Targeting angiogenesis: Lessons from 25 years of normalizing tumor vasculature.

INTRODUCTION
Malignant cells require blood vessels to grow and spread to distal organs. They recruit new blood-vessels by multiple mechanisms. However, the most investigated mechanism is known as angiogenesis. Derived from the Greek angeîon (vessel) and génesis (creation), angiogenesis refers to the sprouting of new vessels from pre-existing ones. While in normal physiology, this underpins developmental growth, wound healing, and the menstrual cycle, in cancer this process fuels tumor growth and progression1,2.
The observation that new blood vessels form around tumors dates back to1800s when Karl Thiersch, a German surgeon, showed that the tumor vessels in carcinomas originate from the host vessels. Subsequently, the German pathologist Rudolf Virchow showed their importance in tumor growth. In 1939, by growing Brown-Pearce carcinoma in the glass-windows implanted in rabbit ears, Ide et al observed that tumor fragments are released into the circulation via tumor blood vessels3. In 1940s, in a series of elegant studies using the glass windows implanted in the dorsal skin of mice, Glenn Algire and colleagues provided direct comparison of angiogenesis between wound healing and tumor growth4,5. In 1968, Greenblatt and Shubik first isolated a secretable diffusible ‘angiogenic’ factor from tumors grown in the hamster cheek pouch6. These foundational observations culminated in Judah Folkman’s landmark 1971 hypothesis, that tumor growth and metastasis are angiogenesis-dependent and that blocking this neovascularization could be an effective anticancer strategy1,7.
Uncontrolled tumor cell proliferation leads to an increase in metabolic demands that outpace the rate at which local oxygen is supplied by existing vasculature, thereby creating regions of hypoxia. This resulting hypoxia, acidosis and oncogenic signaling converge to trigger an ‘angiogenic switch’, which is characterized by upregulating pro-angiogenic factors such as vascular endothelial growth factor (VEGF), fibroblast growth factors (FGFs), angiopoietins and platelet-derived growth factor (PDGF) derived from both cancer and stromal cells, collectively shift the microenvironment toward active vessel genesis and remodeling2,8 (see Box 1 for historical milestones). In the early era of anti-angiogenic therapy, the prevailing paradigm was clear: “starve the tumor by cutting off its blood supply” – based on the seminal hypothesis put forward by the late Dr. Judah Folkman7. This strategy spurred the development of various antiangiogenic agents targeting pro-angiogenic molecules (e.g., VEGF) as well as vascular-disruptive agents (e.g., combretastatin)9–11. Yet preclinical and clinical studies revealed a paradox: excessively pruning tumor vessels often worsened hypoxia, compromised drug delivery, and even promoted metastasis. Yet, when combined with chemotherapy, led to improved survival in some cases leading to a conundrum – how can drugs accumulate in tumors if the vessels are destroyed? In 2001, one of us (Jain) resolved this paradox, by proposing a new and seemingly counterintuitive paradigm: vascular normalization - repairing tumor vessels and improving their function using agents that were originally developed to block their formation12. The resulting vasculature will lead to improved perfusion and drug delivery, lower hypoxia, enhanced immune cell infiltration and function, and thereby sensitize tumors to chemotherapy, radiotherapy and immunotherapy12–17. Although highly controversial in the beginning, with critics arguing that blocking angiogenesis could not simultaneously improve vascular function, this concept was validated in a number of preclinical and clinical studies14,15,18–21. The evolution of angiogenesis research has progressed through several conceptual eras: from the discovery of tumor angiogenesis dependency to the identification of VEGF and other anti-angiogenesis therapy targets, to vessel normalization paradigms and more recently, to multi-omics dissection of endothelial states22–24.
In this review, we map the remarkable journey from foundational insights into tumor angiogenesis and other mechanisms of vessel formation to the current frontier of vessel normalization and its therapeutic implications. We will also revisit the cellular and molecular underpinnings of tumor vascularization15,17,22; explore the structural, functional, physical, and metabolic abnormalities of tumor vessels and how they sculpt the microenvironment17,25,26; examine how the tumor vasculature influences inflammation, immunosuppression, metastasis, metabolism, invasion and therapeutic resistance and influenced by neural, microbial, hormonal, and chronological factors; and discuss how emerging strategies aim to reprogram the tumor vasculature to enhance precision oncology, especially immunotherapy. While previous reviews from our group and others have addressed aspects of tumor angiogenesis and vascular normalization2,15,22, the current review provides a detailed look into changes and future directions in tumor angiogenesis and vascular normalization by integrating spatial and single cell omics findings, newly identified neural, microbiome, and other host-related regulators of tumor vasculature, the emerging tumor metabolism-angiogenesis-immune axis27 and updated clinical evidence.

FOUNDATIONS AND MODES OF TUMOR VASCULARIZATION

Normal vs. Tumor Angiogenesis: The Angiogenic Switch
Angiogenesis is a highly regulated physiological process essential for embryonic development, wound healing, and reproduction. In healthy tissues, it involves the orchestrated balance of pro- and anti-angiogenic signals that give rise to mature, hierarchically organized vascular networks with intact basement membranes, appropriate pericyte coverage, and stable junctions2,28. This architecture is maintained through endothelial cell (EC) specialization along the tip-stalk-phalanx axis. Tip ECs sense VEGF gradients, express high VEGFR2 and DLL4, and extend filopodia to guide sprouting8,29. DLL4 presented by tip cells activates Notch signaling in adjacent stalk cells, suppressing tip fate and promoting stalk-cell proliferation and sprout elongation. Phalanx ECs, in contrast, form the quiescent, barrier-stabilizing monolayer that maintains vascular homeostasis30,31. Tip, stalk, and phalanx ECs represent transient and interchangeable endothelial phenotypes (or differentiation states) rather than distinct and fixed EC subsets32. In physiological contexts, angiogenesis resolves once the tissue’s oxygen and nutrient needs are met.
By contrast, tumor angiogenesis is aberrant, spatially disorganized, and functionally inefficient. Driven by hypoxia, nutrient deprivation, oncogenic signaling, and inflammation, cancer cells sustain chronically elevated VEGF and other pro-angiogenic factors including Galectins, which are recognized as important regulators that stimulate endothelial cell migration, tube formation, immune evasion, and resistance to therapy33. This condition overwhelms endogenous anti-angiogenic cues and disrupts the molecular balance that otherwise maintains vascular quiescence1,7,34. The resulting “angiogenic switch” transforms an initially avascular or dormant lesion into a tissue compartment that is highly vascularized, and paradoxically still poorly perfused. Tumor vessels exhibit disordered tip–stalk–phalanx patterning, excessive and misdirected sprouting, basement membrane fragmentation, abnormal pericyte coverage, and increased leakiness34,35. Recent studies show that tumor endothelial cells reactivate embryonic gene programs including Fbn2, Emilin2, Lox and Serpine1 that are absent in healthy adult vessels and that these tumor EC (TEC)-specific genes promote pathological angiogenesis while creating an immunosuppressive vascular environment36. In addition, TECs acquire distinct immunomodulatory phenotypes, characterized by downregulation of antigen-presentation machinery, reduced leukocyte-adhesion molecule expression, and induction of T-cell dysfunction, collectively generating an endothelial barrier that restricts immune infiltration and blunts anti-tumor immunity37. Moreover, long non-coding and circular RNAs have been shown to influence tumor angiogenesis by modulating endothelial cell differentiation, affecting sprouting and migration, and shaping communication between endothelial cells and the tumor microenvironment through their interactions with VEGF, Notch, and hypoxia related signaling pathways38,39. Collectively, these abnormalities produce a vascular network that is spatially chaotic, inefficient at oxygen delivery, and permissive to tumor invasion and metastasis.

Abnormal Structure and Function of Tumor Vessels
Unlike the well-organized, hierarchically branched vasculature in healthy tissues, tumor vessels are irregular in shape, lacking uniform diameter or directionality, which makes them structurally abnormal and functional compromised. They frequently exhibit blind ends, arterio-venous shunts, and abrupt changes in caliber13,40. These structural aberrations are the direct result of imbalanced angiogenic signaling, most notably excessive VEGF-A, which increases endothelial permeability and suppresses the expression of junctional proteins such as claudins, occludins, and VE-cadherin13,41. Functionally, this leads to high vascular permeability, sluggish and heterogeneous blood flow, and plasma leakage into the interstitium. Vessel leakiness contributes to the uniformly elevated interstitial fluid pressure (IFP) within the tumor that reaches the levels of microvascular pressure and thus, eliminates pressure gradients and convective transport of drugs across the vessel walls. At the tumor margin, IFP drops steeply to normal values, creating a steep pressure gradient that drives convective transport of drugs, growth factors and cells from the tumor into the surrounding tissue42,43. Vessel leakiness and the chaotic structure of the tumor vasculature can drastically reduce tumor blood flow, i.e., perfusion, exacerbating hypoxia and acidosis, which are the two key drivers of tumor aggressiveness and immune evasion44. At the network level, tumor angiogenesis generates a chaotic vascular bed lacking the arteriolar-capillary-venular hierarchy of normal tissues, characterized by irregular branching, excessive loops, and non-uniform vessel spacing45.
Elevated mechanical forces within tumors, such as solid stress generated by the uncontrolled proliferation of cancer within the confined space of the host tissue but also forces developed within structural components of the tumor, such as cancer and stromal cells and ECM components, physically deform and even collapse intratumoral blood vessels, contributing further to hypo-perfusion and hypoxia42,46–48. ECM stiffening, driven by collagen crosslinking (e.g., via LOX) and fibroblast activation49,50, adds further resistance to vessels becoming normalized. These biomechanical barriers reinforce a vicious cycle: hypoxia and acidosis triggers VEGF51,52, VEGF induces leaky vessels, and leakiness plus mechanical stress elevation further fuels hypoxia and acidosis. Moreover, regional heterogeneity in vessel structure and function arises directly from mechanical compression of vessels in the tumor interior, while peripheral vessels remain less affected and better perfused 42. This spatial heterogeneity affects not only oxygenation and pH but also immune microenvironment, setting the stage for the tumor’s “immune-excluded” phenotype53. Beyond these biomechanical abnormalities, TEC (and cancer cells) undergo metabolic reprogramming characterized by heightened glycolytic flux, altered fatty acid oxidation, and impaired nitric oxide signaling27,30,32,54–58. These changes weaken junctional integrity, blunt responses to pro-stabilizing cues, and further reduce perfusion efficiency59.

Classical Sprouting Angiogenesis
Sprouting angiogenesis remains the most thoroughly studied mechanism of neovascularization in normal and pathological conditions. Initiated primarily by VEGF-A/VEGFR2 signaling, the process begins with EC activation and basement membrane degradation. Tip cells respond to VEGF gradients by extending filopodia and leading sprouts, while trailing stalk cells proliferate and form the lumen of new vessel. Notch-DLL4 signaling ensures proper tip-stalk patterning60. In tumors, excess VEGF disrupts this feedback, causing excessive tip cell formation, irregular sprouts, and immature pericyte-deficient vessels with poor connectivity and perfusion61. Brain tumors, such as glioblastoma, exhibit unique sprouting behavior with enriched expression of neural guidance cues and developmental programs36,62. This tip-cell specificity prompted the clinical development of targeted therapies aimed at restricting pathological angiogenesis without impairing physiological angiogenesis, including anti-DLL4 antibodies and DLL4/VEGF bispecific agents, though clinical translation has been challenging owing to cardiovascular toxicities63. More recently, in vitro angiogenesis on a chip model combined with single cell RNA sequencing has revealed previously unrecognized endothelial subpopulations within sprouts, including stalk-like cells with high autophagy activity and distinct transcriptional programs that are linked to tip cell competition and branch formation64. Emerging work in mechanobiology further shows that matrix stiffness, cell generated forces, and YAP and TAZ dependent signaling modulate Notch activity and tip selection, adding a physical layer of control to the classical VEGF and Notch driven patterning of sprouting angiogenesis65,66. In parallel, genome wide nascent RNA profiling of sprouting endothelial cells has expanded the catalogue of tip cell markers and identified new transcriptional modules (such as RND3) that may provide selective targets to restrain pathological sprouting in tumors67.

Alternative Modes of Tumor Neovascularization
Tumors exhibit remarkable plasticity in how they ensure vascular support, often adopting non-sprouting mechanisms when sprouting angiogenesis is blocked by therapy or limited by microenvironmental constraints2,68 (Figure 1).

Intussusceptive Angiogenesis:
In this process, existing blood vessels split to form new ones without endothelial cell sprouting69. In this mechanism, opposing endothelial cells protrude into the lumen of the blood vessel to form a transcapillary pillar, followed by perforation of this pillar and fibroblast and pericytes subsequently invade. This mechanism has been documented in several solid tumors, including metastatic melanoma, glioma, and colorectal tumors, where it supports rapid adaptation of the vascular tree and is often less dependent on classical VEGF-driven sprouting70–72. However, most evidence derives from animal models, and its quantitative contribution relative to sprouting angiogenesis in human tumors, and whether it is truly VEGF-independent or simply less sensitive to VEGF inhibition, remains unclear. More recent work in human metastatic melanoma and experimental tumor models has underscored intussusceptive angiogenesis as a prominent, therapy-resistant mode of vascular remodeling, with abundant intraluminal pillars persisting despite VEGF pathway inhibition73. Building on this, microvessel-on-a-chip systems have recently demonstrated that endothelial delamination can initiate the formation of transluminal pillars, revealing a key biomechanical trigger of intussusceptive angiogenesis and offering new opportunities to therapeutically disrupt this VEGF-independent mode of vascular expansion74.

Vessel Co-option:
Vessel co-option enables tumors to hijack pre-existing blood vessels without inducing angiogenesis and is particularly common in tumors located in the brain, liver, and lung—both in primary and metastatic settings75. Because tumor cells in vessel co-option align themselves along native vasculature, they often exhibit resistance to VEGF-targeted therapies76,77. In colorectal cancer liver metastases, this process is driven by tumor cell motility involving actin cytoskeletal remodeling through the actin-related protein (Arp2/3) complex78. In lung metastases, angiopoietins promote vessel co-option by disrupting endothelial–pericyte interactions, thus increasing vascular accessibility to tumor cells79,80. In glioma, we showed that Olig2–Wnt7 signaling mediates vessel co-option in astrocyte-like GBM tumors77,81. Across tumor types, vessel co-option is driven by convergent programs that enhance tumor-cell motility, weaken endothelial–pericyte interactions, and enable tumor cells to migrate along native vascular basement membranes75,80. These shared biomechanical and adhesive cues explain how co-option can arise independently of angiogenic signals.

Vasculogenic mimicry:
Although controversial, vasculogenic mimicry (VM) is a striking phenomenon in which aggressive tumor cells form matrix-rich, channel-like structures capable of conducting blood independently of endothelial cells82. Concerns over its validity stem from the two main approaches used to identify VM: in vivo reliance on PAS staining combined with CD31 exclusion without tumor-specific marker co-staining to confirm that tumor cells are actually lining the channels, and in vitro use of Matrigel tube-formation assays as proxies for functional blood vessel formation82. These channels, first characterized in melanoma, have since been observed in glioblastoma and triple-negative breast cancer83–85. VM is particularly prominent in hypoxic tumor regions, where oxygen gradients select for stem-like, plastic tumor cell states capable of adopting endothelial-like transcriptional programs. These VM-competent cells upregulate VE-cadherin to establish junction-like contacts and organize into patterned networks resembling vascular channels. Key regulators, including EphA2, Nodal, and HIF-1α, promote cancer stemness, metabolic adaptation, and autophagy, enabling tumor cells to maintain perfusion and invasive growth even under anti-angiogenic therapy86–89.

Tumor-to-Endothelial Transdifferentiation:
Distinct from vasculogenic mimicry, some cancer stem-like cells undergo tumor-to-endothelial transdifferentiation, acquiring bona fide endothelial phenotypes and integrating directly into the tumor vasculature. This process was first documented in glioblastoma90,91 and subsequently in breast cancer92. It involves transcriptional reprogramming driven by EMT-related pathways and endothelial lineage factors such as ETV293–95. The resulting tumor-derived endothelial cells express classical markers including CD31 and CD34, contribute structurally to tumor vessels, and frequently exhibit therapeutic resistance. By generating endothelial-like cells from within the tumor, this mechanism may also facilitate immune evasion and support vascular maintenance under conditions in which angiogenesis is pharmacologically blocked96. However, this concept remains controversial; while in vivo lineage tracing by Cheng et al. demonstrated that glioblastoma stem cells predominantly differentiate into pericytes rather than endothelial cells97, subsequent scRNA-seq profiling of human glioblastoma has failed to provide robust transcriptomic evidence of a shared lineage.

Vasculogenesis from Bone Marrow–Derived Progenitors:
De novo vasculogenesis involves the mobilization of bone marrow–derived endothelial progenitor cells (EPCs), originally defined by expression of CD34, CD133, VEGFR-2, CD31, and VE-cadherin, which were proposed to home to tumors and incorporate into nascent vessels. Early studies suggested that EPCs structurally contribute to the tumor endothelium and support angiogenesis through secretion of proangiogenic mediators98. EPC recruitment is driven by tumor-derived VEGF and guided by SDF-1/CXCR4 gradients, enabling their trafficking to hypoxic tumor regions99. However, while initial reports in glioblastoma and hepatocellular carcinoma supported direct EPC incorporation into tumor vessels, subsequent genetic lineage-tracing studies challenged this view, showing minimal contribution of classical CD34+CD133+ EPCs to the endothelial lining in most tumors98,100–103. In fact, the majority of bone marrow-derived cells recruited to tumors differentiate into macrophages, myeloid-derived suppressor cells, and other non-endothelial lineages that promote angiogenesis indirectly102,104, and any rare incorporation of progenitors would be substantially diluted by conventional endothelial cells expanding through sprouting angiogenesis101. These findings shifted the field toward a more nuanced understanding in which EPCs act predominantly through paracrine, stromal, and immunomodulatory functions rather than by forming endothelial tubes. Recent work further expands the concept of bone marrow-derived vascular progenitors. Studies indicate that myeloid-lymphatic progenitors can support tumor lymphangiogenesis, although this claim requires caution as the primary evidence relies on LYVE-1 as a lymphatic marker without co-validation using established markers such as Prox1, VEGFR3, and podoplanin combined with CD31105. More broadly, a wider repertoire of bone-marrow-derived vascular and endovascular progenitors, including myeloid-biased vascular progenitor cells, influence tumor vascular remodeling and metastatic progression largely through cytokine secretion, ECM remodeling, and modulation of immune-vascular interactions106,107. Together, these insights highlight a shift from the classical EPC-centric paradigm toward a more diverse and context-dependent landscape of bone marrow–derived vascular support cells.
Across these diverse mechanisms, sprouting angiogenesis and vessel co-option stand out as the two dominant types of tumor vascularization, though their relative contribution varies based on tumor type, anatomical location and disease stage. The remaining mechanism, intussesception, vasculogenic mimicry, endothelial transdifferentiation and BMDC-derived vasculogensis, appear to make more limited, context-dependent contributions. A critical gap in this field is that most quantitative data on vascularization mechanisms are based on animal models and the extent to which they play a role in human tumors, or dynamically shift during disease progression and in response to therapy, remains incompletely understood. Tumors can therefore engage multiple vascularization programs simultaneously or sequentially, each with distinct biological features and therapy resistance consequences (summarized in Table 1).

Maturation, Stabilization, and Loss of Quiescence
In physiological angiogenesis, newly formed vessels are stabilized through endothelial–pericyte interactions, deposition of extracellular matrix (ECM), and suppression of pro-angiogenic signals. This maturation involves PDGF-B-mediated pericyte recruitment, Ang1–Tie2 signaling to promote vessel quiescence, and TGF-β to enhance vessel wall integrity28,41,61. In tumors, these stabilization signals are often disrupted. PDGF-B expression may be low, leading to incomplete pericyte coverage. Alternatively, Ang2 levels dominate over Ang1, which destabilizes vessels by blocking Tie2 activation, particularly under hypoxic conditions22,41. The resulting immature vasculature exhibits pericyte dropout, reduced basal lamina deposition, and enhanced endothelial plasticity.
Loss of endothelial cell quiescence is another hallmark of tumor angiogenesis. In normal vasculature, endothelial cells maintain a non-proliferative, barrier-forming phenotype which is regulated by metabolites which are produced after activation of forkhead box O (FOXO) transcription factor108–110. Recent studies have revealed that FOXO1 plays a dual role in endothelial biology: it not only maintains quiescence but also acts as a pioneer transcription factor that directly binds to and stimulates the expression of tip cell-enriched genes (such as ESM1 and ANGPT2) in response to VEGF, thereby promoting vessel sprouting and tip cell specification111. This dynamic regulation involves FOXO1 shuttling between the cytoplasm and nucleus, where nuclear FOXO1 is essential for defining tip cell identity, coordinating sprouting angiogenesis, and exerting mutual inhibition with DLL4-Notch signaling, such that FOXO1 knockdown shifts endothelial cells toward a stalk cell phenotype and underscores its complex, context-dependent role in vascular patterning.
Recent work also suggests that SMCs (smooth muscle cells) surrounding arterioles and larger vessels undergo phenotypic modulation in tumors. Tumor-derived factors such as TGF-β and PDGF can induce SMC dedifferentiation, impairing their contractility and thereby altering blood flow regulation22,112,113. Pericytes, too, exhibit remarkable plasticity in the tumor microenvironment114. They can undergo phenotype switching from a contractile to a synthetic state, a process regulated by signaling pathways converging on Rho kinase (ROCK) activity115. RGS5, a regulator of G protein signaling, is upregulated in tumor-associated pericytes and promotes a synthetic, proliferative phenotype, whereas its downregulation drives expression of contractile markers, supports vessel normalization, and can be exploited therapeutically to enhance vascular normalization and sensitize tumors to immunotherapy116,117. More recently, activation of STING (stimulator of interferon genes) in pericytes has emerged as a complementary innate immune mechanism that reinforces pericyte stabilization and promotes vascular normalization118,119.
Finally, perivascular niche remodeling affects vessel stability. Tumor-associated macrophages (TAMs), cancer-associated fibroblasts (CAFs), and myeloid cells infiltrate the perivascular space and secrete MMPs, cytokines, and angiogenic mediators that further destabilize the vasculature40,120–123. The perivascular niche is now recognized as a critical signaling hub that orchestrates tumor growth, immune evasion, and metastatic dissemination. It harbors a complex network of cellular interactions that regulate vascular integrity, stemness, and therapy resistance and an active regulator of cancer evolution, with dynamic crosstalk between endothelial cells, pericytes, immune cells, and stromal components shaping the tumor microenvironment124.

STRUCTURAL AND FUNCTIONAL HETEROGENEITY OF TUMOR VASCULATURE

Spatial and Regional Heterogeneity of Vasculature Within Tumors
Tumor vasculature shows pronounced spatial heterogeneity, driven by and contributing to local gradients in oxygenation, mechanical stress, stromal composition, and metabolic demand14. Peripheral tumor regions often contain relatively stabilized, perfused vessels with partial pericyte coverage, whereas central zones exhibit tortuous, leaky, and intermittently perfused vessels or entirely avascular necrotic areas marked by severe hypoxia and acidosis13,16,112. These regional differences in maturation, permeability, and flow lead to uneven drug distribution, immune exclusion, and microenvironmental niches that promote tumor evolution and confer treatment resistance (see Box 2 for experimental platforms enabling analysis of vascular heterogeneity). Recent studies have further resolved whole-tumor maps of oxygenation and hemodynamics, revealing complex, patient-specific patterns where hypoxic cores can still contain focal islands of vascularization and oxygenation intermingled with severely hypoxic zones125–127.
High-resolution single-cell transcriptomic studies have shown that endothelial cells exist in multiple transcriptionally and functionally distinct states rather than a single angiogenic phenotype128–130. Classical sprouting angiogenesis generates VEGFR2+CXCR4+DLL4+ tip cells and proliferative Notch1/JAG1-driven stalk cells, while quiescent phalanx-like cells express junctional markers such as Ve-cadherin (CDH5) and claudin-5 (CLDN5) to maintain barrier integrity131–133. Single-cell and bulk studies describe additional endothelial states, including PLVAP+ cells associated with increased permeability and metabolic activity, inflammatory ECs with upregulated adhesion molecules and cytokine signaling, ECs with senescence-associated gene signatures, and high endothelial venule (HEV)-like phenotypes enriched for trafficking and addressin molecules130,134,135. These endothelial states are associated with gradients in angiogenic and inflammatory cues, such as VEGF and cytokines, and with regional differences in perfusion and oxygenation136. Spatial multi-omics further shows that specific endothelial states co-localize with distinct immune and fibroblast compartments, indicating that vascular heterogeneity is embedded within broader tumor, stromal, and immune architecture137–139. Specialized endothelial cell subsets such as high endothelial venule (HEV)-like cells organize immune niches to diverse phenotypes defined by spatial transcriptomics that can correlate with immune infiltration and therapeutic response140–142. These HEV-like cells frequently associate with tertiary lymphoid structures (TLS), ectopic lymphoid aggregates first linked to favorable prognosis in lung cancer143 and subsequently shown to predict immunotherapy response across melanoma, sarcoma, and renal cell carcinoma144–146.

Tissue Context and Lineage-Dependent Vascular Programs
Tumor vasculature is strongly influenced by the anatomical site in which cancer cells grow and by their lineage. Each organ provides a distinct vascular architecture, blood flow pattern, stromal composition, and immune landscape, which together determine how tumors acquire or appropriate vessels (Figure 2)23,147. As a result, primary tumors and metastases deploy different combinations of angiogenesis, vessel co-option, and vasculogenic mimicry. Lesions in the brain, liver, and lung often use hybrid programs that combine sprouting with co-option of pre-existing vessels, whereas tumors in hypovascular tissues such as pancreas or bone frequently face stiff, fibrotic microenvironments that restrict angiogenesis and promote desmoplasia148–150.
Spatial-omics datasets summarized in Table 2 highlight the diversity of endothelial states across tissues. Brain tumors contain arterial, arteriolar, capillary, venular, and large-vein endothelial subsets, along with distinct angiogenic capillaries marked by PLVAP, ESM1, and ANGPT2 that align with hypoxia-driven and fetal-like transcriptional programs23. Liver and lung metastases show specialized sinusoidal-like or breach-like endothelial modules enriched for extracellular matrix remodeling, scavenger receptors, or antigen-presenting functions151,152. Breast cancers contain venular, arterial, stalk-like, and tip-like endothelial subtypes that form spatial units with perivascular fibroblasts, smooth muscle cells, and immune infiltrates153–155. Pancreatic cancers display angiogenic endothelial cells at invasive fronts, perivascular niches populated by myofibroblast-like cancer-associated fibroblasts, and chemokine networks such as CXCR4 and CXCL12 that organize stromal and immune positioning149,156. These patterns reveal that endothelial phenotypes are tightly coordinated with organ-specific stromal and immune ecosystems23,147. Whether these endothelial specializations are shaped primarily by the tissue of metastasis or by the cancer cell lineage of origin, and whether vascular normalization strategies effective in primary tumors extend equivalently to vascular niches in all metastatic sites, remain important open questions in the field.
Lineage-specific programs further shape vascular behavior within a given tissue. Highly angiogenic cancers such as melanoma, glioblastoma, and clear-cell renal carcinoma often rely on robust VEGF signaling, while pancreatic ductal adenocarcinoma and prostate cancer are largely hypovascular with dense fibroblast-rich stroma147,157–159. Even within the same organ, distinct tumor types induce characteristic endothelial states through lineage-specific secretion of VEGF, angiopoietins, TGFβ, chemokines, and metabolic cues. These differences influence whether tumors respond to therapy through vascular normalization, vascular pruning, or a switch toward vessel co-option. For example, anti-angiogenic therapy promotes invasive vessel co-option in glioblastoma81,160 and co-option-mediated resistance in liver metastases78, while normalization responses have been documented in glioblastoma and rectal cancer patients19,20.
Collectively, spatial-omics studies reveal that host tissue context and tumor cell lineage generate a wide landscape of endothelial and perivascular states with distinct immune interactions, stromal architecture, and therapeutic responses. This diversity underscores the need for precision vascular profiling that accounts for metastatic site, spatial organization, and cell-of-origin rather than treating tumor vasculature as a uniform target.

Functional Consequences of Vascular Abnormality on Tumor Growth, Invasion, and Metastasis
The structural abnormalities of tumor blood vessels, including tortuosity, leakiness, discontinuous endothelial lining, and inadequate pericyte coverage, create disorganized vascular network and heterogenous perfusion2. These defects generate regions of transient as well as chronic hypoxia that stabilize HIF transcription factors and induce pro-survival and pro-angiogenic programs such as VEGF, PDGF, and CXCL12161. Hypoxia also promotes epithelial to mesenchymal transition and drives extracellular matrix remodeling through proteases like MMPs, thereby enhancing tumor cell motility, invasiveness, and metastatic seeding162.
Leaky vessels and aberrant basement membranes provide permissive sites for tumor cell entry into the circulation, while shear forces and endothelial activation in downstream microvasculature facilitate arrest and extravasation. In parallel, aberrant tumor-associated lymphatic vessels, often enlarged and hyperpermeable due to VEGF-C/VEGFR3 signaling, serve as major conduits for metastatic spread, especially to regional lymph nodes163,164. Lymphatic endothelial cells also regulate dendritic-cell trafficking and local immune activation, linking vascular abnormalities to impaired antitumor immunity (see Box 2)165. More broadly, the hypoxic, acidic, and high-interstitial-pressure environment created by abnormal vasculature suppresses T cell and NK cell recruitment, promotes immunosuppressive cell infiltration, and generates an immune-excluded phenotype that limits antitumor immunity (see Section 4 for detailed discussion). Endothelial to mesenchymal transition further contributes to vascular dysfunction and fibrosis, reinforcing the mechanical stresses and stromal changes that support invasion162.
These vascular abnormalities also impair delivery of molecular, nano- and cellular medicines. Elevated interstitial fluid pressure reduces convective transport, and uneven perfusion creates regions where cytotoxic and targeted agents do not reach effective levels, enabling tumor persistence and relapse13. Moreover, the resulting hypoxia and low pH can confer resistance to multiple therapies and promote tumor progression and metastasis15. Importantly, our preclinical and clinical studies show that vascular normalization can counter these adverse effects14. Restoring vessel structure improves perfusion and drug delivery, reduces hypoxia, and diminishes hypoxia-driven epithelial to mesenchymal transition and matrix remodeling, and enhances immune cell infiltration and function, collectively leading to improved treatment outcomes in animal models as well as cancer patients14 (see Section 4). Indeed, GBM, NSCLC and TNBC patients whose tumor vessels were normalized in response to anti-angiogenic therapy survived longer than the patients whose tumor vessels did not166–168. Moreover, Strategies that combine vascular normalization with stromal decompression further enhance perfusion and limit metastatic dissemination169,170.

MOLECULAR SIGNALING AND METABOLIC CONTROL OF TUMOR ANGIOGENESIS

Pro-angiogenic growth factor signaling
The VEGF, FGF, PDGF, EGF/HER, HGF, and IGF families constitute the core pro-angiogenic receptor tyrosine kinase (RTK) systems in tumors8. Among these, the VEGF–VEGFR axis remains the dominant driver of endothelial activation in tumor angiogenesis. Originally discovered as Vascular Permeability Factor (VPF) by Harold Dvorak and colleagues in 1983171, it was cloned and named as VEGF by Napoleone Ferrara and colleagues in 19899,172. VEGFA binding to VEGFR2 initiates robust signaling cascades including PLCγ–ERK and PI3K–AKT pathways, promoting endothelial proliferation, survival, migration, and increased vascular permeability60,173,174. VEGFR2 signaling is shaped not only by neuropilin co-receptors but also by the fact that much of the VEGF in tumors is bound to the extracellular matrix and need to be released by proteases before it can activate endothelial cells37,175. Additionally, VEGFR1 acts mainly as a decoy receptor limiting VEGFR2 activation176 but also has a pro-angiogenic role through its expression in tumor-infiltrating myeloid cells supporting their recruitment and angiogenic functions, including premetastatic niche formation8,177–179. VEGF signaling is also a negative regulator of pericyte function and vessel maturation180.
The FGF family, particularly FGF2–FGFR1, continues to provide a potent parallel angiogenic stimulus181. FGFs support endothelial proliferation and sprouting and have been confirmed as important bypass pathways during resistance to VEGF blockade181. Notably, FGF signaling also regulates endothelial metabolism, which is critical for angiogenesis-related endothelial functions such as sprouting and migration60. Recent evidence demonstrates that inhibition of FGF receptors can reduce vessel density and restore tumor sensitivity to anti-VEGF therapies, highlighting its therapeutic relevance in overcoming VEGF resistance182.
PDGF signaling plays a dual role where tumor- or endothelial cell-derived PDGFB activates PDGFRβ on pericytes to promote mural cell recruitment and vessel stabilization183,184. PDGFRα/β activation in stromal or immune cells induces VEGF and other angiogenic mediators, amplifying angiogenic signals183. PDGF-BB regulates endothelial proliferation and VEGFR2 signaling to prevent vascular abnormalities caused by high VEGF levels185,186. Emerging research links PDGF/PDGFR overactivation with increased tissue fibrosis, angiogenesis, and drug resistance, underscoring its crucial contribution to tumor progression and anti-angiogenic therapy failure187,188.
HGF–c-Met signaling remains a potent inducer of endothelial motility and is frequently upregulated in tumors refractory to anti-VEGF treatments, supporting angiogenesis through enhanced endothelial migration and invasion189. IGF1–IGF1R signaling promotes angiogenesis indirectly, largely via induction of VEGF and stabilization of HIF1α190,191. Together, these RTKs form a highly redundant and multilayered signaling network that enables tumors to sustain angiogenesis despite therapeutic inhibition of individual pathways.

Guidance, Patterning, and Vessel-Stabilization Signals
Beyond mitogenic growth factors, endothelial sprouting and vessel maturation are governed by morphogenic cues that pattern vascular architecture. The angiopoietin–Tie system is a central regulator of vascular stability. Ang1-Tie2 signaling, predominantly derived from pericytes, maintains endothelial quiescence, strengthens endothelial–pericyte contacts, and supports barrier integrity192–194. In contrast, endothelial Ang2-induced by hypoxia, inflammation, and VEGF-antagonizes Ang1, destabilizes EC–pericyte interactions, and sensitizes vessels to VEGF-driven sprouting195,196. The balance of Ang2 versus Ang1 dictates whether vessels regress, normalize, or undergo sprouting. Recent work further reveals that therapeutic reinforcement of Tie2 signaling (for example via Tie2 agonists or Src inhibition) can promote vascular normalization197 and barrier restoration in tumors, and that broader angiopoietin family members, including ANGPTLs, contribute additional layers of control over vascular stability and metabolism in the tumor microenvironment198.
Notch signaling, especially Dll4-Notch1, is the key determinant of tip–stalk patterning during angiogenesis199,200. VEGF induces Dll4 in tip cells, which laterally inhibits neighboring cells through Notch to enforce stalk identity and prevent excessive, non-productive sprouting. Jagged ligands (Jag1/2), by contrast, antagonize Dll4 in certain contexts and can promote stalk cell proliferation and vessel expansion. More recent studies indicate that the magnitude and timing of Dll4-Notch signaling, as well as regulated Dll4 degradation, fine-tune tip–stalk ratios and can shift tumor vasculature toward either hyper-sprouting, poorly perfused networks or more normalized, hierarchically organized vessels with improved perfusion and drug delivery65.
Ephrin-Eph receptor signaling also shapes vascular identity, with ephrinB2/EphB4 guiding arterial-venous specification, sprout extension, and endothelial polarity201. These patterning signals ensure orderly vascular development under physiological conditions but become dysregulated in tumors, yielding non-hierarchical, tortuous, and poorly perfused vessels202,203. Recent in vivo tumor studies show that disrupting ephrinB2 in vascular and tumor compartments can normalize vessels and limit growth, while modulating EphB4 (including via soluble EphB4) not only alters aberrant angiogenesis but can also enhance responses to immunotherapies, underscoring the Eph–ephrin axis as a dual regulator of vascular patterning and anti-tumor immunity204–206.

Transcriptional Regulators of Angiogenic Programs
Transcription factors integrate the diverse extracellular cues that impinge on ECs. HIF1α is the principal driver of hypoxia-induced angiogenesis in tumors207. Under oxygen deprivation or via VHL loss, HIF1α stabilizes and induces VEGFA, PDGFB, ANG2, CXCL12, MMPs, and metabolic enzymes that support endothelial adaptation to stress208,209. HIF1α activation occurs not only in cancer cells but also in endothelial and myeloid cells, creating a multi-compartmental pro-angiogenic program. Recent studies highlight HIF1α’s role in counteracting TGFβ1-mediated anti-angiogenic signals (e.g., TSP1 suppression) in endothelial cells and driving metabolic reprogramming toward glycolysis in tumor-associated endothelium, further amplifying hypoxia-responsive angiogenesis across compartments210. Interestingly, stabilization of HIF-2α in migratory endothelial cells (ECs) that reach hypoxic regions in response to a VEGF gradient induces the expression of VE-cadherin and soluble VEGFR1. These changes counteract the motile phenotype and promote a quiescent, phalanx endothelial state, supporting vessel formation and perfusion in anticipation of tissue re-oxygenation31.
NF-κB serves as another major integrator of cytokine, RTK, and mechanical signals. In endothelial cells, NF-κB upregulates adhesion molecules, inflammatory cytokines (e.g., IL-6, CXCL8), and pro-angiogenic mediators, contributing both to sprouting and to vascular inflammation211–213. Crosstalk between VEGFR2, Src, and NF-κB amplifies endothelial activation and vascular permeability, with VEGF-induced VEGFR2 signaling activating c-Src via the adaptor TSAd to phosphorylate VE-cadherin and disrupt endothelial junctions, while converging NF-κB activity drives adhesion molecule expression and sustains vascular inflammation213,214. Emerging evidence links endothelial NF-κB to ESM1-mediated angiogenesis via PI3K/AKT crosstalk and positions it as a key regulator of pyruvate-driven tumor proliferation and vascular remodeling in hypoxic contexts215,216. Other transcriptional programs, including MYC activation, TEAD/YAP responses to mechanical stress, and FOXO-mediated quiescence, also contribute to the dynamic phenotypic states of tumor ECs217–219. Advances reveal YAP-TEAD-MYC networks integrating mechanical and mitogenic inputs to sustain endothelial metabolic flexibility, while FOXO factors enforce vascular quiescence and oppose pathological angiogenesis in tumors220.

Adhesion and ECM-Remodeling Pathways
Endothelial migration, barrier regulation, and sprout invasion require coordinated adhesion and proteolytic programs. Integrins such as αvβ3, αvβ5, and α5β1 cooperate with VEGF and FGF signals to promote cytoskeletal remodeling, directional migration, and survival221,222. These integrins sense matrix stiffness and transmit mechanical forces that are abnormally elevated in tumors, reinforcing pro-angiogenic signaling and barrier disruption. Recent work shows that hypoxia enhances αvβ3 activity through VEGFR2–NRP1 complexes, and that tumor-derived ectosomes deliver αvβ5 to endothelial cells to stimulate tube formation223. Notably, αvβ3 can exert either pro- or anti-angiogenic effects depending on the ECM ligands present, and stiffness-induced activation of α5β1 has been implicated as a driver of tumor angiogenesis and a potential therapeutic target 224–226.
Matrix metalloproteinases (MMPs), including MMP2, MMP9, and MT1-MMP (MMP14), remodel the basement membrane, degrade interstitial matrix, and liberate sequestered VEGF and TGF-β, thereby propagating sprouting227,228. MMP9-mediated mobilization of ECM-bound VEGF is a key event in enabling the angiogenic switch in multiple models229. However, proteolysis also generates endogenous angiostatic fragments (e.g., endostatin, angiostatin), illustrating the dual pro- and anti-angiogenic outputs of ECM remodeling230. MMP2 and MMP9 can activate latent TGF-β through CD44 interactions and release additional growth factors such as bFGF, thereby sustaining angiogenic signaling231. At the same time, emerging ECM-remodeling nanoparticles that degrade HA-rich matrices improve drug penetration and reduce stiffness-induced hypoxia, underscoring the therapeutic relevance of targeting tumor ECM dynamics232,233.

Anti-angiogenic and angiostatic signals
Endothelial quiescence is maintained by a tightly regulated balance of angiostatic and angiogenic cues, many of which tumors downregulate or overwhelm to sustain pathological vessel growth (see Table 3 for an overview of pro- and anti-angiogenic factors). Thrombospondin-1 (TSP1) suppresses VEGF signaling, stabilizes p53, and engages CD36 to inhibit endothelial migration234–236. Endostatin and angiostatin, proteolytically derived ECM fragments, compete with integrins or disrupt cytoskeletal organization to block sprouting230,237. Interferons exert direct anti-proliferative effects on endothelial cells and reprogram macrophages toward anti-angiogenic phenotypes, thereby coupling vascular control to immune activation238–243. Additional angiostatic regulators include guidance molecules such as Semaphorin 3A, which can normalize vessels by restoring endothelial-pericyte alignment, and inhibition of LRG1, which suppresses aberrant TGF-β/ALK1 signaling frequently observed in pathological angiogenesis244–247. These endogenous inhibitors constitute a multi-layered “braking system” that counterbalances pro-angiogenic stimuli. Notably, therapeutic exploitation of this system, for example via TSP1-CD36 agonist mimetics or active TSP1-derived peptides, has been shown to potently suppress VEGF-driven angiogenesis in preclinical models, underscoring its relevance in anti-angiogenic therapy strategies248.

Metabolic reprogramming of endothelial and perivascular cells
Endothelial cells undergo marked metabolic rewiring in tumors that is now recognized as a fundamental determinant of angiogenesis, vessel abnormality, and therapeutic response (Figure 3). Whereas quiescent endothelial cells rely primarily on fatty acid oxidation (FAO) and oxidative phosphorylation, sprouting endothelial cells switch to high glycolytic flux driven by PFKFB3, MYC, and VEGF signaling59,249,250. This glycolytic reprogramming supplies rapid ATP and biomass for tip-cell migration and filopodia dynamics even under hypoxic conditions. Stalk cells, by contrast, depend more heavily on FAO for nucleotide synthesis and proliferation, with CPT1A supporting DNA replication and sprout elongation251. Recent studies refine this model by linking sphingosine kinase 1 (SPHK1)-S1P signaling to PFKFB3 upregulation in tumor endothelial cells, driving pathological glycolysis and vessel dysfunction, while PFKFB3 inhibition normalizes tumor vessels, reduces metastasis, and enhances chemotherapy response in preclinical models (including hepatocellular carcinoma)252. Therefore, PFKFB3 and CPT1A are promising metabolic regulators of pathological angiogenesis in preclinical models but, so far, their inhibitors have not shown clinical efficacy or gained regulatory approval.
Redox metabolism also shapes endothelial behavior: elevated NADPH consumption and ROS production contribute to junctional instability and barrier dysfunction, features characteristic of tumor vessels250,253. Lactate from hypoxic cancer cells acts as a pseudo-hormonal cue, stabilizing HIF1α in endothelial cells and stimulating angiogenesis254. Pericytes likewise undergo metabolic adaptation, relying on PDGFRβ signaling, glycolysis, and FAO to maintain contractility and vessel stabilization in nutrient-poor environments255,256. Endothelium-derived lactate has also been shown to fuel pericyte metabolism (TCA cycle, amino acid synthesis), preserving blood-brain barrier integrity and angiogenesis, with GLUT1 deficiency disrupting this axis; meanwhile, CPT1A-driven FAO in pericytes and TECs supports vessel elongation and confers immune evasion in tumors257.
Collectively, metabolic rewiring operates in parallel with growth factor and morphogen signaling to specify tip-, stalk-, and quiescent endothelial states. Because metabolic pathways are less redundant than RTK pathways, they represent increasingly attractive therapeutic targets for enhancing vascular normalization and overcoming resistance to anti-angiogenic therapies. Multiple, often overlapping mechanisms, including angiogenic redundancy, vascular mimicry, ECM-mediated stabilization, metabolic rewiring, and immune-vascular suppression, drive resistance to VEGF pathway inhibitors (reviewed in Box 3).

ENDOTHELIAL CELL INTERACTIONS WITH THE TUMOR MICROENVIRONMENT

Vascular Regulation of Immune Exclusion and Immunosuppression
Tumor blood vessels are active regulators of immune surveillance, governing both the accessibility and the tumor immune microenvironment (Figure 4). Structural abnormalities, such as poor perfusion and chaotic flow, limit immune cell trafficking into the tumor parenchyma17. In parallel, tumor endothelial cells frequently adopt an anergic, immunosuppressive phenotype characterized by reduced ICAM1 and VCAM1 expression, impaired antigen presentation, and upregulation of inhibitory or death-inducing ligands (including FasL), a program driven by VEGF, Ang2, TGF-β, and hypoxia161,258–263. Through selective FasL-mediated apoptosis of CD8+ effector T cells while sparing regulatory T cells, endothelial cells help establish an imbalance that favors immunosuppressive T cell subsets262. Endothelial-derived angiocrine mediators, including IL-6, CXCL12, and endothelin, further promote recruitment of myeloid-derived suppressor cells and skew tumor-associated macrophages toward pro-angiogenic or immunosuppressive states264,265. In contrast, the IFN-γ-inducible chemokines CXCL9, CXCL10, and CXCL11 exert dual roles as T cell-recruiting and angiostatic factors through endothelial CXCR3 signaling, creating a feed-forward loop in which T cell influx reinforces vascular normalization and further amplifies immune infiltration266,267. Endothelial-to-mesenchymal transition (EndMT) contributes additional matrix deposition and fibrosis, reinforcing physical barriers to immune infiltration268.
Despite this suppressive baseline, the tumor vasculature retains the capacity to support antitumor immunity under certain conditions. High endothelial venules (HEVs), resembling those in secondary lymphoid organs, enable efficient T- and B-cell trafficking and can nucleate tertiary lymphoid structures, correlating with heightened immunotherapy responsiveness140,141,269,270. HEV differentiation is mechanistically driven by lymphotoxin, LIGHT, and TNF signaling through LTβR and the non-canonical NF-κB pathway, and anti-angiogenic and immune-modulating therapies can engage this axis to promote HEV-like vessel and TLS formation140,270,271.
Therapeutic vascular reprogramming can reverse endothelial anergy and restore immune infiltration. Vascular normalization, achieved through calibrated VEGF inhibition or targeted modulation of the Ang–Tie2 axis, improves perfusion, decreases hypoxia, and reinstates endothelial cell expression of adhesion molecules and antigen-presenting machinery16,272. Recent work extends this paradigm by showing that metabolic reprogramming of tumor endothelial cells, specifically through modulation of serine biosynthesis and redox balance, can normalize vascular function, relieve hypoxia, enhance immune cell infiltration, and markedly improve responsiveness to immune checkpoint blockade273. These vascular and metabolic changes enhance dendritic cell maturation, CD8+ T-cell priming, and effector cell recruitment while diminishing the suppressive activity of myeloid-derived cells53,274. Interferon-mediated transcriptional reprogramming of endothelial cells further boosts immune-supportive features265.
Recent studies extend these principles to adoptive T-cell therapies: disorganized or VEGF-stimulated vasculature restricts CAR T-cell homing and persistence, whereas combining CAR T cells with anti-VEGF or Ang2 blockade markedly improves infiltration and tumor control in preclinical models263,275. Together, these insights position the tumor endothelium as a central, targetable interface that governs immune exclusion and as a critical determinant of successful immunotherapy responses.

Stromal and Biomechanical Regulation of Tumor Vasculature
Tumor vasculature is deeply embedded in a web of cellular interactions involving pericytes, smooth muscle cells (SMCs), fibroblasts/cancer associated fibroblasts (CAFs), and tumor cells all embedded in an extracellular matrix. These interactions collectively influence vessel stability, immune infiltration, and therapeutic response15. Classically, endothelial–pericyte communication is mediated via PDGF-B/PDGFRβ and Ang–Tie2 signaling pathways. In tumors, dysregulation of these axes leads to pericyte detachment, impaired vessel maturation, and enhanced endothelial proliferation276,277. Pericyte dropout not only destabilizes vessels but also exposes the endothelial layer to inflammatory and pro-angiogenic cues, amplifying vascular dysfunction. Recent study shows that deletion of soluble guanylate cyclase (sGC) in pericytes disrupts EC-pericyte interactions, impairs Notch-mediated intercellular communication, and compromises vascular stability, demonstrating the importance of pericyte intracellular signaling beyond classical growth-factor pathways278.
Simultaneously, stromal fibroblasts/CAFs in tumors have emerged as highly heterogeneous and functionally diverse. Pan-cancer single-cell and spatial transcriptomic analyses revealed multiple CAF subtypes (e.g., inflammatory CAFs [iCAFs], myofibroblastic CAFs [myCAFs], vascular-associated CAFs [vCAFs], etc.), each with distinct spatial localization and functional roles across cancer types279,280. CAFs produce extracellular matrix (ECM) components (e.g., collagen, fibronectin) and remodeling enzymes (e.g., MMPs, LOX) that stiffen the tumor tissue, These components transmit contractile forces generated by tumor cells and CAFs to compress blood vessels46. CAF-derived cytokines, growth factors (e.g., VEGF, PDGF, TGF-β), and exosomes/metabolites initiates endothelial activation and aberrant sprouting281. Additionally, tumor-EC interactions involving tumor-derived soluble factors can reprogram endothelial gene expression toward angiogenesis and immune-evasive phenotypes. Increasing tissue stiffness has been proposed as a key barrier to immunotherapy delivery (e.g., impeding T cell, CAR-T cell, or oncolytic virus infiltration), and reprogramming CAFs or targeting ECM remodeling (e.g., via MMPs or LOX inhibition) is now under active consideration to boost immunotherapy efficacy46.
Overall, biomechanical forces in the tumor microenvironment, including solid stress, and matrix stiffness, profoundly shape endothelial behavior, vascular architecture, and immune infiltration282. Excessive deposition and cross-linking of ECM proteins (e.g., collagen, fibronectin, proteoglycans) by activated stromal cells such as CAFs leads to tissue stiffening, vessel compression or collapse, impaired perfusion, and physical barriers to drug and immune-cell delivery. Besides their mechanical effect, a stiff ECM engages mechanotransduction pathways in endothelial, stromal, tumor and immune cells283. In particular, the transcriptional co-activators YAP and TAZ, central effectors of the Hippo pathway, respond to increased ECM rigidity by translocating to the nucleus and driving gene expression programs that promote proliferation, matrix remodeling, ECM production, and cytoskeletal contractility, thereby reinforcing desmoplasia in a positive feedback loop284.
Importantly, tissue-stiffness induced mechanobiological changes have significant immunological and therapeutic consequences. Recent data demonstrate that stiff ECM can suppress tumor immunogenicity: in a murine model, high matrix stiffness activated a mechanosignaling axis (ROCK–myosin IIA–F-actin) in tumor cells, leading to downregulation of innate immune sensing (via cGAS), reduced cGAMP production, diminished type I interferon signaling, poor dendritic cell activation, and impaired CD8+ T cell priming285. Pharmacologic inhibition of myosin IIA or ROCK partially restored cGAS signaling, enhanced immune activation, and improved anti-tumor responses in vivo. Similarly, stiff ECM influences immune cell behavior, for instance, it modulates phenotype and polarization of tumor-associated macrophages via mechanosensitive pathways286. Given these effects, targeting stromal–vascular interactions, whether by stabilizing pericytes and mural cells, reprogramming CAFs, or blocking mechanotransduction signaling (e.g., YAP/TAZ or ROCK–myosin IIA axes), is increasingly recognized as a promising strategy to normalize vasculature and restore perfusion. Additionally, remodeling ECM (e.g., inhibiting cross-linking enzymes) improves immune infiltration and enhances efficacy of immunotherapy or drug delivery in solid tumors46.

Neural, Microbial, and Other Host Factors in Vascular-Immune Crosstalk
In addition to immune and stromal cells, neural cues, the microbiome, sex hormones, and aging are emerging as critical modulators of tumor vasculature and its immune interface. Tumors are innervated by sympathetic, parasympathetic and sensory fibers, and neurotransmitters such as norepinephrine can directly influence angiogenesis via β-adrenergic signaling in ECs, enhancing VEGF production and vessel sprouting287–289. In prostate cancer, adrenergic nerves trigger an angiogenic switch via metabolic reprogramming of the endothelium290. Neuro-vascular-cancer interactions are plausible, considering the striking overlap of neuro- and angiotrophic factors like BDNF, VEGF, and NGF, and the close anatomical relationships of nerves and blood vessels291. Neural inputs also affect immune cell homing and function, suggesting a neurovascular-immune axis that may be exploited therapeutically292.
The gut and intratumoral microbiome contribute to systemic immune tone and local vascular phenotypes293. Microbiome-derived metabolites, including short-chain fatty acids and bile acids, can modulate EC inflammation and permeability, and influence the polarization of tumor-infiltrating myeloid cells294,295. Disruption of microbial–vascular communication can impair immune infiltration and responsiveness to checkpoint blockade296–299.
Sex and age further influence vascular structure and function. Female sex hormones such as estrogen exert generally vasoprotective, anti-inflammatory effects on normal endothelium, but in hormone-responsive tumors they can also enhance VEGF-driven angiogenesis and remodel the vascular–immune interface in a context-dependent manner300,301. Aging leads to EC senescence, increased oxidative stress, and loss of endothelial quiescence, all of which reduce perfusion and immune accessibility in older hosts302. Physical aerobic exercise is emerging as a non-pharmacological modulator of tumor vasculature, improving pericyte coverage, perfusion, and immune infiltration thereby enhancing the efficacy of chemotherapy and immunotherapy303–305.
Such multi-dimensional influences of neural, microbial, hormonal, and chronological factors underscore the deep integration of tumor vasculature with host physiology. Understanding and targeting these axes may offer powerful strategies to reprogram the tumor immune landscape, particularly in patient populations that have traditionally shown poor response to vascular and immunotherapies.

FROM ANTI-ANGIOGENESIS TO VASCULAR NORMALIZATION

The Vascular Normalization Paradigm and Anti-Angiogenic Therapies
The introduction of anti-angiogenic therapy in oncology marked a watershed moment. The foundational hypothesis proposed by Folkman in 1971 was that tumors are angiogenesis-dependent and thus “starving” them by cutting off blood supply could be exploited therapeutically. Based on this concept, agents targeting the VEGF pathway were developed and rapidly advanced to the clinic (Summarized in Table 4). The monoclonal antibody bevacizumab, which binds VEGF-A, became the first FDA-approved anti-angiogenic drug in 2004, following its demonstrated ability to improve progression-free survival (PFS) in metastatic colorectal cancer when combined with chemotherapy (median PFS increased by 4.7 months)306–308.
However, bevacizumab’s clinical success in 2004 posed an interesting paradox: How could chemotherapeutic drug be delivered to cancer cells in the bevacizumab was supposedly cutting off tumor’s blood supply? The resolution came from insights from our preclinical study301 showing that regressing tumor blood vessels begin to exhibit normal phenotype after a decrease in VEGF due to hormone-withdrawal from a hormone-dependent tumor. These observations led to the vascular normalization hypothesis that judicious anti-VEGF dosing transiently restores vessel integrity, improving perfusion, oxygenation, and drug delivery12,15. This concept transformed anti-angiogenic therapy from a vessel-destroying strategy into vessel normalizing one.
In 2004, we revealed the mechanism of vascular normalization – blocking VEGFR2 signaling with an antibody (DC101) increased pericyte coverage of tumor vessels via upregulation of Ang1 and degraded their pathologically thick basement membrane via MMP activation. The tumor pO2 increased during the window of normalization and radiation given during this window improved the outcome of radiation therapy in a GBM model in mice18. In parallel, we demonstrated that DC101 upregulated Ang2, increased pericyte coverage and pruned immature vessels in preclinical models of breast (MCaIV) and colorectal cancer (LS174T) resulting in reduced interstitial fluid pressure and increased the delivery of therapeutics21. Of interest, we confirmed these preclinical findings in patients with rectal cancer and showed that a single dose of bevacizumab decreased the interstitial fluid pressure and increased the fraction of vessels with pericyte coverage in rectal carcinoma patients19.
Collectively, these studies established that vascular normalization is achieved by pruning immature vessels, reducing vessel leakiness and increasing pericyte coverage, enhancing flow uniformity, and lowering interstitial fluid pressure (IFP), all of which facilitate delivery of oxygen, drugs, and immune cells into the tumor (Figure 5). Mechanistically, VEGF blockade reduces aberrant endothelial proliferation and downregulates Ang2, while promoting maturation pathways such as Ang1-TIE2 and PDGF-B–PDGFRβ signaling22. Normalization also resets EC gene expression programs associated with immunosuppression and metabolic dysregulation17,22,309.
We subsequently showed that vascular normalization can also reprogram the immune microenvironment of tumors from immunosuppressive to immunostimulatory by improving the delivery of CD8+ T cells and reprogramming the tumor-associated macrophages (TAMs) from pro-tumor to anti-tumor phenotype leading to improved outcome of vaccine therapy in murine models of breast cancer310. Moreover, dual blockade of VEGF and Ang2 extended the window of normalization, reprogrammed the TAMs and improved the survival beyond VEGF-blockade311,312. More recently, we demonstrated that vascular normalization can also improve the efficacy of CAR-T cells by improving the delivery and effector function of both CAR-T cells and endogenous CD8+T cells, while decreasing Tregs275. This has spurred the development of CAR-T cells that can produce single chain antibodies (sFv) against VEGF and other angiogenic molecules313.
These preclinical findings have been confirmed by other laboratories using a varieties of anti-angiogenic agents alone and in combination with radiation, chemo- and immune checkpoint blockers17,22. Recent efforts have focused on bifunctional therapeutics that target both VEGF and immune-checkpoint blockers. For example, Ivonescimab (AK112), a bispecific antibody that targets both VEGF-A and PD-1, represents a next-generation agent designed to capitalize on this convergence314,315. By coupling vascular reprogramming with immune checkpoint blockade, these drugs aim to unlock synergistic efficacy and overcome resistance to monotherapies. Ivonescimab has since become the first PD-1xVEGF bispecific antibody to receive regulatory approval in China and is now being tested in multiple global phase III trials across lung and other solid tumors, establishing this class as a clinically validated embodiment of the normalization-immunotherapy convergence316,317. These emerging antibody-based, multi-targeted, and modular platforms exemplify the next generation of vascular therapeutics designed to enhance immunomodulation and overcome resistance mechanisms (see Box 4)313.

Clinical Translation: Combination Strategies and Therapeutic Synergy
The most impactful realization of the vascular normalization concept has come through its integration into combination regimens, where anti-angiogenic agents serve as adjuvants that enhance the outcomes of radiation-, chemo- and immunotherapies. By improving vessel perfusion and reducing hypoxia, normalization facilitates drug delivery, immune cell infiltration, and reversal of immunosuppression.
The clinical benefits of this approach have been borne out in multiple pivotal trials. In non–small cell lung cancer (NSCLC), hepatocellular carcinoma (HCC), renal cell carcinoma (RCC), and endometrial cancer, combinations of anti-VEGF(R) agents with immune checkpoint inhibitors (ICIs) have demonstrated superior efficacy over monotherapies. Examples include atezolizumab plus bevacizumab (HCC), nivolumab plus cabozantinib (RCC), and pembrolizumab plus lenvatinib (endometrial cancer), among others17,318. As of 2025, seven VEGF–ICI combinations have received regulatory approval globally, and ~200 clinical trials are actively testing such strategies across a wide range of tumor types319. Subsequent systematic analyses and FDA-led meta-analyses confirm that VEGF-ICI combinations have become standard of care in multiple first-line settings (e.g., RCC, HCC), while also revealing that not all VEGF-TKI plus ICI regimens outperform chemotherapy in later-line NSCLC, underscoring the need for rational partner and dose selection as well as clinical stage i.e., neoadjuvant or adjuvant 320,321.
These successes underscore the importance of therapeutic timing and dosing. Excessive VEGF inhibition can prune too many vessels and worsen hypoxia, while low-dose or intermittent regimens may more effectively promote normalization127. Biomarkers such as pericyte coverage, circulating sVEGF1 and Ang2 levels, and dynamic imaging of perfusion can guide the normalization window and personalize therapy14. During this window, improved vessel integrity and reduced hypoxia upregulate endothelial ICAM1/VCAM1 expression, promote CXCL9/CXCL10 gradients that guide T cell extravasation, and reduce immunosuppressive adenosine and lactate accumulation, thereby shifting the tumor microenvironment toward immune permissiveness and enhancing the efficacy of co-administered checkpoint inhibitors and CAR T cells17,275,310.
Building on this success, several next-generation bifunctional agents have entered Phase II/III development. For example, BNT327, a modular fusion protein that combines a VEGF trap with a PD-L1 antagonist, has shown promise in early trials for NSCLC and gastric cancer322,323. Similarly, Ivonescimab, a bispecific antibody co-targeting PD-1 and VEGF, has demonstrated superior efficacy over anti-PD-1 or anti-VEGF monotherapy in NSCLC, with particular benefit observed in high VEGF-expressing, immune-excluded tumors324. These agents aim to overcome the limitations of sequential or additive drug combinations by offering coordinated, spatially targeted reprogramming of the tumor microenvironment.
In summary, the vascular normalization paradigm has not only resolved longstanding paradoxes in anti-angiogenic therapy but also redefined the role of tumor vasculature from a liability to an opportunity. By embracing this shift, modern oncology has entered a phase where targeting the vasculature can amplify the efficacy of both cytotoxic and immune-based therapies, heralding a new era of rational, multi-modal intervention.

CONCLUSION AND FUTURE DIRECTIONS
Over the past three decades, our understanding of tumor vasculature has undergone a profound transformation, from viewing it as a passive scaffold to recognizing it as an active and programmable regulator of tumor progression, immune surveillance, and therapeutic response. The concept of vascular normalization has been pivotal in this shift, revamping anti-angiogenic therapy not as a means to starve tumors, but as a strategy to restore perfusion, reprogram the immune landscape, and enhance treatment efficacy.
Despite these advances, significant challenges remain. Resistance to anti-angiogenic agents continues to limit durable responses, often through vascular co-option, revascularization, or activation of alternative pro-angiogenic pathways. Emerging evidence also suggests that not all patients benefit equally from normalization-based approaches, underscoring the need for biomarkers to define the “normalization window” and tailor dose and schedules accordingly.
Looking ahead, the integration of spatial transcriptomics, single-cell multi-omics, and functional imaging holds enormous promise for refining vascular targeting strategies. These technologies can reveal endothelial cell subpopulations, spatial niches of immune exclusion, and context-specific expression of actionable molecules—thereby enabling precision vascular therapy. In particular, endothelial tip-cell programs, metabolic states, and immunomodulatory signals are emerging as rich therapeutic entry points that could be modulated independently or synergistically with VEGF pathway inhibition.
Importantly, the utility of vascular normalization extends beyond oncology. Preclinical and clinical studies have demonstrated its benefit in non-malignant diseases such as macular degeneration 325, neurofibromatosis and schwannomas326–328, plaque-stabilization329, inflammatory disorders like rheumatoid arthritis330 and tuberculosis331. The vasculature’s role in cardiac remodeling and fibrosis is now a growing focus of normalization-based interventions, suggesting translational potential across diverse pathologies15.
As we stand at the intersection of vascular biology, immunotherapy, and precision medicine, the next decade will be defined by multimodal strategies by combining vascular normalization with immune-therapies (e.g., checkpoint blockers, CAR-T cells, vaccines), metabolic targeting, and stromal modulation. These integrative approaches, grounded in a deeper mechanistic understanding of endothelial plasticity and tissue context, have the potential to redefine the therapeutic landscape for cancer.