Introduction

Insulin is a peptide hormone produced by the beta cells of the pancreatic islets. Upon binding to the insulin receptor (IR) in peripheral tissues, insulin activates the receptor, triggering intracellular signaling cascades through the phosphoinositide 3-kinase (PI3K)–AKT and mitogen-activated protein kinase (MAPK) pathways1,2,3. Therefore, insulin-mediated IR activation plays a pivotal role in regulating glucose homeostasis, protein synthesis, lipogenesis, cell growth, proliferation and development4,5. Dysregulation of IR signaling leads to chronically elevated blood glucose levels, resulting in diabetes mellitus6. Individuals with diabetes are prone to a wide range of complications and have an increased risk of cancer7.

The IR is a homodimer, with each protomer consisting of an α chain and a β chain8. The α-subunit and N-terminal portion of the β-subunit are extracellular, whereas the β-subunit spans the membrane and contains a cytoplasmic tyrosine kinase domain responsible for insulin-dependent phosphorylation. The α-subunit mediates ligand binding via a leucine-rich repeat (L1), a fibronectin type III domain (FnIII-1) and an α-helical C-terminal domain (αCT). Biochemical and structural studies of insulin binding to IR suggest that each IR protomer contains two insulin-binding sites—primary site (site 1) and secondary site (site 2)9,10,11,12,13. Site 1 is formed by the L1 and αCT′ domains (with the prime indicating structural elements from the opposite protomer), and interactions at this site are crucial for insulin binding and IR activation12,13. Site 2 comprises residues on the surface of the FnIII-1 β-sheet9,10,11.

Structural studies have revealed various conformations of the extracellular domain of IR in both the apo state and when bound to different numbers of insulin molecules9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24. In the apo state, the two αβ monomers form a symmetric, Λ-shaped dimer in an antiparallel orientation, with multiple interfaces between protomers17,18. A single insulin molecule binds simultaneously to the L1 and αCT′ domains at site-1 of the IR, resulting in an asymmetric gamma (Γ)-shaped structure12,13,21,24,25. In this Γ-shaped structure, the L1, CR, L2 and FnIII-1 domains in one protomer form an upshifted ‘head’, while the corresponding domains in the opposite protomer form a downshifted head. The stalks are composed of the FnIII-1, FnIII-2 and FnIII-3 domains in both protomers. Upon insulin binding, the L1 and αCT′ domains at site 1 are lifted toward FnIII-1′ relative to their position in the apo state. When multiple insulin molecules bind, the IR adopts a series of tilted T-shaped structures, reaching a fully T-shaped conformation when all four insulin-binding sites are occupied9,10,11,18,22,23,24. Insulin binds to IR with negative cooperativity, and under physiological conditions, the binding of a single insulin molecule is sufficient to fully activate the receptor26,27.

Although insulin has been the most commonly issued treatment for diabetes, considerable effort has been directed toward developing alternative therapeutics. These include peptides that selectively activate the AKT signaling pathway20,21,22,28, antibodies that enhance IR activation in the presence of insulin29,30, and DNA aptamers24,31,32. In a previous study, we developed the A62 aptamer, a selective partial agonist for IR31. The A62 aptamer induces monophosphorylation of the receptor by binding at the interface between the L1 and FnIII-1′ domains of the opposite IR protomer24. Two A62 aptamers bind symmetrically to the IR dimer, inducing an arrowhead-shaped conformation. In addition, a single A62 aptamer can bind IR together with insulin. In both cases, the membrane-proximal ends of the FnIII-3 domains are separated by ~40 Å, which may explain the aptamer’s ability to induce only monophosphorylation of the IR.

One major obstacle to the clinical use of aptamers is their small molecular size, which can limit their stability in vivo. To address this issue, we aimed to enhance the in vivo stability of the A62 aptamer agonist for IR by creating dimeric forms. We crosslinked two A62 molecules using linkers of various lengths. The resulting A62 dimers, in which two A62 monomers (A62M) were connected by linkers ranging from 8 to 19 nucleotides, induced full phosphorylation of the IR. To understand the basis of this aptamer-induced full phosphorylation, we determined the structure of the IR bound to an A62 dimer linked by eight thymine nucleotides (A62D-8T) using cryo-electron microscopy (cryo-EM). Structural analysis revealed three distinct IR conformations: arrowhead-shaped, pseudo-arrowhead-shaped and pseudo-gamma-shaped, which we refer to as IRarrowhead, IRpseudo-arrowhead and IRpseudo-gamma, respectively. Unlike the symmetric IRarrowhead dimer, IRpseudo-arrowhead and IRpseudo-gamma form asymmetric dimers. In these configurations, each A62M module of the A62D binds to the IR dimer, bringing the IR protomers into close proximity to facilitate trans-phosphorylation. Our findings suggest a strategy for inducing full IR phosphorylation by crosslinking aptamer modules that bind to the IR dimer, thereby promoting intermolecular trans-phosphorylation.

Materials and methods

Reagents and antibodies

The aptamers used in this study were synthesized by Aptamer Science. The anti-IR β-subunit (CT-3) antibody was purchased from Santa Cruz Biotechnology. Antibodies specific for phosphorylated IR at tyrosine residues Y1146, Y1150 and anti-phospho-tyrosine (4G10) were purchased from Millipore. Additional antibodies against phosphorylated IR at residues Y1322, Y1316, Y1150/Y1151 and Y960 were purchased from Invitrogen (Carlsbad). The anti-phospho-ERK1/2 (T202/Y204), anti-phospho-AKT (T308) and anti-phospho-AKT (S473) antibodies were purchased from Cell Signaling Technology. Secondary antibodies, goat anti-rabbit IgG and anti-mouse IgG conjugated to DyLight 800 were also purchased from Invitrogen.

IR phosphorylation assay

Rat-1 cells overexpressing human IR (Rat-1/hIR) were cultured in high-glucose Dulbecco’s modified Eagle medium supplemented with 10% (v/v) fetal bovine serum (FBS; Gibco) and antibiotic–antimycotic solution (Gibco). All cells were incubated at 37 °C in a humidified atmosphere with 5% CO2 until used for experiments. Aptamers and insulin were prepared in Krebs–Ringer HEPES buffer (25 mM HEPES, pH 7.4) containing 120 mM NaCl, 5 mM KCl, 1.2 mM MgSO4, 1.3 mM CaCl2 and 1.3 mM KH2PO4. To reconstitute the tertiary structure, all aptamer samples were heated at 95 °C for 5 min and then slowly cooled to room temperature. Before stimulation with insulin or aptamers, cells were serum-starved for 3 h. They were then treated with the indicated concentrations of insulin or aptamers for the specified duration. After stimulation, the cells were washed with cold phosphate-buffered saline and lysed in a cell lysis buffer (50 mM Tris–HCl, pH 7.4) containing 150 mM NaCl, 1 mM EDTA, 20 mM NaF, 10 mM glycerophosphate, 2 mM Na3VO4, 1 mM phenylmethylsulfonyl fluoride, 10% glycerol, 1% Triton X-100, 0.1% SDS, 0.5% sodium deoxycholate and a protease inhibitor cocktail. The lysates were clarified by centrifugation at 14,000 rpm for 15 min at 4 °C, and the supernatant was mixed with 5× Laemmli sample buffer. Proteins were separated by SDS–PAGE, transferred to a nitrocellulose membrane and blocked with 5% skim milk for 30 min. The membrane was then probed with the appropriate primary antibody overnight at 4 °C. Protein detection was performed using the Odyssey infrared imaging system (LI-COR).

Expression and purification of IR and A62D complex

To express apo-IR, stable cells were cultured in suspension in Freestyle293 medium at 37 °C with 8% CO2. When the cell density reached ~4.0 × 106 cells/ml, cells were collected by centrifugation and resuspended in a buffer containing 20 mM HEPES (pH 7.5), 400 mM NaCl and 5% glycerol. The cells were lysed on ice using a Dounce homogenizer (Kimble), followed by solubilization in a buffer consisting of 20 mM HEPES (pH 7.5), 400 mM NaCl, 1% (w/v) n-dodecyl β-d-maltoside (DDM; Anatrace), 0.1% (w/v) cholesterol hemisuccinate (CHS; Sigma), 5% glycerol and a protease inhibitor cocktail (Roche). Solubilization was performed for 2 h on ice. Solubilized membranes were then isolated via ultracentrifugation using a Ti45 rotor (Beckman) at 130,000g for 1 h at 4 °C.

The supernatant was collected and incubated with anti-Flag affinity G1 resin (GenScript) for 2 h at 4 °C. The resin was batch-washed using a washing buffer containing 20 mM HEPES (pH 7.5), 400 mM NaCl, 5% glycerol, 0.1% DDM and 0.01% CHS. Proteins were eluted with a buffer composed of 20 mM HEPES (pH 7.5), 400 mM NaCl, 5% glycerol, 0.03% DDM, 0.003% CHS and 0.4 mg/ml Flag peptide. To remove the Flag peptide, eluted proteins were concentrated using an Amicon Ultra centrifugal device (100 kDa cutoff; Millipore) and diluted in a buffer containing 20 mM HEPES (pH 7.5), 105 mM NaCl, 5% glycerol, 0.03% DDM, 0.003% CHS, 5 mM KCl and 5 mM MgCl2. The proteins were concentrated again using an Amicon Ultra device (100 kDa cutoff; Millipore).

To form the complex with the aptamer, the proteins and preactivated A62D aptamer were mixed at a 1:1 molar ratio and incubated for 1 h on ice. The mixture was then injected onto a Superose 6 10/300 column preequilibrated with a buffer containing 20 mM HEPES (pH 7.5), 105 mM NaCl, 5 mM KCl, 5 mM MgCl2, 0.03% DDM and 0.003% CHS. Eluted fractions were pooled and concentrated to 8 mg/ml using a Vivaspin device (100 kDa cutoff; GE Healthcare) for subsequent cryo-EM analysis.

Cryo-EM sample preparation and data collection

Cryo-EM grids were prepared by applying 3 μl of the sample to glow-discharged holey carbon grids (C-flat 1.2/1.3 Au, 400-mesh; EMS). The grids were then plunge-frozen in liquid ethane using a Vitrobot Mark IV (Thermo Fisher Scientific) with a blot force of 4 for 5 s at 100% humidity and 4 °C. Data collection was performed on a Titan Krios G4 microscope operated at 300 kV, equipped with a Gatan K3 summit direct electron detector in fast mode at a nominal magnification of 79,000×. A total of 9,096 micrographs were acquired. The dataset was collected with a pixel size of 1.0902 Å and a defocus range of −1.0 to −2.0 μm. Each micrograph was dose-fractionated over 50 frames, with a total accumulated dose of 50 electrons/Å2.

Data processing for the IRarrowhead structure

Dose-fractionated image stacks from 9,096 movies were imported into CryoSPARC v4.5.333. The images underwent dose-weighting using patch motion correction, followed by contrast transfer function (CTF) parameter calculation via patch CTF estimation. Low-resolution micrographs were excluded, resulting in 8,664 micrographs being selected for further data processing. A total of 1,312,592 particles were extracted using a template picker. After multiple rounds of two-dimensional (2D) classification, 38,995 particles were used for ab initio reconstruction to generate an initial 3D model. Heterogeneous refinement identified 11,658 particles from one class with an arrowhead shape, which were subsequently used to train Topaz34.

The trained model was then applied to extract 626,038 particles. After 2D classification, 48,970 particles exhibiting an arrowhead shape were subjected to homogeneous refinement and nonuniform refinement, as well as local refinement. This processing yielded a map with a global resolution of 5.02 Å, based on a Fourier shell correlation (FSC) criterion of 0.143.

Data processing for the IRpseudo-arrowhead structure

To process the pseudo-arrow conformation, we used the same dataset as for the arrowhead shape. A total of 1,421,993 particles were extracted using a template picker. After multiple rounds of 2D classification, 147,154 particles were used for ab initio reconstruction to generate an initial 3D model. Heterogeneous refinement identified 15,052 particles from one class with an asymmetric shape, which were subsequently used to train Topaz34.

The trained model was then applied to extract 1,886,306 particles. After ab initio reconstruction and heterogeneous refinement, 303,989 particles exhibiting a pseudo-arrowhead shape were subjected to additional particle sorting. After two rounds of ab initio reconstruction and heterogeneous refinement, two distinct pseudo-arrowhead maps were obtained, consisting of 41,455 and 33,653 particles, respectively. Although the training focused on particles with asymmetric shapes, some particles still adopted an arrowhead shape. Among the 346,320 particles with an arrowhead configuration, further sorting was conducted to isolate those with symmetrical shapes. After several rounds of particle sorting, 18,306 particles were selected for the final 3D sorting process.

The three particle sets (41,455 and 33,653 from the asymmetric class, and 18,306 from the arrowhead shape) were combined for final ab initio and heterogeneous refinement. After excluding 6,559 particles corresponding to junk data, a total of 86,854 particles underwent ab initio, nonuniform refinement and local refinement of the extracellular domain. This processing yielded a map with a global resolution of 6.26 Å, based on a FSC criterion of 0.143.

Data processing for the IRpseudo-gamma structure

To process the pseudo-gamma conformation, we used the same dataset as for the arrowhead shape. Using a template picker, we extracted 3,495,704 particles. After several rounds of 2D classification, 356,821 particles were selected for ab initio reconstruction to generate an initial 3D model. After heterogeneous refinement, 24,688 particles belonging to one class with an asymmetric shape were subjected to Topaz training34. From this trained model, 759,607 particles were extracted. Subsequent ab initio reconstruction and heterogeneous refinement resulted in 100,204 particles exhibiting the pseudo-gamma shape, which were then subjected to additional particle sorting. After three rounds of ab initio reconstruction and heterogeneous refinement, 42,469 pseudo-gamma particles were identified. Further filtering via 2D classification removed nonrelevant particles, leaving 37,078 particles. Ultimately, we achieved a map with a global resolution of 9.35 Å, determined using a FSC criterion of 0.143.

Model building

The atomic model building of the IRarrowhead conformation began with the docking of the human IR bound to A62M (PDB: 7YQ6) into the map, utilizing UCSF Chimera v1.17.124,35. Manual adjustments to the model were performed in Coot36. As the electron density map of A62D covered only the monomer, we fit the A62M model accordingly. The model underwent real-space refinement in PHENIX 1.14, applying rigid body and secondary structure restraints37. The final validated model achieved a MolProbity score of 2.29 with a clash score of 17.7 (ref. 38).

For the IRpseudo-arrowhead structures, model building began with the docking of the human IR bound to A62M (PDB: 7YQ6) into the map, utilizing UCSF Chimera v1.17.124,35. Manual adjustments to the model were performed in Coot36. As the electron density map of A62D covered only the monomer, we fit the A62M model accordingly. The model underwent real-space refinement in PHENIX 1.14, applying rigid body and secondary structure restraints37. The final validated model achieved a MolProbity score of 2.32 with a clash score of 21.47 (ref. 38).

For the IRpseudo-gamma structures, model building was performed by docking the IR bound to A62M and the IRA43+Ins structure (PDB: 7YQ6, 7YQ3)24. The models were manually refined in Coot36 and then subjected to real-space refinement in PHENIX 1.1437, using rigid body and secondary structure restraints. The refined models achieved a MolProbity score of 2.29 and a clash score of 21.76 (ref. 38).

Oligomerization analysis of IR by size exclusion chromatography

To verify whether the A62D aptamer induces intermolecular interactions between IR dimers, we divided purified IR protein into two fractions. Each fraction was incubated with either the A62M or A62D aptamer at a molar ratio of 1:2 and 1:1, respectively, for 1 h on ice. The mixtures were then injected into a Superose 6 10/300 column preequilibrated with buffer containing 20 mM HEPES (pH 7.5), 105 mM NaCl, 5 mM KCl, 5 mM MgCl2, 0.03% DDM and 0.003% CHS. Peaks corresponding to the void volume and the A62M:IR dimer (2:1) or A62D:IR dimer (1:1) complexes were collected. The samples were subsequently concentrated to 0.04 mg/ml using a Vivaspin device (100 kDa cutoff; GE Healthcare) in preparation for negative-stain electron microscopy (EM).

Negative-stain EM analysis

To prepare negative-stain EM grids, 3.5 μl of the sample was applied to glow-discharged holey carbon grids (Formvar/Carbon, Cu 400-mesh; EMS). The sample was allowed to adsorb onto the grid for 30 s, after which excess liquid was blotted off with filter paper. The grid was then washed twice with deionized water and stained with uranyl acetate solution for 20 s. After another blotting step, the grid was allowed to dry overnight at 18 °C. Images were acquired using a BIO TEM JEM-1011 instrument, operated at an acceleration voltage of 80 kV and equipped with a Gatan ES1000W CCD camera at a magnification of 200,000×.

Immunofluorescence assay

Rat-1/hIR cells were cultured in Dulbecco’s modified Eagle medium (Serena) supplemented with 10% FBS (Serena), 100 U/ml penicillin and 100 µg/ml streptomycin. Cultures were maintained at 37 °C in a humidified incubator with 5% CO2. Before the immunofluorescence assay, cells were seeded on coverslips coated with poly-l-lysine (10 µg/ml). Serum starvation was performed using FBS-free medium for 2 h before aptamer treatment. Each aptamer was applied at the indicated concentrations for 1 h. After aptamer treatment, cells were subjected to an immunofluorescence assay as described in previous studies39,40,41,42. In brief, cells were fixed with 4% formaldehyde, prepared by diluting a 16% methanol-free formaldehyde solution (Thermo Scientific), and then permeabilized with Tris-buffered saline containing 0.5% Triton X-100. The cells were subsequently incubated overnight with a primary antibody targeting the IR β subunit (Cell Signaling, #3025S) in an antibody dilution buffer (Tris-buffered saline containing 0.1% Triton X-100, 0.1% sodium azide and 2% bovine serum albumin). After primary antibody incubation, cells were treated with a secondary antibody (goat anti-rabbit IgG (H + L) cross-adsorbed secondary antibody, Alexa Fluor 488, Thermo Fisher) and phalloidin (Alexa Fluor 568 Phalloidin, Invitrogen). Finally, cells were mounted using Vectashield antifade mounting medium (Vector Laboratories).

Structured illumination microscopy (SIM) and data analysis

Super-resolution imaging was performed using an AP DeltaVision OMX Ultra High-Resolution Fluorescence Microscope. 3D-SIM was utilized to acquire images with a resolution below the diffraction limit43. The system used a 60× oil-immersion objective lens with a numerical aperture of 1.42 and a refractive index of 1.518 (Olympus). Structured illumination images were processed for alignment, reconstruction and deconvolution using the SoftWorx software. DAPI, IR and phalloidin signals were detected using 405-, 477- and 568-nm excitation lasers, respectively. Optical sections were obtained at 125-nm spacing. Representative images presented are maximum intensity projections of the acquired images for each sample. Maximum projection was generated using ImageJ (National Institutes of Health).

For IR puncta analysis, particle analysis in ImageJ was performed. Particles ranging in size from 0.5 µm2 to 4 µm2 to were analyzed to exclude nonspecific signals from the dataset. All graphs and statistical analyses were generated using GraphPad Prism 5 software. P values were calculated using Student’s t-test, and the results are expressed as mean ± standard error of the mean (s.e.m.), unless otherwise indicated.

Results

Design of a full agonist aptamer for IR

In our previous study, we developed IR-A62 (referred to here as A62M) which is a partial agonist that specifically phosphorylates the Tyr1150 residue of IR, and a single administration of A62M temporarily reduced blood glucose levels in diabetic mice31. However, to assess A62M’s potential in diabetes, its effects after repeated administration over an extended period need to be evaluated. One of the major challenges in developing aptamers for clinical use is their small size. A62M has a molecular weight of ~9.1 kDa, which is insufficient to avoid rapid clearance from the bloodstream via renal excretion.

A common strategy to improve the pharmacokinetic properties of aptamers is to create multimeric forms44. However, multivalency can sometimes hinder aptamer activity by causing steric clashes with the target45,46. To address this, we examined the activity of a dimeric form of IR-A62. We constructed a series of dimers by linking two A62M aptamers with thymine nucleotide linkers of varying lengths and assessed their ability to activate the IR (Fig. 1a–c). These linkers consisted of 4 T, 6 T, 8 T, 10 T, 12 T, 14 T, 19 T, and 24 T thymine nucleotides. They are in various range of lengths, from 4 T linker with ~2.5 nm that connect two A62 aptamers in close proximity to 24 T linker with ~15.2 nm that are approximately three times the diameter of the A62 aptamer, ~5 nm. We used 400 nM A62M and 200 nM A62D-nTs (where ‘n’ represents the number of thymine nucleotides) to compare their activities, as each A62D-nT aptamer contains two functional ligands. Surprisingly, A62 dimers linked by 8 T to 19 T induced dual phosphorylation at the Tyr1150 and Tyr1151 residues and considerably enhanced AKT signaling (Fig. 1b). By contrast, dimers linked by 4 T, 6 T or 24 T showed reduced and limited phosphorylation of the IR. Thus, we conclude that A62 dimers linked by a moderately sized linker can induce full phosphorylation of the IR, whereas linkers that are too short or too long impair this ability.

Fig. 1: Sequence and agonistic effect on IR phosphorylation of IR-A62 monomer and IR-A62 dimers.
figure 1

a Sequence comparison between the IR-A62 monomer and IR-A62 dimers linked by the 8 T linker. For clarity, only IR-A62D-8T is shown. Z and P represent the benzyl modified- and naphthyl modified-dU, respectively. Red- and green-colored nucleotides represent 2′-fluoro ribose and 2′-O-methyl ribose, respectively. The linker between the two A62M aptamers is underlined. b Activity comparison of the IR-A62 monomer and IR-A62 dimers. Rat-1/hIR cells were stimulated with 50 nM insulin for 10 min, 400 nM A62M for 1 h or 200 nM A62 dimers for 1 h. Bar graphs are presented as the mean ± s.d. of three independent replicates and normalized by A62D-8T stimulated samples. c, d Phosphorylation of IR (c) and downstream signaling (d) were analyzed using specific phospho-antibodies. Rat-1/hIR cells were stimulated with 50 nM insulin for 10 min, 400 nM A62M for 1 h or 200 nM A62M-8T. Bar graphs are presented as the mean ± s.d. of three independent replicates and normalized by 50 nM insulin-stimulated samples.

Full size image

To further characterize the A62 dimer, we selected A62D-8T and measured its effects on the phosphorylation of IR, AKT and ERK after 1 h of stimulation and compared its activity with A62M. A62D-8T was found to induce phosphorylation of all tyrosine residues, similar to insulin (Fig. 1c). Despite the full phosphorylation of IR by A62D-8T, activation of the MAPK pathway remained significantly lower than that induced by insulin (Fig. 1d).

Dose and time characterization of the A62D-8T aptamer

We next evaluated the efficacy and potency of A62D-8T in comparison with A62M (Fig. 2a–f and Supplementary Fig. 1). A62D-8T demonstrated significantly higher efficacy in promoting IR and AKT phosphorylation compared with A62M. Specifically, A62D-8T showed greater potency and efficacy in inducing mono- (Y1150), dual- (Y1150/Y1151) and total Tyr (4G10) phosphorylation of the IR. Consistent with previous findings (Fig. 1a–d), A62D-8T increased AKT phosphorylation in a dose-dependent manner, surpassing A62M. However, even at saturation, A62D-8T did not significantly affect ERK phosphorylation.

Fig. 2: Effects of the dose and duration of A62M and A62D-8T on IR phosphorylation and downstream signaling.
figure 2

a–f Comparison of dose-dependent phosphorylation of IR and downstream signaling by A62M or A62D-8T. Rat-1/hIR cells were incubated with varying concentrations of A62M or A62D for 1 h. The relative band intensities, compared with 100 nM insulin, are shown for pIR Y1150 (a) pIR Y1150/Y1151 (b) pIR pY (c) pAKT S473 (d) pAKT T308 (e) and pERK T202/Y204 (f) presented as the mean ± s.d. of three independent replicates. g–k Time-dependent phosphorylation of IR and downstream signaling. Rat-1/hIR cells were stimulated with 50 nM insulin for 10 min or 200 nM A62M-8T for 5 min to 4 h. The kinetics of pIR Y1150 (g) pIR Y1150/Y1151 (h) pAKT T308 (i), pAKT S473 (j) and pERK T202/Y204 (k) are presented as the mean ± s.d. of three independent experiments, normalized to each negative control (NT) to determine the fold change from basal levels.

Full size image

A notable property of A62M, which differentiates it from insulin, is its slower onset of IR phosphorylation, taking up to 2 h to increase and persisting for over 4 h. Therefore, we also investigated the time-dependent effects of both insulin and A62D-8T (Fig. 2g–k and Supplementary Fig. 1b). Similar to A62M, A62D-8T exhibited a slow and sustained activation of the receptor. Although A62D-8T induced up to a twofold increase in IR and AKT phosphorylation compared with insulin, it did not significantly alter ERK phosphorylation (Fig. 2a–f). These findings indicate that, while A62D-8T fully induces Tyr phosphorylation of the IR, it continues to selectively activate the AKT pathway without triggering the MAPK pathway.

Structures of the A62D-8T and IR complex

Previously, two A62M aptamers were demonstrated to bind to the IR dimer, inducing an arrowhead-shaped conformation. This structural shift from the apo state resulted in monophosphorylation of the IR24,31. To explore how A62D-8T induces full phosphorylation of the IR, we determined the structure of the IR bound to A62D-8T. Full-length IR was purified and incubated with A62D-8T at a 1:1 molar ratio. After further purification via size-exclusion chromatography, we resolved the complex structure using cryo-EM (Supplementary Figs. 2–5 and Supplementary Table 1). Our analysis revealed three distinct conformations of the A62D-8T–IR complex: arrowhead-shaped (IRarrowhead), pseudo-arrowhead-shaped (IRpseudo-arrowhead) and pseudo-gamma-shaped (IRpseudo-gamma) (Fig. 3a–i and Supplementary Figs. 2–5). These conformations were distributed in a ratio of 0.29:0.30:0.12, with resolutions ranging from 6.26 to 9.35 Å (Supplementary Table 1).

Fig. 3: Three distinct conformations of the A62D-8T-bound IR dimer.
figure 3

a–c, Cryo-EM map (a) and structure (b and c) of the IRarrowhead in two different views. The surface representation is shown in c. d–f, Cryo-EM map (d) and structure (e and f) of the IRpseudo-arrowhead in two orientations, shown in the same views as in a–c by aligning the stalk of protomer A to that of IRarrowhead. g–i, Cryo-EM map (g) and structure (h and i) of IRpseudo-gamma in the same views as those of a–c.

Full size image

The structure of the IRarrowhead bound to A62D-8T closely resembles that of the A62M-bound IR (PDB 7YQ6)24. In this conformation, the two protomers of the Λ-shaped apo IR are rotated clockwise relative to each other, pivoting at Cys524, which forms one of the disulfide linkages between the protomers (Fig. 3a–c and Supplementary Fig. 6a,b). In this symmetrical structure, the interactions between FnIII-2 and L2′ (or FnIII-2′ and L2), referred to as auto-inhibition sites, are disrupted17. Within the binding site of the A62D-8T aptamer, we observed density corresponding to the size of only one A62M molecule in each protomer (Supplementary Fig. 6c). The reason why only one A62M binds to each IR protomer is unclear, but we speculate that the two A62M aptamers in A62D are too bulky to fit simultaneously between the L1 domain of one protomer and the FnIII-1′ domain of the other protomer. Each A62M molecule binds to opposite sides of the IR dimer, positioning itself between the L1 domain of one protomer and the FnIII-1′ domain of the other (Fig. 3b). The head of the IRarrowhead is composed of the L1, CR, a portion of L2, and A62M, while the FnIII-1~3 domains are located perpendicularly in the center, forming the stalks of the IRarrowhead. The distance between the membrane-proximal ends of FnIII-3 and FnIII-3′ is ~42 Å (Fig. 3c).

The IRpseudo-arrowhead structure

The second class of particles forms a modified arrowhead-like shape, which we designate as the IRpseudo-arrowhead. In this asymmetrical structure, the head of one protomer (protomer A) is lifted from the symmetric IRarrowhead conformation, while the opposing protomer (protomer B) remains unchanged (Fig. 3d–f). This structure was resolved at a 6.26 Å resolution (Supplementary Fig. 4).

To analyze this structure, we performed rigid-body fitting of the IR and A62M from the A62M-bound IR structure (PDB 7YQ6), which aligned well with the density. In the density map for A62D-8T, only one A62M module was visible between L1 and FnIII-1′ (or, alternatively, between L1′ and FnIII-1; Supplementary Fig. 6d). In the IRpseudo-arrowhead, two A62D-8T (A62M) aptamers bind to different sites on the IR dimer, which we refer to as site u (upper side) and site l (lower side), respectively (Figs. 3d–f and 4a, b). At site u, the L1 domain of protomer A and the FnIII-1′ domain of protomer B simultaneously interact with the A62M aptamer. At site l, a similar interaction is observed between the opposite protomer and a second A62M aptamer (Supplementary Fig. 6e–g). However, we hypothesize that subtle differences exist between the interactions at site u and site l, particularly between the residues and aptamers. In the IRarrowhead conformation, both the L1 and FnIII-1 domains are involved in aptamer binding. It is possible that the A62M in the IRpseudo-arrowhead conformation is slightly rotated or reoriented, leading to slight variations in the binding interface.

Fig. 4: Structure of the IRpseudo-arrowhead and comparison with other IR structures.
figure 4

a, b The overall structure of the IRpseudo-arrowhead in two views. b is rotated by 180° from (a) about the Y-axis. c Comparison of protomers A and B by aligning the stalks. d Comparison of the two protomers by aligning the L2 domain. e Comparison between the A protomers of IRarrowhead and IRpseudo-arrowhead, aligned by stalks. f Alignment of the A protomers by the L2 domain. g Comparison between the B protomers of IRarrowhead and IRpseudo-arrowhead, aligned by stalks. h Alignment of the B protomers by the L2 domains. IRpseudo-arrowhead and IRarrowhead are shown in orange and blue, respectively.

Full size image

We superimposed the two protomers of the IRpseudo-arrowhead by aligning their stalk regions (FnIII-1 to FnIII-3) (Fig. 4c). The head groups of the protomers were arranged differently, with a 13° difference when measured at Q177–L459–V604 (L1–L2–FnIII2), as protomer A is raised relative to protomer B. The distance between the E204 residues (in the CR domain) of the two aligned protomers is 32.7 Å, while the distance between their Q177 residues (in the L1 domain) is 30 Å.

To investigate the hinge flexibility between the CR and L2 domains, we aligned the L2 domains of both protomers (Fig. 4d). In protomer A, the L1–CR domain shifts toward the L2 domain by 6°, as measured at Q177–E204–L459 (L1–CR–L2), whereas no significant changes were observed in protomer B. Consequently, the head domain of protomer A becomes more compact, with the distances between the Q177 residues and E204 residues reduced to 16 Å and 9.9 Å, respectively, in the L2-aligned protomers. In the aligned structures, the aptamers occupy slightly different positions (Fig. 4c). However, due to the shift in the L1–CR domain, the aptamer position also shifts upward while maintaining a similar interface to that between the A62M and L1 domains in the IRarrowhead structure.

Comparison of IRarrowhead and IRpseudo-arrowhead structures

We compared the structures of site u and site l in IRpseudo-arrowhead with the corresponding regions in A62M-bound IRarrowhead (PDB ID: 7YQ6) by aligning their FnIII domains (Fig. 4e–h). Three major conformational differences were identified in the head arrangement of IRpseudo-arrowhead relative to IRarrowhead. Specifically, the head domains of protomer A and B were shifted upward by 9° and downward by 4°, respectively, as measured at the Q177–L459–V604 segment (L1–L2–FnIII2; Fig. 4e, g).

In the alignment of L2, the Q177–E204–L459 (L1–CR–L2) angle differences between protomer A and protomer B were 5° and 1°, respectively, suggesting that movement at the L1–CR hinge (between CR and L2) is not substantial (Fig. 4f, h). However, the downshift of protomer A, indicated by the L1–L1′ distance (21.5 Å between Q177 residues and 18.5 Å between E204 residues), and the upshift of protomer B (9.7 Å for L1 and 11.5 Å for CR at E204) reveal that protomer A undergoes more significant changes from the IRpseudo-arrowhead to IRarrowhead conformation.

In addition, both protomers display rotation relative to the IRarrowhead protomer (Supplementary Fig. 6h,i). In the IR dimer, the two protomers are connected by a Cys524–Cys524′ disulfide bond, which serves as the center of rotation. Upon aligning the stalk of protomer A between IRpseudo-arrowhead and IRarrowhead, we observed that protomer B rotates clockwise by ~11.2° (Supplementary Fig. 6h). Due to these structural changes in the L2–FnIII1 and CR–L2 hinges, along with the rotation of protomers A and B, the distance between the membrane-proximal ends at D907–D907′ is reduced from 43.3 Å to 30.1 Å (Fig. 3b, d and Supplementary Fig. 6i).

The IRpseudo-gamma structure

The third class of particles displayed more pronounced conformational differences from the IRarrowhead structure compared with the IRpseudo-arrowhead structure (Figs. 3g–i and 5a, b and Supplementary Fig. 5). In this conformation, the L1 domain of protomer A is shifted upward, positioned almost perpendicular to the stalk, while the L1′ domain of protomer B is shifted downward, closely aligned with the stalk. This third-class conformation resembles the fully active Γ-shape observed in the insulin-bound or insulin and A43 aptamer-bound forms of the receptor (Fig. 5b, c).

Fig. 5: Structure of IRpseudo-gamma and comparison with other IR structures.
figure 5

a, b Two views of IRpseudo-gamma. Protomers A and B are shown in blue and orange, respectively. a, b are related by a 180° rotation about the Y-axis. c Structure of IRgamma in the same orientation as in b, d, Superimposed structures of the A protomers of IRpseudo-gamma (orange), IRpseudo-arrow (blue) and IRarrowhead (gray), aligned by the A62M, showing the relative positions of the FnIII-1 domains between the two protomers. e Comparison of protomer A and B of IRpseudo-gamma by aligning the stalks. IRpseudo-gamma and IRgamma are shown in orange and blue, respectively. f Comparison of protomer A and B of IRpseudo-gamma by aligning the L2 domain. g, h Superimposed structures of the A (g) and B (h) protomers of IRpseudo-gamma and IRgamma, aligned by the FnIII stalks. i, j Superimposed structures of the A (i) and B (j) protomers of IRpseudo-gamma and IRgamma, aligned by the L2 domains.

Full size image

Unlike the IRarrowhead or IRpseudo-arrowhead conformations, A62D binds at only one site, specifically at the interface between the L1 domain of protomer B and the FnIII-1′ domain of protomer A (Fig. 3g–i). The interaction between the L1 domain and the aptamer is similar to the binding observed between IRarrowhead and A62M (Figs. 3b and 5b). However, the orientation of the FnIII-1 domain in the IRpseudo-gamma conformation differs from that in the IRpseudo-arrowhead and IRarrowhead structures (Fig. 5d). Comparison of the structures between IRpseudo-gamma and IRarrowhead or IRpseudo-arrowhead is shown in Supplementary Fig. 7.

A structural comparison of protomers A and B, aligned by their stalk regions, reveals that the head of protomer A is tilted upward by 30° compared with that of protomer B, as measured by the angles formed by Q177–L459–V604 (L1–L2–FnIII2); 58° versus 27°, respectively (Fig. 5e). When the L2 domains are aligned, protomer A exhibits a more compact head domain, with an angle of 52° (Q177–LE204–L459), while the head of protomer B adopts a more extended conformation, with a corresponding angle of 86.5° (Fig. 5f). The relative distances between the Q177 (L1) residues and the E204 (CR) residues of the two protomers are 60.1 Å and 41 Å, respectively. Thus, aptamer binding induces a more extended conformation of the head domain in protomer B.

Structural comparison of IRpseudo-gamma and IRgamma

The fully activated conformation of IRgamma is achieved when bound by a single insulin molecule, or by insulin and A4324,47, with one side of the receptor’s head (protomer A) lifted to accommodate insulin between the L1, CR and FnIII-1′ domains, while the opposite side (protomer B) remains insulin-free and shifts downward to form a compact head structure (Fig. 5c). Although it is uncertain whether IRpseudo-gamma represents an active conformation, the proximity of the membrane-proximal ends of the FnIII-stalks (~20 Å apart) suggests it might correspond to a fully phosphorylated state (Fig. 3i).

When aligning the stalk of protomer A in IRpseudo-gamma with that of IRgamma, we observed a downward shift in the aptamer-free head of IRpseudo-gamma and an upward shift in the aptamer-bound head (or vice versa) compared with IRgamma (Fig. 5g,h). This movement is further demonstrated by changes in specific domain angles. The Q177–L459–V604 (L1–L2–FnIII2) angle decreased from 58° in IRpseudo-gamma to 40° in IRgamma (Fig. 5g). In addition, the L1 domain in protomer A of IRpseudo-gamma was displaced by 26.4 Å, measured at residue Q177, compared with its position in IRgamma.

In protomer B of IRpseudo-gamma, the Q177–L459–V604 angle was 27.5°, much larger than the 6.5° observed in IRgamma (Fig. 5h). Similarly, the E204–L459–V604 angle in IRpseudo-gamma was 39.1°, significantly greater than the 13.2° in IRgamma. These differences indicate that the head domain of protomer B in IRpseudo-gamma adopts a more extended conformation compared with that of IRgamma.

When superimposing the L2 domains, the Q177–E204–L459 angle in protomer A of IRpseudo-gamma was 52°, slightly wider than the 42° angle seen in IRgamma (Fig. 5i). For protomer B, the head domain in IRpseudo-gamma was notably more open and extended, with an angle of 86.5° compared with 57.7° in IRgamma (Fig. 5j). The distance between L1 domains at Q177 was measured at 44 Å for protomer A and 60 Å for protomer B when the L2 domains were aligned. These findings indicate significant conformational changes in both the CR–L2 and L2–FnIII1 hinges.

A62D-induced crosslinking of IR dimers

Our structural analysis demonstrated that A62D induces the formation of three distinct IR conformations, one of which closely resembles the fully active IR conformation. This structure probably contributes to the full phosphorylation of IR. Given that the IRpseudo-gamma dimer binds only a single A62M module of A62D, we hypothesize that each A62M unit binds separately to an IR dimer, facilitating cross-interaction between the dimers. This interaction positions the dimers in close proximity, promoting trans-phosphorylation. This hypothesis is supported by the data presented in Fig. 1b, which shows that A62D linked with a connector of 8 T to 19 T efficiently phosphorylated IR. By contrast, shorter or longer linkers failed to achieve full phosphorylation. This may be due to short linkers bringing the IR dimers too close together, limiting the flexibility required for trans-phosphorylation, while longer linkers keep the dimers too far apart to efficiently facilitate intermolecular trans-phosphorylation. Therefore, we conclude that A62D aptamers with optimally sized linkers can effectively position two IRs in close proximity, inducing phosphorylation of the IR dimers.

To further validate this model, we mixed A62M or A62D with IR dimers in 4:1 and 2:1 molar ratios, respectively, and conducted size-exclusion chromatography analysis (Fig. 6a, b). We observed substantial fractions of oligomeric A62D–IR complexes in the void volume, whereas relatively small fractions of A62M–IR appeared in this region. Similar amounts of 1:1 A62M–IR complexes were observed. In addition, negative staining analysis revealed numerous oligomeric A62D–IR particles in the void volume fractions (Fig. 6c–f).

Fig. 6: Intermolecular interactions of A62D-induced IR dimers.
figure 6

Comparison of the profiles of IR mixed with A62M (a) and A62D (b) aptamers. a, b Comparison of the profiles of IR mixed with A62M and A62D aptamers. c–f Comparison of negative staining images of fractions from the void volume peak (c) and normal peak (d) of the A62M–IR complex, and fractions from the void volume peak (e) and normal peak (f) of the A62DIR complex. g Representative images from SIM. Rat-1/hIR cells treated with indicated aptamers at various concentrations were co-immunostained with anti-IR, DAPI and phalloidin (F-actin). Scale bar, 10 mm. h A model of the A62D complex with two IR dimers. i A model of the ‘beads on a string’-like cluster formed by A62D complexes with IR dimers.

Full size image

We also examined whether A62D induced the oligomerization of IR in live cells. We stained the IR using a fluorescence-labeled antibody, and observed recruitment of IR by A62D in live cells using SIM. Rat-1/hIR cells expressing IR were treated with A62M and A62D at various concentrations (100 nM, 200 nM and 400 nM for A62M; 50 nM, 100 nM and 200 nM for A62D), and IR oligomerization was monitored (Fig. 6g). While we did not observe noticeable IR clusters in A62M-treated cells, a significant number of IR clusters was observed in A62D-treated cells (Fig. 6g and Supplementary Fig. 8a). Although IR oligomerization in A62D-treated cells was more prominent in the plasma membrane, marked by F-actin, clustering events were also elevated in the cytoplasm and nucleus (Fig. 6g and Supplementary Fig. 8b–d). Collectively, these results demonstrate that A62D aptamer induces full IR phosphorylation by facilitating intermolecular interactions between the IR dimers (Fig. 6h–i).

Discussion

In this study, we aimed to enhance the previously reported A62M aptamer, a partial agonist known to induce Tyr1150 mono-phosphorylation of IR24,31. Our findings revealed that crosslinking two A62M aptamers with linkers of varying lengths resulted in full phosphorylation of the IR and increased potency. To our knowledge, this represents the first example of aptamer-mediated full activation of the IR.

The primary question raised by our study is why the A62D-8T (and other A62D variants with different linkers) behaves differently from A62M, leading to full phosphorylation of the IR. We hypothesized that A62D induces a distinct structural change in the IR compared with A62M. To explore this, we determined the structure of the A62D-bound IR and compared it with the A62M–IR complex. We observed three conformations induced by A62D: the previously identified arrowhead conformation, along with pseudo-arrowhead and pseudo-gamma conformations. In all three structures, only one A62M module of the A62D binds to the IR (Fig. 3b, e, h). The A62M modules bind at the interface between the L1–CR region of one IR protomer and the FnIII-1′ region of another protomer. It appears that the full A62D structure is too bulky to fit between these regions.

While the IRpseudo-arrowhead structure closely resembles the IRarrowhead conformation, the IRpseudo-gamma conformation is distinct from IRarrowhead and more closely resembles the IRgamma structure observed in the presence of a single insulin molecule or a single insulin molecule with the A43-positive allosteric modulator24,47. Although the activation state of IRpseudo-gamma remains unclear, its structural similarity to IRgamma and the proximity of the membrane-proximal ends of the FnIII stalks suggest that IRpseudo-gamma could contribute to full activation of the IR. Initially, based on structural analysis, we considered a possibility that IRpseudo-gamma plays a key role in the A62D-induced full phosphorylation of the IR. The particle distribution for IRarrowhead, IRpseudo-arrowhead and IRpseudo-gamma in the A62D complex was observed at a ratio of 0.29:0.3:0.12, respectively. Despite the relatively small proportion of IRpseudo-gamma, assuming that this conformation is fully phosphorylated, we estimated that the presence of 12% fully phosphorylated IR in the presence of A62D, compared with none in the A62M-bound state, could be sufficient for detection in our system (Fig. 1c). To determine whether IRpseudo-gamma is exclusive to the A62D-bound state, we reanalyzed the A62M-bound IR data and found that ~7% of the particles also exhibited the IRpseudo-gamma conformation. Thus, although we cannot rule out the possibility that IRpseudo-gamma contributes to full phosphorylation, it is unlikely to be the primary mechanism by which A62D induces full IR phosphorylation.

We then considered an alternative explanation: A62D, containing two A62M modules, may bring two or more IR dimers into close proximity, facilitating intermolecular trans-phosphorylation. This hypothesis is supported by increased oligomerization observed in the presence of A62D compared with A62M, as shown by size exclusion chromatography and negative staining. The length of the linker between the two A62M modules is crucial in this crosslinking process. If the linker is too short, each A62M module may not efficiently capture the IR dimers; if it is too long, the bound IR dimers may not interact efficiently enough for optimal intermolecular phosphorylation.

We note that our single-particle cryo-EM analysis revealed the structures of the A62D-bound IR dimer because we focused on the fraction containing homogeneous particles representing a single IR dimer. Although the structural analysis identified three distinct conformations of the A62D-bound single IR dimer, biochemical and negative staining analyses suggest that intermolecular interactions between IR dimers are essential for full phosphorylation and activation of the IR.

A62D may promote the oligomerization of IR dimers through two potential mechanisms. First, each A62D could recruit two IR dimers, one at each end. In this scenario, only one protomer of an IR dimer interacts with the A62M module of A62D, probably positioning the IR in the IRpseudo-gamma conformation (Fig. 6h). Alternatively, multiple IR dimers could be continuously linked via several A62D aptamers in a ‘beads-on-a-string’ configuration (Fig. 6i). In this case, each A62D would bind to IRpseudo-arrowhead or IRarrowhead conformations at both ends. The latter model predicts a higher proportion of oligomerized IR dimers in the void volumes during size-exclusion chromatography (Fig. 6b, e, f, i). However, it is possible that both models coexist in A62D-induced IR oligomers.

Conventional models have depicted IR activation as a simple switch dependent on insulin binding. However, recent studies highlight the importance of spatial dynamics of IR at the plasma membrane in receptor activation and signal transduction. Super-resolution microscopy has shown that IR molecules are incorporated into dynamic clusters48,49. In this model, IR dimers are recruited into clusters at the plasma membrane, cytoplasm,and nucleus, with insulin stimulation further promoting cluster formation in their active states. This results in an increase in both the number of IR-containing clusters and the number of IR molecules per cluster. In addition, rod-like insulin–DNA origami nanostructures have demonstrated that increasing insulin valency enhances both IR and AKT phosphorylation50. These findings suggest that the formation of IR dimers or oligomers plays a critical role in insulin-induced IR activation and downstream signaling, supporting our interpretation of A62D-induced IR activation. However, the precise mechanisms by which IR clustering enhances receptor phosphorylation and signaling remain unclear.

Over the past few decades, various agonists that activate the IR have been developed using antibodies, peptides and aptamers10,12,17,20,21,22,28,31,32,51,52,53,54,55. These agonists exhibited high potency in activating IR. However, because their potencies have been studied through different approaches with A62D, direct comparisons of their potency on IR with that of A62D are not straightforward. Our study indicates that A62D-8T exhibits a slow and sustained activation of IR, while it induces an up to twofold increase in IR and AKT phosphorylation compared with insulin. A common feature of several agonists in a manner distinct from insulin is that they selectively activate the metabolic functions and ATK pathway of the IR. Such functional selectivity appears to be associated with site-specific phosphorylation and specific structural state of IR. In particular, S597, an insulin-mimetic peptide, and A62M induce selective phosphorylation at Y1150 of the IR and nearly identical arrowhead structures24,53. Based on these studies, the stepwise activation model has been proposed, suggesting that the function of the IR is selectively regulated according to its conformational states14.

Despite these advances, previous studies on the agonists have described that a single IR acts in a stand-alone state and have not explained the functional role of IR clustering on the cell membrane. In this study, we artificially induced clustering of IR using A62D and demonstrated that IR activation can be regulated by interreceptor interactions. Collectively, these studies suggest that spatial clustering strengthens intermolecular interactions between IR molecules, leading to full receptor phosphorylation and the selective enhancement of AKT pathway. Furthermore, A62D suggests a new possibility for diabetes treatment. Investigating the potential clinical effects of artificially induced IR clustering in patients with diabetes is an intriguing topic for future research. However, optimizing the aptamer to effectively induce IR clustering under in vivo conditions while maintaining sufficient pharmacokinetics and pharmacodynamics remains a major challenge.

While the structural transition in IR activation involves a shift from the inactive, Λ-shaped dimer to the active, single-insulin-bound IRgamma conformation, the structures of the intermediate states between these two conformations remain poorly understood. Recent studies have proposed potential intermediates, such as the tilted T-shaped IR, resulting from binding two or more insulin molecules or other ligands9,10,22,23,24. However, these structures require further investigation. All conformations observed in this study arise from hinge-bending motions between the L2 and FnIII-1 domains and between the CR and L2 domains, along with rigid-body rotations of the two protomers in the Λ- and gamma-shaped IR dimer. Therefore, the A62D-induced IR structures presented here may offer insights into the intermediate states involved in IR activation.

In summary, by crosslinking A62M, we enhanced the efficacy of the IR aptamer, achieving full activation of the receptor. This study presents a potential strategy for designing more effective IR agonists as therapeutics. By increasing intermolecular interactions between IR dimers, agonist-induced IR signaling could be more efficiently transduced.