Anti-diabetic activity of insulin-degrading enzyme inhibitors mediated by multiple hormones

The discovery of a selective, physiologically active inhibitor of insulin-degrading enzyme (IDE) illuminates the therapeutic potential of IDE inhibitors for the treatment of diabetes and reveals that IDE regulates in vivo glucagon and amylin, in addition to insulin. Established treatments for type-2 diabetes include drugs that increase insulin biosynthesis and secretion or improve insulin sensitivity. In theory, inhibition of endogenous insulin degradation could also be effective therapeutically, but despite the identification of the zinc metalloprotease insulin-degrading enzyme (IDE) as the product of a diabetes susceptibility gene, and decades of research, the relationship between IDE activity and glucose homeostasis has remained unclear. Here David Liu and colleagues report the first physiologically active IDE inhibitor, a 20-membered macrocycle that engages a binding pocket away from the enzyme's catalytic site. Treatment of lean and obese mice with this inhibitor shows that IDE not only regulates insulin abundance and signalling, but also that of glucagon and amylin. Acute IDE inhibition with the new inhibitor leads to substantially improved glucose tolerance. This work demonstrates the potential of IDE as a new target for anti-diabetes therapeutics. Despite decades of speculation that inhibiting endogenous insulin degradation might treat type-2 diabetes1,2, and the identification of IDE (insulin-degrading enzyme) as a diabetes susceptibility gene3,4, the relationship between the activity of the zinc metalloprotein IDE and glucose homeostasis remains unclear. Although Ide–/– mice have elevated insulin levels, they exhibit impaired, rather than improved, glucose tolerance that may arise from compensatory insulin signalling dysfunction5,6. IDE inhibitors that are active in vivo are therefore needed to elucidate IDE’s physiological roles and to determine its potential to serve as a target for the treatment of diabetes. Here we report the discovery of a physiologically active IDE inhibitor identified from a DNA-templated macrocycle library. An X-ray structure of the macrocycle bound to IDE reveals that it engages a binding pocket away from the catalytic site, which explains its remarkable selectivity. Treatment of lean and obese mice with this inhibitor shows that IDE regulates the abundance and signalling of glucagon and amylin, in addition to that of insulin. Under physiological conditions that augment insulin and amylin levels, such as oral glucose administration, acute IDE inhibition leads to substantially improved glucose tolerance and slower gastric emptying. These findings demonstrate the feasibility of modulating IDE activity as a new therapeutic strategy to treat type-2 diabetes and expand our understanding of the roles of IDE in glucose and hormone regulation.

Despite decades of speculation that inhibiting endogenous insulin degradation might treat type-2 diabetes 1,2 , and the identification of IDE (insulin-degrading enzyme) as a diabetes susceptibility gene 3,4 , the relationship between the activity of the zinc metalloprotein IDE and glucose homeostasis remains unclear. Although Ide -/mice have elevated insulin levels, they exhibit impaired, rather than improved, glucose tolerance that may arise from compensatory insulin signalling dysfunction 5, 6 . IDE inhibitors that are active in vivo are therefore needed to elucidate IDE's physiological roles and to determine its potential to serve as a target for the treatment of diabetes. Here we report the discovery of a physiologically active IDE inhibitor identified from a DNAtemplated macrocycle library. An X-ray structure of the macrocycle bound to IDE reveals that it engages a binding pocket away from the catalytic site, which explains its remarkable selectivity. Treatment of lean and obese mice with this inhibitor shows that IDE regulates the abundance and signalling of glucagon and amylin, in addition to that of insulin. Under physiological conditions that augment insulin and amylin levels, such as oral glucose administration, acute IDE inhibition leads to substantially improved glucose tolerance and slower gastric emptying. These findings demonstrate the feasibility of modulating IDE activity as a new therapeutic strategy to treat type-2 diabetes and expand our understanding of the roles of IDE in glucose and hormone regulation.
To discover small-molecule modulators of IDE, we performed in vitro selections on a DNA-templated library of 13,824 synthetic macrocycles 7,8 for the ability to bind immobilized mouse IDE, resulting in six candidate IDE-binding molecules (Extended Data Fig. 1). The 20-membered macrocycle 6b (Fig. 1a, half-maximum inhibitory concentration IC 50 5 60 nM) potently inhibited IDE activity in three complementary assays (Extended Data Fig. 2) 9 . We synthesized and biochemically assayed 30 analogues of 6b in which each building block was systematically varied to elucidate the structural and stereochemical requirements (Extended Data Fig. 1), and based on the results we identified the inhibitor 6bK (IC 50 5 50 nM, Fig. 1b) as an ideal candidate for in vivo studies.
The selectivity of 6bK in vitro was $1,000-fold for inhibition of IDE over all other metalloproteases tested, a substantial improvement over the previously reported substrate mimetic hydroxamic acid inhibitor Ii1 10 (Fig. 1d, e). The selectivity of 6bK, coupled with its ability to inhibit IDE in a synergistic, rather than competitive, manner with Ii1 (Extended Data Fig. 2), led us to speculate that the macrocycle engages a binding site distinct from the enzyme's catalytic site (Supplementary Discussion).
We determined the X-ray crystal structure of catalytically inactive cysteine-free human IDE 11 bound to 6b at 2.7 Å resolution (Fig. 1g, Extended Data Fig. 3). Macrocycle 6b occupies a binding pocket at the interface of IDE domains 1 and 2, and is positioned 11 Å away from the catalytic zinc ion (Fig. 1h). This distal binding site is a unique structural feature of IDE compared to related metalloproteases 12 , and does not overlap with the binding site of Ii1 10 . IDE mutations predicted by the structure to impede macrocycle binding led to losses of 6bK potency (Fig. 1f), and complementary changes in 6b analogues rescued inhibition (Supplementary Discussion, Extended Data Fig. 4). The structure predicts that by engaging this distal site the macrocycle precludes substrate binding and abrogates key interactions that are necessary to unfold peptides for cleavage (Supplementary Video) 13,14 .
We characterized the stability, and the physicochemical and pharmacokinetic properties, of 6bK formulated in Captisol 15 , a b-cyclodextrin agent used to improve delivery through intraperitoneal (i.p.) injection at 2 mg 6bK per animal (Supplementary Discussion, Extended Data Fig. 5). The long half-life in mouse plasma (.2 h) and in circulation (.1 h) of 6bK suggested that it was suitable for in vivo studies (Extended Data Fig. 5). Injection of 6bK resulted in high levels of the inhibitor (.100-fold IC 50 ) in peripheral circulation and in the liver and kidneys, the main insulin-degrading organs. In contrast, 6bK was undetectable in brain tissue, where IDE is known to degrade amyloid peptides 5 (Extended Data Fig. 5), and levels of Ab(40) and Ab(42) peptides in mice injected with 6bK were unchanged (Extended Data Fig. 5). Taken together, these findings suggested the viability of 6bK as an in vivo IDE inhibitor.
To evaluate the ability of 6bK to inhibit IDE activity in vivo, we subjected non-fasted mice to insulin tolerance tests 30 min after a single injection with 6bK (2 mg per animal, 80 mg kg 21 ). Mice treated with 6bK experienced lower hypoglycaemia and higher insulin levels compared to vehicle controls (P , 0.01, see below and Extended Data Fig. 5). These experiments establish that a selective and physiologically stable pharmacological IDE inhibitor can augment the abundance and activity of insulin in vivo.
Next we determined the physiological consequences of acute IDE inhibition for glucose tolerance. We used two methods of glucose delivery (oral gavage or i.p. injection 16 ) and two different mouse models (lean or diet-induced obese (DIO) mice 17,18 ). These four conditions were chosen to survey the role of IDE activity under a broad range of endogenous insulin levels and insulin sensitivities 16,17 . Oral glucose administration results in greater insulin secretion compared to injected glucose delivery (Extended Data Fig. 6) that arises from the 'incretin effect' 17,19 . DIO mice display hyperinsulinaemia and insulin resistance, and serve as a model for early type-2 diabetes in humans 18 .
During oral glucose tolerance tests (OGTTs, Fig. 2a and b), both lean and DIO mice treated with 6bK displayed significantly improved glucose tolerance compared to control groups treated either with vehicle alone, or with the inactive stereoisomer bisepi-6bK (Figs 1c, 2a and b; Extended Data Fig. 7). Effects of similar magnitude on oral glucose tolerance in mice have been observed from several human diabetes therapeutics (Extended Data Fig. 8) [19][20][21] . The vehicle and bisepi-6bK control groups exhibited similar blood glucose profiles, indicating that that the effects of 6bK on glucose tolerance are lost when the stereochemistry of 6bK is altered in a way that abolishes IDE inhibition. Collectively, these observations represent the first demonstration that transient IDE inhibition improves blood glucose tolerance 1 .
We repeated the above experiments with 6bK using i.p.-injected glucose tolerance tests (IPGTTs). In contrast to the observed improvement in oral glucose tolerance on 6bK treatment, IDE inhibition with 6bK followed by a glucose injection (1.5 g kg 21 i.p.) surprisingly resulted in impaired glucose tolerance in both lean and obese mice compared to vehicle alone or bisepi-6bK-treated controls ( Fig. 2d and e). DIO mice treated with 6bK followed by glucose injection displayed a biphasic response: glucose levels were lower over the initial 30 min of the IPGTT, followed by a hyperglycaemic 'rebound' starting 1 h after glucose injection ( Fig. 2e and Extended Data Fig. 7). To further test if the effects of 6bK are specific to its interaction with IDE, we repeated these experiments using Ide -/knockout mice 5,6 . Mice lacking IDE were not affected by 6bK treatment and exhibited OGTT and IPGTT blood glucose responses comparable to that of vehicle-treated cohorts ( Fig. 2c and f), suggesting that blood glucose profile improvement during OGTT and impairment during IPGTT of 6bK-treated wild-type mice are mediated by IDE.
The dependence of the physiological response to 6bK on the route of glucose administration cannot be easily explained by a simple model in which IDE only degrades insulin. Instead, these results strongly suggest a role for IDE in modulating other glucose-regulating hormones in vivo beyond insulin. The biochemical properties of IDE and its substrate recognition mechanism 12,13 enable this enzyme to cleave a wide range of peptide substrates in vitro for which experimental validation in vivo has not been previously possible (Supplementary Table 1). Two glucoseregulating hormones, beyond insulin, that are potential candidates for physiological regulation by IDE are glucagon and amylin. Whereas purified IDE can cleave these two peptides in vitro 22,23 , neither hormone is known to be regulated by IDE in vivo or ex vivo.
To probe the possibility that glucagon and/or amylin are regulated in vivo by IDE, we measured plasma levels of these hormones in DIO mice treated with 6bK or vehicle alone following IPGTTs (Fig. 3a), and observed substantially higher levels of insulin, glucagon and amylin. Next we injected each of these three putative substrates 30 min after treatment with 6bK or vehicle alone ( Fig. 3b-d). The 6bK-treated cohorts exhibited significantly stronger blood glucose responses to each of these hormones; mice treated with 6bK experienced hypoglycaemia during insulin tolerance tests (Fig. 3b) and hyperglycaemia following challenges with either amylin 24 (Fig. 3c) or glucagon (Fig. 3d) compared to control animals (via non-physiological activation of the Cori cycle, and gluconeogenesis, respectively) 24-26 . Moreover, in each case the plasma level of the hormone injected remained elevated at the end of the procedure in 6bK-treated mice relative to control animals (Figs 3b-d insets). Collectively these results reveal that IDE regulates the abundance and physiological effects of glucagon and amylin, in addition to insulin.
Amylin is co-secreted with insulin, and is involved in glycaemic control by inhibiting gastric emptying, promoting satiety and antagonizing

LETTER RESEARCH
glucagon secretion 24,25 . Pramlintide (Symlin) is a synthetic amylin derivative used clinically to control post-prandial glucose levels 20,21 . To determine the effects of IDE inhibition on endogenous amylin signalling, we measured gastric emptying efficiency, an amylin-mediated process 25 . Mice treated with 6bK exhibited twofold slower gastric emptying compared to vehicle and bisepi-6bK-treated controls (Fig. 3e). Importantly, coadministration of AC187, a 25-mer peptide amylin receptor antagonist 25 , blocked the effects of 6bK on gastric emptying (Fig. 3e). Collectively, these data reveal that IDE inhibition can slow post-prandial gastric emptying by modulating amylin signalling in vivo at physiologically relevant levels.
Amylin-mediated effects on gastric emptying and satiety have been recognized to have therapeutic relevance 20,21 , and our results demonstrate a small-molecule-based strategy to modulate amylin signalling (Extended Data Fig. 10).
Higher glucagon levels on 6bK treatment provide a possible explanation for impaired glucose tolerance during IPGTT experiments. Substrates are processed by IDE at rates dependent on their relative concentrations (Supplementary Discussion). We observed two-to fourfold higher insulin levels during OGTTs than IPGTTs, consistent with the incretin effect [16][17][18][19] (Extended Data Fig. 6). During an OGTT, IDE inhibition therefore results primarily in an increase in insulin signalling and lower blood glucose levels. In contrast, during an IPGTT, insulin secretion levels are lower [16][17][18][19] , IDE processes proportionally more glucagon, and the loss of IDE activity thus results in higher glucagon and glucose levels (Figs 2d, e and 3a; Supplementary Discussion) [16][17][18][19] .
To test this hypothesis, we repeated the glucose tolerance experiments using Gcgr -/mice that lack the G-protein-coupled glucagon receptor ( Fig. 3f and g) 26 . Treatment of Gcgr -/mice with 6bK followed by an OGTT resulted in improved glucose tolerance as expected (Fig. 3f), consistent with a model in which insulin signalling in these mice is intact and regulated by IDE. In contrast, glucose tolerance in Gcgr -/mice following IPGTT was not impaired by 6bK treatment, consistent with a model in which glucagon signalling is responsible for driving higher glucose levels in wild-type mice on IDE inhibition during IPGTTs (compare Figs 3g and 2d). In addition, 6bK treatment augmented the expression of liver gluconeogenic markers regulated by endogenous glucagon signalling following a pyruvate injection (Supplementary Discussion, Extended Data Fig. 9). These results collectively show that the ability of IDE to regulate glucagon in vivo can account for the impaired glucose tolerance observed during IPGTTs (see the Supplementary Discussion on the hyperglycaemic rebound phase in DIO mice).
The discovery and application of the physiologically active IDE inhibitor 6bK reveals that transient IDE inhibition can improve glucose tolerance under conditions that mimic the intake of a meal. In the context of recent genetics studies identifying IDE as a diabetes susceptibility gene 3,4 , our findings establish the potential of IDE as a target for the treatment of diabetes 1,2 . Our study also reveals a new requirement for therapeutic strategies that target IDE-namely, that transient IDE inhibition during meals, rather than chronic treatment, is desirable to minimize elevation of glucagon signalling (Extended Data Fig. 10

RESEARCH LETTER
may result in synergistic effects 28 ; indeed 6bK treatment results in stronger improvement in oral glucose tolerance when co-administered with the DPP4 inhibitor sitagliptin (Extended Data Fig. 8) 19,20 . In addition, the IDEN6b structure raises the possibility of developing IDE inhibitors that selectively impede insulin degradation without affecting glucagon degradation. These findings collectively inform and motivate additional studies to progress these discoveries towards new diabetes therapeutics.

METHODS SUMMARY
The in vitro selection of the DNA-templated library 8 used 20 mg His 6 -tagged mouse IDE immobilized on cobalt magnetic beads (Invitrogen). IDE inhibition was assayed using the fluorogenic peptide Mca-RPPGFSAFK(Dnp)-OH (R&D), confirmed using an anti-insulin antibody time-resolved FRET assay (Cysbio), and a LCMS assay for CGRP cleavage fragments in plasma 9 . Macrocyclic inhibitors were synthesized by Fmoc-based solid-phase synthesis and purified by HPLC. LCMS quantitation of 6bK in biological samples was performed using 6bK synthesized with 13 C 6 , 15 N 2 lysine (Sigma-Aldrich). Wild-type lean and DIO C57BL/6J age-matched male mice (Jackson Laboratories) were used at 14-16, and 24-26 weeks respectively (.20 weeks of high-fat diet). Gcgr -/and Ide -/mice were fully backcrossed to the C57BL/6J line, bred from heterozygous mice, and used between 11 and 21 weeks. Animals were fasted overnight 14 h for all experiments, except for the insulin tolerance test, which required 5 h of fasting during the morning. Blood glucose was measured from tail nicks using AccuCheck (Aviva) meters. Trunk blood was obtained for plasma hormone measurements using the Multiplexed Mouse Metabolic Hormone panel (Milliplex, EMD Millipore) on a Luminex FlexMap 3D instrument.
Online Content Any additional Methods, Extended Data display items and Source Data are available in the online version of the paper; references unique to these sections appear only in the online paper.   , 1a and 1b). The two isomers were synthesized as previously reported 18,19 and separated by HPLC. c, IDE inhibition potency of trans hits 1b to 6b compared to 30 structurally related 6b analogues in which the linker, scaffold and the three building blocks were systematically varied.   Extended Data Figure 4 | Small-molecule/enzyme mutant complementation study to confirm the macrocycle binding site and placement of the benzophenone and cyclohexyl building-block groups. a, IDE mutant A479L is inhibited by 6b .600-fold less potently compared to wild-type IDE. b, Analogue 9, in which the p-benzoyl ring is substituted for a smaller t-butyl group, inhibits A479L IDE and WT IDE comparably.
c, Similarly, IDE mutant G362Q is inhibited 77-fold less potently by 6b compared with WT IDE. d, Analogue 13, in which the L-cyclohexyl alanine side chain was substituted with a smaller L-leucine side chain, inhibits G362Q IDE and WT IDE comparably. The full list of IDE mutants investigated is shown in Supplementary Table 5. e, Synthetic scheme for 6bK, also used for synthesis of 6bK analogues. Oral glucose, DIO mice co-administered 6bK and DPP4 inhibitor sitagliptin OGTT blood glucose profile areas # Extended Data Figure 8 | Effects of co-administration of 6bK and dipeptidyl peptidase 4 (DPP4) inhibitor sitagliptin followed by oral glucose challenge. a, DIO mice were first treated with either oral gavage of sitagliptin (4 mg kg 21 , 5 ml kg 21 in sterile saline, n 5 14) or saline alone (n 5 12). After 30 min, each group of mice were treated either with a low dose of 6bK (40 mg kg 21 , n 5 7) or vehicle alone (n 5 5), and after an additional 30 min all mice were given a bolus of glucose by gavage (3.0 g kg 21 , 10 ml kg 21 ). Mice treated with the combination of sitagliptin and 6bK displayed glucose levels lower than baseline (t 5 0) after 60 min. b, Blood glucose profile areas of sitagliptin and 6bk were similarly reduced by 60-64% compared to vehicle alone, and further 15% lower when sitagliptin and 6bK were co-administered. Symbol and bar colour-coding: black, vehicle alone; blue, sitagliptin alone; green, 6bK alone; purple, sitagliptin 1 6bK. All data points and error bars represent mean 6 s.e.m. Statistics were performed using a one-tail Student's t-test. Significance levels shown in the figures are: *P , 0.017 (Bonferroni correction) versus vehicle control group, # P , 0.017 versus the 6bK cohort. See Supplementary Methods for a description of the AUC calculation. The data shown are from a study performed once. calculations, and pyruvate tolerance test for gluconeogenesis during IDE inhibition. a, AUC calculations for Fig. 2 data show that during OGTT, lean and DIO mice treated with 6bK display improved glucose tolerance, compared to vehicle controls and inactive bisepi-6bK. b, In contrast, during IPGTT both lean and DIO mice treated with 6bK display impaired glucose tolerance compared to vehicle or bisepi-6bK controls. The AUCs for 6bK versus vehicle treatments using Ide -/mice are similar for both OGTT and IPGTT. c, Wild-type mice fasted overnight were injected i.p. with pyruvate (2.0 g kg 21 ) 30 min after treatment with 6bK, inactive analogue bisepi-6bK, or vehicle alone. d, Plasma hormone measurements 60 min post-PTT reveal elevated glucagon (top) but similar insulin (bottom) levels for the 6bK-treated cohort relative to bisepi-6bK or vehicle controls. e, RT-PCR analysis of liver samples 60-min post-PTT revealed elevated gluconeogenesis transcriptional markers for the 6bK-treated group relative to vehicle controls. All data points and error bars represent mean 6 s.e.m. Statistics were performed using a two-tail Student's t-test, and significance levels shown in the figures are *P , 0.05 versus vehicle control group; **P , 0.01 versus vehicle control group. See Supplementary Methods for a description of the AUC calculation. Data shown not involving knockout mice are representative of two or more independent studies; studies on knockout mice and PTT measurements were performed once.