No Ligand, No Kinase, No Problem

HER2 has no ligand. HER3 has no kinase. Together they are the most potent signalling pair in human biology. That single sentence explains most of what is strange about this receptor family and all of what is strange about drugging it. The HER family is usually drawn as four parallel receptors, and that drawing is the source of nearly every misunderstanding about it — because the receptors do not act independently. They pair, and the identity of the pair decides which tyrosines are phosphorylated, which adaptors are recruited, and whether the signal runs to proliferation or to survival.

The EGFR/HER family — 39 clickable nodes from seven ligands and four receptors through the therapeutic antibodies and their epitopes, both downstream arms, the proliferative output and the bypass routes. Open the interactive version to click any protein through to its ELISA kit, biosimilar or In Vivo antibody.

One receptor with no ligand, one with no kinase

EGFR holds a tethered, autoinhibited conformation: its dimerisation arm is folded back against the receptor until ligand binding extends it. That is why EGFR can be blocked by competing with the ligand. HER2 cannot, because its extracellular region already sits in the extended, dimerisation-ready conformation with nothing to compete against. Overexpression alone is therefore sufficient to drive signalling — the only thing limiting HER2 was finding a partner, and there is no threshold of ligand to cross.

HER3 is the mirror-image problem. Its kinase domain is a pseudokinase and contributes essentially no catalytic activity, so it cannot be phosphorylated without a partner. What it does contribute is six docking sites for the PI3K regulatory subunit, more than any other family member. EGFR has none at all and has to borrow Gab1 to reach the survival arm. Put the two deficits together and the HER2–HER3 heterodimer pairs an unlimited dimerisation partner with the strongest PI3K recruiter in the family, which is why it is the most transforming combination in this system and why NRG1 gene fusions are oncogenic on their own, entirely invisible to any EGFR-directed treatment.

HER4 is left out of most HER-family drawings and deserves better. It is the only member that is regularly tumour-suppressive, and it is cleaved to release an intracellular fragment that travels to the nucleus. Its omission is part of why HER-family expression signatures are so hard to interpret.

Seven ligands, and they are not interchangeable

Ligand identity changes the fate of the receptor as well as the strength of the signal. EGF binds with high affinity and drives the receptor towards degradation; TGF-α dissociates in the acidifying endosome, so the receptor recycles instead. The result is a longer, lower signal from an identical binding event, which is about as clean a demonstration as this field offers that duration rather than amplitude is what the cell reads.

Amphiregulin and epiregulin are the predictive pair. High amphiregulin in colorectal cancer means the tumour is genuinely EGFR-driven and an anti-EGFR antibody has something to block; low means it is being driven from elsewhere. Epiregulin binds several HER pairs rather than only EGFR homodimers, and its breadth is why it correlates with response better than receptor expression does. HB-EGF is worth knowing about for a different reason: it is made as a membrane precursor and shed by ADAM17, which is how G-protein-coupled receptors transactivate EGFR with no EGF present anywhere. A substantial share of EGFR signalling in vivo starts at a different receptor entirely. Betacellulin crosses the two halves of the family, binding both EGFR and HER4.

Where the antibody binds decides what it does

This is the part that justifies drawing the antibodies as nodes rather than as footnotes, because the differences between them are mechanistic rather than incremental. Cetuximab, panitumumab and necitumumab all bind domain III and sterically block the ligand-binding site. What separates them is the constant region: cetuximab is chimeric IgG1 and recruits ADCC, panitumumab is fully human IgG2 with almost none, necitumumab is human IgG1. Three antibodies to one epitope region with three different isotypes is the cleanest available dissection of how much of an anti-EGFR antibody's effect is receptor blockade and how much is effector function.

Nimotuzumab makes a different point. It was designed with deliberately intermediate affinity, so it only remains bound where EGFR density is high enough to permit bivalent attachment. Avidity rather than affinity does the targeting, and that single design choice is why it spares normal skin — a rare instance of therapeutic index being engineered at the binding step rather than at the dose.

Pertuzumab is the direct consequence of HER2 having no ligand: it binds domain II, the dimerisation arm, because there is no ligand site to occupy. That is why it adds to trastuzumab rather than duplicating it — the two bind different domains and do different things. Patritumab goes further still and attacks the catalytically dead partner, removing the PI3K docking sites that HER2 cannot supply itself.

The same five adaptors, whichever pair formed

Below the membrane the components are shared, which is both why the pathway is robust and why downstream read-outs are uninformative about which receptor was engaged. Shc binds specific phosphotyrosines on the receptor tail and routes towards Ras; Grb2 is the constitutive bridge to SOS1, the guanine exchange factor that is now directly druggable — SOS1 inhibitors act upstream of KRAS and therefore cover mutations that receptor blockade cannot touch. Gab1 scaffolds PI3K onto a receptor with no PI3K site of its own.

The MAPK arm then runs KRAS to BRAF to MEK1 to ERK1/2. Testing KRAS before giving an anti-EGFR antibody in colorectal cancer is the founding example of a predictive biomarker, and it is a negative one — it identifies who cannot benefit. BRAF mutation has the same consequence in a smaller group, and BRAF inhibition alone fails there because EGFR is rapidly reactivated, a feedback loop that only makes sense once the receptor is drawn above the kinase.

The survival arm runs PI3Kα to AKT to mTOR, restrained by PTEN. PIK3CA and PTEN status are read for the same reason KRAS is: they say whether the lesion is still above them. PTEN loss is the harder of the two to use, because unlike a point mutation it is often partial, which is why it correlates with response less cleanly.

The pathway defends itself at every level

Two feedback loops are worth committing to memory because they explain most unexpected clinical results. ERK phosphorylates SOS1 and the receptor itself as a brake, so inhibiting ERK or MEK releases that brake and the signal rebounds — adaptive resistance appearing within days rather than months. And inhibiting mTOR relieves S6K-mediated suppression of IRS-1, reactivating PI3K upstream of the drug.

The bypass routes are the same lesson at a slower tempo. Amplified MET phosphorylates HER3 directly, restoring the survival arm while the EGFR antibody remains fully bound to a receptor the tumour no longer needs — the drug still works, and the pathway has stopped depending on it. AXL does the same by expression change rather than mutation, which makes it reversible and invisible to sequencing, and associates it with the mesenchymal state. IGF1R offers a third route into PI3K that passes through no HER receptor at all, and STAT3 is activated by EGFR and IL-6 alike, which is one reason an inflamed microenvironment blunts receptor blockade.

SHP2 is where nearly all of these converge above Ras, and that convergence is precisely why it became a drug target: blocking it closes several bypass routes at once rather than one at a time. The alternative strategy is to intervene below all of them, which is the logic of pairing receptor blockade with CDK4 inhibition once cyclin D1 is the limiting node.

What to measure, and what it will and will not tell you

ERK and AKT both converge on the same output, which is convenient clinically and inconvenient experimentally: both arms stabilise MYC, so a change in MYC says the pathway moved without saying which arm moved it. Ki-67 is the practical on-treatment endpoint, and it shifts in biopsies well before tumour size does.

Which reagents this map is made of

It is worth being exact. Only one node on this pathway — EGFR itself — carries a functional-grade In Vivo antibody, and it is the only blue dot on the page. This is a biosimilar map, and the biosimilar range covers it unusually well: nine anti-EGFR antibodies against a single receptor, two on HER2, two on HER3 and one on MET, all supplied as research-grade biosimilars. They carry their clinical identity in the tooltip and deliberately do not carry the In Vivo dot, which is reserved for the functional-grade range.

The practical value of that depth is that the mechanistic comparisons described above are actually runnable. Domain III blockade against dimerisation-arm blockade is cetuximab against pertuzumab. Blockade against effector function is cetuximab against panitumumab. Affinity against avidity is either of those against nimotuzumab. Driver against partner is trastuzumab against patritumab. These are controlled comparisons, not catalogue variety.

Said plainly: everything below the membrane on this map is measurable and not blockable. Every adaptor, both kinase arms and the whole proliferative output are ELISA and research-antibody territory here, which for a signalling pathway is the expected division. PI3K/AKT/mTOR, JAK/STAT and the IL-6 axis each have their own dedicated map in this library and appear here as entry nodes rather than being redrawn.

For research use only; not for use in diagnostic or therapeutic procedures. Explore the full library of interactive pathway diagrams.