Anti-Angiogenics Do Not Starve Tumours. They Open Them.

Anti-angiogenic therapy does not work by starving tumours, and the trials that assumed it did are the reason the field spent a decade disappointed. Cutting off the blood supply was the original pitch: hypoxia stabilises HIF-1α, HIF drives VEGF-A, VEGF-A drives VEGFR2, so block the ligand and the vessels regress. They do regress — and survival barely moves. What actually helps is the opposite of starvation: pruning the worst vessels so the remaining ones work properly, which improves perfusion, drug delivery and, most importantly, immune access. Everything on this map follows from one asymmetry that took the field twenty years to take seriously: the same signal that builds the tumour vessel is the signal that keeps T cells out of it. Click any protein for the matching Assay Genie ELISA kit, In Vivo antibody or biosimilar.

The VEGF axis end to end: oxygen sensing, the ligand family divided by receptor, the escape routes that make anti-VEGF therapy fail, vessel maturation, and the immune-exclusion layer. Open the interactive version to click any protein for its role and the matching validated reagent.

Oxygen is the switch, and VHL is why it works at all

In normal oxygen, prolyl hydroxylases tag HIF-1α and HIF-2α for destruction and VHL carries it out, so HIF protein is nearly undetectable even though the gene is transcribed constantly. This is worth internalising because it dictates how you assay it: HIF regulation is entirely post-translational, so HIF mRNA tells you almost nothing and a Western blot or ELISA on protein tells you everything. Drop the oxygen and protein accumulates within minutes.

Delete VHL and the switch is stuck on permanently, which is exactly what happens in clear-cell renal carcinoma. That single genetic fact explains why kidney cancer became the proving ground for every anti-angiogenic drug: it is the one tumour type where the pathway is constitutively maximal rather than hypoxia-dependent. CA9 and GLUT1 are on the map not as mechanism but as practical readouts — both are HIF targets you can stain in a section when you cannot measure HIF protein in situ.

The ligand family is not redundant — it is divided by receptor

This is where most experimental designs go wrong. The VEGF family looks redundant and is not. VEGF-B binds essentially only VEGFR1 and is therefore close to non-angiogenic. VEGF-C reaches VEGFR3 and drives lymphangiogenesis — the metastatic route rather than the nutrient route. PlGF is VEGFR1-selective. And VEGFR1 itself, despite binding VEGF-A with higher affinity than VEGFR2 does, signals weakly and functions largely as a decoy that sequesters ligand.

VEGFR2 carries almost all of the pro-angiogenic signal. The practical consequence is that an anti-VEGF-A antibody and an anti-VEGFR2 antibody are not interchangeable interventions: bevacizumab removes one ligand from a family with alternatives, whereas ramucirumab closes the channel every family member has to use. Neuropilin-1 is the co-receptor that raises VEGFR2's effective affinity without signalling itself, which is why NRP1 loss blunts responses rather than abolishing them — a pattern that has confused more than one knockout study.

Resistance is built into the network, not acquired by mutation

Anti-VEGF therapy almost never fails because VEGFR2 mutates. It fails because the tumour switches ligand. FGF2 signalling through FGFR1 reconstitutes a full sprouting programme that no anti-VEGF agent touches. PDGF-BB through PDGFR-β recruits pericytes, and a pericyte-covered vessel becomes largely VEGF-independent — α-SMA coverage predicts survival off VEGF far better than vessel count does. And angiopoietin-2 displaces angiopoietin-1 at TIE2, so the ANG-2:ANG-1 ratio matters more than either measured alone.

Those three routes are why the field moved to multi-target kinase inhibitors and to ANG2/VEGF bispecifics, and why measuring VEGF-A alone in a progressing patient is close to uninformative. If you are running a resistance model, the useful experiment is a VEGFR2 block with an FGF2 and ANG-2 time course alongside it: the ligand switch is detectable well before tumour regrowth is.

DLL4 and NOTCH1 do the opposite job and are on the map as a caution. Notch signalling from the tip cell suppresses VEGFR2 in its neighbours so that only one endothelial cell per sprout leads. Block DLL4 and you get dramatically more vessels, all of them non-functional. It is the cleanest illustration on this map that vessel count is not a measure of anything useful.

The vessel is an immune barrier, and this is the part worth measuring

VEGFR2 signalling loosens VE-cadherin junctions. The intuition that leaky vessels deliver drugs better is backwards: leak raises interstitial pressure, collapses perfusion and makes delivery worse. VEGF also induces PD-L1 on endothelium, so the vessel wall itself becomes a checkpoint surface. And — the detail that surprises people — it downregulates ICAM-1, so leukocytes cannot arrest and extravasate at all. This is endothelial anergy, and it is a large part of why some tumours are immunologically cold despite plentiful vasculature.

Add TGF-β signalling through endoglin on proliferating endothelium, and CXCL12 retaining suppressive myeloid cells through CXCR4, and the tumour vasculature stops being a passive conduit and becomes an active exclusion zone. ALCAM (CD166) is the junctional molecule leukocytes actually use to cross, and MMP-2 and MMP-9 clear the basement membrane — MMP-9 also liberating matrix-bound VEGF, so proteolysis and ligand supply are the same event. Thrombospondin-1 is the endogenous brake the tumour has usually silenced, and it is worth measuring before concluding that the pro-angiogenic drive is high.

Two structural nodes are worth a line each because they are what you will actually count. CD31 is the standard endothelial marker — when a paper reports microvessel density, it is almost always counting CD31. And eNOS is the vasodilation and permeability arm of VEGFR2 signalling, which makes it the most direct functional readout that the receptor fired at all, as opposed to merely being occupied.

This is the mechanistic basis for anti-angiogenic plus checkpoint combinations, and it is why those combinations succeeded where anti-angiogenics alone plateaued. Normalise the vessel, restore ICAM-1, let the T cells in.

What to measure, and in what order

The In Vivo angle — and an honest note on this map

This is a more ELISA-led map than most in this library, and that is deliberate rather than a gap. Nine of the 38 nodes carry a functional-grade In Vivo antibody: the VEGFR2 arm both as ramucirumab for human work and as a mouse CD309 antibody for models, PDGFR-α as olaratumab, plus endoglin (CD105), ALCAM (CD166), CXCR4, ICAM-1, PD-L1 and pan-TGF-β. That is enough to block the receptor axis and the immune-exclusion layer in vivo.

The ligands are a different matter, and the map says so rather than implying otherwise. VEGF-A, VEGF-B, VEGF-C, PlGF, FGF2, PDGF-BB, ANG-1, ANG-2 and DLL4 are quantitative analytes here, covered by ELISA rather than by blocking antibodies. Given that the whole resistance story is a ligand-switching story, that is arguably the more useful configuration: you block the receptor and you measure which ligand rose to meet it. The mouse VEGFR2 tool is what makes that a runnable in vivo experiment rather than a correlative one.

The one-line version

Do not think of this pathway as a nutrient supply line. Think of it as a barrier the tumour builds and maintains, using a signal you can block at the receptor and track at the ligand. Measure HIF as protein, never as mRNA; read the ligand panel rather than VEGF-A alone; judge success by perfusion and ICAM-1, not by vessel count.

For the myeloid cells this vasculature retains, see the TAM and MDSC reprogramming map; for the hypoxia-driven purine axis that runs alongside it, see the adenosine axis; and for the TGF-β arm in full, see TGF-β / SMAD signalling.

For research use only. Not for use in diagnostic or therapeutic procedures.