An Isoform-Selective PTP1B Inhibitor Derived from Nitrogen-Atom Augmentation of Radicicol

Abstract

A library of natural products and their derivatives was screened for inhibition of protein tyrosine phosphatase (PTP) 1B, which is a validated drug target for the treatment of obesity and type II diabetes. Of those active in the preliminary assay, the most promising was compound 2 containing a novel pyrrolopyrazoloisoquino-lone scaffold derived by treating radicicol (1) with hydrazine. This nitrogen-atom augmented radicicol derivative was found to be PTP1B selective relative to other highly homologous nonreceptor PTPs. Biochemical evaluation, molecular docking, and mutagenesis revealed 2 to be an allosteric inhibitor of PTP1B with a submicromolar K_i. Cellular analyses using C2C12 myoblasts indicated that 2 restored insulin signaling and increased glucose uptake.

Protein tyrosine phosphatases (PTPs) and kinases (PTKs) are signaling enzymes that modulate phosphorylation of proteins and play crucial roles in many essential cellular pathways including mitosis, metabolic regulation, cell proliferation, and apoptosis.¹ PTKs catalyze the phosphorylation of tyrosine residues of proteins while PTPs reverse this action by hydrolyzing the tyrosine phosphate. This addition/subtraction mechanism allows for careful regulation of cellular signaling events. Because of their critical biological functions, both the PTKs and PTPs have been pursued as drug targets for a variety of pathologic states including diabetes, cancer, and neurological and immune disorders.² However, to date, only molecules targeting PTKs have entered the clinic,³ but discovery of clinically viable PTP modulators remains an area of active research.⁴

The nonreceptor PTP family member, PTP1B, is a validated target for the treatment of diabetes and obesity.⁵ PTP1B is ubiquitously expressed in the cytoplasm of all cells and localizes to the outer face of the endoplasmic reticulum.⁶ PTP1B-mediated dephosphorylation inactivates both the insulin and leptin signaling pathways.⁷ In the case of insulin signaling, PTP1B dephosphorylates the active insulin receptor (IR) and its downstream targets including insulin receptor substrates 1, 2, and 3.⁸ In leptin signaling, PTP1B antagonizes JAK2 activity,⁹ which is immediately downstream of the leptin receptor (ObR). PTP1B is therefore a negative regulator of both the insulin and leptin signaling pathways, interfering with both lipid and glucose homeostasis. In addition, PTP1B levels have been shown to be increased in patients with metabolic syndrome.¹⁰ Genetic ablation of PTPN1, the gene encoding PTP1B, leads to an increase in insulin sensitivity and resistance to obesity in animal models on a high fat diet.¹¹

Many proteins that are desirable drug targets remain without viable clinical leads because they are “undruggable”.¹² This can be due to a variety of reasons, but in the case of PTP1B, this is because of a charged, shallow active site that is conserved among all PTPs.¹³ Although the issue of active site selectivity has been addressed through the use of complex synthetic molecules,^1b,14 often these molecules suffer from poor pharmacological properties.¹⁵ Roadblocks to human trials with previously discovered molecules have arisen primarily due to lack of PTP1B specificity and/or poor pharmacology. Allosteric inhibitors are an attractive alternative to active site inhibitors, as allosteric sites vary across PTPs, giving heightened selectivity, and these sites are potentially more druggable than the active site. This approach was employed by Wiesmann et al. in 2004, who reported the PTP1B inhibitory activity of benzofuran derivatives. These molecules were shown to bind to a hydrophobic pocket, discovered through X-ray crystallography, located ~20 Å from the active site.¹⁶ Computational studies suggest that these derivatives keep PTP1B in an inactive state, in which the WPD loop (a conserved loop in PTPs essential for activity) is in an open conformation, by interrupting the triangular interaction among helix 7α, helix 3α, and loop 11.¹⁷ Other approaches have also been employed to tackle the inherent difficulties of drugging PTP1B. Krishnan et al. recently identified trodusquemine, also known as MSI-1436, which targets the disordered C-terminus of PTP1B.¹⁸ In addition, there has been a flurry of publications reporting both natural product and synthetic small-molecule PTP1B inhibitors.^14,19 While these results show encouraging progress and offer critical insight into PTP1B as a viable therapeutic target, more isoform-selective PTP1B inhibitors with improved pharmacological properties and enhanced potency are still needed, since a clinically effective drug targeting PTP1B has yet to be discovered.

Natural products constitute a rich source of biologically active molecules.²⁰ Many of the drugs that are available on the market either mimic naturally occurring molecules or consist of structures that are fully or in part taken from such natural motifs.²¹ The value of natural products as platforms for developing front-line drugs is in part due to the fact that their chemical diversity is complementary or superior to that found in synthetic libraries. It is noteworthy that compared with synthetic and combinatorial counterparts, natural products are stereochemically more complex and have a broader diversity of ring systems, structural features that have resulted from a long evolutionary selection process.²² However, less than 20% of the ring systems found in natural products are represented in current commercial drugs;²³ whereas drugs and combinatorial molecules contain a higher number of heteroatom-containing groups and/or ring systems.²⁴ In continuing our work on structural diversification of natural products,²⁵ we have recently utilized a chemical strategy involving hydrazine-mediated nitrogen (N)-atom augmentation of carbonyl compounds abundant in nature to obtain more druglike compounds.²⁶ Generation of natural product-like compound libraries by chemical modification of compounds in complex natural mixtures of unknown composition has been described previously.²⁷

In an effort to discover isoform selective PTP inhibitors, we cloned, expressed, and purified a panel of highly conserved nonreceptor PTPs. Using this panel, we screened a small library (500 members) of natural products and their N-atom augmented derivatives using a well characterized p-nitrophenol phosphate (pNPP) assay²⁸ with a Z-factor >0.8 (Supporting Information, Table S1). The library was made up of purified marine and terrestrial natural products with diverse scaffolds in addition to a much smaller subset of derivatized natural products. From this screen, we found 14 hits that were reconfirmed in triplicate and in dose−response giving an ~2.8% hit rate. Four of the hits were from our N-atom augmented compounds. Interestingly, of those tested, the most promising, based on both potency and specificity, was the dihydropyrrolopyrazoloisoqionolone (2) resulting from N-atom augmentation of the well-known HSP90 inhibitor, radicicol (1) (Scheme 1). Reaction of the macrocyclic benezenediol lactone, radicicol (1) with hydrazine hydrate in EtOH afforded a mixture of N-augmented analogues 2−5 containing a hitherto unprecedented carbon skeleton (Scheme 1). The structures of compounds 2−5 were characterized by extensive NMR analysis (see discussion in the Supporting Information; pages S6 and S7, Figures S1–S24). However, compound 2 could be isolated as the major product in 71% yield if the reaction was conducted for 16 h exposed to air. In this case, the reduction of 3 leading to 2 is probably mediated by hydrazine.²⁹ Compound 2 could also be accessed indirectly by Pd/C-catalyzed hydrogenation of 3 in quantitative yield (Scheme 1). The structure of 2 was unambiguously confirmed by X-ray crystallography (Figure 1; CCDC: 1880627). Aiming to improve the bioactivity of radicicol (1), attempts to modify its structure have previously drawn attention. However, all of the efforts have focused on finding more druglike HSP90 inhibitors and have left the core structure unchanged.³⁰ By contrast, we completely reshaped radicicol (1) with simple hydrazine treatment and shifted the bioactivity from HSP90 to PTP1B.

Scheme 1.

N-Atom Augmentation of Radicicol (1) Resulting in Compounds 2, 3, 4, and 5

Figure 1.

X-ray crystallographic structure of 2, showing displacement ellipsoid plots (50% probability).

We propose the formation of compounds 2−5 begins with the Michael addition of hydrazine to radicicol (1), leading to intermediate A. Then A can undergo an intramolecular transesterification affording B. Intermediate B could be transformed to intermediate C through hydrazinolysis of the lactone. An S_N2 attack of the epoxide by hydrazine, resulting in the reversal of the absolute configuration at the reactive carbon center, would lead to intermediate D. Subsequently, the tetracyclic compound 5 can be generated via hydrazine addition to the ketone. Elimination of water from 5 would provide compounds 3 and 4. Compound 3 was likely reduced by another equivalent of hydrazine to furnish compound 2 (Scheme 2).

Scheme 2.

Proposed Mechanism for the Formation of 2, 3, 4, and 5 from Radicicol Treated with Hydrazine Hydrate

Compound 2 was found to inhibit PTP1B with very good to excellent selectivity versus six other nonreceptor PTPs (for SDS-PAGE of PTPs see the Supporting Information, Figure S25). These hits were confirmed in triplicate, and then a 10-point dose−response was carried out to determine the IC₅₀ values for each compound. Once validated as bona fide PTP1B inhibitors, the IC₅₀ values relative to the other PTPs used in the initial screen were also measured. Compound 2 showed >10-fold selectivity for PTP1B relative to TCPTP, which shares 70% structural identity with PTP1B, and is considered the greatest off-target liability. SHP1 was the other PTP inhibited by 2, but there was >5-fold selectivity. For each of the other PTPs, 2 showed >20-fold selectivity (Figure 2a and Figures S26–S28). In order to determine the mechanism of inhibition and perhaps gain insight into the source of the selectivity, an enzymological evaluation was carried out by varying the concentrations of substrate and inhibitor and measuring the initial rates. Based on the Dixon plot of these data, 2 showed mixed inhibition of PTP1B with K_i = 0.43 ± 0.15 μM (Figure 2b). Mixed inhibition is indicative of an allosteric mechanism of inhibition.

Figure 2.

Compound 2 is an isoform-selective PTP1B inhibitor. (a) Relative IC₅₀ values of 2 for PTPs normalized to PTP1B IC₅₀ value (* relative IC₅₀ values >20); (b) Dixon plot for PTP1B showing 1/rate versus varying concentrations of compound 2 at p-NPP concentrations of 1 mM (black dots), 2 mM (blue dots), 2.5 mM (green dots), and 3 mM (red dots). The x-coordinate of the intersection point gives a K_i of 0.43 ± 0.15 μM. Data are representative of three independent experiments.

Next, to gain insight into the structural details of PTP1B inhibition, molecular docking experiments were performed on 2 (CCDC 1880627) and PTP1B (1T4J) with Autodock 4.2.³¹ By comparing to an aryl diketoacid derivative, a known competitive inhibitor that binds to the active site of PTP1B, we could see compound 2 docked to a reported allosteric site of PTP1B (Figure 3a).³² Based on these docking results, five residues were predicted to interact with 2 (Figure 3b). Of these five, F280, F196, and L192 were predicted to produce stacking interactions with compound 2, while K197 and N193 were predicted to form hydrogen bonds with the phenolic hydroxy groups, the carbonyl group of the amide, and the two hydroxy groups on the aliphatic side chain. To test the derived computational model, a C-terminal truncation of PTP1B-WT_1–320, containing the first 320 amino acids, and five alanine mutants were constructed through site-directed mutagenesis.³³ The K_i of compound 2 against each of these PTP1B variants was measured (Supporting Information, Figures S29–S34 and Table S2). As shown in Figure 3c, there is no statistically significant inhibitory difference between PTP1B-WT₁₋₃₂₀ and PTP1B-WT₁₋₄₀₁, indicating that compound 2 does not bind to the C-terminus of PTP1B. Three of the mutants tested, F280A, K197A, and N193A, were observed to have K_i values which were 8, 10, and 6-fold less potent, respectively. Therefore, these data support that compound 2 binds to this allosteric site and these three residues participate in the interaction. The other two mutants, F196A and L192A, did not cause a statistically significant increase in K_i values. This can potentially be explained by the docking model which shows residues F196 and L192 pointing away from the ligand. Presently, we cannot rule out a severe structural perturbation caused by these mutations that leads to rearrangement of a distal pocket,³⁴ but our differential scanning calorimetry shows no loss of protein stability (Supporting Information, Figure S35) and we observe no deleterious effects to biochemical function based on hydrolysis of p-NPP.

Figure 3.

Molecular modeling of the binding of compound 2 to PTP1B. (a) Surface structure of PTP1B generated in pymol showed a predicted allosteric binding pocket of compound 2 (green, PDB ID 1T4J) and a known active site binding aryl diketo acid derivative (blue, PDB ID 3eb1). (b) Close up of specific interactions predicted between PTP1B and compound 2. (c) Mutations in the predicted compound 2 binding site of PTP1B compromise its inhibitory activity. The K_i values of each mutant and PTP1B-WT₁₋₃₂₀ challenged with compound 2 are shown relative to PTP1B-WT₁₋₄₀₁. * P < 0.05.

Subsequently, we tested the cellular activity of compound 2. Immunoblot analyses were performed to detect changes in insulin signaling upon stimulation of C2C12 myotubes treated with 0.5, 25, or 50 μM 2. This was accomplished by assessing the phosphorylation state of IR and AKT with phospho-protein specific antibodies. DMSO and sodium orthovanadate were used as vehicle and positive controls, respectively.³⁵ As presented in Figure 4a, the phosphorylation levels of IR and AKT increase in a dose-dependent manner after treatment with 2. These data corroborate our biochemical data and argue that PTP1B is a cellular target of 2. Moreover, given 2 is semisynthetic, it is an attractive lead that can be optimized through medicinal chemistry efforts. We next asked if the effect of 2 on the insulin pathway resulted in increased glucose uptake in C2C12 cells. To answer this question, we treated differentiated C2C12 cells with 2, DMSO (NC), 200 nM insulin (PC), and used 2-(N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)amino)-2-deoxyglucose (2-NBDG) to evaluate the glucose uptake level of the treated cells. 2-NBDG is a fluorescent glucose mimetic that allows for quantitation of cellular glucose uptake without using a radioisotope.³⁶ We observed an increase in glucose uptake with as little as 10 μM compound 2, and the increase was ~5-fold at 100 μM compound 2 (Figure 4b).

Figure 4.

(a) Compound 2 increases insulin signaling in C2C12 myotubes in a dose-dependent manner. C2C12 myotubes were treated with 0.5 μM, 25 μM, and 50 μM compound 2, DMSO (NC), and 20 μM sodium orthovanadate (PC) for 6 h. Then cells were stimulated with 10 nM insulin for 10 min. The lysates were subjected to immunoblot analysis with indicated antibodies. (b) Compound 2 increases glucose uptake in differentiated C2C12 myotubes in a dose-dependent manner. Differentiated C2C12 myotubes were treated with 0.1 μM, 1 μM, 10 μM, and 100 μM compound 2, DMSO (vehicle control), and 200 nM insulin (positive control) for 6 h. The cells were then treated with 200 μM 2-NBDG for 24 h and then stimulated with 10 nM insulin for 2 h. 2-NBDG was detected at Ex/Em = 485/535 nm. *P < 0.05.

In conclusion, we report a PTP1B inhibitor with very good to excellent isoform selectivity relative to other highly similar nonreceptor PTPs. The promising compound was derived from our N-atom augmentation strategy of the HSP90 inhibitor, radicicol (1). Based on molecular docking experiments, mutagenesis, and enzymological evaluation, this inhibitor is predicted to bind to a previously described allosteric site of PTP1B. The computational model was tested by measuring the K_i of alanine mutants of PTP1B and C-terminally truncated PTP1B. The results showed that the C-terminus was not the binding site of 2, whereas F280, K197, and N193 were likely involved in the binding interaction. The cellular activity of the inhibitor was evaluated in insulin-stimulated C2C12 myotube cells. Treatment with compound 2 resulted in upregulation of insulin signaling and ultimately glucose uptake in a dose-dependent manner. To advance 2 toward a more viable clinical lead and to understand the details of binding, we have initiated structural investigations. These will be used to inform future medicinal chemical optimization of 2. The structure of 2 suggests that it can be readily modified at several positions for the optimization of potency and selectivity. In addition, we will evaluate an optimized analogue of 2 in a series of murine models of obesity and diabetes.