An aptamer agonist of the insulin receptor acts as a positive or negative allosteric modulator, depending on its concentration

Identification of the IR-A62 aptamer

SELEX was performed to identify aptamers that bind to the extracellular domain of IR (His 28-Lys 944). The single-strand DNA library used contained a 40-mer random region flanked on both sides by 20-mer constant regions that were used for PCR amplification of the library. To improve the specificity and affinity of the aptamer–protein interaction, Nap-dU was used instead of thymine bases in the 40-mer random regions¹⁶. In this way, we obtained 41 different aptamers containing Nap-dU modifications.

To evaluate the autophosphorylation of IR induced by the aptamers, Rat-1 cells overexpressing human IR (Rat-1/hIR) were stimulated with 500 nM aptamers for 1 h. We used the IR-A48 agonistic aptamer as a positive control to compare the efficacy of the novel aptamers¹¹. Although most of the aptamers had no effect or a smaller effect than the IR-A48 agonistic aptamer on IR autophosphorylation, one aptamer, IR-A62-F, significantly induced IR autophosphorylation to a similar extent to IR-A48. Full-length IR-A62-F is a 79-mer that contains a 39-mer variable region and two 20-mer constant regions (Fig. 1a). Furthermore, we identified a core sequence (IR-A62-T) of IR-A62-F that is essential for its agonistic activity by comparing the effects of IR-A62-F truncation variants containing 3′ or 5′ sequential deletions (data not shown). IR-A62-T consists of 25 nucleotides, of which seven are Nap-dUs, and forms a small stem-loop structure (Fig. 1b). IR-A62-T showed biased agonism, preferentially phosphorylating a specific tyrosine residue of IR, similar to IR-A48 (Fig. 1c). In contrast to insulin, which increased phosphorylation at the Y960, Y1146, Y1150, Y1151, Y1316, and Y1322 residues, IR-A62-T preferentially stimulated the phosphorylation of Y1150, which is in the kinase domain of IR. Moreover, the reversed sequence of IR-A62-T (IR-A62-R) did not stimulate the phosphorylation of Y1150, which indicates that the agonistic effect of IR-A62-T is not caused by a nonspecific interaction with oligonucleotides.

Post-SELEX optimization of the IR-A62 aptamer

A critical limitation of the in vivo use of aptamers is their rapid degradation by serum nucleases¹⁷. Therefore, it is essential to improve their stability by chemically modifying the nucleotides, such as by adding a methoxy (2’-OMe) or fluoro (2’-F) group at the 2’-sugar position of the ribose. However, such chemical modifications can seriously affect the binding of the aptamer to its target. To determine whether the efficacy of IR-A62-T was affected by the incorporation of modified nucleotides, we prepared IR-A62-T variants in which each dA, dC, and dG nucleotide was substituted by the corresponding 2’-OMe derivative (mA, mC, and mG) (Supplementary Fig. 1a). The effects of the IR-A62-T variants on IR phosphorylation were then evaluated by comparison with IR-A62-T in Rat-1/hIR cells, and the results showed that 11-mG, 12-mA, 13-mG, 19-mA, 21-mC, 22-mC, and 25-mC had no effect or positive effects on the activity of IR-A62-T (Fig. 2a, Supplementary Fig. 1b). We finally chose the 11-mG, 13-mG, 21-mC, and 25-mC modifications to lengthen the distances between each modification because consecutive 2’-OMe modifications significantly disturb the activity of IR-A62-T (Supplementary Fig. 1c). We also performed a similar screen of IR-A62-T variants containing the corresponding 2’-F-derivatives (fA, fC, and fG) in place of each nucleotide, except at the four 2’-OMe modifications sites (Supplementary Fig. 2a). The results showed that 2-fA, 6-fC, 8-fC, 12-fA, 19-fA, and 22-fC had no effect on the activity of IR-A62-T (Supplementary Fig. 2b). To evaluate the combined effects of both the 2’-OMe and 2’-F modifications on IR-A62-T activity, we then tested three IR-A62-T variants containing both 2’-OMe and 2’-F modifications and found that the IR-A62-T variants showed slightly higher activity than the original IR-A62-T (Supplementary Fig. 2c).

The placement of hydrophobic side chains at the 5-position of uracil improves the success of SELEX and increases the affinity of aptamers by adding hydrophobicity to aptamer-target interactions¹⁶. However, these hydrophobic sites also increase the plasma clearance of the molecules in vivo, which has a negative effect on the pharmacokinetic properties of therapeutic aptamers¹⁸. Therefore, to reduce the hydrophobicity of IR-A62-T, seven IR-A62-T variants were synthesized, in which each Nap-dU was replaced by 5-(N-benzylcarboxamide)-2’-deoxyuridine (Bn-dU) (Supplementary Fig. 3a). Substitution with Bn-dU at 10-Nap, 16-Nap, and 20-Nap significantly reduced the activity of IR-A62-T (Supplementary Fig. 3b). Therefore, we ultimately selected 3-Bn, 4-Bn, 14-Bn, and 24-Bn as the most appropriate Bn-dU substitutions for IR-A62-T.

The results of the testing of IR-A62-T variants with chemical modifications are summarized in Fig. 2a. The most favorable combination of substitutions was in a derivative named IR-A62, which contained four 2’-OMe groups, six 2’-F groups, four Bn-dU side chains, and three Nap-dU side chains (Fig. 1a, Fig. 2a). The affinity (K_d) and maximal binding capacity (B_max) of IR-A62 were slightly superior to those of unmodified IR-A62-T (Fig. 2b). Consistent with the results of the aptamer binding assay, IR-A62 was a more potent inducer of Y1150 phosphorylation on IR than IR-A62-T (Fig. 2c, Supplementary Fig. 4). Moreover, we assessed the nuclease resistance of IR-A62 using an in vitro serum stability assay, in which IR-A62-T and IR-A62 were labeled with a 3′-inverted dT (3′-idT) to protect the aptamers from degradation by 3′-exonucleases in the serum and were incubated with 90% human serum at 37 °C for up to 48 h. The degradation of the aptamers at various time points was then analyzed using denaturing polyacrylamide gel electrophoresis, which demonstrated the stability of IR-A62 (serum half-life [t_1/2]=24.9 h) was significantly superior to that of IR-A62-T (t_1/2 = 7.4 h) (Fig. 2d). These results indicate that the combination of chemical modifications successfully improved the nuclease stability of IR-A62-T without causing any loss of agonistic activity. Therefore, all subsequent experiments were performed using IR-A62 containing these modifications.

IR-A62 demonstrates binding cooperativity that differs in a concentration-dependent manner

To determine whether the binding site of IR-A62 is allosteric or orthosteric, we next studied the effect of IR-A62 on the binding of insulin to IR on the plasma membrane. Rat-1/hIR cells were incubated with FITC-labeled insulin (100 nM) and various concentrations of IR-A62 (3.2 nM, 16 nM, 80 nM, 400 nM, 2 µM, or 10 µM), and insulin binding was measured using flow cytometry. FITC-labeled insulin alone caused a 6.24% shift in the peak fluorescent intensity (Fig. 3a). At low IR-A62 concentrations (3.2–80 nM), coincubation of FITC-labeled insulin with IR-A62 gradually increased the peak shift, up to 64.2%. However, as the concentrations of IR-A62 were increased further (400 nM–10 µM), the binding of FITC-labeled insulin gradually decreased to 2.67%, which was a lower level than with FITC-labeled insulin alone.

To verify that the binding cooperativity between insulin and IR-A62 varies depending on concentration, we also measured the binding of FITC-labeled IR-A62 to IR in the presence of various concentrations of insulin (3.2 nM, 16 nM, 80 nM, 400 nM, 2 µM, and 10 µM). Consistent with the results of the insulin-binding assay, coincubation with low insulin concentrations (3.2–16 nM) gradually increased FITC-labeled IR-A62 binding compared to incubation with FITC-labeled IR-A62 alone (Fig. 3b). Moreover, as the concentration of insulin was further increased (80 nM–10 µM), the binding of FITC-labeled IR-A62 gradually decreased. These results imply that insulin and IR-A62 cooperatively bind in a concentration-dependent manner. At low concentrations, insulin and IR-A62 act as mutual PAMs, with the binding of one promoting the binding of the other to IR. However, at high concentrations, IR-A62 and insulin act as mutual NAMs, inhabiting each other’s binding to IR.

In a previous study, we demonstrated that the enhancement of insulin binding to IR by a PAM aptamer potentiates the phosphorylation of tyrosine residues in the intracellular domain of IR¹⁴. As shown in Fig. 1c, insulin binding to IR leads to the autophosphorylation of tyrosine residues, and IR-A62 preferentially induces monophosphorylation of the Y1150 residue of IR. Thus, we can distinguish between insulin- or IR-A62-induced IR phosphorylation by comparing the levels of phosphorylation of Y1150 and other tyrosine residues. To determine whether the concentration-dependent cooperativity between insulin and IR-A62 affects IR autophosphorylation, we evaluated the phosphorylation of IR in the presence of 50 nM insulin and various concentrations of IR-A62 (30 nM, 100 nM, 300 nM, 1 µM and 3 µM). The IR Y1146 phosphorylation induced by insulin increased at low IR-A62 concentrations (30–300 nM) and decreased at higher IR-A62 concentrations (1–3 µM). Because IR Y1146 phosphorylation is induced by insulin but not by IR-A62, this implies that insulin-induced IR phosphorylation can be potentiated or inhibited by concentration-dependent cooperativity with IR-A62 (Fig. 3c). Although a low level of IR Y1150 phosphorylation was induced by IR-A62 alone at low IR-A62 concentrations (30–100 nM), the level induced by coincubation with insulin and IR-A62 was significantly higher. However, as the IR-A62 concentration was increased further (300 nM–3 µM), the IR Y1150 phosphorylation induced by coincubation with insulin and IR-A62 gradually decreased to a similar level to that induced by 3 µM IR-A62 alone. These findings demonstrate that the concentration-dependent differences in the mutual cooperativity displayed by insulin and IR-A62 directly affect the autophosphorylation of IR.

IR signaling is induced by IR-A62

We have shown that IR-A62 is a biased agonist that preferentially induces Y1150 phosphorylation of IR, similar to IR-A48 (Fig. 1c). Moreover, in our previous study, we showed that IR-A48 is characterized by slower and more sustained phosphorylation kinetics of IR and downstream proteins than insulin¹¹. To further investigate the signaling kinetics of IR-A62, we first compared the kinetics of the Y1150 phosphorylation of IR induced by insulin and IR-A62. In contrast to insulin, IR-A62 slowly increased the phosphorylation of IR at Y1150 over 2 h, and this phosphorylation was sustained for 8 h (Fig. 4a), which indicates that IR-A62 also induces signaling slowly but sustains it over a relatively long period of time. However, IR-A62 had a 4.7-fold lower EC₅₀ (18.4 nM) for IR phosphorylation (Y1150) than insulin (86.4 nM) (Fig. 4b). Furthermore, IR-A62 did not bind to IGF-1 receptor (IGF-1R), despite the high degree of structural similarity between IR and IGF-1R (Fig. 2b). Consistent with this binding specificity, IR-A62 had no effect on the phosphorylation of IGF-1R (Fig. 4c).

To characterize the downstream signaling activated by IR-A62, we treated fully differentiated 3T3-L1 adipocytes with IR-A62 for 5 min, 1 h, or 2 h and measured the phosphorylation of IR, AKT, and extracellular signal-regulated kinase (ERK) (Fig. 4d). IR-A62 (200 nM) stimulation for 5 min only slightly increased the phosphorylation of ERK, and the level of AKT phosphorylation induced by IR-A62 was lower than that induced by insulin, even though the level of IR Y1150 phosphorylation induced by IR-A62 was significantly higher than that induced by insulin (Fig. 4e–h). Moreover, the AKT phosphorylation induced by IR-A62 was sustained for up to 2 h, but the ERK phosphorylation induced by IR-A62 was not. Taken together, these results imply that although IR-A62 induces IR Y1150 phosphorylation more potently than insulin, its effects on signaling downstream of IR are minor and less than those of insulin. However, the activation of the AKT pathway by IR-A62 was sustained over a longer period of time than the activation induced by insulin, which is consistent with the Y1150 phosphorylation kinetics of IR.

The effects of IR-A62 on glucose uptake and cell proliferation

IR is a critical regulator of metabolic processes, such as glucose uptake, fat synthesis, gluconeogenesis, and glycogenolysis¹⁹. Many previous studies have shown that the metabolic effects of insulin and IR mainly involve the AKT pathway rather than the MAPK pathway²⁰. Because IR-A62 stimulated AKT phosphorylation in 3T3-L1 adipocytes, we next quantified the time- and dose-dependent effects of IR-A62 on 2-deoxy-glucose uptake. The timing of the effect of IR-A62 on glucose uptake was similar to the timing of its effect on IR and AKT phosphorylation (Fig. 5a): glucose uptake increased slowly over 30 min, in contrast to the rapid effect of insulin, and was sustained for up to 4 h. Moreover, in contrast to the glucose uptake following insulin stimulation, which decreased rapidly after 30 min, the glucose uptake induced by IR-A62 remained greater than half-maximal after 8 h.

To compare the potency of the effects of IR-A62 and insulin on glucose uptake, we next measured 2-deoxy-glucose uptake in 3T3-L1 adipocytes after stimulation with various doses of each (Fig. 5b). The maximal glucose uptake induced by insulin or IR-A62 at concentrations >500 nM did not differ significantly. However, similar to the effects of each on the phosphorylation of AKT at S473 and T308, the glucose uptake induced by insulin was higher than that induced by IR-A62 at low concentrations (5–100 nM). Moreover, IR-A62 exponentially increased glucose uptake at concentrations of 100–500 nM (Hill coefficient: 4.9) compared with insulin, which increased glucose uptake more gradually in the range 5–500 nM (Hill coefficient: 1.27). Consequently, although IR-A62 has a lower EC₅₀ value for IR Y1150 phosphorylation than insulin, the EC₅₀ value of IR-A62 for glucose uptake was higher (177.6 nM) than that of insulin (36.5 nM). These results indicate that IR-A62 alone increases glucose uptake to a level that is comparable with the effect of insulin.

In Fig. 3, we show that the cooperative binding of insulin and IR-A62 to IR is mutual and depends on the concentration of each. To determine whether the effect of IR-A62 on increasing insulin binding by IR-A62 potentiates glucose uptake, we measured the glucose uptake induced by IR-A62 in the absence or presence of insulin. Figure 5b shows that glucose uptake in the presence of IR-A62 began to increase at ~100 nM. Therefore, 50 nM, 100 nM, and 150 nM IR-A62 were used to stimulate 3T3-L1 adipocytes in the absence or presence of a low concentration of insulin (12.5 nM) to determine the effect of IR-A62 on insulin-induced glucose uptake (Fig. 5c). Fifty nanomolar IR-A62 alone did not induce glucose uptake, but cotreatment with insulin potentiated insulin-induced glucose uptake. In addition, the glucose uptake induced by IR-A62 alone was greater at concentrations of 100 nM or 150 nM than 50 nM, and the level of glucose uptake that was induced by insulin and IR-A62 together was greater than that induced by a combination of 50 nM IR-A62 and insulin. These results indicate that IR-A62 cooperatively increases glucose uptake when coadministered with insulin.

Insulin is also a growth factor: it induces the proliferation and growth of cancer cells, principally via the MAPK pathway²¹. In contrast to insulin, IR-A62 had little effect on the MAPK pathway (Fig. 4d, f). However, we also performed a cell proliferation assay in MCF-7 human breast cancer cells to determine whether IR-A62 affected cell proliferation (Fig. 5d). In this assay, insulin-stimulated cell proliferation by up to 1.76-fold, but IR-A62 did not significantly change the number of cells, even at a concentration of 1 µM. Given that the glucose uptake induced by IR-A62 was maximal at ~500 nM, this implies that IR-A62 is a biased agonist that selectively induces the metabolic effects of IR, similar to IR-A48.

IR-A62 reduces glycemia in diabetic mice

Our in vitro data demonstrate that IR-A62 is a biased agonist that induces glucose uptake but not cellular proliferation. Moreover, IR-A62 is stable when exposed to serum nucleases in vitro (t_1/2 = 24.9 h). Therefore, to investigate the effect of IR-A62 on blood glucose in vivo, we compared the effects of subcutaneous injections of insulin or IR-A62 on the blood glucose levels of diabetic mice.

We established a model of type 1 diabetes by administering STZ to C57BL/6 mice, in which the basal glucose levels were maintained at ~450 mg/l (Fig. 6a). The subcutaneous injection of either insulin or IR-A62 markedly reduced the blood glucose levels within 1 h, and these gradually returned to their basal levels over the next 2 h. The kinetics and magnitudes of the effects of insulin and IR-A62 on the blood glucose levels did not differ significantly.

Next, we administered insulin or IR-A62 subcutaneously to ob/ob and db/db mice, which are well-established models of type 2 diabetes²². Both insulin and IR-A62 markedly reduced their blood glucose levels within 1 h (Fig. 6b, c), but the blood glucose levels of mice administered insulin returned to their resting levels within the following 3 h, whereas those of mice administered IR-A62 did not. These results imply that IR-A62 lowers blood glucose to a similar extent to insulin, but the kinetics of its effects on blood glucose differ according to the mouse model used.

To determine whether IR-A62 increases IR and AKT phosphorylation in peripheral tissues in the same way as insulin, we administered insulin (1.5 U/kg) or IR-A62 (10 mg/kg) subcutaneously to normal C57BL/6 mice (n = 3). After 30 min, adipose tissue samples were collected, and AKT (S473) and IR (Y1150) phosphorylation were measured (Fig. 6d). Consistent with the in vitro findings, both insulin and IR-A62 increased AKT and IR phosphorylation. This implies that the ability of IR-A62 to reduce glucose in vivo may be the result of the activation of IR in peripheral tissues.