Designed switch from covalent to non-covalent inhibitors of carboxylesterase Notum activity

N-Acyl indolines 4 are potent, non-covalent Notum inhibitors developed from a covalent virtual screening hit 2a. The lead compounds were simple to synthesise, achieved excellent potency in a biochemical Notum-OPTS assay and restored Wnt signalling in a cell-based TCF/LEF reporter assay. Multiple high resolution X-ray structures established a common binding mode of these inhibitors with the indoline bound centred in the palmiteolate pocket with key interactions being aromatic stacking and a water mediated hydrogen bond to the oxyanion hole. These N-acyl indolines 4 will be useful tools for use in vitro studies to investigate the role of Notum in disease models, especially when paired with a structurally related covalent inhibitor (e.g. 4w and 2a). Overall, this study highlights the designed switch from covalent to non-covalent Notum inhibitors and so illustrates a complementary approach for hit generation and target inhibition.


Introduction
Carboxylesterase Notum is a serine hydrolase enzyme that catalyses the delipidation of Wnt proteins and so regulates Wnt signalling [1,2].Wnt signalling plays a key role in embryonic development and adult biology, and has become a drug target of considerable interest for human disease [3].Notum may also be a valid drug target through its deactivation of Wnt signalling by removal of an essential palmitoleate group needed for the binding of Wnt proteins to the Frizzled cell surface receptors.Inhibition of Notum activity has the potential to restore Wnt signalling tone and may be a therapeutic approach for disease associated with a decrease in Wnt signalling.Recent studies have shown there is a growing understanding of the role Notum plays in human disease such as osteoporosis [4,5] and colorectal cancer [6,7].Notum has been proposed as a druggable target with reliable screening technologies, multiple structural studies and fit-for-purpose small molecule inhibitors available [8,9].
A noteworthy covalent inhibitor of Notum is carbamate ABC99 (1), which is a selective, irreversible inhibitor developed by activity-based protein profiling (Fig. 1) [10].ABC99 is a valuable research tool that has been used to investigate the link between inhibition of Notum activity with cellular regeneration in aged intestinal epithelium [11] and neurogenesis in the subventricular zone of the brain [12].
The successful development of drugs with a covalent mechanism of action (MoA) has prompted the design of covalent inhibitors for a wide range of proteins through modification of suitable amino acid residues (catalytic or noncatalytic) [13][14][15][16][17]. Covalent inhibitors that react with the serine hydrolase family have long been an important designed modality in the identification of structural and chemical biology tools, and new therapeutics [18,19].A number of 'warheads' have been created that react selectively with the activated, catalytic serine residue in either a reversible or irreversible covalent manner and usually mimic a substrate transition state [18,20].
There has been significant discussion of the risk-benefit of covalent drugs with an evolving understanding that they can offer exceptional potency, selectivity (especially for non-conserved residues), prolonged target engagement, and an approach to treat challenging drug targets [13].Concerns can arise from overt reactivity of the warheads leading to potentially poor selectivity (conserved residues in families) and toxicity.Hence, to balance these risks, it may be advantageous to develop matched tool inhibitors from the same chemotype offering both covalent and non-covalent MoA, along with an inactive control.This aligned trinity of tools would be a valuable toolbox for use in disease association and target validation studies for a protein of interest.
We have previously described a virtual screen (VS) as a new approach to discover hits as inhibitors of Notum activity [21].This VS identified several fragment-sized series and singletons that were experimentally validated as nanomolar inhibitors of Notum.One of these hits, butyric ester 2a, was subsequently shown to act through a covalent mechanism [22].An X-ray structure of the Notum-2a complex showed a covalent adduct had formed between the catalytic Ser232 and the butyric ester (PDB: 7ARG), which was confirmed by mass-spectrometry analysis [22].The corresponding carboxylic acid 2b was a weak inhibitor of Notum that did not display covalent inhibition (PDB: 7B37 and 7B3F (S232A)) [22].This matched pair of inhibitors suggested it could be feasible to switch from covalent to non-covalent MoA within a suitable chemical template.

Results and discussion
Our objective was to develop potent, non-covalent inhibitors of Notum from this indoline scaffold 2 but without the ester or the acid functional groups.Replacement of the ester would prevent acylation of Ser232 and likely improve stability as esters can be vulnerable to plasma and liver esterases [23].Removal of the carboxylic acid would promote passive brain permeability and potentially provide a tool suitable for use in exploring Wnt signalling in disease of the CNS [24,25].In this Article, we describe detailed structure-activity relations (SAR) supported by structure-based drug design (SBDD) that ultimately led to the identification of highly potent, non-covalent inhibitors of Notum activity.
The first round of design was guided by the X-ray crystal structure of Notum-2b (PDB: 7B37) that indicated a disordered state of the pendant butyric chain, likely due to a steric clash with Ser232 [22].A truncated version of 2b could reduce the flexibility of the chain, avoid the clash with Ser232, and potentially achieve more favourable binding; this gave N-acyl indoline 3a as a prototype inhibitor to test this hypothesis.Screening 3a in a high-throughput biochemical assay with Notum (81-451C330S) and OPTS as a synthetic fluorescent substrate showed 3a to be an inhibitor of Notum activity with IC 50 0.40 ± 0.11 μM (n = 6) (Table 1).Notably, 3a is 7-fold more potent than 2b.Indoline 3a is a relatively small molecule with moderate lipophilicity (MW 161.2, HAC 12, clogP 1.4) and proved to be a highly efficient inhibitor (LE 0.75, LLE 5.0).The Notum-3a X-ray structure revealed clear electron density (Fig. S1).The structure of 3a showed the inhibitor bound centred in the Notum palmitoleate pocket sandwiched between Phe268 and Tyr129 with the aromatic B-ring stacking against Phe320 (Fig. 2).While not forming any direct hydrogen bonds with Notum, a water molecule (W122) bridges an interaction with the oxyanion hole.Comparison of the relative positions and interactions of 3a with 2a showed the binding mode of the acyl indoline motifs was highly conserved in these structures with a near identical overlay.Hence, removal of the butyric ester group from 2a had not significantly changed or disrupted the binding of the indoline group in the pocket despite deletion of the covalent tether validating the hypothesis that a truncated version of 2b could achieve a favourable binding mode.
These early findings gave encouragement that our objective was realistic and that 3a was an attractive hit as a starting point for further optimisation.The successful structure determination of 3a in Notum provides valuable insight on the binding mode of these non-covalent inhibitors to guide further design in a SBDD enabled manner.
Notum inhibitors 3-5 (Tables 1-4) were either purchased or prepared using established synthetic methods from readily available starting materials (Scheme 1).Where necessary, the synthesis of many of these amines (e.g. 6) have been reported previously and full details of their preparation have been included in the Experimental Section.Notum inhibition activity was determined using the standard Notum-OPTS biochemical assay using IC 50 values.These values were used to compare relative activities with covalent inhibitors under carefully regulated assay conditions [26].Confirmed inhibitors were routinely screened in in vitro ADME assays.Selected inhibitors were screened for inhibition of Notum activity in a TCF/LEF Reporter (Luciferase) cell line with exogenous WNT3a and Notum, and soaked into Notum crystals to determine their binding modes and evidence of MoA.

B.N. Atkinson et al.
were detrimental for inhibition of Notum activity compared to 3a.The modest structural change of an indoline 3a to the corresponding indole 3b caused a 10-fold decrease despite being very similar in size and likely to occupy a similar position in the pocket (Fig. S2).Binding affinity in this region may be driven by lipophilicity (over π-stacking) with a preference for sp 3 carbons at C2 and C3.Tetrahydroquinoline 3c was 6fold weaker then 3a, which is due to the larger A-ring (Fig. S2).Overlay of 3c with 3a would suggest that the expanded A-ring of 3c extends forward and may clash with Ala342 at the front of the pocket and so 3c would be required to sit deeper in the pocket disrupting the interaction with the water bridging to the oxyanion hole.Loss of activity was observed for a variety of A-ring sizes (5,6) and Npositions (α, β) that resulted in an unfavourable change in the acyl vector (3a vs 3c-3e).Deconstructing the [6,5]-ring system into monocyclic (3f, 3g) or acyclic (3h) systems resulted in a substantial loss in activity.Oxidation of the A-ring through introduction of a carbonyl group at the 2-or 3-positions was tolerated with a preference for the 3-position (3i) showing some polarity could be incorporated into the template.Exchanging the indoline for alternative N-acyl benzo heterocycles (3k-3o) was explored and were inferior to 3a.
Hence, 3a was selected as the preferred scaffold for further investigation through sequential substitution at each of the positions on the indoline rings; preferred groups at these positions were then combined to see if the effects were additive to produce inhibitors with greatly improved activity (Table 2).Introduction of small alkyl groups onto the A-ring gave a significant boost in activity.A single 3-Me (4b) gave a 2fold increase but this gain was lost when geminal 3,3-Me 2 (4c) were present.Constraining the two Me into a 3,3-spiro cPr (4d: IC 50 0.061 μM) gave the most potent inhibitor so far although the corresponding 3,3-spiro cBu (4e) or 2,3-fused cPr (4f) were less well accommodated, probably due to the increased bulk at the opening to the palmiteolate pocket.
Substitution on the aromatic B-ring was initially explored by walking a Cl substituent around the scaffold (4g-4j).A 4-Cl (4g: IC 50 0.040 μM) gave a 10-fold increase in activity over 3a with a 5-Cl (4h: IC 50 0.095 μM) also showing an improvement, and these positions were prioritised (4-Cl > 5-Cl > H ≫ 6-Cl > 7-Cl).The X-ray structure of 3a showed there to be space in the pocket adjacent to the 4-and 5-positions on the indoline to accommodate a substituent, whereas the available space near the 6-and 7-positions was limited.Exploring small substituents as alternatives to the Cl (F, Me, CF 3 , CN, OMe) at 4-and 5-positions (4k-4u) showed the 4-CF 3 (4o: IC 50 0.030 μM) achieved a similar level of activity to 4-Cl but that Cl was favoured in the 5-position.Combining these preferred groups (4-Cl, 4-CF 3 , 5-Cl) as dual substituents gave 4-CF 3 -5-Cl (4v: IC 50 0.0013 μM) and 4,5-Cl 2 (4w: IC 50 0.0025 μM), which achieved still better potency and inhibitors with single digit nanomolar activity.Similar gains in Notum activity were observed when combining a preferred aryl substituent (4-Cl, 4-CF 3 or 5-Cl) with a 3,3-spiro cPr group (4y-4aa) with 4y (IC 50 0.0035 μM) as the most potent inhibitor from this subset.In summary, SAR studies in the N-acyl indoline template has identified inhibitors (4v, 4w, 4y) with a similar level of activity to the most potent Notum inhibitors reported [8,9].Indoline 4w is still a relatively small molecule (MW 230, HAC 14, clogP 2.7) and was a highly efficient inhibitor of Notum (LE 0.85, LLE 5.9) with modest improvements compared to the original lead 3a (LE 0.75, LLE 5.0).In an effort to further understand the binding mode of the indoline series 4, select compounds were subjected to soaking in Notum crystals and their structures solved (Fig. 3).Notum structures were determined for 4g, 4v and 4w in high resolution and unambiguous ligand electron density.
The structures of 4g, 4v, and 4w were very similar and retained some key features of the binding mode of 3a such as the water mediated interaction with the oxyanion hole.However, addition of substituents on the aromatic B-ring led to modest shifts in the pocket positioning of these inhibitors compared to 3a, most likely due to steric interactions with the back of the pocket; the indoline rings move towards Ser232 by Overall, these three inhibitors occupy a larger portion of the pocket volume, explaining the increase in Notum activity.
As the programme evolved, integration of the SAR data with available X-ray structures showed that Notum activity for these N-acyl indoles 3a and 4 was largely dependent upon optimum occupancy of the palmitoleate pocket provided that the water mediated H-bond between the ligand carbonyl through to the oxyanion hole was not disrupted.
Early leads 4v, 4w and 4y were then evaluated in additional pharmacology and ADME screens (Table 3).These three compounds restored Wnt signalling in the presence of Notum in a cell-based TCF/LEF (Luciferase) reporter assay and gave regular concentration-response curves up to 10 μM (Fig. S4).Performing these experiments in the absence of Notum showed that the activation of Wnt signalling was due to direct on-target inhibition of Notum and not by assay interference or cell toxicity (up to 10 μM).Screening 4w and 4y in the biochemical assay using full length Notum and OPTS confirmed potent inhibition of Notum activity.
Throughout the design cycle, confirmed Notum inhibitors (<100 nM) were routinely screened in in vitro ADME assays to assess their aqueous solubility, microsomal stability and cell permeability (Table 3, Tables S1-S5).It was quickly established that most indolines 4 displayed poor stability in mouse and human liver microsomes (MLM and HLM: t 1/ 2 <5 min, Table S4).In contrast, mouse plasma stability of 4w was excellent with no detectable loss (t 1/2 >4 h, Table S5).
Given the poor microsomal stability of 4, the main metabolites and routes of metabolism of 4y were determined in MLM (Scheme 2, Tables S6 and S7).In addition to unchanged parent 4y, four metabolites of 4y were detected and identified by LC-UV-MS.These metabolites M1-M4 were assigned: M1: di-oxygenation metabolite; M2-M3: monooxygenation metabolites; and M4: amide hydrolysis and dehydrogenation metabolite.Unchanged parent, 4y accounted for 30% of the total drug-related components in MLM with M4 as the primary metabolite (61%) and M3 as a minor metabolite (8%).M1 and M2 were trace metabolites only detected by MS-MS.The proposed metabolic pathway of 4y in MLM was amide hydrolysis, dehydrogenation and monooxygenation (Scheme 2).
As the N-acyl group was a defining motif of the indolines 4 for Notum inhibition (cf 5n, 5o), it was likely that microsomal hydrolysis of the amide was a metabolic liability across this series.The next phase of design would need to retain key elements of the pharmacophore for Notum inhibition whilst simultaneously fine-tuning structural features to improve microsomal stability and reduce amide hydrolysis (Table 4).
Increasing the steric bulk of the amide group (5a) caused a significant reduction in Notum activity as did replacement of the carboxamide with a sulfonamide (5b).Revisiting the N-acyl indole (5c, 5d) scaffold with the introduction of preferred aromatic substituents (4-Cl, 4,5-Cl 2 ) to enhance Notum inhibition gave an increase in activity over 3b with the SAR tracking that of the indoline template.However, indole 5d failed to improve MLM stability.An alternative strategy was to tether the N-acyl group onto the indoline/indole at C2 to create [6,5,5]-and [6,5,6]-tricyclic systems (5f-5m).Prototype pyrroloindoline 5f [27] was encouraging as Notum inhibition was 2-fold superior to the non-cyclised analogue 3a and MLM stability had improved (5f: IC 50 0.18 μM, t 1/2 18 min, Cl i 74 μL/min/mg protein).Introduction of a Cl group at the preferred aromatic sites (5j: 7-Cl and 5k: 8-Cl) failed to improve Notum activity and caused a modest decrease compared to 5f indicating a divergence of SAR between these scaffolds.Substitution on the pyrrolidinone ring with geminal 2,2-F 2 5m [28] gave an increase in potency over 5f albeit with a similar level of MLM stability (5m: IC 50 0.070 μM, t 1/2 16 min, Cl i 91 μL/min/mg protein).
X-ray structures of 5f and 5m revealed a shared binding mode very similar to 3a with key interactions being aromatic stacking to Phe268, Tyr129 and Phe320, and a water bridging the interaction with the oxyanion hole (Fig. 4).It was not immediately clear from the binding mode of these pyrroloindolines why introduction of Cl substituents at preferred aromatic sites fails to improve activity for this template (5j, 5k v 5f).It could be possible that the shifted binding of the Cl-substituted inhibitors places the framework of the pyrrolidinone ring in a position that would clash with Ala342 at the opening of the pocket (cf 2-Me of 4a).
From this set of compounds, 4 and 5, it had not proved possible to combine potent inhibition of Notum activity (IC 50 < 10 nM) with acceptable microsomal stability (Cl i < 20 μL/min/mg protein), and so further optimisation of this series to identify a brain penetrant tool for in vivo use was not merited as better tools are currently available [29].

Conclusion
We have described N-acyl indolines 4 as potent, non-covalent Notum inhibitors developed from a covalent virtual screening hit 2a.The lead compounds were simple to synthesise, achieved excellent potency in a biochemical Notum-OPTS assay (IC 50 < 10 nM) and were shown to restore Wnt signalling in the presence of Notum in a in a cell-based TCF/ LEF reporter assay (EC 50 < 40 nM).
Multiple high resolution X-ray structures established the binding mode of these N-acyl indolines.The indoline bound centred in the palmiteolate pocket with aromatic stacking to Phe268, Tyr129, and Phe320 and a water mediated hydrogen bond to the oxyanion hole.Indolines 4v and 4w are two potent inhibitors of Notum with SAR and X-ray structures clearly showing that incorporation of 4,5-Cl 2 and 4-CF 3 -5-Cl substituents were beneficial to binding interactions (cf 3a).
The N-acyl group was an essential motif of the indoline pharmacophore for Notum inhibition although it proved challenging to combine potent inhibition of Notum with acceptable microsomal stability as hydrolysis of the amide bond was the primary metabolic liability in MLM as shown by metabolite identification studies.
We believe these N-acyl indolines 4 will be useful tools for use in in vitro studies to investigate the role of Notum in disease models, especially when paired with a structurally related covalent inhibitor (e.g.4w

and 2a).
Overall, this study highlights the designed switch from covalent to non-covalent Notum inhibitors and so illustrates a complementary approach for hit generation and target inhibition.

Notum protein expression and purification
Methods have been described in detail [21,22].
To prepare the figures of the crystal structures the PDB entry was processed using the Protein Preparation Wizard in Schrödinger Maestro 12.4 using the default settings and optimizing the hydrogen bond assignment.No changes were made to the position of heavy atoms.Final images were prepared using PyMol version 2.4.
Selected compounds were routinely screened for aqueous solubility in PBS (pH 7.4), transit performance in MDCK-MDR1 cell lines for permeability, and metabolic stability in MLM and HLM as a measure of clearance [21,29].The main metabolites and routes of metabolism of 4y were determined in MLM. Assay protocols and additional data are presented in the Supplementary Material (Tables S1-S7).ADME studies reported in this work were performed by WuXi AppTec (Shanghai,

Chemistry 4.4.1. General information
Unless preparative details are provided, all reagents were purchased from commercial suppliers and used without further purification.Solvents were of ACS reagent grade or higher and purchased from commercial suppliers without further purification.Microwave assisted reactions were performed in a Biotage Initiator+.Thin-layer chromatography (TLC) was carried out on aluminium-backed silica plates.The plates were visualized under UV (254 nm) light, followed by staining with phosphomolybdic acid dip or potassium permanganate and gentle heating.Organic solvent layers were routinely dried with anhydrous Na 2 SO 4 or MgSO 4 and concentrated using a Büchi rotary evaporator.Compound purification by column chromatography was performed using a Biotage Isolera using prepacked Biotage SNAP KP-Sil silica cartridges or Biotage SNAP Ultra C18 reverse phase cartridges. 1 H and 13 C NMR spectra were recorded in deuterated (≥99.5%)solvents on either a Bruker Avance 400 (400 MHz), Bruker Avance 600 (600 MHz), or Bruker Avance 700 (700 MHz).Chemical shifts (δ) are reported as parts per million (ppm), coupling constants (J) are reported in Hz, and signal multiplicities are reported as singlet (s), doublet (d), triplet (t), quartet (q), pentet (p), sextet (sext), doublet of doublets (dd), doublet of triplets (dt), triplet of doublets (td), triplet of triplets (tt), multiplet (m), broad (br) or apparent (app).Liquid chromatography− mass spectrometry (LCMS) analysis was performed on a Waters Acquity H-Class UPLC system with either an acidic (HSS C18 Column, H 2 O/MeCN, 0.1% formic acid) or a basic (BEH C18 Column, H 2 O/MeCN, 10 mM NH 4 OH) mobile phase.The observed mass and isotope pattern matched the corresponding theoretical values as calculated from the expected elemental formula.
Most yields for chemical reactions are for single experiments and have not been optimised.The purity of compounds 3-5 was evaluated by NMR spectroscopy and LCMS analysis.All compounds had purity ≥95% unless otherwise stated.
General Method 1.2.Sodium hydride (32 mg, 1.4 mmol, 1.5 equiv.) was added portionwise to the amine (0.9 mmol, 1.0 equiv.) in THF (6.0 mL) at 0 • C (external).Once gas evolution had ceased, acetyl chloride (130 μL, 1.8 mmol, 2.0 equiv.) was added cautiously, the ice bath was removed and the mixture was stirred at RT for 2 h H 2 O (20 mL) and CH 2 Cl 2 (20 mL) were then cautiously added, and vigorously stirred for 5 min.The organic phase was dried, filtered, and concentrated under reduced pressure.The crude material was purified by reverse phase chromatography to give the amide.

Fig. 2 .
Fig. 2. X-ray structure of N-acyl indoline 3a (PDB 8BT8) (salmon).(A) Side view.(B) Top view.(C) Overlay of binding poses of 2a (white) and 3a (salmon).(D) Map of key interactions.Select binding site residues shown within 4 Å of their respective ligands.Polar contacts are indicated as dashed lines.Water molecules not directly interacting with the inhibitor have been removed for clarity.The surface of the Notum binding pocket outline (grey) is based on the respective structures.Hydrogens were added and optimised using the Protein Preparation Wizard in Schrödinger Maestro.
B.N.Atkinson et al.

Table 1
Exploration of N-acyl core scaffolds.

Table 3
Notum inhibition and ADME properties for selected N-acyl indolines 4.
a See Supplementary Material Tables S1-S5.B.N.Atkinson et al.