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

Pigs were chosen deliberately. As a biomedical model, swine share striking anatomical and metabolic similarities with humans, particularly in cardiovascular physiology, body fat distribution, and digestive function. Human diet studies are notoriously difficult to control, because participants cheat, under-report, and vary wildly in adherence. Fifteen post-pubertal, non-pregnant, non-lactating gilts, Landrace by Yorkshire crosses averaging around 170 kilograms, were individually penned and paired by weight before being randomly assigned to one of two diets. Seven animals received a conventional, nutrient-balanced maintenance ration built around corn, while eight received a low-carbohydrate, high-fat formulation in which beef tallow supplied the energy normally delivered by grain. Soybean hulls were added to the high-fat diet to match the physical consistency of the control ration, and both diets were formulated to meet National Research Council requirements for swine of this age, sex, and reproductive stage.

The laboratory analysis confirmed the diets differed dramatically. The control diet contained roughly 3.3 percent crude fat and nearly 69 percent non-fiber carbohydrate on a dry matter basis, while the experimental diet contained 22.5 percent crude fat and only about 15.6 percent non-fiber carbohydrate, with neutral detergent fiber soaring to more than 44 percent thanks to the soy hulls. Interestingly, the calculated gross energy of the two rations diverged: the control diet provided approximately 3,993 kilocalories per kilogram, while the high-fat formulation came in at 3,448 kilocalories per kilogram, a discrepancy the authors attribute largely to the indigestible fiber fraction used to dilute the diet. Both rations were fed daily at 1.75 percent of each animal’s body weight, and feed refusals were weighed to calculate actual intake.

The results on body composition were the study’s most eye-catching finding. Control gilts gained weight steadily across the trial, ending roughly seven percent heavier than they started, while the high-fat group maintained a nearly constant body weight throughout. More striking was what happened beneath the skin. Measured by ultrasound adjacent the tenth thoracic vertebra, subcutaneous fat depth fell by 17.7 percent in the high-fat animals over 32 days, while control animals showed a 14 percent increase. At slaughter, the differences inside the body were even more dramatic: control gilts carried 1.74 times more visceral fat, the metabolically dangerous internal belly fat, than their high-fat counterparts, 2,660 grams versus 1,440 grams on average. Crucially, the high-fat animals did not waste away; they simply stopped depositing fat while presumably continuing to build lean tissue, an outcome the researchers interpret as a favorable shift in the ratio of muscle to fat.

Appetite appears to have played a central role. Although both groups were offered the same amount of feed relative to body weight, the high-fat gilts consumed significantly less, averaging 2.36 kilograms per day between days 19 and 32 compared with 3.09 kilograms in the control group. Calculated energy intake told the same story, with the control animals consuming more than 12,300 kilocalories daily against roughly 8,150 for the high-fat group. The authors suggest the animals reached satiety sooner and stopped eating, consistent with human trials reporting appetite suppression on ketogenic diets. They also caution, however, that the enormous soy hull load in the experimental diet may have contributed to early fullness or restricted digestible energy, a confound that future work should disentangle by examining fiber and fat effects on total tract digestibility directly.

Blood chemistry offered reassurance on one of the diet’s most feared risks: acidosis. Blood pH did not differ between treatments, and fasted values for sodium, bicarbonate, and carbon dioxide remained within normal reference ranges, indicating that four weeks on the high-fat regimen did not disrupt acid-base homeostasis. Urine ketone testing revealed that nutritional ketosis was largely absent, with no detectable ketones in the third week and only trace amounts, about 5 milligrams per deciliter, in three animals during the fourth. Yet the body composition changes occurred anyway, suggesting a metabolic shift away from fat storage that operated independently of measurable ketosis. One unexpected trend did emerge: fasting glucose in the high-fat group crept upward over the trial, ending significantly higher than controls at day 32, a pattern the authors tentatively link to early keto-adaptation and the liver’s efforts to protect glycogen reserves when adequate dietary protein fuels gluconeogenesis.

The microscopic analysis of the pancreas produced perhaps the most provocative data. Using immunofluorescent staining for insulin on formalin-fixed tissue sections, the researchers found that the percentage of active beta cells in the high-fat group was 2.81 times greater than in controls, 90.7 percent versus 42.4 percent, and that the area occupied by beta cells within the pancreatic islets was 1.81 times larger. Islet area as a percentage of total pancreatic tissue was also more than doubled. These findings suggest an enhancement of pancreatic architecture and insulin-producing capacity, which stands in contrast to earlier rodent studies reporting that long-term ketogenic diets can shrink beta-cell mass and impair glucose tolerance. The discrepancy may reflect species differences, the short feeding duration, or the moderate degree of ketosis actually achieved.

In skeletal muscle, the story was more selective. The researchers quantified insulin receptor presence by measuring the relative fluorescence intensity of antibody-labeled receptor proteins in cross-sections of the longissimus thoracis, the semimembranosus, and the kidney. Only the semimembranosus, a powerful muscle of locomotion in the hind limb, showed a statistically significant increase in insulin receptor density in the high-fed animals, with relative intensity of 32.8 versus 21.7 in controls. The longissimus showed a numerical but non-significant rise, and the kidney showed no difference. The authors propose that upregulated insulin receptors on muscle fibers may contribute to a muscle-sparing effect, supporting the anabolic influence of insulin on myofibrillar protein synthesis while ketone bodies spare amino acids from being diverted into gluconeogenesis. Notably, proximate analysis found no difference in crude fat content of the liver or longissimus muscle between groups, and although the high-fat animals had heavier livers relative to body weight, there was no biochemical indication of early fatty liver disease.

The researchers are candid about their limitations. The trial used only one sex, mature females, because intact males were unavailable and castrated males are considered irrelevant to human medicine, restricting generalizations to women. The ex vivo tissue analyses relied on just three biological replicates per group, and the 32-day feeding period is short by dietary-intervention standards; some human evidence suggests full metabolic benefits of ketogenic adherence may require twelve weeks or more. Statistical tendencies, including the glucose trajectory and the energy balance calculation, should be viewed with caution given the small sample size. Still, the authors argue the data provide preliminary justification for pursuing the metabolic impact of low-carbohydrate, high-fat diets on obesity using swine as a model, particularly the observation that animals lost fat mass without losing body weight.

For the millions of people experimenting with ketogenic eating, the study offers a cautiously encouraging signal wrapped in important caveats. In a tightly controlled animal model, replacing carbohydrate energy with animal fat reduced the most harmful fat depots, preserved body weight, enhanced markers of pancreatic insulin-producing capacity, and increased insulin receptor expression in a major locomotor muscle, all without measurable ketosis and without disturbing blood chemistry. Whether the same remodeling occurs in humans over months and years, and whether the fiber dilution that accompanied the high-fat ration contributed to the appetite effects, remains to be determined. What the pig study demonstrates is that the diet’s effects on body composition and insulin biology are real, measurable at the tissue level, and more nuanced than the popular narrative of ketosis alone would suggest.

Subject of Research: Effects of a low-carbohydrate, high-fat diet on muscle insulin receptor concentration and body composition in a swine biomedical model

Article Title: Does a low‐carb, high‐fat diet change muscle insulin receptor concentration and body composition in a swine biomedical model?

Article References: Acosta‐Castellanos, N., Tiede, S., Byrd, C. J., Borowicz, P., Young, J. M., & Berg, E. P. (2026). Does a low‐carb, high‐fat diet change muscle insulin receptor concentration and body composition in a swine biomedical model?. Physiological Reports, 14(19), Article e71124. https://doi.org/10.14814/phy2.71124

Image Credits: AI Generated

DOI: 10.14814/phy2.71124

Keywords: ketogenic diet, low-carbohydrate high-fat diet, insulin receptor, body composition, swine model, pancreatic beta cells, visceral fat, beef tallow, nutritional ketosis, skeletal muscle, metabolism, Physiological Reports