Endothelial origins, reprogramming, and plasticity

TEC heterogeneity arises primarily from the reprogramming and state transition of pre-existing endothelial cells, with a context-dependent contribution from recruited endothelial progenitor-like cells. Endothelial-to-mesenchymal transition (EndMT) represents a phenotypic transition of endothelial cells rather than a distinct cellular origin, whereas vasculogenic mimicry is generated by malignant cells and should not be classified as an endothelial lineage (Figure 1).

Figure 1.

Endothelial sources, phenotypic plasticity, stromal support, and alternative perfusion mechanisms in tumours. Note: Most tumour endothelial cells (TECs) arise through the activation and reprogramming of pre-existing tissue-resident endothelial cells in response to angiogenic, hypoxic, inflammatory, and metabolic cues. Circulating or bone-marrow-derived endothelial progenitor-like cells may provide paracrine support or occasionally contribute to the endothelial compartment, although their contribution is context-dependent. Endothelial-to-mesenchymal transition (EndMT) generates mesenchymal-like endothelial states from pre-existing endothelial cells, whereas fibroblasts and mesenchymal stromal cells provide stromal support rather than serving as the cellular origin of EndMT. Vessel co-option involves tumour cells growing along and exploiting pre-existing endothelial-lined host vessels without inducing conventional sprouting angiogenesis. It should be distinguished from vasculogenic mimicry, in which malignant cells form perfusable channels lacking a bona fide endothelial lining. Together, these processes contribute to vascular heterogeneity, increased permeability, and tumour progression.

Figure 1. Long description

The diagram is organized into three main horizontal sections.

Top Section: Endothelial sources.

* On the left, Pre-existing resident E C s are shown as a row of red cells releasing V E G F, F G F, Angiopoietins, and Inflammatory cytokines. A red dashed arrow labeled Activation and reprogramming points toward a central vessel.

* On the right, Circulating or B M-derived endothelial progenitor-like cells are shown near a bone icon. A green dashed arrow labeled Context-dependent contribution points toward the same central vessel.

* The center shows Tumor endothelial cells T E C s lining tumor vasculature, branching into a cluster of tumor cells.

Middle Section: Three vertical panels.

* Left panel: Endothelial plasticity and stromal support. It shows Pre-existing E C s undergoing End M T to become a Mesenchymal-like endothelial state. Below, Fibroblasts and mesenchymal stromal cells provide Stromal support, indicated by a pink arrow pointing to the tumor microenvironment.

* Center panel: Tumor microenvironment. A dense cluster of tumor cells, vessels, and stromal cells.

* Right panel: Alternative perfusion mechanisms. The top sub-panel shows Vessel co-option, where tumor cells grow along pre-existing host vessels. The bottom sub-panel shows Vasculogenic mimicry, featuring malignant-cell-lined channels without a bona fide endothelial lining.

Bottom Section: Progression flow.

* A box lists Hypoxia, V E G F, F G F, P D G F and inflammatory cues, which promote the next stage.

* A black downward arrow leads to Angiogenesis and increased permeability, illustrated by a vessel with gaps and leaking molecules.

* A final black downward arrow leads to Tumor progression, showing a vessel with tumor cells invading the surrounding tissue.

The most widely supported mechanism involves activation and reprogramming of pre-existing endothelial cells within local vasculature. Persistent exposure to tumour-derived signals – including VEGF, fibroblast growth factor (FGF), angiopoietins, inflammatory cytokines, and extracellular matrix cues – drives resident endothelial cells to acquire tumour-associated phenotypes characterized by enhanced proliferation, impaired junctional integrity, and altered angiocrine signalling (Refs Reference Rohlenova, Goveia, García-Caballero, Subramanian, Kalucka, Treps, Falkenberg, de Rooij, Zheng, Lin, Sokol, Teuwen, Geldhof, Taverna, Pircher, Conradi, Khan, Stegen, Panovska, De Smet, Staal, Mclaughlin, Vinckier, Van Bergen, Ectors, De Haes, Wang, Bolund, Schoonjans, Karakach, Yang, Carmeliet, Liu, Thienpont, Dewerchin, Eelen, Li, Luo and Carmeliet5, Reference Goveia, Rohlenova, Taverna, Treps, Conradi, Pircher, Geldhof, de Rooij, Kalucka, Sokol, García-Caballero, Zheng, Qian, Teuwen, Khan, Boeckx, Wauters, Decaluwé, de Leyn, Vansteenkiste, Weynand, Sagaert, Verbeken, Wolthuis, Topal, Everaerts, Bohnenberger, Emmert, Panovska, de Smet, Staal, Mclaughlin, Impens, Lagani, Vinckier, Mazzone, Schoonjans, Dewerchin, Eelen, Karakach, Yang, Wang, Bolund, Lin, Thienpont, Li, Lambrechts, Luo and Carmeliet12, Reference Larionov, Hammer, Fiedler and Filgueira22–Reference Leone, Malerba, Susca, Favoino, Perosa, Brunori, Prete and Racanelli24).

Bone marrow-derived endothelial progenitor cells (EPCs) have also been proposed to contribute to the tumour vascular compartment (Ref. Reference Salybekov, Kobayashi and Asahara25). Hypoxia and chemokine signalling, particularly VEGF and C-X-C motif chemokine ligand 12 (CXCL12), can mobilize EPC-like populations to tumours. However, direct incorporation into functional tumour endothelium appears limited or context dependent in many human cancers; paracrine support of angiogenesis and perivascular niches may be more important than durable endothelial engraftment (Refs Reference Han, Zhu and Meng23, Reference Anastasiadou, Quesnel, Duran, Filippou and Karagiannis26, Reference Mpekris, Panagi, Charalambous, Voutouri and Stylianopoulos27).

Endothelial plasticity provides an additional source of phenotypic diversity. EndMT can generate mesenchymal-like endothelial states with reduced junctional integrity and enhanced extracellular matrix remodelling (Refs Reference Yin and Wang28, Reference Du, Li, Zhong, Zhang and Zhao29, Reference Bourreau, Navarro, Cotinat, Krejbich, Guillonneau, Guette, Boissard, Henry, Corre, Treps and Clere30). This process should be distinguished from vasculogenic mimicry, in which aggressive malignant cells form perfusable, vessel-like channels that are not lined by bona fide endothelial cell (Ref. Reference Maniotis, Folberg, Hess, Seftor, Gardner, Pe’er, Trent, Meltzer and Hendrix31). Both phenomena can alter vascular function and therapeutic response, but only EndMT directly represents endothelial-state remodelling.

TEC heterogeneity is shaped by both cellular origin and phenotypic plasticity. These processes should be distinguished from alternative perfusion mechanisms, such as vessel co-option and vasculogenic mimicry, which support tumour blood supply without generating conventional TEC-lined neovessels (Table 1). Accordingly, cellular origin, endothelial-state transition, and alternative perfusion should be considered distinct but interacting dimensions of tumour vascular heterogeneity.

Table 1.

Endothelial sources, plasticity programs, and alternative perfusion mechanisms in tumours

Table 1. Long description

The table is organized into five columns: Source or process, Key features, Main drivers, Functional relevance, and Refs.

* Resident E C s: Features E C reprogramming driven by V E G F, F G F, Ang, and inflammatory cytokines, resulting in abnormal angiogenesis.

* B M-derived E P C s: Features E P C recruitment driven by V E G F and C X C L 12, resulting in paracrine angiogenesis.

* Endothelial plasticity: Features End M T generating mesenchymal-like endothelial states driven by T G F-beta, hypoxia, and inflammatory or matrix cues, resulting in barrier loss and E C M remodeling.

* Vasculogenic mimicry (non-endothelial): Features tumor-cell-lined perfusion channels driven by hypoxia, tumor-cell plasticity, E M T-associated programs, and matrix remodeling, providing alternative perfusion.

* Vessel co-option: Features tumor cells growing along pre-existing host vessels without sprouting angiogenesis, driven by tumor-cell invasion and vascular basement-membrane adhesion, providing alternative perfusion and resistance to angiogenesis inhibition.

Abbreviations: BM: bone marrow; EC: endothelial cells; EPCs: endothelial progenitor cells; EndMT: endothelial-to-mesenchymal transition; ECM: extracellular matrix.

Signalling-metabolic programs driving TEC activation

TEC states are generated through the integration of angiogenic signalling, oxygen and nutrient availability, inflammatory cues, and endothelial metabolism (Ref. Reference Rohlenova, Veys, Miranda-Santos, De Bock and Carmeliet36). These inputs do not act independently: angiogenic pathways reshape the metabolic requirements of endothelial cells, whereas metabolic adaptation determines whether endothelial activation can be sustained under hypoxic and nutrient-limited conditions. TEC phenotypes should therefore be understood as dynamic signalling–metabolic states rather than the products of isolated molecular pathways.

VEGF/VEGFR signalling remains a principal driver of endothelial activation in tumours (Refs Reference Lee, Kim, Kumar, Lee, Yang and Kim37–Reference Vimalraj40). VEGF-A binding to VEGFR2 promotes endothelial survival, proliferation, migration, permeability, and sprouting through downstream PI3K–AKT–mTOR, MAPK, and SRC-family signalling pathways (Refs Reference Baek, Kwak and Ahn41, Reference Izadpanah, Izadpanah and Alt42). However, VEGF signalling does not uniformly activate all endothelial cells. High local VEGF exposure promotes tip-cell-like migration and sprouting (Ref. Reference Olsson, Dimberg, Kreuger and Claesson-Welsh43), whereas VEGF–Notch interactions regulate the balance between tip- and stalk-like endothelial programs. VEGF-induced DLL4 expression activates Notch signalling in adjacent endothelial cells, suppressing excessive tip-cell formation and promoting stalk-like proliferation and vessel elongation (Refs Reference Hellström, Phng, Hofmann, Wallgard, Coultas, Lindblom, Alva, Nilsson, Karlsson, Gaiano, Yoon, Rossant, Iruela-Arispe, Kalén, Gerhardt and Betsholtz44, Reference Gerhardt, Golding, Fruttiger, Ruhrberg, Lundkvist, Abramsson, Jeltsch, Mitchell, Alitalo, Shima and Betsholtz45). Tip- and stalk-like phenotypes are therefore functional and reversible endothelial states rather than fixed cell lineages (Ref. Reference Jakobsson, Franco, Bentley, Collins, Ponsioen, Aspalter, Rosewell, Busse, Thurston, Medvinsky, Schulte-Merker and Gerhardt46).

Additional pathways regulate the stability and maturation of these angiogenic states. Ang1-mediated Tie2 activation generally supports endothelial quiescence, junctional integrity, pericyte association, and vascular stabilization (Ref. Reference Suri, Jones, Patan, Bartunkova, Maisonpierre, Davis, Sato and Yancopoulos47). By contrast, increased Ang2 frequently destabilizes endothelial–pericyte interactions and sensitizes tumour vessels to VEGF-dependent remodelling (Ref. Reference Maisonpierre, Suri, Jones, Bartunkova, Wiegand, Radziejewski, Compton, McClain, Aldrich, Papadopoulos, Daly, Davis, Sato and Yancopoulos48). Nevertheless, Ang2 should not be regarded as a universal Tie2 antagonist because its effects vary with endothelial context, VEGF availability, integrin signalling, and the involvement of Tie2-expressing myeloid cells (Refs Reference Felcht, Luck, Schering, Seidel, Srivastava, Hu, Bartol, Kienast, Vettel, Loos, Kutschera, Bartels, Appak, Besemfelder, Terhardt, Chavakis, Wieland, Klein, Thomas, Uemura, Goerdt and Augustin49, Reference De Palma, Venneri, Galli, Sergi Sergi, Politi, Sampaolesi and Naldini50). The functional outcome of the Ang/Tie2 axis therefore depends on the cellular and signalling environment rather than on a simple Ang1-stabilizing/Ang2-destabilizing dichotomy.

Hypoxia further reshapes endothelial signalling and state transitions. Stabilization of HIF-1α and HIF-2α increases the expression of VEGF and other angiogenic mediators, thereby linking oxygen deprivation to endothelial activation and vascular remodelling (Refs Reference Pugh and Ratcliffe59, Reference Tang, Wang, Esko, Giordano, Huang, Gerber, Ferrara and Johnson60) (Table 2). However, persistent hypoxia also promotes abnormal sprouting, impaired endothelial junctions, metabolic stress, and selection of endothelial populations capable of surviving under reduced oxygen and nutrient availability (Refs Reference Cohen, Geck and Toker65, Reference Carmeliet and Jain66). Hypoxia is therefore both a driver of angiogenesis and a selective pressure that contributes to TEC heterogeneity (Refs Reference Cohen, Geck and Toker65, Reference Carmeliet and Jain66).

Table 2.

Functional and potential therapeutic implications of major tumour-associated endothelial cell states

Table 2. Long description

The table consists of five columns and seven rows of data.

* Row 1: Tip-like angiogenic T E C s. Markers: A P L N, E S M 1, K D R-associated programs. Functions: Sprout guidance, migration, and angiogenesis initiation. Therapy: Sensitivity to V E G F / V E G F R or D L L 4–Notch pathway modulation.

* Row 2: Stalk-like/proliferative T E C s. Markers: K D R, N O T C H 1-related programs. Functions: Endothelial proliferation, vessel elongation, and vascular lumen formation. Therapy: Sensitivity to angiogenic or proliferation-targeted therapy.

* Row 3: Capillary-like/quiescent T E C s. Markers: C A 4, E M C N, R G C C and capillary exchange programs. Functions: Solute exchange, barrier maintenance, and tissue perfusion. Therapy: Preservation or enhancement during vascular-normalization strategies.

* Row 4: Venous- or H E V-like T E C s. Markers: A C K R 1, S E L P, I C A M 1 and inflammatory adhesion programs. Functions: Leukocyte adhesion and regulation of immune-cell entry. Therapy: Immune-permissive vascular reprogramming or immunotherapy enhancement.

* Row 5: Immunosuppressive T E C s. Markers: C X C L 1 2, C D 2 7 4 / P D-L 1-associated programs. Functions: Restriction of effector lymphocyte activity and recruitment of suppressive immune populations. Therapy: Chemokine-axis, adhesion-molecule, or immune-checkpoint modulation.

* Row 6: Metabolically adapted T E C s. Markers: P F K F B 3 / glycolytic, H I F-response, P H G D H / serine-biosynthesis, and redox-stress programs. Functions: Survival under hypoxic and metabolic stress. Therapy: Endothelial metabolic targeting combined with angiogenic-pathway inhibition.

* Row 7: Matrix-remodelling/mesenchymal-like T E C s. Markers: C O L 1 A 1-associated, E C M-remodelling or End M T-related programs. Functions: Basement membrane remodelling, altered junctions, and barrier dysfunction. Therapy: Modulation of End M T-associated signalling or E C M remodelling.

Abbreviations: EC: endothelial cells; ECM: extracellular matrix; TECs: tumour-associated endothelial cells.

Note: TEC states may overlap, transition between phenotypes, and differ across tumour types, species, anatomical regions, and treatment conditions. The proposed therapeutic sensitivities are hypothesis-generating and require endothelial-specific functional and longitudinal validation.

Endothelial metabolism provides the energetic and biosynthetic capacity required to sustain these signalling programs. Despite their direct access to circulating oxygen, angiogenic endothelial cells rely predominantly on glycolysis for ATP production (Ref. Reference De Bock, Georgiadou, Schoors, Kuchnio, Wong, Cantelmo, Quaegebeur, Ghesquière, Cauwenberghs, Eelen, Phng, Betz, Tembuyser, Brepoels, Welti, Geudens, Segura, Cruys, Bifari, Decimo, Blanco, Wyns, Vangindertael, Rocha, Collins, Munck, Daelemans, Imamura, Devlieger, Rider, Van Veldhoven, Schuit, Bartrons, Hofkens, Fraisl, Telang, Deberardinis, Schoonjans, Vinckier, Chesney, Gerhardt, Dewerchin and Carmeliet61). Glycolytic regulators such as PFKFB3 support migration and proliferation by increasing local glycolytic flux, particularly in endothelial protrusions and other regions of high energetic demand (Ref. Reference De Bock, Georgiadou, Schoors, Kuchnio, Wong, Cantelmo, Quaegebeur, Ghesquière, Cauwenberghs, Eelen, Phng, Betz, Tembuyser, Brepoels, Welti, Geudens, Segura, Cruys, Bifari, Decimo, Blanco, Wyns, Vangindertael, Rocha, Collins, Munck, Daelemans, Imamura, Devlieger, Rider, Van Veldhoven, Schuit, Bartrons, Hofkens, Fraisl, Telang, Deberardinis, Schoonjans, Vinckier, Chesney, Gerhardt, Dewerchin and Carmeliet61) (Table 2). Additional pathways, including PHGDH-dependent serine biosynthesis, may maintain redox balance and support endothelial survival under hypoxic or nutrient-limited conditions (Refs Reference Vandekeere, Dubois, Kalucka, Sullivan, García-Caballero, Goveia, Chen, Diehl, Bar-Lev, Souffreau, Pircher, Kumar, Vinckier, Hirabayashi, Furuya, Schoonjans, Eelen, Ghesquière, Keshet, Li, Vander Heiden, Dewerchin and Carmeliet62, Reference Zhang, Li, Hu, Huang, Yang, Zhang, Wellen, Xu, Conn, Zou, Kahn, Rhoades, Weljie, Fuchs, Amankulor, Yoshor, Ye, Koumenis, Gong and Fan63) (Table 2). These metabolic dependencies can differ among proliferative, quiescent, immunomodulatory, and stress-adapted TEC states (Ref. Reference Eelen, de Zeeuw, Simons and Carmeliet67).

Collectively, angiogenic signalling and endothelial metabolism form an interconnected regulatory network that determines TEC activation, state stability, and adaptation to the tumour microenvironment. Metabolic programs should therefore be viewed not merely as consequences of angiogenic activation but as potential determinants of endothelial-state persistence and therapeutic sensitivity. This signalling–metabolic integration may help explain why inhibition of a single angiogenic pathway does not uniformly eliminate TECs and why metabolically adapted endothelial states can persist during vascular-directed therapy.

Structural vascular remodelling and barrier dysfunction

The signalling–metabolic programs described above are accompanied by structural changes in tumour-associated endothelial cells and their surrounding vascular support compartments. Persistent VEGF signalling, hypoxia, inflammatory mediators, extracellular matrix remodelling, and altered mechanical forces can disrupt endothelial junctions, basement-membrane organization, and endothelial–pericyte interactions (Refs Reference Subudhi, Lee and Jain68, Reference Guelfi, Hodivala-Dilke and Bergers69, Reference Siddhartha and Garg70). Consequently, tumour vessels frequently exhibit irregular architecture, enlarged endothelial gaps, discontinuous basement membranes, insufficient or abnormal pericyte coverage, and heterogeneous vessel diameters (Refs Reference Subudhi, Lee and Jain68, Reference Guelfi, Hodivala-Dilke and Bergers69).

Disruption of endothelial junctional proteins, including VE-cadherin, claudin-5, and occludin, weakens vascular barrier integrity and increases permeability (Refs Reference Subudhi, Lee and Jain68, Reference Boucher, Da Costa Oliveira, Levesque, Rutault, Savoyaud, Bréchot and Germain71). At the same time, proteolytic remodelling of the basement membrane and extracellular matrix by matrix metalloproteinases and other enzymes can impair vascular stability and facilitate endothelial migration (Refs Reference Liu, Chen, Zheng, Sun and Shi39, Reference Siddhartha and Garg70). These structural abnormalities contribute to plasma leakage, elevated interstitial fluid pressure, irregular blood flow, and spatially heterogeneous perfusion (Ref. Reference Subudhi, Lee and Jain68). Importantly, increased vascular permeability does not necessarily improve drug delivery: severe leakage and elevated interstitial pressure may reduce convective transport and limit the uniform distribution of therapeutic agents within tumours (Ref. Reference Subudhi, Lee and Jain68).

Barrier dysfunction is not uniform across the tumour vasculature. Proliferative tip- and stalk-like TEC populations may display immature junctions and limited pericyte support, whereas capillary-like, quiescent, or more mature endothelial states may retain relatively intact barrier and perfusion functions. Endothelial phenotypes also vary across hypoxic tumour cores, invasive margins, perivascular immune niches, and regions exposed to therapy (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Subudhi, Lee and Jain68). Spatial differences in endothelial junctions, basement-membrane organization, pericyte coverage, and vessel maturity may therefore generate regional variation in permeability, oxygenation, immune-cell trafficking, and drug delivery (Refs Reference Subudhi, Lee and Jain68, Reference Lahooti, Akwii, Zahra, Sajib, Lamprou, Alobaida, Lionakis, Mattheolabakis and Mikelis72).

These structural abnormalities have consequences beyond nutrient supply. Irregular perfusion can worsen hypoxia and promote metabolic adaptation, treatment resistance, and selection of aggressive tumour-cell populations (Refs Reference Subudhi, Lee and Jain68, Reference Jordan, Drake and Hodivala-Dilke73). Defective endothelial barriers may also facilitate tumour-cell intravasation and metastatic dissemination, while abnormal expression of endothelial adhesion molecules can restrict immune-cell entry into the tumour microenvironment (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Subudhi, Lee and Jain68, Reference Boucher, Da Costa Oliveira, Levesque, Rutault, Savoyaud, Bréchot and Germain71). The vascular barrier should therefore be regarded as a dynamic regulator of tumour progression, therapeutic delivery, and antitumour immunity rather than as a passive structural defect (Ref. Reference Subudhi, Lee and Jain68).

The spatial heterogeneity of vascular structure also helps explain why vascular-directed therapies can produce different functional outcomes. Extensive elimination of immature vessels may cause vessel pruning and aggravate hypoxia, whereas appropriately dosed intervention may restore endothelial junctions, pericyte association, and perfusion, producing transient vascular normalization (Refs Reference Subudhi, Lee and Jain68, Reference Jordan, Drake and Hodivala-Dilke73). By contrast, vascular-disrupting strategies are designed to damage established tumour vessels and induce rapid loss of blood flow (Ref. Reference Guelfi, Hodivala-Dilke and Bergers69). These heterogeneous vascular phenotypes therefore provide the structural basis for distinguishing vessel pruning, vascular normalization, and vascular disruption and underscore the need to measure vascular function rather than infer therapeutic effects from changes in vessel density alone (Refs Reference Subudhi, Lee and Jain68, Reference Guelfi, Hodivala-Dilke and Bergers69).

Endothelial-state heterogeneity and spatial specialization

The structural and functional heterogeneity described above reflects the presence of multiple endothelial states within and across tumours. Single-cell RNA sequencing has identified TEC populations associated with angiogenic sprouting, endothelial proliferation, capillary exchange, venous differentiation, immune regulation, extracellular matrix remodelling, and metabolic adaptation (Figure 2) (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Pan, Li, Dong, Liu, Zhang, Zhang, Zhang, Huang, Shi, Sun, Fang, Sun, Huang, Shao, Wang and Yin51). Spatial transcriptomic and in situ approaches further indicate that these states are not randomly distributed but are associated with distinct anatomical and microenvironmental niches, including hypoxic tumour cores, invasive margins, immune-cell-rich regions, and treatment-altered stromal compartments (Refs Reference Pan, Li, Dong, Liu, Zhang, Zhang, Zhang, Huang, Shi, Sun, Fang, Sun, Huang, Shao, Wang and Yin51, Reference Guo, Zhou, Dou, Guo, Li, Qin, Wang, Huai, Xue, Li, Ying, Xue, Gao and He54).

Figure 2.

Single-cell identification of representative TEC states and their functional programs in tumour vasculature. Note: Single-cell RNA sequencing (scRNA-seq) and dimensionality-reduction analysis can resolve representative endothelial states and programs within tumour tissues. The illustrated programs include angiogenic/permeability-associated TECs characterized by PLVAP, APLN, and ESM1; mesenchymal-like TECs associated with COL1A1 expression; immune-excluding TEC programs associated with CXCL12; immune-permissive venous/high-endothelial-venule-like programs associated with ACKR1; and metabolically adapted TECs characterized by PHGDH or related stress-response programs. These endothelial states may differentially contribute to angiogenesis, extracellular matrix remodelling, immune-cell trafficking, vascular permeability, blood flow, and therapeutic resistance. CXCL12- and ACKR1-associated endothelial programs may exert distinct or opposing effects on immune-cell trafficking and should not be interpreted as a single uniform immunoregulatory TEC state. The schematic illustrates representative rather than exhaustive TEC programs; additional tip-like, stalk-like, capillary-like, venous/HEV-like, quiescent, and treatment-associated states may occur in specific tumour contexts. These state–function relationships are hypothesis-generating and require spatial, protein-level, and functional validation.

Figure 2. Long description

The diagram is organized into two main sections. The top section shows a linear workflow from left to right. It begins with a cluster of cells labeled Tumor tissue, followed by an arrow pointing to a microcentrifuge tube labeled s c dash R N A seq. A second arrow points to a U M A P clustering scatter plot with axes U M A P 1 and U M A P 2, showing distinct color-coded cell clusters.

The bottom section details the representative T E C states and programs. On the left, a vertical list identifies five T E C categories with corresponding cell icons and markers.

- Angiogenic or permeability-associated T E C s characterized by P L V A P, A P L N, and E S M 1.

- Mesenchymal-like T E C s associated with C O L 1 A 1.

- Immune-excluding T E C programs associated with C X C L 12.

- Immune-permissive venous or H E V-like programs associated with A C K R 1.

- Metabolically adapted T E C s characterized by P H G D H or stress-response programs.

An arrow from this list points to a large schematic on the right illustrating functional outcomes. At the center of this schematic is a Tumor mass with arrows radiating outward to various processes.

- Upward to Angiogenesis, showing E C s, Tip cells, and E P C s forming new vessels toward the tumor.

- Left to E C M remodeling, showing a cross-section of tissue layers and a blood vessel.

- Right to Immune suppression, showing a cluster of diverse immune cells.

- Downward-right to Permeability, showing a Leaky vessel with arrows for Blood flow.

- Downward-left to Drugs resistance, showing Drugs being blocked or diverted from a secondary tumor mass.

Angiogenic TECs are among the most consistently identified tumour endothelial populations. Tip-like TECs commonly express angiogenesis-associated genes such as APLN, ESM1, and KDR and are characterized by migratory and sprouting activity (Refs Reference Pan, Li, Dong, Liu, Zhang, Zhang, Zhang, Huang, Shi, Sun, Fang, Sun, Huang, Shao, Wang and Yin51, Reference Goveia, Rohlenova, Taverna, Treps, Conradi, Pircher, Geldhof, de Rooij, Kalucka, Sokol, García-Caballero, Zheng, Qian, Teuwen, Khan, Boeckx, Wauters, Decaluwé, De Leyn, Vansteenkiste, Weynand, Sagaert, Verbeken, Wolthuis, Topal, Everaerts, Bohnenberger, Emmert, Panovska, De Smet, Staal, Mclaughlin, Impens, Lagani, Vinckier, Mazzone, Schoonjans, Dewerchin, Eelen, Karakach, Yang, Wang, Bolund, Lin, Thienpont, Li, Lambrechts, Luo and Carmeliet52, Reference Xie, Niu, Du, Chen, Zheng, Dai, Dan, Duan, Dou, Feng, Zhang and Zheng53). Stalk-like or proliferative endothelial populations support vessel elongation through endothelial proliferation and are frequently associated with VEGF–Notch and cell-cycle programs (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Pan, Li, Dong, Liu, Zhang, Zhang, Zhang, Huang, Shi, Sun, Fang, Sun, Huang, Shao, Wang and Yin51). However, tip- and stalk-like phenotypes should not be considered fixed endothelial lineages. They represent dynamic functional states whose abundance and transcriptional features can change in response to local VEGF availability, hypoxia, tissue context, and treatment (Ref. Reference Pan, Li, Dong, Liu, Zhang, Zhang, Zhang, Huang, Shi, Sun, Fang, Sun, Huang, Shao, Wang and Yin51).

Other TEC populations retain features associated with capillary, venous, or quiescent endothelial differentiation. Capillary-like states expressing genes such as CA4, EMCN, and RGCC may contribute to vascular exchange, barrier function, and tissue perfusion, although their functional competence within tumours remains variable (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Goveia, Rohlenova, Taverna, Treps, Conradi, Pircher, Geldhof, de Rooij, Kalucka, Sokol, García-Caballero, Zheng, Qian, Teuwen, Khan, Boeckx, Wauters, Decaluwé, De Leyn, Vansteenkiste, Weynand, Sagaert, Verbeken, Wolthuis, Topal, Everaerts, Bohnenberger, Emmert, Panovska, De Smet, Staal, Mclaughlin, Impens, Lagani, Vinckier, Mazzone, Schoonjans, Dewerchin, Eelen, Karakach, Yang, Wang, Bolund, Lin, Thienpont, Li, Lambrechts, Luo and Carmeliet52, Reference Guo, Zhou, Dou, Guo, Li, Qin, Wang, Huai, Xue, Li, Ying, Xue, Gao and He54). Venous- or high-endothelial-venule-like populations may express ACKR1, SELP, ICAM1, or related inflammatory adhesion programs and can influence leukocyte adhesion and transendothelial migration (Refs Reference Pan, Li, Dong, Liu, Zhang, Zhang, Zhang, Huang, Shi, Sun, Fang, Sun, Huang, Shao, Wang and Yin51, Reference Asrir, Tardiveau, Coudert, Laffont, Blanchard, Bellard, Veerman, Bettini, Lafouresse, Vina, Tarroux, Roy, Girault, Molinaro, Martins, Scoazec, Ortega, Robert and Girard55, Reference Wang, Wang, Liao, Yuan, Mei, Zhang, Wu, Kang, Zheng, Yang, Liu, Shang, Li, Wang, Li, Liu, Hu, Dong, Zhao, Wang, Liu, den, Wang, Han, Chen, Yang, Xu, He, Yue, Sun, Fang and Ying56). These observations indicate that not all tumour endothelial states are uniformly dysfunctional or immunosuppressive; some may retain or acquire functions that support perfusion and immune-cell access (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Asrir, Tardiveau, Coudert, Laffont, Blanchard, Bellard, Veerman, Bettini, Lafouresse, Vina, Tarroux, Roy, Girault, Molinaro, Martins, Scoazec, Ortega, Robert and Girard55, Reference Wang, Wang, Liao, Yuan, Mei, Zhang, Wu, Kang, Zheng, Yang, Liu, Shang, Li, Wang, Li, Liu, Hu, Dong, Zhao, Wang, Liu, den, Wang, Han, Chen, Yang, Xu, He, Yue, Sun, Fang and Ying56, Reference Wen74).

Immunoregulatory and metabolically adapted TEC states add further dimensions to this heterogeneity. Endothelial populations enriched for CXCL12, CD274/PD-L1-associated programs, or other immune-regulatory signals may contribute to immune-cell exclusion, retention of suppressive immune populations, or inhibition of T-cell activity (Refs Reference Cleveland and Fan57, Reference Fang, Lu, Zheng, Jiang, Shen, Shang, Lu and Fu58). Metabolically adapted TECs may exhibit enhanced glycolytic, hypoxia-response, redox, or amino-acid metabolic programs that support survival under oxygen- and nutrient-limited conditions (Refs Reference Cleveland and Fan57, Reference Kane, Edwards and Chen75). Another category consists of mesenchymal-like or invasive TECs, which display transcriptional features associated with extracellular matrix remodelling and endothelial-to-mesenchymal transition (EndMT). For instance, COL1A1+ endothelial cells exhibit mesenchymal characteristics and facilitate tumour invasion through activation of the ANGPTL4-SDC4 signalling pathway, thereby promoting tumour cell migration and stromal remodelling (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Liu, Yu, Lin, Wu, Xia, Zhu, Huang, Wu, Zhang, Li, Zhu and Wang64). These functional programs may overlap with angiogenic or immunoregulatory phenotypes rather than defining mutually exclusive endothelial populations (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Kane, Edwards and Chen75).

The identification of these states has generated hypotheses regarding therapeutic sensitivity (Table 2) (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Cleveland and Fan57, Reference Kane, Edwards and Chen75). Tip- and stalk-like populations may be relatively sensitive to VEGF/VEGFR, DLL4–Notch, or proliferation-associated interventions, whereas capillary-like or quiescent states may be better suited for functional preservation or vascular normalization. Venous- or high endothelial venule (HEV)-like states may provide potential entry points for immune-permissive vascular reprogramming, while immunosuppressive or metabolically adapted states may be vulnerable to chemokine, immune-checkpoint, or endothelial metabolic interventions. However, these associations remain predominantly hypothesis-generating (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Subudhi, Lee and Jain68). They should not be interpreted as established state-specific treatment indications because most have not been prospectively validated across tumour types, patients, and therapeutic settings.

Cross-study comparison is further complicated by inconsistent nomenclature, differences in tissue dissociation and endothelial-cell enrichment, species-specific expression, limited sampling, and treatment-induced state transitions (Ref. Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15). Similar biological programs may receive different cluster labels, whereas cells assigned to the same label may differ across tumour types or anatomical sites. Moreover, transcriptomic similarity does not establish that a TEC state is stable, functionally distinct, or selectively targetable (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Subudhi, Lee and Jain68).

TEC states should therefore be classified according to convergent transcriptional, spatial, protein-level, and functional evidence rather than marker expression alone (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Subudhi, Lee and Jain68). Their therapeutic relevance depends not only on abundance but also on causal contribution to vascular dysfunction, accessibility from the circulation, stability during treatment, prevalence across patients, and selectivity relative to essential normal vessels (Ref. Reference Subudhi, Lee and Jain68). These criteria are necessary to distinguish descriptive endothelial clusters from clinically actionable vascular states.

TECs as regulators of the tumour immune microenvironment

Among the heterogeneous endothelial populations described above, immunoregulatory TEC states are particularly important because the vascular endothelium forms the principal interface between circulating immune cells and tumour tissue. Beyond their role in angiogenesis, TECs regulate immune-cell recruitment, adhesion, transendothelial migration, retention, and function through chemokines, endothelial adhesion molecules, antigen-presentation machinery, and immune-checkpoint pathways (Refs Reference Cleveland and Fan57, Reference Xu, Miller, Tykodi, Akilesh and Warren76).

Abnormal tumour vessels can restrict immune-cell entry through both structural and molecular mechanisms. Irregular perfusion and impaired endothelial barrier organization reduce effective leukocyte access to some tumour regions (Ref. Reference Choi and Jung77) while sustained exposure to VEGF, hypoxia, and inflammatory mediators can suppress or dysregulate endothelial adhesion molecules required for leukocyte arrest and transmigration. Altered expression of ICAM-1, VCAM-1, E-selectin, and related adhesion programs may therefore contribute to endothelial anergy and immune-cell exclusion despite the presence of inflammatory signals within the tumour microenvironment (Refs Reference Subudhi, Lee and Jain68, Reference Xu, Miller, Tykodi, Akilesh and Warren76, Reference Wachholz, Akbari, Huijbers, Jalan, van Beijnum and Griffioen78, Reference Nowak-Sliwinska, van Beijnum, Griffioen, Huinen, Sopesens, Schulz, Jenkins, Dings, Groenendijk, Huijbers, Thijssen, Jonasch, Vyth-Dreese, Jordanova, Bex, Bernards, de Gruijl and Griffioen79).

TEC-derived chemokines and cytokines further shape the composition and spatial distribution of tumour-infiltrating immune cells (Refs Reference Xu, Miller, Tykodi, Akilesh and Warren76, Reference Mempel, Lill and Altenburger80). CXCL12 produced by selected endothelial populations, including tumour-associated lymphatic endothelial cells in some tumours, can restrict the positioning of effector T cells and promote their retention or egress (Ref. Reference Steele, Jaiswal, Delclaux, Dryg, Murugan, Femel, Son, du Bois, Hill, Leachman, Chang, Coussens, Anandasabapathy and Lund81). CCL2 and IL-6 within the vascular and tumour microenvironment may also support the accumulation and functional polarization of tumour-associated macrophages, myeloid-derived suppressor cells, regulatory T cells, or other suppressive immune populations (Refs Reference Cleveland and Fan57, Reference Subudhi, Lee and Jain68, Reference Xu, Miller, Tykodi, Akilesh and Warren76, Reference Lasser, Ozbay Kurt, Arkhypov, Utikal and Umansky82). These effects are context dependent because the same chemokine may produce different outcomes according to its concentration, spatial distribution, receptor expression, and tumour type (Ref. Reference Mempel, Lill and Altenburger80).

TECs can also influence immune-cell activity after recruitment. Endothelial PD-L1 may inhibit antigen-specific T-cell responses through PD-1 signalling, whereas other endothelial checkpoint or death-associated pathways may contribute to T-cell dysfunction or selective exclusion (Refs Reference Cleveland and Fan57, Reference Xu, Miller, Tykodi, Akilesh and Warren76). Endothelial cells can express major histocompatibility complex molecules and participate in antigen presentation, but they generally lack the full co-stimulatory capacity of professional antigen-presenting cells. Depending on endothelial state and inflammatory context, endothelial antigen presentation may therefore support local immune surveillance or contribute to T-cell tolerance and dysfunction (Refs Reference Wen74, Reference Xu, Miller, Tykodi, Akilesh and Warren76).

Importantly, endothelial immune regulation is spatially heterogeneous and is not uniformly immunosuppressive (Ref. Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15). Angiogenic or hypoxia-adapted TEC states may restrict lymphocyte adhesion, entry, or activity, whereas inflamed venous- or high-endothelial-venule-like states may express adhesion molecules and chemokines that support immune-cell recruitment (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Wang, Wang, Liao, Yuan, Mei, Zhang, Wu, Kang, Zheng, Yang, Liu, Shang, Li, Wang, Li, Liu, Hu, Dong, Zhao, Wang, Liu, den, Wang, Han, Chen, Yang, Xu, He, Yue, Sun, Fang and Ying56, Reference Cleveland and Fan57, Reference Subudhi, Lee and Jain68). The presence of ACKR1, SELP, ICAM1, or related inflammatory endothelial programs may indicate increased capacity for leukocyte trafficking, although marker expression alone does not establish the existence of a functionally mature high endothelial venule. Spatial localization and direct evidence of immune-cell adhesion and transmigration are required to define an immune-permissive endothelial state (Refs Reference Asrir, Tardiveau, Coudert, Laffont, Blanchard, Bellard, Veerman, Bettini, Lafouresse, Vina, Tarroux, Roy, Girault, Molinaro, Martins, Scoazec, Ortega, Robert and Girard55, Reference Wang, Wang, Liao, Yuan, Mei, Zhang, Wu, Kang, Zheng, Yang, Liu, Shang, Li, Wang, Li, Liu, Hu, Dong, Zhao, Wang, Liu, den, Wang, Han, Chen, Yang, Xu, He, Yue, Sun, Fang and Ying56).

TECs should therefore not be uniformly characterized as immunosuppressive. Instead, tumours contain a spectrum of endothelial states ranging from immune-excluding or immune-suppressive phenotypes to inflamed and potentially immune-permissive vascular states. These opposing endothelial programs provide a biological rationale for vascular–immune combination therapy, as discussed in section “Vascular-immune therapeutic combinations” (Refs Reference Cleveland and Fan57, Reference Subudhi, Lee and Jain68, Reference Nowak-Sliwinska, van Beijnum, Griffioen, Huinen, Sopesens, Schulz, Jenkins, Dings, Groenendijk, Huijbers, Thijssen, Jonasch, Vyth-Dreese, Jordanova, Bex, Bernards, de Gruijl and Griffioen79).

Broad anti-angiogenic therapy and its limits

VEGF/VEGFR blockade remains the most clinically validated vascular-targeting strategy. Bevacizumab established the clinical feasibility of VEGF-A inhibition when combined with chemotherapy in metastatic colorectal cancer (Table 4) (Ref. Reference Hurwitz, Fehrenbacher, Novotny, Cartwright, Hainsworth, Heim, Berlin, Baron, Griffing, Holmgren, Ferrara, Fyfe, Rogers, Ross and Kabbinavar83), whereas VEGFR-directed multikinase inhibitors expanded this approach across hepatocellular carcinoma (HCC), renal cell carcinoma (RCC), and other solid tumours. Sorafenib improved overall survival in advanced HCC in the SHARP trial (Table 4) (Ref. Reference Llovet, Ricci, Mazzaferro, Hilgard, Gane, Blanc, de Oliveira, Santoro, Raoul, Forner, Schwartz, Porta, Zeuzem, Bolondi, Greten, Galle, Seitz, Borbath, Häussinger, Giannaris, Shan, Moscovici, Voliotis and Bruix104); Lenvatinib demonstrated non-inferior overall survival to sorafenib in the first-line treatment of advanced HCC in the REFLECT trial (Tables 3 and 4) (Ref. Reference Kudo, Finn, Qin, Han, Ikeda, Piscaglia, Baron, Park, Han, Jassem, Blanc, Vogel, Komov, Evans, Lopez, Dutcus, Guo, Saito, Kraljevic, Tamai, Ren and Cheng84); and cabozantinib improved both progression-free and overall survival compared with placebo in previously treated advanced HCC in the CELESTIAL trial (Table 4) (Refs Reference El-Khoueiry, Hanna, Llovet and Kelley85, Reference Kelley, Ryoo, Merle, Park, Bolondi, Chan, Lim, Baron, Parnis, Knox, Cattan, Yau, Lougheed, Milwee, el-Khoueiry, Cheng, Meyer and Abou-Alfa86). Integration with immune-checkpoint blockade has further extended the clinical impact of vascular targeting, as illustrated by atezolizumab plus bevacizumab in unresectable HCC and pembrolizumab plus Lenvatinib in advanced RCC (Table 4) (Refs Reference Motzer, Alekseev, Rha, Porta, Eto, Powles, Grünwald, Hutson, Kopyltsov, Méndez-Vidal, Kozlov, Alyasova, Hong, Kapoor, Alonso Gordoa, Merchan, Winquist, Maroto, Goh, Kim, Gurney, Patel, Peer, Procopio, Takagi, Melichar, Rolland, de Giorgi, Wong, Bedke, Schmidinger, Dutcus, Smith, Dutta, Mody, Perini, Xing and Choueiri87, Reference Kudo, Ren, Guo, Han, Lin, Zheng, Ogasawara, Kim, Zhao, Li, Madoff, Ghobrial, Kawaoka, Gerolami, Ikeda, Kumada, el-Khoueiry, Vogel, Peng, Mody, Dutcus, Dubrovsky, Siegel, Finn and Llovet88, Reference Finn, Qin, Ikeda, Galle, Ducreux, Kim, Kudo, Breder, Merle, Kaseb, Li, Verret, Xu, Hernandez, Liu, Huang, Mulla, Wang, Lim, Zhu and Cheng89).

Table 4.

Selected phase III clinical trials of vascular-directed therapies and combinations

Table 4. Long description

The table consists of seven columns: Trial, Cancer, Investigational regimen, Comparator, Phase, Key outcome, and Refs.

* Trial A V F 2 1 0 7 g: m C R C cancer. Investigational regimen is Bevacizumab plus I F L chemotherapy. Comparator is Placebo plus I F L chemotherapy. Phase I I I. Key outcome: Improved O S, P F S, and O R R.

* Trial S H A R P: H C C cancer. Investigational regimen is Sorafenib. Comparator is Placebo. Phase I I I. Key outcome: Improved O S.

* Trial R E F L E C T: H C C cancer. Investigational regimen is Lenvatinib. Comparator is Sorafenib. Phase I I I. Key outcome: Non-inferior O S; improved P F S and O R R.

* Trial C E L E S T I A L: H C C cancer. Investigational regimen is Cabozantinib. Comparator is Placebo. Phase I I I. Key outcome: Improved O S and P F S.

* Trial I M brave 1 5 0: H C C cancer. Investigational regimen is Atezolizumab plus bevacizumab. Comparator is Sorafenib. Phase I I I. Key outcome: Improved O S and P F S.

* Trial K E Y N O T E dash 4 2 6: R C C cancer. Investigational regimen is Pembrolizumab plus axitinib. Comparator is Sunitinib. Phase I I I. Key outcome: Improved O S, P F S, and O R R.

* Trial C L E A R: R C C cancer. Investigational regimen is Lenvatinib plus pembrolizumab. Comparator is Sunitinib. Phase I I I. Key outcome: Improved O S, P F S, and O R R.

Abbreviations: HCC: hepatocellular carcinoma; IFL: irinotecan, fluorouracil, and leucovorin; mCRC: metastatic colorectal cancer; ORR: objective response rate; OS: overall survival; PFS: progression-free survival; RCC: renal cell carcinoma.

However, clinical efficacy does not establish selective targeting of a defined TEC state (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Leone, Malerba, Susca, Favoino, Perosa, Brunori, Prete and Racanelli24). Anti-VEGF antibodies and VEGFR-directed multikinase inhibitors act on signalling pathways shared across heterogeneous endothelial populations and, in the case of multikinase inhibitors, may also affect tumour cells, pericytes, fibroblasts, and other stromal compartments (Refs Reference Lee, Kim, Kumar, Lee, Yang and Kim37, Reference Liu, Chen, Zheng, Sun and Shi39). Their clinical benefits vary across tumour types, and improvements in progression-free survival do not always translate into durable overall-survival gains (Ref. Reference Kaka, Hafazalla, Samawi, Simpkin, Perry, Sahgal and das90). Responses may be limited by compensatory angiogenic signalling, hypoxia-driven metabolic adaptation, recruitment of pro-angiogenic immune and stromal cells, persistence of treatment-resistant endothelial states, and alternative vascularization mechanisms (Refs Reference Kuczynski and Reynolds34, Reference Lee, Kim, Kumar, Lee, Yang and Kim37, Reference Liu, Chen, Zheng, Sun and Shi39).

These limitations have shifted the therapeutic objective from maximal suppression of tumour vessels towards context-dependent vascular modulation. The functional consequence of VEGF-pathway inhibition depends on treatment intensity, timing, baseline vascular architecture, and the endothelial states present within the tumour. Accordingly, the same pathway-directed intervention may produce vessel regression, transient vascular normalization, or excessive pruning that aggravates hypoxia and compromises drug delivery (Refs Reference Liu, Chen, Zheng, Sun and Shi39, Reference Subudhi, Lee and Jain68). This context dependence provides the rationale for vascular-normalization strategies and for measuring vascular function rather than inferring therapeutic effects solely from drug administration or reduced vessel density (Refs Reference Liu, Chen, Zheng, Sun and Shi39, Reference Subudhi, Lee and Jain68).

Vascular normalization strategies

Vascular normalization refers to the partial restoration of tumour-vessel structure and function rather than the complete elimination of tumour vessels (Ref. Reference Luo, Zou, Wei, Yu, Zhao, Wu, Wang and Lu106). Tumour vasculature commonly exhibits discontinuous endothelial junctions, insufficient or abnormal pericyte coverage, excessive permeability, and heterogeneous blood flow (Ref. Reference Goel, Duda, Xu, Munn, Boucher, Fukumura and Jain91). Normalization may include improved endothelial junctional integrity, greater pericyte association, reduced vascular leakage, more uniform perfusion, and enhanced oxygen or drug delivery. By contrast, excessive vessel pruning can reduce perfusion, aggravate hypoxia, and impair the delivery of partner therapies (Refs Reference Tong, Boucher, Kozin, Winkler, Hicklin and Jain107, Reference Huang, Stylianopoulos, Duda, Fukumura and Jain108). The therapeutic objective is therefore to restore selected vascular functions within a context-dependent window rather than to achieve maximal suppression of the vascular compartment (Refs Reference Guelfi, Hodivala-Dilke and Bergers69, Reference Huang, Stylianopoulos, Duda, Fukumura and Jain108).

Appropriately dosed VEGF/VEGFR blockade is the best-established approach to vascular normalization (Ref. Reference Saharinen, Eklund and Alitalo92). Reduction of excessive VEGF signalling can strengthen endothelial junctions, decrease vascular leakage, improve pericyte association, lower interstitial fluid pressure, and promote more uniform perfusion. Anti-VEGF antibodies such as bevacizumab and VEGFR-directed multikinase inhibitors are used clinically in metastatic colorectal cancer, non-squamous NSCLC, HCC, RCC, and other solid tumours (Refs Reference Hurwitz, Fehrenbacher, Novotny, Cartwright, Hainsworth, Heim, Berlin, Baron, Griffing, Holmgren, Ferrara, Fyfe, Rogers, Ross and Kabbinavar83, Reference Kudo, Finn, Qin, Han, Ikeda, Piscaglia, Baron, Park, Han, Jassem, Blanc, Vogel, Komov, Evans, Lopez, Dutcus, Guo, Saito, Kraljevic, Tamai, Ren and Cheng84, Reference Motzer, Alekseev, Rha, Porta, Eto, Powles, Grünwald, Hutson, Kopyltsov, Méndez-Vidal, Kozlov, Alyasova, Hong, Kapoor, Alonso Gordoa, Merchan, Winquist, Maroto, Goh, Kim, Gurney, Patel, Peer, Procopio, Takagi, Melichar, Rolland, de Giorgi, Wong, Bedke, Schmidinger, Dutcus, Smith, Dutta, Mody, Perini, Xing and Choueiri87, Reference Sandler, Gray, Perry, Brahmer, Schiller, Dowlati, Lilenbaum and Johnson109). However, clinical use of these agents does not by itself demonstrate vascular normalization. Their vascular effects depend on treatment intensity, exposure duration, baseline vascular architecture, and tumour context; higher or prolonged exposure may favour vessel regression rather than functional restoration (Refs Reference Guelfi, Hodivala-Dilke and Bergers69, Reference Chen, Han, Du, Shi and Zhou110, Reference Wang, Chen, Liu, Zhang and Pan111).

Vascular normalization may also be promoted by modulating pathways that regulate endothelial quiescence and vessel maturation. The angiopoietin–Tie2 axis is particularly relevant: Ang1-mediated Tie2 activation generally supports endothelial junctional integrity, pericyte association, and vascular stabilization, whereas increased Ang2 frequently destabilizes these functions under pathological conditions (Table 3) (Refs Reference Goel, Duda, Xu, Munn, Boucher, Fukumura and Jain91, Reference Saharinen, Eklund and Alitalo92, Reference Falcón, Hashizume, Koumoutsakos, Chou, Bready, Coxon, Oliner and McDonald93). Nevertheless, the effects of Ang2 are context-dependent and may vary with VEGF availability, integrin signalling, endothelial state, and the involvement of Tie2-expressing myeloid cells. Ang2 inhibition or Tie2 activation may therefore complement VEGF-pathway blockade and extend or reinforce vascular normalization in preclinical tumour models (Refs Reference Coutelle, Schiffmann, Liwschitz, Brunold, Goede, Hallek, Kashkar and Hacker112, Reference Peterson, Kirkpatrick, Huang, Farrar, Marijt, Kloepper, Datta, Amoozgar, Seano, Jung, Kamoun, Vardam, Snuderl, Goveia, Chatterjee, Batista, Muzikansky, Leow, Xu, Batchelor, Duda, Fukumura and Jain113), although their clinical efficacy, optimal patient selection, and treatment schedule remain to be established (Refs Reference Saharinen, Eklund and Alitalo92, Reference Gillen, Richardson and Moore114).

Endothelial metabolic modulation represents a further potential route to normalization (Table 3) (Ref. Reference Cantelmo, Conradi, Brajic, Goveia, Kalucka, Pircher, Chaturvedi, Hol, Thienpont, Teuwen, Schoors, Boeckx, Vriens, Kuchnio, Veys, Cruys, Finotto, Treps, Stav-Noraas, Bifari, Stapor, Decimo, Kampen, de Bock, Haraldsen, Schoonjans, Rabelink, Eelen, Ghesquière, Rehman, Lambrechts, Malik, Dewerchin and Carmeliet94). In syngeneic Lewis lung carcinoma and B16F10 melanoma models, salvianic acid A (SAA) was reported to bind pyruvate kinase M2(PKM2), reduce endothelial glycolysis, and activate the β-catenin/claudin-5 junctional program (Ref. Reference Qian, Zhou, Zhang, Dong, Song, Tang, Wei, Yu, Shen, Chen, Choi, Yan, Zhong, Wan, Li, Wang, Lu and Zhao115). These changes were associated with tighter endothelial junctions, reduced vascular leakage, improved perfusion, and enhanced intratumoural delivery and antitumour activity of doxorubicin (Ref. Reference Qian, Zhou, Zhang, Dong, Song, Tang, Wei, Yu, Shen, Chen, Choi, Yan, Zhong, Wan, Li, Wang, Lu and Zhao115). Although these findings remain preclinical, they provide proof of concept that vascular normalization can be achieved by modifying endothelial metabolism rather than exclusively through VEGF-pathway inhibition (Refs Reference Cantelmo, Conradi, Brajic, Goveia, Kalucka, Pircher, Chaturvedi, Hol, Thienpont, Teuwen, Schoors, Boeckx, Vriens, Kuchnio, Veys, Cruys, Finotto, Treps, Stav-Noraas, Bifari, Stapor, Decimo, Kampen, de Bock, Haraldsen, Schoonjans, Rabelink, Eelen, Ghesquière, Rehman, Lambrechts, Malik, Dewerchin and Carmeliet94, Reference Qian, Zhou, Zhang, Dong, Song, Tang, Wei, Yu, Shen, Chen, Choi, Yan, Zhong, Wan, Li, Wang, Lu and Zhao115).

Reduced vessel density alone does not demonstrate vascular normalization because it cannot distinguish functional improvement from ineffective or destructive vessel pruning. Evidence of normalization should include measurable changes in perfusion, oxygenation, permeability, endothelial junctions, pericyte coverage, vessel maturity, or delivery of a partner therapy. Because the normalization window is transient and varies among tumour types, agents, doses, and treatment schedules, combination therapy should be timed according to demonstrated vascular improvement rather than inferred from drug administration alone (Refs Reference Goel, Duda, Xu, Munn, Boucher, Fukumura and Jain91, Reference Huang, Stylianopoulos, Duda, Fukumura and Jain108, Reference Winkler, Kozin, Tong, Chae, Booth, Garkavtsev, Xu, Hicklin, Fukumura, di Tomaso, Munn and Jain116, Reference Batchelor, Sorensen, di Tomaso, Zhang, Duda, Cohen, Kozak, Cahill, Chen, Zhu, Ancukiewicz, Mrugala, Plotkin, Drappatz, Louis, Ivy, Scadden, Benner, Loeffler, Wen and Jain117). Dynamic contrast-enhanced MRI and other perfusion- or permeability-based measurements can provide functional vascular readouts, but they do not resolve the molecular identity and heterogeneity of responsive endothelial populations (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Sorensen, Batchelor, Zhang, Chen, Yeo, Wang, Jennings, Wen, Lahdenranta, Ancukiewicz, di Tomaso, Duda and Jain118, Reference Li, Quan, Li, Lu and Cui119). A precision approach will therefore require integration of serial functional imaging with circulating, tissue-based, and spatial biomarkers to identify normalization and guide the timing of chemotherapy, radiotherapy, or immune-checkpoint blockade (Refs Reference Sorensen, Batchelor, Zhang, Chen, Yeo, Wang, Jennings, Wen, Lahdenranta, Ancukiewicz, di Tomaso, Duda and Jain118, Reference Li, Quan, Li, Lu and Cui119).

Vascular-directed immune modulation

Vascular-directed immune modulation aims to alter the immunological functions of the tumour vascular interface rather than solely suppress angiogenesis (Refs Reference Guelfi, Hodivala-Dilke and Bergers69, Reference Lanitis, Irving and Coukos120). Abnormal TEC states can restrict immune-cell trafficking through defective adhesion-molecule expression, chemokine signalling, immune-checkpoint pathways, and impaired vascular perfusion (Ref. Reference Lanitis, Irving and Coukos120). Therapeutic intervention may therefore promote antitumour immunity by restoring immune-cell access, directly eliminating antigen-positive vascular or perivascular cells, or combining vascular inhibition with immune-checkpoint blockade (Refs Reference Guelfi, Hodivala-Dilke and Bergers69, Reference Wachholz, Akbari, Huijbers, Jalan, van Beijnum and Griffioen78, Reference Kuo, Khan and Kerbel121). These outcomes should be distinguished because vascular injury, immune activation, and immune-permissive vascular reprogramming represent different biological effects (Refs Reference Guelfi, Hodivala-Dilke and Bergers69, Reference Kuo, Khan and Kerbel121).

Strategies under investigation include antibodies, vaccines, bispecific constructs, and cellular approaches directed against vascular-associated antigens such as TEM1/CD248, CD105/endoglin, and VEGFR2 (Table 3) (Refs Reference Bagley, Honma, Weber, Boutin, Rouleau, Shankara, Kataoka, Ishida, Roberts and Teicher95–Reference Niethammer, Xiang, Becker, Wodrich, Pertl, Karsten, Eliceiri and Reisfeld100). However, these markers are not uniformly restricted to TECs. TEM1/CD248 is frequently associated with perivascular stromal cells and may also be expressed by subsets of malignant cells, particularly in sarcomas (Refs Reference Bagley, Honma, Weber, Boutin, Rouleau, Shankara, Kataoka, Ishida, Roberts and Teicher95, Reference Lu, Gan, Lu, Zhang, Zhang, Wu, Han, Xu, Liu, Yang, Qin and Wen122, Reference Yee, Vigil, Sun, Torphy, Schulick and Zhu123); CD105 can be expressed by activated endothelial and other cell populations, including CAFs and epithelial tumour cells in some settings (Ref. Reference Hakuno, Janson, Trietsch, de Graaf, de Jonge-Muller, Crobach, Harryvan, Boonstra, Dinjens, Slingerland and Hawinkels124); and VEGFR2 is present in angiogenic vasculature but is not specific to tumour endothelium (Ref. Reference Lee, Kim, Kumar, Lee, Yang and Kim37). Some vascular-associated targets may additionally be detected in pericytes, fibroblast-lineage cells, hematopoietic populations, or normal vascular compartments (Refs Reference Christian, Winkler, Helfrich, Boos, Besemfelder, Schadendorf and Augustin125, Reference Tian, Huang, Golzio, Gao, Hector-Greene, Katsanis and Blobe126). These approaches should therefore be described as vascular-directed rather than TEC-specific unless selective endothelial engagement has been directly demonstrated.

Endothelial- or vascular-associated vaccination strategies have been explored as a means of inducing cytotoxic immune responses against antigen-positive components of the tumour vasculature (Refs Reference Facciponte, Ugel, De Sanctis, Li, Wang, Nair, Sehgal, Raj, Matthaiou, Coukos and Facciabene98, Reference Oladejo, Nguyen, Silwal, Reese, Paulishak, Markiewski and Wood127). Vaccines targeting antigens such as TEM1 or CD105 have reduced vascularization or tumour growth in preclinical models (Refs Reference Hong, Yu, Pai, Chen, Hsieh, Wu, Lin and Huang96, Reference Facciponte, Ugel, De Sanctis, Li, Wang, Nair, Sehgal, Raj, Matthaiou, Coukos and Facciabene98, Reference Oladejo, Nguyen, Silwal, Reese, Paulishak, Markiewski and Wood127, Reference Zhuo, Zheng, Liu, Huang, Wang, Zheng, Mao, Meng, Zhou, Zhong and Zhao128). Nevertheless, the observed effects may involve endothelial cells, perivascular stromal cells, or other antigen-expressing populations (Refs Reference Facciponte, Ugel, De Sanctis, Li, Wang, Nair, Sehgal, Raj, Matthaiou, Coukos and Facciabene98, Reference Lu, Gan, Lu, Zhang, Zhang, Wu, Han, Xu, Liu, Yang, Qin and Wen122, Reference Oladejo, Nguyen, Silwal, Reese, Paulishak, Markiewski and Wood127). Evaluation of each strategy should therefore establish the cell types expressing the target, its accessibility from the circulation, the mechanism of immune-mediated injury, and whether the effect is vascular destruction, angiogenesis inhibition, or immune-permissive vascular reprogramming (Ref. Reference Guelfi, Hodivala-Dilke and Bergers69).

Bispecific antibodies provide a complementary strategy by simultaneously modulating angiogenic and immune-checkpoint pathways. VEGF/PD-1 bispecific antibodies are designed to neutralize VEGF while blocking the PD-1 receptor, thereby coupling inhibition of pro-angiogenic signalling with restoration of T-cell activity within the tumour microenvironment (Refs Reference Wang, Wei, Zheng, Wang, Ying, Chen, Luo, Luo, Yu, Chen, Ma and Xu129, Reference Fang, Zhao, Luo, Yang, Huang, He, Zhao, Li, Li, Song, Du, Sun, Li, Xu, Wang, Yang, Fan, Liu, Zhao, Hu, Jia, Xu, Yi, Lv, Lan, Li, Liang, Wang, Yang, Jia, Chen, Lu, Feng, Liu, Zhou, Zhou, Liu, Zhou, He, Dong, Chen, Chen, Su, Li, Zhang, Yang, Cheng, Chen, Hou, Zhang, Guo, Wang, Lu, Wu, Feng, Li, Huang, Wang, Song, Peng, Liu, Guo, Li, Lu, Hu, Wang, Li, Xia and Zhang130). They should not be described as blocking PD-L1 unless the construct directly binds PD-L1 (Ref. Reference Shi, Yang, Gao, Hou, Lv, Cao, Chen, Zhang and Wang131). Although such agents may enhance the spatial coordination of vascular and immune modulation, their clinical benefit does not by itself demonstrate selective engagement or reprogramming of a defined TEC state (Refs Reference Guelfi, Hodivala-Dilke and Bergers69, Reference Wang, Wei, Zheng, Wang, Ying, Chen, Luo, Luo, Yu, Chen, Ma and Xu129–Reference Shi, Yang, Gao, Hou, Lv, Cao, Chen, Zhang and Wang131).

Across these therapeutic platforms, target enrichment alone is insufficient to establish vascular selectivity (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Guelfi, Hodivala-Dilke and Bergers69). Preclinical and clinical development should demonstrate cell-type-resolved protein expression, luminal or otherwise therapeutically relevant accessibility, endothelial versus stromal target engagement, tumour-to-normal selectivity, pharmacodynamic vascular effects, and an acceptable safety window (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Guelfi, Hodivala-Dilke and Bergers69). Because many of these approaches remain preclinical or in early clinical development, their classification should be based on the demonstrated target cell and vascular consequence rather than on the intended mechanism alone.

Tumour vascular-interface drug-delivery systems

Precision drug-delivery technologies represent another promising strategy to enhance endothelial targeting (Ref. Reference Giordo, Wehbe, Paliogiannis, Eid, Mangoni and Pintus132). Antibody-drug conjugates (ADCs), ligand-modified nanoparticles, aptamer-guided delivery systems and stimuli-responsive nanocarriers can exploit TEC-enriched surface molecules to increase local drug exposure while limiting systemic distribution (Refs Reference Zhang and He133, Reference Liu, Zhang, Shen, Sun, Tang and Chen134, Reference Endo, Ueda, Nagaoki, Sato, Maishi, Hida, Harashima and Nakamura135). Their therapeutic effect can arise from direct injury to TECs, inhibition of endothelial proliferation and migration, or secondary collapse of the vascular support required for tumour growth.

TEM1/endosialin/CD248 illustrates the difficulty of assigning cellular specificity to a vascular-associated delivery target. TEM1 (endosialin/CD248) is a cell-surface glycoprotein detected in tumour-associated vascular and perivascular compartments, particularly activated mesenchymal stromal cells, and has been reported in several solid tumours, including lung cancer and sarcoma-associated microenvironments (Refs Reference Rouleau, Curiel, Weber, Smale, Kurtzberg, Mascarello, Berger, Wallar, Bagley, Honma, Hasegawa, Ishida, Kataoka, Thurberg, Mehraein, Horten, Miller and Teicher136, Reference Ash, Orha, Mole, Dinesh-Kumar, Lee, Turrell and Isacke137). In some sarcomas, TEM1 may also be expressed directly by malignant cells (Ref. Reference Guo, Hu, Wang, Peng, Min, Wang, Matthaiou, Cheng, Sun, Tong, Fan, Zhang, Kandalaft, Irving, Coukos and Li138). A TEM1-directed antibody–drug conjugate combines a target-binding antibody with a linker and cytotoxic payload; following binding and internalization by TEM1-positive cells, intracellular payload release can induce target-dependent cytotoxicity. In preclinical studies, an anti-TEM1 ADC containing monomethyl auristatin E produced durable tumour inhibition in TEM1-positive Ewing sarcoma and neuroblastoma xenografts (Ref. Reference Rouleau, Gianolio, Smale, Roth, Krumbholz, Harper, Munroe, Green, Horten, Schmid and Teicher101), while the humanized hMP-E-8.3 antibody conjugated to a duocarmycin derivative induced marked regression in a TEM1-high SJSA-1 osteosarcoma xenograft model (Table 3) (Ref. Reference Capone, Piccolo, Fichera, Ciufici, Barcaroli, Sala, de Laurenzi, Iacobelli, Iacobelli and Sala102). These studies provide proof of concept for payload delivery to TEM1-positive tumour-associated compartments. However, because TEM1 may be expressed by endothelial, perivascular stromal, and malignant cells, the observed antitumour activity cannot be attributed to selective TEC targeting without cell-type-resolved biodistribution and uptake analyses.

The NGR platform provides an illustrative vascular-interface delivery strategy. Luo et al. developed cyclic NGR-modified sterically stabilized liposomes containing paclitaxel (NGR-SSL-PTX) (Table 3) (Ref. Reference Luo, Huang, Zhao, Zhao, Duan, Du, Yu, Song, Zhao, Zhang and Zhang103). The NGR motif recognizes aminopeptidase N (APN/CD13), a membrane-associated metalloprotease expressed by angiogenic tumour vasculature and by subsets of tumour cells, thereby facilitating the localization of NGR-decorated carriers to the tumour vascular interface (Refs Reference Pasqualini, Koivunen, Kain, Lahdenranta, Sakamoto, Stryhn, Ashmun, Shapiro, Arap and Ruoslahti139, Reference Curnis, Arrigoni, Sacchi, Fischetti, Arap, Pasqualini and Corti140). Under a metronomic administration schedule, NGR-SSL-PTX inhibited endothelial-cell proliferation and migration, reduced tumour microvessel density, and showed greater antitumour and antiangiogenic activity than control formulations in HT1080 human fibrosarcoma xenografts (Ref. Reference Luo, Huang, Zhao, Zhao, Duan, Du, Yu, Song, Zhao, Zhang and Zhang103). These findings provide preclinical proof of concept for peptide-guided delivery to the tumour vascular interface; however, they do not establish clinical efficacy or TEC-specific delivery.

The reported Annexin A1 aptamer–doxorubicin system should be interpreted primarily as tumour-cell-targeted rather than TEC-selective delivery. Bavi et al. conjugated doxorubicin to a 16-nucleotide DNA aptamer directed against Annexin A1 (Ref. Reference Bavi, Hang, Banerjee, Aquib, Jadhao, Rane, Bavi, Bhosale, Kodam, Jeon and Gu141). The construct recognized ANXA1-expressing A549 lung adenocarcinoma, HepG2 hepatocellular carcinoma, and U-87 MG glioblastoma cells in vitro and produced greater tumour-growth inhibition with less body-weight loss than free doxorubicin in an A549 xenograft model (Ref. Reference Bavi, Hang, Banerjee, Aquib, Jadhao, Rane, Bavi, Bhosale, Kodam, Jeon and Gu141). This study primarily demonstrated ANXA1-dependent targeting of malignant cells and should not be cited as evidence of selective delivery to TECs. ANXA1-directed vascular delivery has been reported using other ligands and payloads, but endothelial expression, luminal accessibility, ligand format, cellular uptake, and tumour model should be specified for each platform (Ref. Reference Yoneyama, Hatakeyama, Sutoh Yoneyama, Yoshiya, Uemura, Ishizu, Suzuki, Hachinohe, Ishiyama, Nonaka, Fukuda and Ohyama142).

Integrin-binding ligands provide another example of vascular-interface delivery with incomplete cell-type specificity. RGD- and isoDGR-containing ligands targeting angiogenesis-associated integrins, particularly α_vβ_3, have been used to functionalize liposomal, polymeric, lipid, and inorganic nanocarriers (Refs Reference Murphy, Majeti, Barnes, Makale, Weis, Lutu-Fuga, Wrasidlo and Cheresh143, Reference Corti, Sacchi, Gasparri, Monieri, Anderluzzi, Colombo, Gori, Mondino and Curnis144, Reference Amin, Badiee and Jaafari145). In orthotopic pancreatic and renal carcinoma models, α_vβ_3-targeted nanoparticles carrying doxorubicin localized to α_vβ_3-positive tumour vessels, induced vascular apoptosis, and suppressed metastasis (Ref. Reference Murphy, Majeti, Barnes, Makale, Weis, Lutu-Fuga, Wrasidlo and Cheresh143). However, α_vβ_3 and related integrins may also be expressed by malignant cells, fibroblasts, pericytes, and other stromal populations. Carrier composition, ligand specificity and density, payload, tumour model, cellular biodistribution, and vascular endpoint should therefore be reported for each formulation rather than inferring TEC-selective delivery from the targeting ligand alone.

Collectively, vascular-directed delivery platforms differ substantially in target distribution, cellular uptake, and mechanism of action. TEM1-directed ADCs provide evidence of delivery to TEM1-positive tumour-associated compartments, whereas the cited ANXA1 aptamer–doxorubicin study primarily supports malignant-cell targeting. NGR- and integrin-modified carriers illustrate vascular-interface localization but do not necessarily demonstrate selective TEC uptake. Future studies should distinguish endothelial uptake from uptake by tumour or perivascular stromal cells and determine whether delivery produces vascular normalization, angiogenesis inhibition, vascular disruption, or direct tumour-cell cytotoxicity. Single-cell and spatial omics may assist target discovery, but candidate targets will still require protein-level confirmation, vascular accessibility assessment, cell-type-resolved biodistribution, functional validation, and evaluation of tumour-to-normal selectivity.

Combination with chemotherapy

Combining anti-angiogenic therapy with chemotherapy is one of the earliest clinically established vascular-treatment strategies. Tumour vessels are frequently hyperpermeable yet poorly perfused, and vascular leakiness does not necessarily improve drug delivery. Chemotherapy benefits from vascular modulation only when the intervention produces functional improvement in delivery. Partial vascular normalization may strengthen endothelial junctions, reduce interstitial fluid pressure, improve perfusion, and increase intratumoural drug exposure, whereas excessive vessel regression may aggravate hypoxia, reduce perfusion, and compromise chemotherapy delivery (Refs Reference Qian, Zhou, Zhang, Dong, Song, Tang, Wei, Yu, Shen, Chen, Choi, Yan, Zhong, Wan, Li, Wang, Lu and Zhao115, Reference Terwoord, Norwood Toro, Hader, Hammond, Hockenberry, Linn, Vazirabad, Kong, Kriegel, Liu, Kivelä, Murtagh, Gutterman and Beyer146, Reference Yang, Fang, Ma, Wang, Wang, Jia, Yang, Sun, Zhou and Li147). The therapeutic objective is therefore to administer cytotoxic treatment during a measurable period of improved vascular function rather than to assume that all anti-angiogenic exposure enhances drug delivery.

Clinically, bevacizumab combined with fluoropyrimidine-based chemotherapy remains an established treatment option for metastatic colorectal cancer (Ref. Reference Van Cutsem, Rivera, Berry, Kretzschmar, Michael, DiBartolomeo, Mazier, Canon, Georgoulias, Peeters, Bridgewater and Cunningham148), while bevacizumab plus carboplatin and paclitaxel has demonstrated benefit in non-squamous NSCLC (Ref. Reference Zhou, Wu, Chen, Liu, Zhu, Lu, Feng, He, Han, Wang, Jiang, Hu, Zhang, Cheng, Song, Lu, Pan, Zheng and Yin149). Exploratory studies of anlotinib combined with chemotherapy in gastrointestinal malignancies, oesophageal squamous-cell carcinoma, and HCC have also reported encouraging activity, although larger confirmatory trials are required (Refs Reference Wu, Zhou, Han, Zhang, Yan, Chen, Wang, Qin, Mao, Tang, Zhu, Wei, Cui, Yang, Shi, Zhao, Jiang, Zhu, Wang, Wang, Zhu and Zhang150, Reference Li, Xia, Gao, Zhou, Hong, Cui, Zhao, Wu, Guo, Wang and Luo151, Reference Zhou, Bai, Luo, Bai, Pan, Yang, Gao, Shi, Zhang, Zheng, Hua, Xu, Hu, Zhang, Yang, da, Wang, Ma, Jia, Quan, Peng, Yang, Yin, Qi, Zhang, du, Mao, Meng, Jiao and Fan152). Future clinical studies should relate treatment sequence and timing to serial vascular measurements, including DCE-MRI-derived perfusion or permeability parameters, complementary perfusion imaging, and, where feasible, direct assessment of intratumoural drug delivery, rather than relying on fixed empirical schedules.

Combination with radiotherapy and the bridge to radio-immunotherapy

Radiotherapy efficacy is strongly influenced by tumour oxygenation. Hypoxia reduces the fixation of radiation-induced DNA damage and activates adaptive programs associated with HIF signalling, invasion, and treatment resistance (Refs Reference Beckers, Pruschy and Vetrugno153, Reference Zdrowowicz, Spisz, Hać, Herman-Antosiewicz and Rak154, Reference Chowdhury and PK155). When vascular modulation improves perfusion during irradiation, increased oxygen availability can promote fixation of radiation-induced free-radical damage and enhance tumour-cell killing (Refs Reference Guo, Lei, Zhang, Du and Chen156, Reference Zhao, Shi, Guo, Shang, Sun, Meng, Xiao, Wang and Li157, Reference Horsman and Overgaard158). Improved vascular function may also facilitate immune-cell trafficking into irradiated tumours. Conversely, excessive or prolonged vessel pruning can reduce perfusion, worsen hypoxia, and compromise both radiosensitivity and immune-cell access (Refs Reference Kuo, Khan and Kerbel121, Reference Hartley, Ebert and Cook159).

Radiation itself has context-dependent effects on the tumour microenvironment. It may promote antitumour immunity through immunogenic cell death, tumour-antigen release, and inflammatory signalling, but can also induce VEGF, PDGF, vascular remodelling, and recruitment of suppressive myeloid populations (Refs Reference Hartley, Ebert and Cook159, Reference Bergerud, Berkseth, Pardoll, Ganguly, Kleinberg, Lawrence, Odde, Largaespada, Terezakis and Sloan160, Reference Lin, Choyke and Sato161). Vascular-directed therapy may counteract these effects by limiting pro-angiogenic signalling, improving vascular function, and restoring immune accessibility (Refs Reference Subudhi, Lee and Jain68, Reference Kuo, Khan and Kerbel121). These interactions provide a mechanistic rationale for combining vascular modulation with radiotherapy and immune-checkpoint blockade, although benefit depends on whether treatment generates an immune-permissive rather than severely pruned vascular state (Refs Reference Kuo, Khan and Kerbel121, Reference Hartley, Ebert and Cook159).

Treatment sequence and timing should therefore be guided by measured vascular responses rather than assumed from drug administration (Refs Reference Subudhi, Lee and Jain68, Reference Kuo, Khan and Kerbel121). Candidate readouts include serial perfusion and oxygenation imaging, tissue hypoxia markers, indices of vessel maturity, and spatial assessment of immune-cell access (Ref. Reference Subudhi, Lee and Jain68). Although these measurements may provide pharmacodynamic or prognostic information, no biomarker has been prospectively validated as a universal guide for scheduling vascular-targeted therapy with radiotherapy (Refs Reference Subudhi, Lee and Jain68, Reference Hartley, Ebert and Cook159).

Vascular-immune therapeutic combinations

Abnormal tumour vasculature promotes immune exclusion through irregular perfusion, hypoxia, defective endothelial adhesion programs, and endothelial-derived immunosuppressive signals (Refs Reference Lanitis, Irving and Coukos120, Reference Kuo, Khan and Kerbel121). Vascular–immune combinations aim to reduce these barriers while restoring T-cell activity through immune-checkpoint blockade (Refs Reference Subudhi, Lee and Jain68, Reference Kuo, Khan and Kerbel121). Anti-angiogenic therapy may improve perfusion and cytotoxic-lymphocyte access, whereas checkpoint inhibition reverses inhibitory signalling in tumour-reactive T cells (Refs Reference Subudhi, Lee and Jain68, Reference Kuo, Khan and Kerbel121). Their therapeutic benefit may therefore depend on whether vascular modulation establishes an immune-permissive vascular state without causing excessive vessel pruning (Refs Reference Subudhi, Lee and Jain68, Reference Guelfi, Hodivala-Dilke and Bergers69, Reference Kuo, Khan and Kerbel121).

Clinical benefit has been established for selected regimens. In unresectable HCC, IMbrave150 demonstrated improved overall and progression-free survival with atezolizumab plus bevacizumab compared with sorafenib (Table 4) (Ref. Reference Finn, Qin, Ikeda, Galle, Ducreux, Kim, Kudo, Breder, Merle, Kaseb, Li, Verret, Xu, Hernandez, Liu, Huang, Mulla, Wang, Lim, Zhu and Cheng89). In advanced RCC, KEYNOTE-426 showed improved outcomes with pembrolizumab plus axitinib, while CLEAR demonstrated benefit with lenvatinib plus pembrolizumab (Table 4) (Refs Reference Llovet, Ricci, Mazzaferro, Hilgard, Gane, Blanc, de Oliveira, Santoro, Raoul, Forner, Schwartz, Porta, Zeuzem, Bolondi, Greten, Galle, Seitz, Borbath, Häussinger, Giannaris, Shan, Moscovici, Voliotis and Bruix104, Reference Oladejo, Nguyen, Silwal, Reese, Paulishak, Markiewski and Wood127). These trials establish the clinical efficacy of specific vascular–immune combinations but do not identify the TEC states responsible for response or resistance.

Vascular normalization and immune activation are mechanistically interconnected but are not equivalent outcomes (Refs Reference Subudhi, Lee and Jain68, Reference Kuo, Khan and Kerbel121). Not all vascular–immune combinations provide comparable benefit because response may depend on baseline endothelial state, treatment intensity, the timing of immune-checkpoint blockade relative to the vascular response, and the extent of hypoxia or immune exclusion (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Subudhi, Lee and Jain68, Reference Kuo, Khan and Kerbel121). Angiogenic, immunosuppressive, metabolically adapted, inflamed, and immune-permissive endothelial states may therefore respond differently to the same regimen (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Lanitis, Irving and Coukos120). Radiotherapy may further alter this balance by promoting immune priming while also increasing VEGF expression and vascular dysfunction, as discussed in section “Combination with radiotherapy and the bridge to radio-immunotherapy” (Ref. Reference Hartley, Ebert and Cook159). Future studies should distinguish patients with angiogenic, immune-excluding vasculature from those with inflamed or immune-permissive endothelial states and should test whether state-informed patient selection and treatment timing improve clinical outcomes (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Subudhi, Lee and Jain68, Reference Lanitis, Irving and Coukos120).

Combination with tumour cell-targeted therapy

Simultaneous targeting of tumour-cell-intrinsic drivers and tumour vasculature may suppress complementary mechanisms of cancer progression. Driver-targeted therapies inhibit oncogene-dependent survival and proliferation, whereas vascular-directed interventions can modify perfusion, barrier function, angiocrine signalling, immune-cell access, and metastatic dissemination (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Guelfi, Hodivala-Dilke and Bergers69, Reference Coppé, Mori, Pan, Yau, Wolf, Ruiz-Saenz, Brunen, Prahallad, Cornelissen-Steijger, Kemper, Posch, Wang, Dreyer, Krijgsman, Lee, Chen, Peeper, Moasser, Bernards and van ‘t Veer162–Reference Shaw, Dwivedi, Bhattacharya, Mukherjee and Rao165). The rationale for combining these approaches is strongest when the oncogenic driver also promotes VEGF production, angiogenic signalling, or endothelial remodelling (Refs Reference Lim, Jung, Kim, Kim, Lee, Park, Choi, Ji, Yeo and Lee166, Reference Wang, Zeng, Zhu and Song167).

EGFR-mutant NSCLC provides a representative example. Erlotinib, an EGFR tyrosine-kinase inhibitor, combined with the anti-VEGF-A antibody bevacizumab prolonged progression-free survival compared with erlotinib alone, although an overall-survival benefit has not been consistently demonstrated (Refs Reference Kawashima, Fukuhara, Saito, Furuya, Watanabe, Sugawara, Iwasawa, Tsunezuka, Yamaguchi, Okada, Yoshimori, Nakachi, Seike, Azuma, Kurimoto, Tsubata, Fujita, Nagashima, Asai, Watanabe, Miyazaki, Hagiwara, Nukiwa, Morita, Kobayashi and Maemondo168, Reference Li, Li, Zhang and Li169). EGFR-driven tumour cells can increase VEGF production, providing a biological rationale for simultaneous inhibition of tumour-cell signalling and vascular support (Refs Reference Lim, Jung, Kim, Kim, Lee, Park, Choi, Ji, Yeo and Lee166, Reference Wang, Zeng, Zhu and Song167). However, the clinical benefit of this combination does not establish that bevacizumab selectively targets or durably reprograms a defined TEC state (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Guelfi, Hodivala-Dilke and Bergers69).

The success of tumour–vascular combinations is therefore likely to depend on biological coherence between the tumour driver and the vascular phenotype. EGFR or HER2 activation may enhance angiogenic signalling in some tumours, whereas the relationship between oncogenic signalling and endothelial dependence may be weaker or more heterogeneous in others (Refs Reference Lim, Jung, Kim, Kim, Lee, Park, Choi, Ji, Yeo and Lee166, Reference Wang, Zeng, Zhu and Song167, Reference Agostinetto, Curigliano and Piccart170, Reference Mou, Li, Zheng, Meng and Tang171). Future studies should integrate tumour genotype with baseline and treatment-induced vascular features, including angiogenic activity, perfusion, endothelial immune phenotype, and evidence of vascular normalization. Such dual stratification may identify patients whose tumours are simultaneously dependent on a targetable oncogenic driver and a therapeutically modifiable vascular state (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Guelfi, Hodivala-Dilke and Bergers69, Reference Shaw, Dwivedi, Bhattacharya, Mukherjee and Rao165).

Endothelial plasticity and multidimensional resistance

Resistance to vascular-directed therapy is an ecosystem-level process rather than the consequence of a single pathway alteration. It may involve five interconnected dimensions: signalling bypass, hypoxic and metabolic adaptation, selection or transition of TEC states, myeloid–stromal–pericyte support, and alternative perfusion mechanisms (Table 5). These mechanisms can coexist, vary across tumour regions, and evolve during treatment.

Table 5.

Mechanisms of resistance to vascular-directed therapy

Table 5. Long description

The table consists of five columns and six rows of data.

Row 1: Resistance dimension is Signaling bypass. Pathways include F G F / F G F R, P D G F / P D G F R, H G F / M E T, A X L, and Ang / Tie 2. Effect is compensatory angiogenesis. Strategy is to combine V E G F inhibition with mechanism-based targeting of compensatory pathways.

Row 2: Resistance dimension is Hypoxic and metabolic adaptation. Pathways include H I F signaling, glycolysis, and P H G D H-dependent serine metabolism. Effect is promoting endothelial survival and alternative angiogenic signaling. Strategy is to optimize treatment intensity and target validated hypoxic or metabolic pathways.

Row 3: Resistance dimension is T E C heterogeneity and plasticity. Pathways include tip/stalk, quiescent, immunomodulatory, and metabolically adapted states. Effect is selectively preserving or enriching treatment-resistant T E C states. Strategy is to apply longitudinal and spatial profiling and develop state-specific interventions.

Row 4: Resistance dimension is Myeloid and stromal support. Pathways include T A Ms, neutrophils, pericytes, and fibroblast-lineage cells. Effect is providing V E G F-independent angiogenic signals and remodeling the extracellular matrix. Strategy is to combine vascular therapy with selected myeloid- or stromal-targeted approaches.

Row 5: Resistance dimension is Vessel co-option. Pathways include tumor invasion along pre-existing host vessels. Effect is maintaining perfusion without sprouting angiogenesis. Strategy is to identify co-option-dominant tumors and target invasion or tissue-specific dependencies.

Row 6: Resistance dimension is Vasculogenic mimicry. Pathways include tumor-cell plasticity and matrix remodeling. Effect is forming non-endothelial, vessel-like perfusion channels. Strategy is to target validated V M-associated tumor-cell programs.

Abbreviations: HIF: hypoxia-inducible factor; PHGDH: phosphoglycerate dehydrogenase; TAM: tumour-associated macrophage; TEC: tumour endothelial cell; VEGF: vascular endothelial growth factor; VM: vasculogenic mimicry.

Clinical translation challenges in precision vascular targeting

Beyond biological resistance, clinical translation of precision vascular targeting is constrained by four interrelated challenges: limited target specificity, spatial and temporal endothelial heterogeneity, inadequate pharmacodynamic biomarkers, and uncertainty in dose and treatment scheduling.

First, genuinely TEC-selective targets are uncommon. Markers enriched in tumour vessels may also be expressed by normal endothelial cells, pericytes, fibroblasts, hematopoietic populations, or malignant cells (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Guelfi, Hodivala-Dilke and Bergers69). Candidate targets should therefore be evaluated for cell-type-resolved protein expression, luminal accessibility, internalization capacity, interpatient prevalence, stability during treatment, and distribution across essential normal organs (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Guelfi, Hodivala-Dilke and Bergers69). Spatial transcriptomics and proteomics may assist target discovery, but transcript enrichment or predicted ligand–receptor interactions cannot replace protein-level and functional validation (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Chen, Larsson, Swarbrick and Lundeberg184). Toxicity caused by direct engagement of a candidate target should also be distinguished from the class effects of systemic VEGF-pathway inhibition, including hypertension, proteinuria, bleeding, thrombosis, and impaired wound healing (Refs Reference Dilli Babu, Singh, Thota, Duhan, Sainatham, Gill, Raghavakurup, Tantry, Bliden and Gurbel185–Reference Garcia, Hurwitz, Sandler, Miles, Coleman, Deurloo and Chinot188).

Second, target expression varies across space and time. Endothelial cells in hypoxic cores, invasive margins, metastatic niches, and treatment-altered stromal regions may differ in angiogenic activity, barrier function, metabolism, and immune regulation (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Guelfi, Hodivala-Dilke and Bergers69, Reference Chen, Larsson, Swarbrick and Lundeberg184, Reference Ji, Rubin, Thrane, Jiang, Reynolds, Meyers, Guo, George, Mollbrink, Bergenstråhle, Larsson, Bai, Zhu, Bhaduri, Meyers, Rovira-Clavé, Hollmig, Aasi, Nolan, Lundeberg and Khavari189, Reference Berglund, Maaskola, Schultz, Friedrich, Marklund, Bergenstråhle, Tarish, Tanoglidi, Vickovic, Larsson, Salmén, Ogris, Wallenborg, Lagergren, Ståhl, Sonnhammer, Helleday and Lundeberg190). A target detected in one biopsy may be absent from other clinically relevant vascular compartments, while treatment may reduce target expression, induce state transitions, or select resistant TEC populations (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Guelfi, Hodivala-Dilke and Bergers69). Target assessment should therefore incorporate spatially resolved sampling and, where feasible, longitudinal analysis rather than relying on a single pretreatment bulk specimen (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Chen, Larsson, Swarbrick and Lundeberg184).

Third, no single biomarker adequately captures both vascular function and endothelial identity (Refs Reference Subudhi, Lee and Jain68, Reference da S Mororó, Meira, Bizzo, Altoé, Casotti, Santana, Louro, Louro, Vicente, de Araújo, Groisman, de Carvalho and Louro191). DCE-MRI and other perfusion- or permeability-based measurements can provide indirect functional vascular readouts but cannot identify the endothelial populations responsible for the observed response (Refs Reference da S Mororó, Meira, Bizzo, Altoé, Casotti, Santana, Louro, Louro, Vicente, de Araújo, Groisman, de Carvalho and Louro191, Reference Hoffmann, Masthoff, Kunz, Seidensticker, Bobe, Gerwing, Berdel, Schliemann, Faber and Wildgruber192). Conversely, tissue-based molecular markers may define endothelial states without demonstrating improved perfusion, oxygenation, barrier function, or drug delivery (Refs Reference Subudhi, Lee and Jain68, Reference da S Mororó, Meira, Bizzo, Altoé, Casotti, Santana, Louro, Louro, Vicente, de Araújo, Groisman, de Carvalho and Louro191). Clinical assessment will therefore likely require integration of functional imaging with circulating markers and spatially resolved tissue measurements, with each biomarker linked to a prespecified biological process (Refs Reference Subudhi, Lee and Jain68, Reference da S Mororó, Meira, Bizzo, Altoé, Casotti, Santana, Louro, Louro, Vicente, de Araújo, Groisman, de Carvalho and Louro191, Reference Hoffmann, Masthoff, Kunz, Seidensticker, Bobe, Gerwing, Berdel, Schliemann, Faber and Wildgruber192).

Finally, vascular responses are dose- and time-dependent. Normalization may be transient, whereas excessive or prolonged pathway inhibition can induce vessel regression, hypoxia, and impaired delivery of partner therapies (Refs Reference Subudhi, Lee and Jain68, Reference Guelfi, Hodivala-Dilke and Bergers69). Early-phase trials should incorporate serial pharmacodynamic measurements to determine whether treatment produces target engagement, vascular normalization, vascular disruption, or no functional vascular change (Refs Reference Subudhi, Lee and Jain68, Reference da S Mororó, Meira, Bizzo, Altoé, Casotti, Santana, Louro, Louro, Vicente, de Araújo, Groisman, de Carvalho and Louro191). Adaptive scheduling based on measured vascular responses may therefore be more informative than fixed combination schedules (Refs Reference Subudhi, Lee and Jain68, Reference Guelfi, Hodivala-Dilke and Bergers69).

Emerging technologies and a precision vascular-targeting framework

Addressing the translational challenges described above requires biomarkers that distinguish target engagement from functional vascular response. Vessel density alone cannot differentiate normalization from ineffective or destructive vessel pruning because it does not directly measure perfusion, permeability, oxygenation, blood volume, or vessel maturity (Refs Reference Sorensen, Batchelor, Zhang, Chen, Yeo, Wang, Jennings, Wen, Lahdenranta, Ancukiewicz, di Tomaso, Duda and Jain118, Reference Quiros-Gonzalez, Tomaszewski, Golinska, Brown, Ansel-Bollepalli, Hacker, Couturier, Sainz and Bohndiek193–Reference Shrestha, Stern, Zhou, Dunn and Porter197). Functional imaging, including DCE-MRI, contrast-enhanced ultrasound, and photoacoustic imaging, can provide complementary measurements of vascular function (Refs Reference Sorensen, Batchelor, Zhang, Chen, Yeo, Wang, Jennings, Wen, Lahdenranta, Ancukiewicz, di Tomaso, Duda and Jain118, Reference Shrestha, Stern, Zhou, Dunn and Porter197–Reference Kim, Choi, Kim, Kim and Park201). Circulating and tissue-based markers may additionally reflect angiogenic signalling, endothelial injury, extracellular matrix remodelling, hypoxia, pericyte coverage, or endothelial-state change (Refs Reference Sorensen, Batchelor, Zhang, Chen, Yeo, Wang, Jennings, Wen, Lahdenranta, Ancukiewicz, di Tomaso, Duda and Jain118, Reference Shrestha, Stern, Zhou, Dunn and Porter197–Reference Batchelor, Gerstner, Emblem, Duda, Kalpathy-Cramer, Snuderl, Ancukiewicz, Polaskova, Pinho, Jennings, Plotkin, Chi, Eichler, Dietrich, Hochberg, Lu-Emerson, Iafrate, Ivy, Rosen, Loeffler, Wen, Sorensen and Jain199). Each biomarker should therefore be linked to a prespecified biological process, assessed for analytical and test–retest reproducibility, and prospectively evaluated to determine whether biomarker-guided treatment improves clinical outcome (Refs Reference Shrestha, Stern, Zhou, Dunn and Porter197, Reference Dietrich, Correas, Cui, Dong, Havre, Jenssen, Jung, Krix, Lim, Lassau and Piscaglia200).

Single-cell and spatial technologies provide a complementary route to endothelial-state discovery. Single-cell RNA sequencing can identify endothelial programs associated with angiogenesis, immune regulation, extracellular matrix remodelling, and metabolic adaptation, whereas spatial transcriptomics can localize these programs to hypoxic regions, invasive margins, and immune or stromal niches (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Ji, Rubin, Thrane, Jiang, Reynolds, Meyers, Guo, George, Mollbrink, Bergenstråhle, Larsson, Bai, Zhu, Bhaduri, Meyers, Rovira-Clavé, Hollmig, Aasi, Nolan, Lundeberg and Khavari189, Reference Berglund, Maaskola, Schultz, Friedrich, Marklund, Bergenstråhle, Tarish, Tanoglidi, Vickovic, Larsson, Salmén, Ogris, Wallenborg, Lagergren, Ståhl, Sonnhammer, Helleday and Lundeberg190, Reference Moncada, Barkley, Wagner, Chiodin, Devlin, Baron, Hajdu, Simeone and Yanai202). Integration with proteomic, epigenomic, imaging, and ligand–receptor analyses may identify candidate surface targets, metabolic dependencies, and cellular interactions that are not detectable in bulk-tissue measurements (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Horvath and Coscia203).

However, transcriptomic clusters should not be treated as therapeutic entities without orthogonal validation. Tissue dissociation can induce stress-response programs, preferentially deplete fragile or tightly adherent cell populations, and distort estimates of cell-state abundance; limited sampling, batch effects, tumour-type differences, and treatment-induced transitions further reduce cross-study reproducibility (Refs Reference Zeng, Mousa, Nadukkandy, Franssens, Alnaqbi, Alshamsi, Safar and Carmeliet15, Reference Denisenko, Guo, Jones, Hou, de Kock, Lassmann, Poppe, Clément, Simmons, Lister and Forrest204, Reference Luecken and Theis205). Moreover, RNA enrichment does not establish protein abundance, cell-surface localization, luminal accessibility, internalization, causal function, or tumour-to-normal selectivity (Refs Reference Goveia, Rohlenova, Taverna, Treps, Conradi, Pircher, Geldhof, de Rooij, Kalucka, Sokol, García-Caballero, Zheng, Qian, Teuwen, Khan, Boeckx, Wauters, Decaluwé, De Leyn, Vansteenkiste, Weynand, Sagaert, Verbeken, Wolthuis, Topal, Everaerts, Bohnenberger, Emmert, Panovska, De Smet, Staal, Mclaughlin, Impens, Lagani, Vinckier, Mazzone, Schoonjans, Dewerchin, Eelen, Karakach, Yang, Wang, Bolund, Lin, Thienpont, Li, Lambrechts, Luo and Carmeliet52, Reference Corti, Sacchi, Gasparri, Monieri, Anderluzzi, Colombo, Gori, Mondino and Curnis144, Reference Nanda and Croix206).

We therefore propose a four-stage framework for precision vascular targeting.