The insulin receptor is a receptor tyrosine kinase, but it breaks the rules that govern many of its relatives. Rather than dimerizing upon ligand binding, IR exists as a preformed dimer at the cell surface, held in an autoinhibited conformation in which the ectodomain geometry keeps the intracellular kinase domains separated and poorly oriented for the trans-phosphorylation that ignites signaling. Insulin activates the receptor through a multistep trajectory involving two binding sites: site 1, formed by the leucine-rich repeat domain L1 of one protomer together with the alpha-chain C-terminal segment of the opposing protomer, and site 2, located on the lateral surface of the first fibronectin type III domain. When the receptor is fully liganded, up to four insulin molecules can engage it, stabilizing a T-shaped conformation that brings the kinase domains into productive proximity.

Once activated, the receptor autophosphorylates across its activation loop, juxtamembrane region, and C-terminal tail, creating a docking platform for adaptor proteins. The classical branches follow: insulin receptor substrate proteins recruit PI3K to drive the AKT pathway and metabolic regulation, while the adaptor Shc channels signaling into the MAPK cascade, more closely tied to mitogenic and proliferative outputs. Crucially, these are not simply parallel wires but competing engagement routes, and their relative utilization shapes the overall signaling balance. That balance, the review emphasizes, can be distorted in disease, most famously in selective insulin resistance, where the metabolic AKT arm fails while the mitogenic MAPK arm persists, a combination implicated in endothelial dysfunction and in the uncoupling of hepatic glucose production from lipogenesis in type 2 diabetes.

Beyond pathway choice, the review highlights two underappreciated dimensions of IR regulation: trafficking and spatial organization. Internalized receptors were long thought to be simply silenced and recycled, but accumulating evidence shows that endosomal IR can remain signaling-competent, and that insulin-stimulated AKT2 phosphorylation depends on specific endosomal compartments enriched in the protein WDFY2. Genetic mouse models drive the point home. Disrupting the interaction between IR and the spindle checkpoint protein MAD2 delays receptor endocytosis, reduces insulin clearance, elevates circulating insulin, and reshapes both membrane and intracellular signaling, with measurable consequences for lipid metabolism. Conversely, hepatic overexpression of EphB4 drives receptor degradation and insulin resistance, while loss of the beta-cell scavenger inceptor improves glucose tolerance. Precise control of receptor traffic, these studies show, is required for systemic metabolic homeostasis.

Super-resolution imaging adds yet another layer: IR dimers are organized into nanoclusters roughly 74 nanometers across, and insulin stimulation modestly enlarges these assemblies without fundamentally reorganizing them. To test whether ligand architecture matters, researchers built a DNA origami platform presenting insulin at defined valency and spacing. Multivalent insulin NanoRods bound the receptor more tightly, dissociated more slowly, and induced stronger autophosphorylation and downstream transcriptional responses than monovalent insulin. Receptor activation, in other words, is sensitive to nanoscale geometry, not merely to ligand identity, and even the temporal pattern of insulin exposure matters: pulsatile delivery elicits stronger hepatic signaling than constant infusion at an equivalent dose.

The most striking section of the review concerns engineered insulin mimetics, which the authors treat not as insulin substitutes but as molecular probes exposing signaling states inaccessible to the native hormone. Phage display screens originally identified the site 1 and site 2 hotspots, and fusing peptides targeting them in different orientations produced either agonists or antagonists. The 31-amino-acid peptide S597 activates the receptor through a non-native geometry, cross-linking site 1 and site 2 across protomers, and preferentially stimulates metabolic over mitogenic signaling, with in vivo data suggesting possible benefits in atherosclerosis. Reversing the fusion orientation yields antagonists such as S961, which cryo-electron microscopy reveals trap the receptor in an intermediate conformation between the inactive and active states. Fusion geometry, not binding affinity, determines whether a synthetic ligand stabilizes an active, biased, or antagonistic configuration.

Computational protein design has now made that principle fully explicit. De novo designed binders, roughly 7 kilodaltons each and directed against individual extracellular domains, were fused in defined orientations with tunable linker flexibility. Reversing the topology flipped the function: site 2 to site 1 constructs act as agonists, while site 1 to site 2 constructs act as antagonists. Linker rigidity further tunes the output. Rigid constructs such as RF-405 stabilize a compact, fully active conformation and signal in a balanced, insulin-like fashion, whereas flexible constructs such as S2-F1-S1 produce partial autophosphorylation, attenuated mitogenic signaling, and altered trafficking. In animals, these designer agonists lower glucose with slower onset but longer duration than insulin, bypass certain disease-causing IR mutations, and show reduced stimulation of cancer cell proliferation compared with the native hormone.

Aptamers and antibodies round out the modulator toolkit. The DNA aptamer IR-A43 does not activate the receptor alone but potentiates insulin signaling by stabilizing an asymmetric, singly occupied conformation, while IR-A62 behaves as a concentration-dependent allosteric modulator that can lower glucose even under insulin-deficient conditions. Monoclonal antibodies such as 83-7 and 83-14, and the engineered XMet series, demonstrate that epitope selection can tune the receptor from activation to inhibition: XMetA acts as a partial agonist with preferential AKT signaling, XMetS potentiates insulin action, and XMetD reverses hypoglycemia in hyperinsulinemic mouse models. Together these ligands show that IR activation can be synthetically programmed.

The disease implications are broad. In cancer, hyperinsulinemia is associated with increased risk and progression, tumors preferentially express the IR-A isoform that binds IGF2 and sustains mitogenic signaling, and the super-mitogenic analog Asp(B10), which induced mammary tumors in rodents, stands as a cautionary example of what mitogenic bias can do. In polycystic ovary syndrome, skeletal muscle shows post-receptor defects centered on the IRS-PI3K-AKT axis, while ovarian adaptor remodeling promotes androgen excess. In Alzheimer disease, postmortem tissue reveals brain insulin resistance with preserved signaling proteins but elevated inhibitory IRS1 serine phosphorylation, and the genetic risk factor APOE4 impairs neuronal signaling by trapping the receptor in endosomes. Severe insulin resistance syndromes caused by germline IR mutations, meanwhile, may one day be treated with mimetics that activate the receptor through mechanisms distinct from insulin.

The authors are candid about the challenges ahead: it remains unclear which signaling biases are beneficial versus pathological, how bias is maintained in complex physiological environments, and how trafficking, ligand geometry, and adaptor selection integrate in vivo. Translational hurdles, from peptide stability to CNS delivery, are substantial, though the engineering playbook proven by GLP-1 receptor agonists offers a template. Still, the conceptual shift is hard to miss. If insulin action is a programmable network rather than a simple switch, then the future of therapy may lie not in replacing insulin but in reprogramming its receptor, selectively and precisely, in the tissues and disease contexts where it matters most.

Subject of Research: Non-canonical insulin receptor signaling and the design of insulin mimetics

Article Title: Insulin mimetics and non-canonical insulin receptor signaling: mechanisms and disease relevance

Article References: Park, J., & Choi, E. (2026). Insulin mimetics and non-canonical insulin receptor signaling: mechanisms and disease relevance. Experimental & Molecular Medicine. https://doi.org/10.1038/s12276-026-01862-5

Image Credits: AI Generated

DOI: 10.1038/s12276-026-01862-5

Keywords: insulin receptor, insulin mimetics, biased agonism, receptor trafficking, endosomal signaling, protein design, type 2 diabetes, insulin resistance, cancer, PCOS, Alzheimer disease, receptor tyrosine kinase