Massive Variant Map of the Insulin Receptor Unlocks Precision Therapy for Severe Insulin Resistance
Scientists catalogued ~14,000 insulin receptor mutations, pinpointing which impair function and which can be rescued by experimental antibody agonists.
Summary
Researchers performed deep mutational scanning of the entire extracellular domain of the human insulin receptor, generating functional scores for roughly 14,000 missense variants. Using a specially engineered mouse cell line stripped of endogenous insulin and IGF-1 receptors, they measured each variant's effect on cell surface expression, insulin binding, and signaling triggered by either insulin or therapeutic monoclonal antibodies. The resulting map strongly correlated with known clinical severity — from lethal Donohue syndrome to milder Type A insulin resistance. Crucially, it revealed which loss-of-function mutations retain enough receptor structure to be activated by non-insulin antibody agonists, opening a rational path to precision therapy. Novel potential gain-of-function variants were also identified, offering new insights into insulin binding dynamics and receptor conformation.
Detailed Summary
Severe insulin resistance caused by mutations in the INSR gene — encoding the insulin receptor — spans a devastating clinical spectrum. Near-complete loss of receptor function produces Donohue syndrome, a condition lethal within the first months of life, while retaining even 10–20% of normal function can permit survival into adulthood with Rabson-Mendenhall syndrome. Because the relationship between receptor function and clinical outcome is steep at the low end, even a modest pharmacological boost to a crippled receptor could be life-changing. Anti-insulin-receptor monoclonal antibodies (mAbs 83-7 and 83-14) have already been shown to activate some severely dysfunctional mutant receptors, but until now it was unknown which of the hundreds of possible INSR mutations would respond to such treatment. This study was designed to answer that question at scale.
The team performed saturation mutagenesis of residues 28–955 of the INSR ectodomain (the entire extracellular, ligand-binding region), generating a barcoded plasmid library covering 15,996 single missense and stop-codon variants — 86% of all theoretically possible substitutions, with 81% of mutations tagged by more than one independent barcode for reliability. This library was introduced into mouse embryo fibroblasts engineered to lack endogenous IGF-1 receptor (Igf1r knockout) and with doxycycline-inducible shRNA knockdown of mouse Insr, providing a clean background in which only the transfected human INSR variant was expressed. Four parallel flow-cytometry-based assays then sorted millions of cells per bin: cell surface expression (antibody binding), insulin binding (fluorescently labeled insulin), and AKT phosphorylation as a proxy for signaling induced by either insulin or the two therapeutic mAbs.
Barcode sequencing across sorted bins produced function scores for each variant. Synonymous variants and wild-type barcodes showed superimposable score distributions (Mann–Whitney p = 0.8 for insulin binding; p = 0.052 for expression), validating the assay's specificity, while missense variants were significantly shifted toward lower scores (p < 10⁻³⁰⁸ for both). Known pathogenic variants — A119V and S350L — scored low as expected. Critically, function scores correlated strongly with clinical syndrome severity: variants from patients with Donohue syndrome clustered at the lowest scores, those from Rabson-Mendenhall syndrome at intermediate values, and Type A IR variants at higher scores, providing a quantitative genotype-phenotype bridge that has been elusive with one-variant-at-a-time in vitro studies.
The antibody-stimulated signaling assays provided the translationally pivotal data layer. By comparing insulin-stimulated versus mAb-stimulated signaling scores for each variant, the map identifies which loss-of-function mutations retain the structural epitope needed for antibody engagement and downstream activation — even when they cannot bind insulin. Several classes of variants emerged: those impaired at the level of expression (receptor never reaches the cell surface), those that reach the surface but cannot bind insulin yet remain activatable by mAb, and those so structurally disrupted that neither insulin nor mAb can signal through them. This stratification directly informs which patients might benefit from non-canonical INSR agonist therapy. The study also flagged novel potential gain-of-function variants with signaling scores above wild-type, offering new hypotheses about receptor activation dynamics and sites of autoinhibition.
The clinical and scientific implications are substantial. For diagnostics, the dataset resolves the growing problem of variants of uncertain significance (VUS) identified in routine genetic testing, especially in patients from underrepresented ethnic groups absent from large population databases. For precision medicine, the mAb-signaling layer provides a rational, pre-clinical basis for patient selection in future trials of antibody-based insulin receptor agonists. Caveats include the use of a mouse fibroblast rather than a human metabolic tissue context, coverage gaps in the most difficult-to-mutagenize regions, and the use of AKT phosphorylation as a single downstream readout which may not capture all signaling branches relevant to metabolic versus mitogenic outcomes. Nonetheless, this is the most comprehensive INSR variant function resource yet produced and will accelerate both rare-disease diagnosis and the development of therapies for one of medicine's most intractable conditions.
Key Findings
- 86% coverage of all possible INSR ectodomain missense variants achieved: 15,996 variants characterized across 928 residues, with 81% tagged by more than one independent barcode
- Synonymous variant scores were statistically indistinguishable from wild-type (insulin binding p = 0.8; expression p = 0.052), confirming assay specificity, while missense variants were significantly shifted lower (p < 10⁻³⁰⁸ for both assays)
- Function scores stratified patients by clinical severity: Donohue syndrome variants scored lowest, Rabson-Mendenhall syndrome variants intermediate, and Type A IR variants highest, creating a quantitative genotype-phenotype map
- A distinct subset of loss-of-function variants that cannot respond to insulin retain the structural epitopes required for activation by therapeutic anti-receptor monoclonal antibodies 83-7 and 83-14, identifying them as candidates for antibody agonist therapy
- Novel potential gain-of-function variants were identified with signaling scores above wild-type, providing new structural hypotheses about insulin receptor activation and autoinhibition
- Variants impaired at cell surface expression, insulin binding, or downstream signaling were resolved as distinct mechanistic classes, enabling precise molecular diagnosis beyond a binary 'pathogenic/benign' classification
- At least 13 million cells were sorted per bin across five parallel assays, providing sufficient statistical power to produce reliable scores for the vast majority of single-amino-acid substitutions in the ectodomain
Methodology
Deep mutational scanning was performed on INSR residues 28–955 (the full ectodomain) using nicking mutagenesis to generate a barcoded plasmid library verified by PacBio long-read sequencing; 80,956 unique barcodes tagged 15,996 variants. The library was integrated via Bxb1 recombinase into a landing pad in Igf1r-knockout mouse embryo fibroblasts with doxycycline-inducible Insr shRNA knockdown, providing a receptor-null background. Five flow-cytometry assays (cell surface expression, insulin binding, mAb binding, insulin-stimulated AKT phosphorylation, mAb-stimulated AKT phosphorylation) sorted ≥13 million cells into four bins each, with barcode frequencies quantified by Illumina sequencing and converted to weighted-average function scores. Statistical comparisons used two-sided Mann–Whitney U tests, with p < 10⁻³⁰⁸ for missense versus wild-type distributions.
Study Limitations
The assays were conducted in mouse embryo fibroblasts rather than human hepatocytes, adipocytes, or skeletal muscle cells, so tissue-specific processing, glycosylation, and signaling context may differ from physiologically relevant settings. AKT phosphorylation captures only one branch of the insulin signaling network and may not fully reflect metabolic versus mitogenic signaling divergence. Coverage gaps remain in difficult-to-mutagenize regions (~14% of variants), and the study cannot fully distinguish primary signaling defects from secondary effects on receptor trafficking or recycling without additional biochemical follow-up.
Enjoyed this summary?
Get the latest longevity research delivered to your inbox every week.
Enter your email to subscribe: