|Home | About | Journals | Submit | Contact Us | Français|
Insulin-degrading enzyme (IDE, insulysin) is the best characterized catabolic enzyme implicated in proteolysis of insulin. Recently, a peptide inhibitor of IDE has been shown to affect levels of insulin, amylin, and glucagon in vivo. However, IDE−/− mice display variable phenotypes relating to fasting plasma insulin levels, glucose tolerance, and insulin sensitivity depending on the cohort and age of animals. Here, we interrogated the importance of IDE-mediated catabolism on insulin clearance in vivo. Using a structure-based design, we linked two newly identified ligands binding at unique IDE exosites together to construct a potent series of novel inhibitors. These compounds do not interact with the catalytic zinc of the protease. Because one of these inhibitors (NTE-1) was determined to have pharmacokinetic properties sufficient to sustain plasma levels >50 times its IDE IC50 value, studies in rodents were conducted. In oral glucose tolerance tests with diet-induced obese mice, NTE-1 treatment improved the glucose excursion. Yet in insulin tolerance tests and euglycemic clamp experiments, NTE-1 did not enhance insulin action or increase plasma insulin levels. Importantly, IDE inhibition with NTE-1 did result in elevated plasma amylin levels, suggesting the in vivo role of IDE action on amylin may be more significant than an effect on insulin. Furthermore, using the inhibitors described in this report, we demonstrate that in HEK cells IDE has little impact on insulin clearance. In total, evidence from our studies supports a minimal role for IDE in insulin metabolism in vivo and suggests IDE may be more important in helping regulate amylin clearance.
Insulin-degrading enzyme (IDE2 or insulysin) is an evolutionarily conserved zinc metalloprotease belonging to the cryptidase family. Members of this protease family contain a large active site (~15,700 Å3) referred to as a crypt, which can fully enclose substrates (1). Biochemical characterization and crystallography studies have revealed mechanisms whereby IDE binds and cleaves a diverse array of substrates, including insulin, glucagon, amyloid β-peptide (Aβ(1–40) and Aβ(1–42)), ubiquitin, amylin, insulin-like growth factor II, atrial natriuretic peptide, and transforming growth factor α (2,–9). Of its many substrates, IDE is exceptionally effective at degrading insulin (Km = 85 nm and kcat/Km = 2.42 min−1 μm−1) (4). The IDE active site arises from a clamshell-like structure of the enzyme that consists of two concave halves connected by a linker (5). This creates an overall structure similar to a hollow sphere where the catalytic zinc atom is contained in the cavity (Fig. 2A). Because the linker region is flexible, the enzyme has the ability to adopt an “open state” in which insulin can enter into the active site, allowing it access to the catalytic zinc (10). The IDE·insulin complex then adopts a “closed state” where IDE completely encloses insulin. Insulin bound to the closed form of the enzyme undergoes unfolding and an ordered series of cleavage events that produce a mixture of fragments that are then released when the enzyme opens (11).
In addition to having a unique architecture as well as substrate binding and recognition processes, IDE can exist in solution as a monomer and in aggregates with the most active form being the dimer (12,–15). Hersh and co-workers (16) have shown that the individual proteins in the dimer can exert influence on the rate of catabolism when small peptides are present and that these peptides can activate or inhibit substrate catabolism in both a substrate- and ligand-dependent fashion. Additionally, the enzyme is known to be activated by ATP, which destabilizes the closed form of the enzyme (5). Both Türkay and co-workers (17) and Leissring and co-workers (18) have reported the identification of additional unique small molecules that show the ability to activate the enzyme with substrate specificity.
Although much is understood about the biochemical processes utilized by IDE to cleave and inactivate insulin, the cellular and physiological roles of this enzyme are less clear. IDE is mainly found in the cytosol, but it has also been reported to be associated with the plasma membrane, within endosomes and peroxisomes, and IDE has been detected in media from cultured cells (19,–31). Attempts to evaluate the cellular role of IDE have shown that overexpression of IDE in heterologous cell lines increases intracellular degradation of 125I-insulin (32). Similarly, IDE overexpression in cells reduces insulin receptor autophosphorylation (25). Furthermore, fluorescence microscopy experiments using CHO cells expressing the insulin receptor, FITC-labeled insulin, and the peptide hydroxamate-derived IDE inhibitor Ii-1 suggest inhibition of IDE activity decreases intracellular insulin degradation; data from parallel assays suggest the inhibitor may enhance cellular insulin action (33).
Significant efforts aimed at trying to understand the in vivo role of IDE have been carried out using gene deletion studies. Several reports have evaluated IDE−/− mice, but the described phenotype of the knockouts generated by different groups has varied. The initial characterization of IDE knock-out mice indicated that the animals have elevated levels of circulating insulin and are mildly glucose-intolerant (34). Leissring and co-workers (35) later presented evidence indicating IDE-mediated insulin degradation plays a role in glucose homeostasis. In these studies, IDE null mice showed improved glucose tolerance as a result of 3-fold higher fasting serum insulin levels in 2-month-old animals. However, when mice reached 6 months of age, animals developed mild glucose intolerance and insulin resistance. Tissue sample analysis showed the change in glucose metabolism and insulin sensitivity over time likely results from insulin receptor down-regulation due to sustained hyperinsulinemia. In contrast to these studies, characterization of IDE knock-out mice by Steneberg et al. (36) found fasting insulin levels were not significantly changed nor was insulin resistance observed in IDE-deficient animals. Interestingly, in intraperitoneal glucose tolerance tests, these IDE−/− mice displayed suppressed glucose-stimulated insulin secretion. If confirmed, these studies identify a new regulatory role of IDE in insulin secretion whereby IDE forms stable complexes with α-synuclein to reduce α-synuclein oligomerization.
Recently, a cyclic peptide-based IDE inhibitor (compound 6bk, insulin hIDE degradation homogeneous time-resolved fluorescence assay IC50 = 50 nm) has been shown to produce pharmacological effects consistent with IDE being involved in the clearance of glucagon, amylin, and insulin (37). Maianti et al. (37) report several observations from animals treated with inhibitor 6bk. Compound treatment improved glucose clearance during OGTT experiments in lean and DIO mice. In these animals they also observed raised plasma glucose during intraperitoneal glucose tolerance tests. Lean mice treated with inhibitor also showed elevated insulin, amylin, or glucagon levels in trunk blood 60 min after a bolus hormone injection. Enhanced insulin action in an ITT with lean mice treated with compound was also observed. Finally, the researchers also found that compound treatment slowed gastric emptying in mice. Although various roles for IDE in glucose metabolism have been suggested by studies using 6bk, additional questions remain regarding its impact on insulin catabolism. Studies herein identify structurally distinct inhibitors of IDE that allowed evaluating the role of IDE in insulin catabolism in vivo. These studies utilized biochemical and crystallography methods to identify weak ligands that were then refined through molecular modeling into potent small molecule IDE inhibitors. Because IDE has historically been proposed to primarily regulate insulin catabolism and insulin receptor signaling (see above), we tested the hypothesis that inhibiting IDE activity would reduce insulin breakdown and prolong insulin action. This report identifies a new series of N-terminal exosite (NTE) ligands as potent IDE inhibitors and describes the effects of these molecules both in vitro and in vivo. Our results are consistent with the main findings of Maianti et al. (37) but also provide additional insight into the relative importance of IDE for insulin clearance. Furthermore, we investigate the potential of IDE inhibition on enhancing insulin sensitivity in rodents.
Experimental methods and analytical data for the synthesis of NTE-1 and NTE-2 are provided in the supplemental material.
All IDE proteins used in this work were expressed in E. coli and purified by nickel-nitrilotriacetic acid, Mono Q, and size exclusion chromatography (Lilly). Insulin was biosynthetic human insulin (Lilly).
The cysteine-free human IDE-CF-E111Q mutant (IDE-CF: C110L, C171S, C178A, C257V, C414L, C573N, C590S, C789S, C812A, C819A, C904S, C966N, and C974A) was created as described previously (11). A complex with inhibitor was produced by adding 0.25 mm ligand to 15 mg/ml protein 1 h prior to crystallization. Crystallization was set up at 295 K in a 24-well VDX hanging-drop format containing 1 μl of protein (15 mg/ml IDE, 50 mm Tris, pH 8, 150 mm NaCl, 1 mm tris(2-carboxyethyl)phosphine, and 0.5% DMSO) + 1 μl of crystallization solution (20% PEG3350 and 0.2 mm sodium thiocyanate) suspended over 500 μl of crystallization solution. Crystals (100 × 100 × 50 μm cube) grew to full size within 1 week and were frozen in 25% glycerol for data collection. X-ray diffraction data were collected at beam line LRL-CAT at Advanced Photon Source (APS). The structures were solved by molecular replacement (Phaser) using the IDE portion of Aβ-bound IDE-E111Q structure as a search model (Protein Data Bank code 2G47 (38)). Although our experiments used cysteine-free hIDE for structural studies, the reported activity of cysteine-free human IDE and the data herein are very comparable with that of the wild-type enzyme (39). However, kinetic differences between wild-type and cysteine-free rat IDE have been observed (40). Importantly, there are minimal structural differences between the published wild-type and cysteine-free structures (root mean square deviation of Cα ~0.3 Å) (39).
In accordance with recommendations from the National Institutes of Health, we report relative IC50 values for assays used in support of compound screening and optimization (41). Here, the relative IC50 value is defined as “the molar concentration of a substance that inhibits 50% of the v versus [I] curve (top to bottom) for that particular substance. It can also be described as the concentration at which the inflection point is determined, whether it is from a three- or four-parameter logistic fit.”
The degradation of insulin by IDE was followed using an AlphaLISA format (PerkinElmer Life Sciences). In this approach, antibodies that detect total insulin were used; thus, in the absence of inhibition, a decrease in signal due to degradation of insulin is observed, whereas inhibition of IDE maintains a higher signal similar to a no enzyme control. The hydroxamate-based IDE inhibitor Ii1 (33) was used as a positive control. In this assay, Ii1 had an IC50 of 83 nm. The enzymatic reactions were carried out in 50 mm HEPES, 150 mm NaCl, 0.0025% Triton X-100, 10 mm CaCl2, 0.1% human serum albumin, 5% DMSO, pH 7.5. IDE (375 pm) was mixed with 500 pm insulin in the presence and absence of compound, and the reactions were allowed to proceed for 4 h at room temperature. The anti-insulin acceptor beads (10 μg/ml) and anti-insulin streptavidin donor beads (40 μg/ml) were added to the wells, and plates were sealed and read using an Envision plate reader. Measured values were converted to insulin using a standard curve. The relative IC50 values were determined using a four-parameter fit as shown in Equation 1,
In this equation, bottom and top are defined as the plateaus of the curve, and H is the Hill Slope.
No inhibitory effect was observed from addition of Triton X-100, DMSO, or CaCl2 to the assay. The substrate concentration used for the screening assay was purposely selected to be below the published Kd value for insulin (see “Results” and “Discussion” for rationale).
Following the screening campaign and compound optimization, enzyme reactions using both WT-IDE (2 nm) and CF-IDE (0.5 nm) were performed in assay buffer containing PBS, pH 7.4, and 0.1% BSA at 37 °C for 2 h. Insulin (60 nm for WT-IDE or 30 nm for CF-IDE) was then added. After 2 h, the remaining insulin was measured using the PerkinElmer Life Sciences AlphaLISA human insulin assay kit. The AlphaLISA measurements were converted to insulin values using a calibration curve. Data were plotted as insulin versus inhibitor concentration using the single site four-parameter fit model in XLFit software (Surrey, UK). (Note: because this method relies on measuring remaining substrate rather than product formation, the reaction was allowed to progress sufficiently (~30% substrate depletion) such that the amount of insulin degraded would be large enough relative to the assay's precision to allow for reproducible determinations.)
Ex vivo assays using liver lysates from Sprague-Dawley rats were performed to confirm the ability of inhibitors to block insulin degradation. In these experiments, exogenous insulin at a concentration of 29 ng/ml was added to liver homogenates, prepared using T-PER (Thermo Scientific, Rockford, IL), that contained various concentrations of inhibitor. Lysates were incubated at 37 °C for 60 min. Intact insulin remaining in the lysate was measured using an electrochemiluminescence assay (Meso Scale Discovery, Gaithersburg, MD) (43). For comparison, Ii-1 had an IC50 value of 103 nm in this assay.
Varying concentrations of compound were incubated with 1 nm glucagon and 20 pm IDE for 30 min at 37 °C in assay buffer containing PBS, pH 7.4, and 0.1% BSA. After incubation, the amount of glucagon in each reaction was detected using the Meso Scale multi-array mouse/rat glucagon assay kit. The MSD signals were converted to glucagon concentrations by establishing a glucagon standard curve. Percentage of inhibition was calculated using Equation 2,
where Δ[glucagon]buffer = the amount of glucagon degraded in 30 min in the absence of compound; Δ[glucagon]compound = the amount of glucagon degraded in 30 min in the presence of compound. The data were plotted using GraphPad Prism.
The degradation of insulin by IDE was followed by mass spectrometry as described in the supplemental material. Inhibitors were preincubated (100 nm) with 100 nm IDE for 1 h at room temperature in assay buffer (50 mm HEPES, 150 mm NaCl, 0.1% human serum albumin, 10 mm CaCl2, and 1% DMSO, pH 7.5). The preincubation mixture was diluted 100-fold into an assay mixture containing 200 nm insulin. The reaction was stopped at various time points with the addition of 10% trifluoroacetic acid solution, and the degradation of insulin by IDE was followed using mass spectrometry. The dissociation rate constants were determined using GraFit 7.0 (Erithacus Software, Horley, Surrey, UK) and Equation 3:
where [P] is the product concentration at any time (t); υ0 and υs are the initial and final steady-state rates, respectively; and k−1 is the apparent first order rate constant for the establishment of the final steady-state velocity. Under the conditions of this assay, the value of υ0 and the effective inhibitor concentrations were considered to be approximately 0. Thus, the rate of activity regenerated will provide the dissociation rate constant k−1. The final steady-state rate, υs, was determined from a control incubated without inhibitor.
The system consisted of Shimadzu 20 AD pumps, autosampler, controller, and a 20-μl micro-mixer connected to a Thermo Fisher Q-Exactive operated in positive ion mode. The column used in this experiment was an EMD Chromolith C18 (2.0 × 50 mm). Solvent A was water, solvent B was acetonitrile both containing 0.1% formic acid (v/v), and rinse solution contained 50:50 water/acetonitrile with 0.1% formic acid (v/v). The flow rate was 500 μl/min, and the injection volume was 10 μl. The Q-Exactive was operated in full scan mode 250–3000 m/z at 35,000 resolution with an AGC target of 1e6. The instrument was calibrated in positive ion mode using the Thermo Fisher positive ion calibration solution on the same day the samples were run. The S-lens was set to 50 V; the ESI voltage was 3500; the sheath and auxiliary gases were set at 60/40, and the sweep gas was set to 5. The capillary and sheath temperature were set to 400 °C. The gradient profile of the LC system was time (min)/%B 0.0/5, 0.25/5, 7.0/75, 7.01/95, 10.0/95, 10.01/5, and 15/5. The fragments used for determination of dissociation rates are contained in the supplemental material. These fragments were mapped to human insulin from work done by Grasso et al. (44) on bovine insulin degradation by IDE.
To generate preformed IDE·NTE-2 complexes, 1.2 nm IDE in PBS containing 0.1% BSA was preincubated for 30 min at 37 °C with or without 60 nm NTE-2 in a total volume of 1 ml. The IDE·NTE-2 complex was purified by gel filtration on a PD-10 column containing SephadexTM G-25 medium (GE Healthcare) as follows: after column equilibration, the 1-ml sample containing 1.2 nm IDE and 60 nm NTE-2 was loaded on the column, and the IDE·NTE-2 complex was eluted in a total volume of 2 ml. As a control, the 1-ml sample containing only 1.2 nm IDE was purified the same way. Total time for the purification process was 1–2 min.
To start the insulin digestion reaction, 0.5 ml of the purified IDE:NTE-2 solution or the purified IDE solution was mixed with 0.5 ml of 5 nm insulin in PBS containing 0.1% BSA in the presence or absence of 60 nm NTE-2. The IDE digestion reactions were carried out at 37 °C. The amount of undigested insulin at each time point was determined with the Meso Scale assay.
The time course of insulin digestion by IDE can be described by a Michaelis-Menten equation (Equation 4)
where S is insulin concentration; Kcat is the catalytic constant of IDE; and E is the IDE concentration.
where the ProductLog(z) (i.e. Lambert function) is defined as a principal solution for x in Equation 6,
The catalytic constant of the dissociation of the IDE·NTE-2 complex under conditions of no rebinding can be described by Equation 7,
where Kcat (0) is a catalytic rate constant of the IDE·NTE-2 complex, and Kcat (app) is an apparent catalytic rate constant that is the sum of the catalytic rates of the IDE·NTE-2 complex and free IDE.
The time course of insulin digestion by the dissociating the IDE·NTE-2 complex can be obtained by substituting Kcat (app) from Equation 7 for the Kcat in Equation 5. Fitting of the data were performed using Mathematica software (Wolfram Research). The Km value for IDE used for fitting the data was 85 nm.
HEK cells (ATCC, CRL-1573) with and without overexpression of insulin receptor, as wells as HEK cells with insulin receptor overexpression and IDE silencing or IDE overexpression were used for the study. 200,000 cells were seeded per well in a volume of 350 μl in 24-well poly-d-lysine coated plates (BD Biocoat, #354414) and grown overnight in their regular culture medium. The cells were thereafter grown in 225 μl fresh culture medium containing 1 nm insulin and depending on the experiment with 10 μm NTE-2 or vehicle. After 0, 3, 7 and 24 h incubation periods, 25 μl of each sample was collected for insulin measurement. Nondegraded insulin in the samples was measured with MSD insulin Kit (MSD, # K112BZC).
Twenty micrograms of protein cell lysates were loaded onto 4–20% gradient SDS Tris-HCl polyacrylamide gels, separated by electrophoresis, and immobilized onto nitrocellulose membranes by dry blot transfer. After 1 h of incubation in Odyssey blocking buffer (Li-Cor), IDE was detected by a rabbit anti-IDE antibody (Abcam, catalog no. 25970) and the IRDye 800-conjugated goat anti-rabbit IgG secondary antibody (Rockland, catalog no. 926--32211). β-Actin was detected on the same membrane using a mouse anti-β-actin antibody (Abcam, catalog no. ab6276-100) and the Alexa Fluor 680-conjugated goat anti-mouse IgG secondary antibody (Molecular Probes, catalog no. A21058). Fluorescence was captured using the Odyssey Infrared Imaging System (Li-Cor).
400,000 HEK cells/well were seeded in a 24-well poly-d-lysine-coated plate (BD Biocoat, catalog no. 354414) and grown overnight in regular culture medium consisting of DMEM (Hyclone, catalog no. SH30022), FBS (Gibco, 16,000), HEPES (Invitrogen, catalog no. 15630), l-glutamine (Invitrogen, catalog no. 25030), and PGE/streptomycin (Invitrogen, catalog no. 15140). Thereafter, the cells were incubated for 30 min in the culture medium containing various concentrations of NTE-2 and washed 10 times with PBS. The last wash was collected to check for leftover NTE-2. Then cells were lysed in 400 μl of M-per extraction reagent (Thermo Scientific, catalog no. 78501). For the insulin degradation assay, 80 μl of cell lysate was mixed with 20 μl of washing buffer, and 5 nm insulin was added to the reaction mixture. As a control, 300 pm recombinant IDE was added to 5 nm insulin in PBS. The reaction mixtures were incubated at 37 °C for various periods of time. The amount of nondegraded insulin was measured with the MSD assay.
All animals were maintained in accordance with the Institutional Animal Care and Use Committee of Eli Lilly and the National Institutes of Health Guide for the Care and Use of Laboratory Animals.
Male, C57Bl/6 diet-induced obese (DIO) mice were purchased from Taconic Farms (Germantown, NY) and were acclimated to a 12:12 h, light/dark cycle 2 weeks prior to initiation of studies. All animals were maintained ad libitum on a 40% high fat diet (TD95217, Teklad, Madison WI) and water. Mice (n = 6/group) were dosed s.c. two times (16 h apart) with NTE-1 at 15 mg/kg or vehicle (20% Captisol w/v NaPO4 buffer 25 mm, pH 3). After an overnight fast and 2 h following a second administration of NTE-1 or vehicle, whole blood glucose was determined in duplicate using Aviva Accu-check Plus glucose strips, and 30–40 μl of blood was collected for determination of serum hormone levels. Immediately following, all animals were given an oral glucose load of 2 g/kg (50% dextrose, Hospira). Whole blood glucose was determined at 15, 30, 60, 120, and 180 min, and additional blood was collected at the 15- and 30-min time points for serum hormone measurements. All hormones were determined using a mouse metabolic hormone magnetic bead panel metabolism multiplex assay kit (EMD Millipore, Billerica, MA). The data for each analyte were analyzed by repeated measures ANOVA in SAS 9.3 after applying the log transform to satisfy assumptions. The Bonferroni correction was applied in determining the statistical significance.
Male C57Bl/6 DIO mice were acclimated to a 12:12 h, light/dark cycle 2 weeks prior to initiation of studies. All animals were maintained ad libitum on water and 60% high fat diet (D12492, Research Diets, New Brunswick, NJ). DIO mice (n = 8/group) were fasted for 4 h. Two h after the initiation of the fast, whole blood glucose was determined, and immediately following, the mice were dosed subcutaneously with NTE-1 (15 mg/kg) or vehicle. Two h later, insulin was administered at 0.6 units/kg intraperitoneally (i.p.), and glucose was determined 15, 30, 60, and 90 min later.
Male Sprague-Dawley rats (Harlan Laboratories, Indianapolis, IN) were housed in standard cages and light cycle (12:12 h). One week prior to the studies, chronic indwelling catheters were placed in the left carotid artery and advanced to the aortic arch. Two catheters were also placed in the right jugular vein and advanced to the right atrium. Animals were fasted overnight and allowed to acclimate to the study cages for 2 h. All studies were 4 h in length. A bolus/infusion of vehicle (20% Captisol) or NTE-1 was initiated at −120 min at a concentration and flow rate that would yield a steady-state plasma concentration of ~2 μm NTE-1 for the duration of the clamp. At −60 min, a bolus/continuous infusion of [3-3H]glucose (PerkinElmer Life Sciences, 6 μCi bolus and 0.1 μCi/min) was initiated and maintained throughout the rest of the clamp. At time 0 min, a primed-continuous infusion of 0.5 milliunits/kg/min insulin was administered, and a variable infusion of 20% glucose was started and periodically adjusted to maintain blood glucose concentration at 5.5–6.0 mm. Somatostatin (American Peptide, Vista, CA, 5 μg/kg/min) was administered to inhibit endogenous insulin secretion. Arterial blood samples were obtained during the experiment to monitor hematocrit, plasma insulin, rat C-peptide, NTE-1 concentrations, and to determine basal and clamp hepatic glucose production. At 120 min, animals were sacrificed by intra-arterial administration of pentobarbital. Animals were perfused with saline; liver samples were rapidly excised, washed in ice-cold saline, snap-frozen with tongs cooled in liquid nitrogen, and stored at −80 °C.
Male Sprague-Dawley rats (Harlan Laboratories; housed as described above) were fasted for 2 h and then administered either vehicle (20% Captisol w/v NaPO4 buffer 25 mm, pH 3) or NTE-1 (15 mg/kg) via subcutaneous injection. Thirty min later, animals were injected intravenously with a single bolus of the hormone (insulin at 1 nmol/kg, amylin at 250 μg/kg, or glucagon at 100 μg/kg). Plasma hormone levels were measured periodically over the next 60 min using the respective Meso Scale assays discussed above, and amylin levels were determined using an ELISA (EMD Millipore).
Compound concentrations in plasma or tissue matrix were determined by homogenization, extraction with acetonitrile, and quantification using LC/MS.
One-way ANOVAs followed by Dunnett's test was used to determine compound-related changes relative to vehicle control. These methods were used for all studies unless otherwise noted.
In undertaking a search for small molecule inhibitors of insulin degradation by IDE, there are several aspects of this enzyme that we considered in selecting an assay to support compound library screening. For example, the regulation of the biochemical activity of IDE by small molecule activators has been shown in several cases to be substrate-specific (16,–18). Substrate specificity among IDE small molecule inhibitors has also been reported (45). Furthermore, the binding of IDE to insulin has been shown by isothermal calorimetry to fit a two binding site model (probably due to the propensity of IDE to exist as a dimer) with Kd values of 10 and 140 nm (11). CF-IDE was found to give comparable Kd values (20 and 280 nm) (11).
In the search for inhibitors of IDE-mediated catabolism of insulin, we wanted to ensure that molecules capable of inhibiting formation of the IDE·insulin complex could be detected in the screening assay. Therefore, we used a medium through-put screen-compatible biochemical assay designed to measure the degradation of insulin by IDE, which could be performed at insulin concentrations below the Kd value.
A screening campaign testing >210,000 structurally diverse compounds identified two modest inhibitors that were leveraged as chemical starting points for design of the potent IDE inhibitors disclosed in this study. We also developed rodent hepatocyte and liver tissue lysate assays to assess the ability of compounds to inhibit wild-type rat and mouse IDE.
A series of dipeptide aniline amides (e.g. 1, Fig. 1) was identified as IDE inhibitors. These compounds had relative IC50 values of 1–10 μm against CF-hIDE. These compounds did not produce 100% inhibition in the IDE screening assay. Initial attempts to co-crystallize the molecules with IDE provided structures that suggested the compounds bind near an exosite that interacts with the N terminus of the A-chain of insulin (Fig. 2B) (11). Importantly, our x-ray crystallography studies of the apo enzyme also showed partially resolved density near this exosite implying that endogenous peptide carried over from the expression and purification process might have been retained in this exosite. To ensure the density observed was due to the compound, selenomethionine amide 1 was synthesized (Fig. 1). Incorporation of the Se atom enabled unequivocal determination of the binding mode of the ligand (Fig. 2C). The x-ray structure of 1 confirmed that the dipeptides bind to the insulin A-chain N-terminal exosite (Fig. 3A). This is the same IDE exosite occupied by the peptide activators described by Hersh et al. (16).
The x-ray co-crystal structure of IDE and selenomethionine amide 1 identified several interactions that drive binding of the ligand to the enzyme (Fig. 2C). Under physiological conditions, the primary amine of the ligand should be protonated and make hydrogen bonds with the Gly-339 and Leu-359 carbonyls as well as a charge-charge interaction with the Glu-341 side chain. The x-ray structure also shows hydrogen bonding interactions between the selenomethionine carbonyl and the Val-360–Gly-361 amide bond. The aniline proton is also engaged in a strong hydrogen bonding interaction with the Gly-361 carbonyl. Finally, there are two hydrophobic interactions between the enzyme and the ligand; the first is with the phenyl ring and the Val-360 and Ile-374 side chains. The other is between the cyclohexyl and the protein surface created by the Gly-335, His-336, and His-332 residues. Significant structural activity relationship assessment of the scaffold was carried out by rapid synthesis of over 100 analogues using Lilly's Automated Synthesis Lab (46). Characterization of the compounds showed that elimination of any of the polar or hydrophobic interactions caused significant loss in potency. Furthermore, we were unable to identify analogues that produced full inhibition in the screening assay.
Considering the mode of binding between selenomethionine amide 1 and IDE, it is clear that it could act as an inhibitor of insulin that competes with substrate for the N-terminal exosite. Given that the maximal inhibition for this series was only ~70% in the IDE screening assay, the relative IC50 is 1–2 μm for the most potent compounds, and the scaffold showed low stability in plasma due to protease degradation (data not shown). This series was not an optimal starting point for generating compounds suitable for in vivo studies.
A second inhibitor series identified via screening is represented by quinoline 2 (Fig. 1). This compound also did not produce 100% inhibition in the screening assay (maximal inhibition was ~60%) with a relative IC50 of 2 μm against CF-hIDE (Fig. 4A). This molecule readily co-crystallized with the enzyme (Fig. 2D) to reveal a unique binding mode. Key interactions of this ligand series are primarily hydrophobic with the quinoline ring being deeply buried within a hydrophobic exosite and the difluorophenyl engaging in an edge to face interaction with Tyr-314. The x-ray structure also suggests a hydrogen bonding interaction between the side chain of Glu-205 and the carbamate.
It is important to note that the hydrophobic exosite occupied by quinoline 2 is not occupied by insulin in any of the published insulin-IDE co-crystal structures (Fig. 3B). Furthermore, it is far removed from the catalytic zinc residue (>18 Å, see Fig. 2, B and D). This same hydrophobic exosite is also occupied by inhibitor 6bk (37). Synthesis of over 50 analogues failed to provide molecules that could produce 100% inhibition in the screening assay. Furthermore, compound 2 was the most potent inhibitor identified within the series.
In overlays of quinoline 2 with published IDE-insulin co-crystals, we observed no negative steric interactions that might explain how quinoline 2 inhibits insulin degradation (Fig. 3B). However, given the highly dynamic process of IDE transitioning between “open” and “closed” states, insulin binding, insulin unfolding, and subsequent multiple amide bond cleavage events, it is clear that all of the reported insulin-IDE co-crystal structures reveal only a single moment within a complicated ballet of molecular movements and interactions. Therefore, with the available data, it is not possible at this time to fully explain mechanistically how quinoline 2 inhibits IDE-mediated degradation of insulin.
Although neither of the inhibitors presented a clear opportunity to develop a potent IDE inhibitor, we hypothesized that linking together the ligand systems would produce a hybrid molecule with improved binding affinity. To explore this concept, computer modeling was used to audition a number of different sites of connection as well as linker element composition and length. In addition, to enhance plasma stability, the peptide backbone was masked with a cyclic lactam motif that allows retention of the hydrogen bonding interactions of the open chain form but retards proteolysis in plasma. From the set of the most promising opportunities generated in silico, several compounds were synthesized and assessed for their IDE inhibitory and pharmacokinetic performance characteristics. The compound designated as NTE-1 (Fig. 1) had the best combination of enzyme activity and rodent pharmacokinetic properties. Importantly, NTE-1 is easily prepared on a reasonable scale producing quantities of material suitable for in vivo pharmacology experiments.
Co-crystallization of NTE-1 with IDE confirmed the molecule occupies both exosites simultaneously. NTE maintains all the key interactions noted within each of the individual ligands 1 and 2 with the exception of the carbamate hydrogen bond-Glu-205 interaction (Fig. 2E).
NTE-1 is a potent inhibitor of CF-hIDE with a relative IC50 of 4 nm in the screening assay (Fig. 4A and Table 1). Because the screening assay used insulin concentrations below the Kd value of insulin at room temperature, we tested NTE-1 in the insulin degradation assay using 30 nm insulin at 37 °C (insulin Km = 85 nm) (Fig. 4B) (11). Under these conditions, NTE-1 was a potent inhibitor of CF-hIDE with an IC50 of 11 nm (Table 2). It has been reported that significant kinetic differences between wild-type and cysteine-free rat IDE have been observed (40). Therefore, NTE-1 was also tested against WT-hIDE (Fig. 4C). NTE-1 also demonstrated robust inhibition in this assay with an IC50 of 15 nm. Furthermore, in an ex vivo rat liver lysate insulin degradation assay, NTE-1 shows a relative IC50 of 18 nm, suggesting nonspecific protein binding does not significantly affect the ability of the compound to inhibit the enzyme (Fig. 4D).
The close structural analogue NTE-2 also showed similar IDE inhibitory activities (WT-hIDE IC50 = 18 nm, rat lysate relative IC50 = 68 nm, see Tables 1 and and2),2), and x-ray co-crystal structure determination shows it binds IDE similarly to that of NTE-1 (Fig. 2F). Both NTE compounds also showed potent inhibition of insulin degradation in mouse liver lysates (data not shown). In addition, extensive internal profiling of NTE-1 as well as compounds 1 and 2 did not show cross-reactivity against other enzymes (kinases, proteases, or other enzyme classes) or cell surface receptor targets. The NTE compounds were not tested against other cryptidase enzymes such as neprilysin, presequence peptidase, or pitrilysin.
Although the new NTE series was identified and optimized using assays to measure the ability of compounds to inhibit cleavage of insulin, other peptide substrates have also been shown to be degraded by IDE (see above). To assess inhibition of the NTE compounds on other substrates, compounds were tested for their ability to inhibit CF-hIDE-mediated degradation of glucagon. In these assays, NTE-1 and NTE-2 are potent inhibitors of glucagon degradation with IC50 values of 3 and 150 nm, respectively (Table 1 and Fig. 5).
To determine whether a structural basis for the inhibitory activity of IDE-mediated degradation of the new ligands versus glucagon could be postulated, we compared the NTE-1-IDE co-crystal to published x-ray co-crystal structures of IDE with glucagon (Protein Data Bank code 2G49) and amylin (Protein Data Bank code 2G48) (38). Fig. 6A shows an overlay of the resolved segments of glucagon from this structure with inhibitor NTE-1. This overlay reveals the common element of binding to the N-terminal exosite between NTE-1 and glucagon. This commonality was also observed when comparing the NTE-1 x-ray to the amylin-IDE co-crystal (Fig. 6B) (38). The important nature of the N-terminal exosite interaction with the substrate hormones is further supported by the fact that peptide 1 shows modest inhibition of glucagon degradation, similar to its activity in the insulin assay (Table 1). Neither the amylin-IDE or glucagon-IDE co-crystal structures provide clear evidence of engagement with the hydrophobic exosite occupied by compound 2. However, in contrast to compound 2 showing modest inhibition in the insulin degradation screening assay, no inhibition of glucagon degradation was observed.
Kinetic studies show NTE-1 has a slow dissociation rate and a long residence time. To determine koff and t½, a 1:2 mixture of IDE and NTE-1 (200 nm) or DMSO was incubated for 2 h (Fig. 7A). The reaction was then diluted (100-fold) in the presence of a high concentration of insulin (10× Km). The changes in rate observed reflect the dissociation of the IDE·NTE-1 complex. The koff and t½ were determined to be 0.0047 min−1 and 2.45 h, respectively.
An alternative approach was used to determine the dissociation rate of NTE-2 with IDE. Here, we preformed the IDE·NTE-2 complex and then purified it by rapid gel filtration. The complex was diluted in the presence of insulin, and the rate of insulin degradation versus time was plotted, and the data were fit to a modified Michaelis equation (Fig. 7B). In these experiments, the koff was determined to be 0.011 min−1, and the t½ was calculated to be 63 min. Thus, the IDE·NTE-2 complex also displayed slow dissociation kinetics similar to NTE-1.
The pharmacokinetic properties of NTE-1 were measured in mice and rats to determine whether the molecule is suitable for conducting in vivo pharmacology experiments (Table 3). In C57Bl/6 mice, a 5 mg/kg subcutaneous dose of NTE-1 resulted in a compound exposure AUC of 14,400 ng·h/ml with a Cmax of 1880 ng/ml and t½ of 9 h. In Sprague-Dawley rats, NTE-1 has a lower AUC of 1200 ng·h/ml with a Cmax of 230 ng/ml and a t½ of 16 h after subcutaneous administration. With a potent IDE inhibitor capable of sustaining concentrations of compound in vivo well in excess of its biochemical IC50, NTE-1 can be used to explore the pharmacological consequences of inhibiting IDE in animals.
Because NTE-1 possesses pharmacokinetic properties suitable for in vivo studies, experiments in rodents were performed to evaluate the effect of pharmacologically inhibiting IDE on insulin action. DIO mice receiving NTE-1 (15 mg/kg s.c.) showed improved glucose tolerance, consistent with the findings of Maianti et al. (37) with 6bk (Fig. 8). NTE-1 treatment also resulted in an increase in plasma insulin at 15 and 30 min post-glucose challenge. Plasma amylin was found to be higher at 0, 15, and 30 min. Glucagon concentrations were not changed in these experiments. However, basal glucagon levels are typically low during OGTT experiments (37). It is important to note that in contrast to the report indicating Ide null animals display impaired insulin secretion (37), insulin levels were not negatively affected by NTE-1 treatment, possibly suggesting a difference between loss-of-function versus pharmacological inhibition of IDE.
To further assess the ability of NTE-1 to mediate the degradation of insulin in vivo, we performed ITT experiments in DIO mice (Fig. 9). In these assays, there was no improvement in insulin-induced glucose lowering by the compound, suggesting plasma insulin concentrations were not elevated. To specifically address this, insulin clearance studies were performed in Sprague-Dawley rats (Fig. 10A). Here, insulin concentrations in plasma were not increased in animals administered NTE-1. To confirm IDE was inhibited, at the conclusion of the insulin clearance experiments livers were collected at both 5 and 60 min following the insulin bolus. The data in Fig. 10B show strong inhibition of insulin degradation in liver lysates from animals administered NTE-1 compared with vehicle-treated controls.
We also examined the ability of NTE-1 to block amylin and glucagon degradation in lean Sprague-Dawley rats (Fig. 10, C and D). Consistent with the DIO mouse study shown in Fig. 8, we observed that NTE-1 treatment produced sustained amylin levels after an i.v. amylin challenge. In contrast, NTE-1 treatment did not show enhanced plasma glucagon levels when rats were given an i.v. glucagon challenge.
To further explore the importance of IDE on insulin action and clearance in vivo, euglycemic clamp experiments using Sprague-Dawley rats were performed (Fig. 11). To ensure endogenous glucagon, insulin, and amylin release was suppressed, animals were infused throughout the entire experiment with somatostatin (42, 47,–49). At the start of the study, NTE-1 was infused at high concentrations (plasma levels of NTE-1 were sustained at >2 μm) over 2 h while the glucose levels in the animals were stabilized. After the stabilization period, a low fixed-rate insulin infusion that achieved post-prandial insulin levels in plasma was initiated, and the glucose infusion rate was adjusted to maintain euglycemia. Consistent with the lack of insulin sensitization observed in the ITT in DIO mice, NTE-1 treatment did not alter the glucose infusion rate or suppress hepatic glucose production in either the basal period or after insulin infusion (Fig. 11, A and B). Furthermore, no enhancement of plasma concentration of insulin with compound treatment was observed.
Following the euglycemic clamp experiments, livers were harvested from animals to confirm the NTE-1 concentrations achieved were sufficient to inhibit insulin degradation. Liver lysates were prepared; exogenous insulin was added, and homogenates were incubated at 37 °C for 60 min. Insulin levels were dramatically decreased in lysates from animals administered vehicle, but livers from NTE-1-treated animals showed high concentrations of insulin, indicating sufficient levels of the compound were achieved to assess the pharmacodynamic effect of inhibiting IDE (Fig. 11B, inset). Consistent with these pharmacokinetic assessments, we determined the average concentration of NTE-1 in the livers to be ~70 mg/g.
Because the in vivo studies failed to show a robust effect of IDE inhibition on insulin clearance or action, studies were undertaken to evaluate the role of IDE inhibition on the cellular clearance of insulin. Our initial investigations used HEK293 (HEK) cells (Fig. 12). HEK cells were incubated in media containing insulin in the absence or presence of NTE-2. As shown in Fig. 12A, little insulin was cleared from media of normal HEK cells over 24 h. To establish that NTE-2 was cell-penetrant, we treated HEK cells with various concentrations of NTE-2 for 30 min and then washed the cells 10 times with PBS. The washed cells were lysed, and the degradation of exogenously added insulin in the lysate was assessed (Fig. 12B). NTE-2 treatment significantly blunted insulin degradation at all doses. To ensure that all extracellular NTE-2 had been removed by washing, recombinant IDE was added to the final PBS wash; insulin was added, and the course of the reaction was monitored (Fig. 12C). The data in Fig. 12, A–C, suggest that in HEK cells inhibition of IDE (both extra- and intracellular) has little effect on insulin clearance.
The rate of insulin clearance in HEK cells overexpressing the insulin receptor (HEK-IR) was increased relative to the parent HEK cells (Fig. 12D). However, addition of NTE-2 at 10 μm did not significantly affect insulin clearance in HEK-IR cells. Similar results were found in studies using HEK-IR cells where IDE levels were either reduced by >80% using shRNAs targeting IDE (Fig. 12E) or increased by 3-fold via IDE overexpression (Fig. 12F). Relative expression levels of IDE in HEK-IR cells used in these experiments are shown in Fig. 12G. In total, the data in Fig. 12 suggest that insulin clearance by HEK cells is mediated primarily by the insulin receptor and that this process is not impacted by IDE inhibition.
The studies in this report describe the identification and characterization of a novel class of IDE inhibitors. An integrated discovery strategy utilizing screening with insulin as the IDE substrate, protein crystallography and three-dimensional modeling enabled design of the NTE series. Key to the success of this work was co-crystallization studies showing the binding interactions of two distinct small molecule ligands (both identified from the screening campaign) with separate exosites within IDE. These data informed molecular docking and linker geometry design that led to constructing molecules with substantially improved potencies (>1000-fold). Interestingly, the act of linking the molecules also produced compounds with slow dissociation rates and increased the maximal IDE inhibition. Finally, cyclizing the N terminus of the ligand improved plasma stability. As a result, NTE-1 has pharmacokinetic properties and potency sufficient to support pharmacology studies in rodents to assess IDE inhibition. Furthermore, NTE-1 acts as an inhibitor of the IDE-mediated catabolism of glucagon in vitro.
Consistent with the observations of Maianti et al. (37), when NTE-1 was dosed to DIO animals prior to an oral glucose challenge, we observed increases in glucose clearance and plasma amylin levels. Similarly, NTE-1 also inhibited clearance of plasma amylin in cannulated Sprague-Dawley rats.
However, in contrast to our observation of increased insulin levels during OGTT experiments, neither ITT experiments in DIO mice or euglycemic clamp studies in rats showed increased plasma insulin levels or improvements in insulin sensitization following NTE-1 administration. This is in contrast to the findings of Maianti et al. (37) where studies show plasma insulin was increased in trunk blood of lean mice following treatment with the 6bk inhibitor during an ITT experiment.
Based on these data, it appears that IDE catabolic activity plays an in vivo role in amylin clearance such that inhibition of catabolism brings significant changes in amylin levels in plasma. However, with regards to insulin clearance and action, the data suggest IDE catabolism plays a more limited role. It should be acknowledged that it is impossible to exclude the explanation that NTE-1 and the cyclic peptide IDE inhibitor 6bk act differently in vivo due to differences in biochemical mechanisms of action or biodistribution. The compounds have some structural differences in how they bind to the enzyme. They were also dosed at different levels and via different routes (80 mg/kg i.p. for 6bk versus 15 mg/kg s.c. or infusion for NTE-1). Further experimentation will be needed to adequately explore these hypotheses.
Our cellular studies suggest HEK cells process insulin primarily through an IR-mediated mechanism that is not highly impacted by inhibition of IDE activity or IDE expression levels (extracellular or intracellular). To the best of our knowledge, the relationship of IDE to amylin and glucagon clearance at the cellular level has yet to be investigated. Hopefully, research to assess the plasma and tissue levels of IDE and how those levels might be regulated by nutrients such as glucose will help to further put the results in this report into context.
The ability to catabolize a multitude of substrates and its broad tissue expression have implicated IDE in a variety of physiological processes and clearance mechanisms, well beyond those directly associated with insulin. However, establishing a primary catabolic role in vivo for this unique enzyme has been difficult for over 50 years, in part due to a paucity of potent and selective inhibitor molecules. Experiments to better understand the primary physiological role of IDE and its action at the cellular level should be facilitated by the availability of a potent IDE inhibitor such as NTE-1.
T. B. D. and K. W. S. were involved in the design of all experiments and wrote the manuscript. The NTE molecules were synthesized by J. L. T. and T. B. D. T. B. D., V. J. K., and M. J. C. designed the NTE molecules. V. J. K. carried out the molecular modeling. Y. W. and S. Y. G. conducted the x-ray crystallography experiments and solved the structures. In vivo studies in rodents were designed by K. W. S., M. D. M., A. M. S., J. A.-C., M. C., D. K. S., and T. B. D. The cell lysate assay was developed by K. W. S. and J. X. C. C. Cellular studies were designed by V. V. K. and J. X. C. C. Biochemical assays using NTE molecules were developed by V. V. K. and G. Y. W. Off-rate determination of NTE-1 with IDE was completed by S. J. T., J. S., and R. N. C. Off-rate determination of NTE-2 with IDE was completed by V. V. K. and G. Y. W. Pharmacokinetics experiments were designed by N. A. C., T. B. D., and K. W. S. The screening assay was developed by J. E. M. and J. T. D. All authors approved the manuscript.
*All authors were employees and shareholders of Eli Lilly and Company at the time this work was completed.
This article contains supplemental Schemes 1–12, Fig. 1, and additional references.
2The abbreviations used are: