The Cancer Target That Wasnt: How HER2 and HER3 Broke a Decades-Old Model

For decades, the textbook model was clean: HER2 and HER3 were inseparable partners. HER2, a well-known oncogene amplified in breast and other cancers, couldn't signal on its own. Neither could HER3. Together, they heterodimerized, and that pairing drove the downstream pathways that made cells grow. The model had it in diagrams, in review articles, in drug design documents. It was one of the tidy stories biology told itself about how receptor tyrosine kinases worked.

It was wrong.

A team from the Broad Institute of MIT and Harvard published the evidence in Cell on April 28. Using a new single-molecule imaging technique built around upconverting nanoparticle probes, they watched HER2 and HER3 in living human cells and found something the field hadn't seen before: both receptors form homodimers on their own, without needing each other, and they do it constitutively. The activation model that guided a generation of cancer drug development needs a rewrite.

"We were surprised," Sam Peng, the study's lead and an assistant professor of chemistry at MIT, told MIT News. "The field has long viewed HER2 and HER3 as obligate partners." What Peng and his colleagues observed was different in kind, not just degree: these receptors were pairing with themselves, persistently, in the absence of any external signal.

The finding matters because HER2 is one of the most heavily studied targets in oncology. Drugs like trastuzumab and pertuzumab were designed with the heterodimerization model in mind — they were meant to block HER2's ability to partner with HER3 and kick off the signaling cascade that drives tumor growth. If HER2 homodimerizes on its own, the logic of those drugs deserves another look. Not necessarily that they're wrong, but that the mechanism they've been attributed to may be incomplete.

The same paper also tracked what happens to EGFR — one of the most frequently mutated oncogenes in lung cancer, occurring in roughly 10 to 15 percent of non-small cell lung cancers — when it carries those oncogenic mutations. Normal EGFR receptors pair up when stimulated by an external ligand, hold together for a few minutes, and then let go. It's a transient, regulated event. Mutated EGFR behaves differently: the dimers are more stable, persist longer, and in some cases form without any ligand present at all. The more stabilizing the mutation, the more potent the cancer it drives in patients. This is resistance, finally observable in real time rather than inferred from endpoint assays.

The nanoparticle technique is what made both findings possible. Traditional organic fluorophores used in single-molecule tracking photobleach within seconds under laser excitation — you get a few snapshots, not a movie. Peng's team used upconverting nanoparticles doped with rare-earth ions that continue to luminesce for minutes to hours, allowing continuous observation of receptor movement and interaction across the cell membrane. The team tracked three receptor types simultaneously in the same cell and watched them navigate the surface, find partners, unpair, and find new ones. It's the kind of dynamic, living data that static biochemistry can't produce.

Peng's lab is now planning studies on how the findings might affect drug mechanism research and improving the probes themselves. The technique, described as having potential for drug screening in living cells, remains research-only for now — no commercial path exists yet.

The honest version of that sentence is worth stating plainly: this is a tool in the hands of biologists, not a tool in the hands of clinicians. It generates hypotheses and disrupts models. It does not yet generate drugs. The revised HER2/HER3 activation model will take time to filter into the drug design process, and whether it changes anything about how existing HER2-targeted therapies work will require years of follow-on work. The biology is interesting; the clinical translation is not here yet.

That's the right place to hold this story. The nanoparticle imaging platform is genuine and the findings are real, but the paper is a method paper with biological implications, not a clinical result. What it tells us is that the map we've been using for two major cancer targets was incomplete, and that a technology capable of watching receptors in real time can expose gaps that snapshot biochemistry missed. For anyone working on HER2-directed therapies or EGFR resistance mechanisms, the model revision is worth knowing. For everyone else, the story is the microscope — and what happens when you can finally watch the machinery run.