72. Cascone T, Xu L, Lin HY, Liu W, Tran HT, Liu Y, et al. The HGF/c-MET pathway is a driver and biomarker of VEGFR-inhibitor resistance and vascular remodeling in non–small cell lung cancer. Clin Cancer Res. (2017) 23:5489–501. doi: 10.1158/1078-0432.CCR-16-3216
73. Carbone C, Moccia T, Zhu C, Paradiso G, Budillon A, Chiao PJ, et al. Anti-VEGF treatment-resistant pancreatic cancers secrete proinflammatory factors that contribute to Malignant progression by inducing an EMT cell phenotype. Clin Cancer Res. (2011) 17:5822–32. doi: 10.1158/1078-0432.CCR-11-1185
74. Huang W, Zhang C, Cui M, Niu J, and Ding W. Inhibition of bevacizumab-induced epithelial-mesenchymal transition by BATF2 overexpression involves the suppression of Wnt/beta-catenin signaling in glioblastoma cells. Anticancer Res. (2017) 37:4285–94. doi: 10.21873/anticanres.11821
75. Piao Y, Liang J, Holmes L, Zurita A, Henry V, Heymach J, et al. Glioblastoma resistance to anti-VEGF therapy is associated with myeloid cell infiltration, stem cell accumulation, and a mesenchymal phenotype. Neuro Oncol. (2012) 14:1379–92. doi: 10.1093/neuonc/nos158
76. Prakash J and Shaked Y. The interplay between extracellular matrix remodeling and cancer therapeutics. Cancer Discov. (2024) 14:1375–88. doi: 10.1158/2159-8290.CD-24-0002
77. Coursier D and Calvo F. CAFs vs. TECs: when blood feuds fuel cancer progression, dissemination and therapeutic resistance. Cell Oncol. (2024) 47:1091–112. doi: 10.1007/s13402-024-00931-z
78. Crawford Y, Kasman I, Yu L, Zhong C, Wu X, Modrusan Z, et al. PDGF-C mediates the angiogenic and tumorigenic properties of fibroblasts associated with tumors refractory to anti-VEGF treatment. Cancer Cell. (2009) 15:21–34. doi: 10.1016/j.ccr.2008.12.004
79. Zahedipour F, Hosseini SA, Astaneh M, Kesharwani P, Jaafari MR, and Sahebkar A. Application of VEGF/VEGFR peptide vaccines in cancer: A systematic review of clinical trials. Crit Rev Oncol Hematol. (2023) 187:104032. doi: 10.1016/j.critrevonc.2023.104032
80. Zhou Y, Liu Z, Yu A, Zhao G, and Chen B. Immune checkpoint inhibitor combined with antiangiogenic agent synergistically improving the treatment efficacy for solid tumors. Immunotargets Ther. (2024) 13:813–29. doi: 10.2147/ITT.S494670
81. Wei C, Yang C, Wang S, Shi D, Zhang C, Lin X, et al. Crosstalk between cancer cells and tumor associated macrophages is required for mesenchymal circulating tumor cell-mediated colorectal cancer metastasis. Mol Cancer. (2019) 18:64. doi: 10.1186/s12943-019-0976-4
82. Jiang M, Chen J, Zhang W, Zhang R, Ye Y, Liu P, et al. Interleukin-6 trans-signaling pathway promotes immunosuppressive myeloid-derived suppressor cells via suppression of suppressor of cytokine signaling 3 in breast cancer. Front Immunol. (2017) 8:1840. doi: 10.3389/fimmu.2017.01840
83. Hamidi A, Song J, Thakur N, Itoh S, Marcusson A, Bergh A, et al. TGF-beta promotes PI3K-AKT signaling and prostate cancer cell migration through the TRAF6-mediated ubiquitylation of p85alpha. Sci Signal. (2017) 10:eaal4186. doi: 10.1126/scisignal.aal4186
84. Kim BG, Malek E, Choi SH, Ignatz-Hoover JJ, and Driscoll JJ. Novel therapies emerging in oncology to target the TGF-β pathway. J Hematol Oncol. (2021) 14:55. doi: 10.1186/s13045-021-01053-x
85. Zhou L, Liu XD, Sun M, Zhang X, German P, Bai S, et al. Targeting MET and AXL overcomes resistance to sunitinib therapy in renal cell carcinoma. Oncogene. (2016) 35:2687–97. doi: 10.1038/onc.2015.343
86. Marona P, Górka J, Kwapisz O, Jura J, Rys J, Hoffman RM, et al. Resistance to tyrosine kinase inhibitors promotes renal cancer progression through MCPIP1 tumor-suppressor downregulation and c-Met activation. Cell Death Dis. (2022) 13:814. doi: 10.1038/s41419-022-05251-4
87. Rini BI and Vogelzang NJ. Future challenges for drug development in renal cell carcinoma. J Clin Oncol. (2017) 35:577–9. doi: 10.1200/JCO.2016.71.0673
88. Desai A and Small EJ. Treatment of advanced renal cell carcinoma patients with cabozantinib, an oral multityrosine kinase inhibitor of MET, AXL and VEGF receptors. Future Oncol. (2019) 15:2337–48. doi: 10.2217/fon-2019-0021
89. Frentzas S, Simoneau E, Bridgeman VL, Vermeulen PB, Foo S, Kostaras E, et al. Vessel co-option mediates resistance to anti-angiogenic therapy in liver metastases. Nat Med. (2016) 22:1294–302. doi: 10.1038/nm.4197
90. Rahbari NN, Kedrin D, Incio J, Liu H, Ho WW, Nia HT, et al. Anti-VEGF therapy induces ECM remodeling and mechanical barriers to therapy in colorectal cancer liver metastases. Sci Transl Med. (2016) 8:360ra135. doi: 10.1126/scitranslmed.aaf5219
91. Gjerdrum C, Tiron C, Høiby T, Stefansson I, Haugen H, Sandal T, et al. Axl is an essential epithelial-to-mesenchymal transition-induced regulator of breast cancer metastasis and patient survival. Proc Natl Acad Sci USA. (2010) 107:1124–9. doi: 10.1073/pnas.0909333107
92. Xu Z, Goel HL, Burkart C, Burman L, Chong YE, Barber AG, et al. Inhibition of VEGF binding to neuropilin-2 enhances chemosensitivity and inhibits metastasis in triple-negative breast cancer. Sci Transl Med. (2023) 15:eadf1128. doi: 10.1126/scitranslmed.adf1128
93. Antony J and Huang RY. AXL-driven EMT state as a targetable conduit in cancer. Cancer Res. (2017) 77:3725–32. doi: 10.1158/0008-5472.CAN-17-0392
94. Engelsen AST, Lotsberg ML, Abou Khouzam R, Thiery JP, Lorens JB, Chouaib S, et al. Dissecting the role of AXL in cancer immune escape and resistance to immune checkpoint inhibition. Front Immunol. (2022) 13:869676. doi: 10.3389/fimmu.2022.869676
95. Zhao XS, Joshi JJ, Aird D, Karr C, Yu K, Huang CT, et al. Combined inhibition of FGFR4 and VEGFR signaling enhances efficacy in FGF19 driven hepatocellular carcinoma. Am J Cancer Res. (2022) 12:2733–43.
96. Li X, Lu W, Kharitonenkov A, and Luo Y. Targeting the FGF19-FGFR4 pathway for cholestatic, metabolic, and cancerous diseases. J Intern Med. (2024) 295:292–312. doi: 10.1111/joim.13767
97. Xie M, Lin Z, Ji X, Luo X, Zhang Z, Sun M, et al. FGF19/FGFR4-mediated elevation of ETV4 facilitates hepatocellular carcinoma metastasis by upregulating PD-L1 and CCL2. J Hepatol. (2023) 79:109–25. doi: 10.1016/j.jhep.2023.02.036
98. Akli A, Takam Kamga P, Julie C, Capron C, Costantini A, Dumenil C, et al. Role of the HGF/c-MET pathway in resistance to immune checkpoint inhibitors in advanced non-small cell lung cancer. Cancer Immunol Immunother. (2025) 74:58. doi: 10.1007/s00262-024-03882-4
99. Huang Y, Yuan J, Righi E, Kamoun WS, Ancukiewicz M, Nezivar J, et al. Vascular normalizing doses of anti-angiogenic treatment reprogram the immunosuppressive tumor microenvironment and enhance immunotherapy. Proc Natl Acad Sci USA. (2012) 109:17561–6. doi: 10.1073/pnas.1215397109
100. Kuczynski EA, Vermeulen PB, Pezzella F, Kerbel RS, and Reynolds AR. Vessel co-option in cancer. Nat Rev Clin Oncol. (2019) 16:469–93. doi: 10.1038/s41571-019-0181-9
101. Ferrara N, Hillan KJ, Gerber HP, and Novotny W. Discovery and development of bevacizumab, an anti-VEGF antibody for treating cancer. Nat Rev Drug Discov. (2004) 3:391–400. doi: 10.1038/nrd1381
102. Scott LJ. Apatinib: A review in advanced gastric cancer and other advanced cancers. Drugs. (2018) 78:747–58. doi: 10.1007/s40265-018-0903-9
103. Yakes FM, Chen J, Tan J, Yamaguchi K, Shi Y, Yu P, et al. Cabozantinib (XL184), a novel MET and VEGFR2 inhibitor, simultaneously suppresses metastasis, angiogenesis, and tumor growth. Mol Cancer Ther. (2011) 10:2298–308. doi: 10.1158/1535-7163.MCT-11-0264
104. Grüllich C. Cabozantinib: a MET, RET, and VEGFR2 tyrosine kinase inhibitor. Recent Results Cancer Res. (2014) 201:207–14. doi: 10.1007/978-3-642-54490-3_12
105. Jain RK. Normalization of tumor vasculature: an emerging concept in antiangiogenic therapy. Science. (2005) 307:58–62. doi: 10.1126/science.1104819
106. Goel S, Duda DG, Xu L, Munn LL, Boucher Y, Fukumura D, et al. Normalization of the vasculature for treatment of cancer and other diseases. Physiol Rev. (2011) 91:1071–121. doi: 10.1152/physrev.00038.2010
107. Li S, Zhang Q, and Hong Y. Tumor vessel normalization: A window to enhancing cancer immunotherapy. Technol Cancer Res Treat. (2020) 19:1533033820980116. doi: 10.1177/1533033820980116
108. Feng Q, Zhang C, Lum D, Druso JE, Blank B, Wilson KF, et al. A class of extracellular vesicles from breast cancer cells activates VEGF receptors and polarizes macrophages. Nat Commun. (2017) 8:14450. doi: 10.1038/ncomms14450
109. Socinski MA, Nishio M, Jotte RM, Cappuzzo F, Orlandi F, Stroyakovskiy D, et al. IMpower150 final overall survival analyses for atezolizumab plus bevacizumab and chemotherapy in first-Line metastatic nonsquamous NSCLC. J Thorac Oncol. (2021) 16:1909–24. doi: 10.1016/j.jtho.2021.07.009
110. Rini BI, Plimack ER, Stus V, Gafanov R, Waddell T, Nosov D, et al. Pembrolizumab plus axitinib versus sunitinib for advanced clear cell renal cell carcinoma: 5-year survival and biomarker analyses of the phase 3 KEYNOTE-426 trial. Nat Med. (2025) 31(10):3475–84. doi: 10.1038/s41591-025-03867-5. Online ahead of print.
111. Choueiri TK, Penkov K, Uemura H, Campbell MT, Pal S, Kollmannsberger C, et al. Avelumab + axitinib versus sunitinib as first-line treatment for patients with advanced renal cell carcinoma: final analysis of the phase III JAVELIN Renal 101 trial. Ann Oncol. (2025) 36:387–92. doi: 10.1016/j.annonc.2024.12.008
112. Motzer RJ, Porta C, Eto M, Hutson TE, Rha SY, Merchan JR, et al. Biomarker analyses from the phase III randomized CLEAR trial: lenvatinib plus pembrolizumab versus sunitinib in advanced renal cell carcinoma. Ann Oncol. (2025) 36:375–86. doi: 10.1016/j.annonc.2024.12.003
113. Motzer RJ, Escudier B, Burotto M, Powles T, Apolo AB, Bourlon MT, et al. Final analysis of nivolumab plus cabozantinib for advanced renal cell carcinoma from the randomized phase III CheckMate 9ER trial. Ann Oncol. (2025) 37(1):33–43. doi: 10.1016/j.annonc.2025.09.006. Online ahead of print.
114. Tirosh I and Suva ML. Cancer cell states: Lessons from ten years of single-cell RNA-sequencing of human tumors. Cancer Cell. (2024) 42:1497–506. doi: 10.1016/j.ccell.2024.08.005
115. Cilento MA, Sweeney CJ, and Butler LM. Spatial transcriptomics in cancer research and potential clinical impact: a narrative review. J Cancer Res Clin Oncol. (2024) 150:296. doi: 10.1007/s00432-024-05816-0
116. Chow SC. Adaptive clinical trial design. Annu Rev Med. (2014) 65:405–15. doi: 10.1146/annurev-med-092012-112310
Keywords: anti-VEGF therapy, direct signal remodeling, paradox of aggression, therapeutic resistance, vascular endothelial growth factor
Citation: Wei P, Li J, Cheng X, Lu Y, Xia Q and Wang Z (2026) “Brake” and “accelerator”: revisiting tumor cell direct responses and the paradox of aggression in anti-VEGF therapy. Front. Immunol. 16:1718894. doi: 10.3389/fimmu.2025.1718894
Received: 04 October 2025; Accepted: 29 December 2025; Revised: 27 December 2025;
Published: 16 January 2026.
Edited by:
Chunsik Lee, GEMCRO, Republic of KoreaCopyright © 2026 Wei, Li, Cheng, Lu, Xia and Wang. 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.
*Correspondence: Zhiyong Wang, d2FuZ3poaXlvbmdAem11emguZWR1LmNu