Design of a Potent, Selective, and Brain-Penetrant Inhibitor of Wnt-Deactivating Enzyme Notum by Optimization of a Crystallographic Fragment Hit

Notum is a carboxylesterase that suppresses Wnt signaling through deacylation of an essential palmitoleate group on Wnt proteins. There is a growing understanding of the role Notum plays in human diseases such as colorectal cancer and Alzheimer’s disease, supporting the need to discover improved inhibitors, especially for use in models of neurodegeneration. Here, we have described the discovery and profile of 8l (ARUK3001185) as a potent, selective, and brain-penetrant inhibitor of Notum activity suitable for oral dosing in rodent models of disease. Crystallographic fragment screening of the Diamond-SGC Poised Library for binding to Notum, supported by a biochemical enzyme assay to rank inhibition activity, identified 6a and 6b as a pair of outstanding hits. Fragment development of 6 delivered 8l that restored Wnt signaling in the presence of Notum in a cell-based reporter assay. Assessment in pharmacology screens showed 8l to be selective against serine hydrolases, kinases, and drug targets.


■ INTRODUCTION
The Wnt signaling pathway is a complex, evolutionarily conserved system from Drosophila and Caenorhabditis elegans through to mammals, which plays key roles in both embryonic development and the adult animal. These roles include cell fate determination and maintenance, and cell division in many different tissues. 1 The "canonical Wnt signaling pathway" is characterized by the secreted activator (or agonist) Wnt proteins, which initiate an intracellular signaling cascade by binding to the cell surface receptor (a member of the Frizzled family of cell surface proteins, complexed with the coreceptor LDL-receptor-related protein family 5/6). 1 The intracellular signaling cascade utilizes β-catenin and GSK3β to activate gene transcription via T-cell factor (TCF)/lymphoid enhancing factor (LEF) transcription factors. This signaling system is tightly regulated by a complex network of modulators and feedback inhibitory proteins, including the family of dickkopf (Dkk) proteins 2 and the carboxylesterase Notum. 3,4 Given the pleiotropic and essential function of Wnt signaling, it is unsurprising that aberrant Wnt signaling can contribute to disease in humans, including certain cancers, diabetes, and osteoporosis. 1,5 In the mammalian brain, Wnt signaling performs several important roles, including the maintenance of synapse function, 6 modulating microglial signaling and survival, 7 neurogenesis, 8 and maintaining the integrity and function of the blood−brain barrier (BBB). 9,10 In Alzheimer's disease (AD), when the Wnt signaling tone is disrupted by an increase in the inhibitory Dkk proteins, this leads to synapse loss, which may contribute to cognitive impairment. 11 In stroke, loss of Wnt signaling leads to BBB breakdown, contributing to the pathology. 12,13 In a model of multiple sclerosis, reactivation of Wnt signaling at the BBB partially restores functional BBB integrity and limits immune cell infiltration into the CNS. 14 Therefore, augmenting Wnt signaling in these disorders of the central nervous system (CNS) has therapeutic potential.
Notum is a negative regulator of the Wnt signaling pathway. It is a secreted protein that functions as a carboxylesterase, removing a palmitoleate ester from a conserved serine residue in Wnt proteins that is essential for its binding and activation of the cell surface coreceptors. 3 While first identified in Drosophila, its function in mammals, in both healthy and diseased, is being increasingly explored. The key tools to enable this progress are fit-for-purpose small-molecule inhibitors of Notum carboxylesterase activity. 15,16 First-generation inhibitors of Notum activity were LP-922056 (1) 17 and the irreversible inhibitor ABC99 (2) 18 that have proved to be valuable research tools in exploring the link between pharmacological inhibition of Notum activity and disease association in rodent models ( Figure 1). LP-922056 is a potent inhibitor of Notum suitable for chronic, oral dosing in rodents discovered through the optimization of a highthroughput screening (HTS) hit. 17 LP-922056 was used to show that inhibition of Notum activity is a potential novel anabolic therapy for strengthening the cortical bone and preventing non-vertebral fractures. 19 Notum has recently been identified as a key mediator for people at high risk of developing colorectal cancer. 20,21 Pharmacological inhibition of Notum with LP-922056 abolished the ability of Apc-mutant cells to expand and form intestinal adenomas. 21 ABC99 is a potent and selective covalent inhibitor of Notum activity developed by competitive activity-based protein profiling (ABPP) from a library of activated carbamates. 18 Inhibition of Notum produced by Paneth cells in mice with ABC99 restored the regeneration of aged intestinal epithelium. 22 In the brain, Notum regulates neurogenesis in the subventricular zone (SVZ) by modulating Wnt signaling, and inhibition of Notum with ABC99 leads to an activation of Wnt signaling and increased proliferation in the SVZ. 8 A number of structurally distinct Notum inhibitors have been reported from hit-to-lead programs ( Figure 2). 23−27 However, most of these inhibitors have limited brain penetration as directly measured in mouse pharmacokinetic (PK) studies; 24,26−28 this restricts their use to peripheral (non-CNS) indications. Only ABC99 has been reported to reach the brain and activate the Wnt pathway as assessed by qRT-PCR of Axin2 in the ventricular-SVZ tissue following an intraperitoneal (i.p.) injection to Nestin-CFP mice. 8 Our objective was to identify a screening hit that could be optimized to a potent, selective, and brain-penetrant inhibitor of Notum activity suitable for chronic, oral delivery. Such an inhibitor would complement LP-922056 and ABC99 as it could be administered in rodent models of neurodegenerative disease.
Fragment-based drug discovery (FBDD) is an established method for hit finding of biologically active compounds and is now supported by multiple complementary screening technologies. 29,30 The concept of a "fragment" has essentially remained consistent with an emphasis on small, low-molecularweight organic molecules with high aqueous solubility as described by Congreve et al. 31 In contrast, there is an increase in the application of FBDD to more challenging, non-traditional targets such as GPCRs and ion channels, and to phenotypical screens including parasites. 29 The creation of fragment libraries with a clearly defined purpose beyond chemical diversity has also received considerable attention. Examples of such libraries include reactive covalent fragments, 32 3D fragments, 33 natural product-like compounds, 34 very small fragments such as FragLites and Minifrags, 35 halogen-enriched fragments, 36 and synthetically poised fragment libraries. 37 The original Diamond-SGC Poised Library (DSPL) was created by Cox et al. as a 768-membered fragment library that was designed to allow rapid follow-up synthesis to provide quick structure−activity relationships (SARs) around fragment screening hits. Each member of the collection contained at least one functional group that could be created with reliable chemical reactions, that is, the fragment library was created to be structurally diverse and synthetically enabled. 37,38 Herein, we describe the discovery of crystallographic fragment screening hits 6a and 6b from the DSPL and their optimization to identify 1-phenyl-1,2,3-triazoles as a new Notum inhibitor chemotype. We present the profile of lead compound 8l that supports it to be a potent, selective, and brainpenetrant inhibitor of Notum activity.

■ RESULTS AND DISCUSSION
In order to find hit compounds for the development of new Notum inhibitors, we chose to conduct a crystallographic fragment screen. The structure of Notum has a clearly defined lipophilic pocket adjacent to the catalytic triad that accommodates a C16-lipid palmitoleate group of Wnt. 3,39 The size and topography of this pocket gave encouragement that suitable fragments could be found to bind as the pocket has been assessed to be classically druggable. 40 An X-ray crystallography-based fragment screen was performed at the XChem platform of Diamond Light Source (Oxford, U.K.). The highly automated XChem platform, in  combination with synchrotron beamline I04-1, enabled us to efficiently screen 768 compounds from the DSPL fragment library for binding to Notum. Compounds were individually soaked into crystals of Notum and high-resolution structures were determined for each of these putative Notum−fragment complexes. The electron density maps provided evidence of compound binding associated with the active site of Notum for 59 of the crystal structures (hit rate: 7.7%). 23,26,41 The complete output of the hits from this crystallographic screen with the DPSL will be published separately. 42,43 Hit validation was then performed by demonstration that these binders were able to inhibit Notum carboxylesterase activity in a biochemical assay (up to 1 mM).
A pair of outstanding hits were 4-(hydroxymethyl)triazoles 6a (IC 50 12 μM) 26 and 6b (IC 50 0.5 μM), which also highlighted some early beneficial SAR (Figure 3). A 3,4-Cl 2 disubstituted phenyl ring conferred a substantial improvement in Notum inhibition when compared to a single 4-Cl (23-fold). Both the Notum-6a and -6b X-ray structures showed that the ligand formed a H-bond between N3 of the triazole with the backbone of Trp128. The increase in potency of 6b over 6a can be attributed to the improved occupancy of the palmitoleate pocket by the 3,4-Cl 2 disubstituted phenyl ring (Figures 4 and S1). Specific interactions in the pocket include the potential stacking of the ligand with Trp128, Tyr129, and Phe268 as well as interactions by the phenyl ring substituents with Phe320.
Triazole 6b has molecular and physicochemical properties consistent with lead-like chemical space 44 45,46 including a favorable prediction of brain penetration (CNS MPO 5.6/6.0; BBB Score 4.9/6.0). 47,48 Triazole 6b was further evaluated in standard in vitro ADME assays and showed good aqueous solubility, moderate stability in liver microsomes, and excellent cell permeability (Table 1).
Hits 6 were an attractive starting point to improve Notum activity and develop PK properties for oral dosing to achieve good plasma exposure and brain penetration.
The optimization of 6 to deliver this profile was directed at three principle areas of these hit structures: (1) aryl ring 7 (Table 1); (2) the C4 substituent on triazole 8 ( Table 2); and (3) further fine-tuning of the triazole 9 ( Table 3). Inhibition of Notum activity was routinely measured in a biochemical assay with human Notum (81−451 Cys330Ser) as the enzyme and trisodium 8-octanoyloxypyrene-1,3,6-trisulfonate (OPTS) as the substrate. 23−26 This assay was used to establish SAR and to identify preferred leads for additional studies. As Notum activity was developed, representative examples were screened in standard in vitro ADME assays to assess their suitability for progression to mouse PK studies.
As before, trusted design metrics were calculated to aid compound design and assess progress, supported by structurebased drug design (SBDD) through inhibitor docking and X-ray structures. However, a significant change in our strategy was to discount any weakly acidic functional groups or heterocycles as these had failed to give sufficient brain penetration in mouse PK experiments (e.g., 4 and 5). 26,27 Instead, lead compounds would need to be non-ionized, or weakly basic, at physiologically relevant pH to be advanced through the screening sequence.  General synthetic methods to prepare inhibitors 7−9 (Tables 1−3) are shown in Schemes 1−5. Our initial efforts focused on exploration of the aryl ring SAR, thus a reliable synthesis of substituted (1-aryl-1H-1,2,3-triazol-4-yl)methanols 7 was required (Scheme 1). Target compounds 7 were readily prepared in two steps from commercially available starting materials. Either the prerequisite boronic acid 10 or aniline 11 was converted to the required aryl azide 12 and then subsequent copper(I)-catalyzed azide−alkyne cycloaddition (CuAAC) with propargyl alcohol afforded the desired triazoles 7.
This approach (11 → 12 → 7) proved to be versatile and reliable and was applied to the synthesis of a number of inhibitors. 2,4-Dichloro-3-(trifluoromethyl)aniline (11a) was converted to the corresponding azide 12a, and reaction with the appropriate 1-alkyne 13 under CuAAC conditions gave a variety of 4-substituted 1,2,3-triazoles (8, 14−17) (Schemes 2 and 3). A number of inhibitors were prepared by functional group conversion and/or protecting group strategies. Methyl ester 14 was converted to the acid 8f by base hydrolysis, and to amides 8g and 8h by amination with NH 3 and MeNH 2 , respectively. Acid deprotection of the Boc group from 15 gave methyl amine 8i, and K 2 CO 3 /TBAF deprotection deprotection of 16 furnished 4-methyl triazole 8k. CuAAC-promoted [2 + 3] cyclization of 12a with Me 3 SiCCH gave the corresponding 4-TMS triazole 17, and desilylation with K 2 CO 3 /MeOH afforded 8l. Desilylation of 17 with CD 3 OD and D 2 O as the solvents promoted deuteration at the C4 and C5 positions of the triazole and afforded d 2 -8l with high-isotope incorporation.
Additive effects from multiple substituents were then investigated by preparing a set of dichlorophenyl triazolematched molecular pairs (7q−t). This set of inhibitors was prioritized due to their synthetic accessibility from available commercial reagents and the precedent from 6b. The original hit 6b was preferred over all other isomers with activity following the order: 3,4-Cl 2 (6b, IC 50 0.5 μM) > 3,5-Cl 2 (7t, IC 50 2.4 μM) > 2,3-Cl 2 (7q, IC 50 5.1 μM) ≈ 2,4-Cl 2 (7r, IC 50 5.6 μM) > 2,5-Cl 2 (7s, IC 50 23 μM). Improvements could be achieved with alternative 3,4-disubstitutions when combinations of preferred substituents (Me, Cl, and CF 3 ) were used with a clear preference for 3-CF 3 -4-Cl (7w, IC 50 0.065 μM). Addition of a 2-Cl as a third substituent onto 7w as 2-Cl-3-CF 3 -4-Cl (7y) gave a further boost in potency with 7y (IC 50 0.0095 μM) being 50-fold more active than hit 6b and 10,000-fold more active than the unsubstituted phenyl ring 7a. At this point in the program, mouse PK data for 7y was generated in vivo to determine the extent of plasma and brain exposure and to verify the correlation of the in vitro ADME data with in vivo outcomes in this series. Triazole 7y was selected as it combined excellent Notum activity (IC 50 9.5 nM) with an attractive ADME profile as it possessed good aqueous solubility, microsomal stability, and cell permeability (Table 1 and Supporting Information). Following a single oral dose of 10 mg/kg, plasma exposure for 7y was low (C max 300 ng/mL; AUC (0→inf) 150 ng h/mL), which we attributed to higher clearance than predicted by MLM and highlighted the need to further improve metabolic stability ( Figure S2). Our hypothesis for this disconnection between MLM stability and plasma clearance of 7y was due to phase 2 metabolism/conjugation of the OH group, and this was subsequently supported by the higher clearance of 7y in mouse hepatocytes (Cl i 48 μL/min/ 10 6 cells). Hence, we designed inhibitors that explored direct modification and/or replacement of the −CH 2 OH (Table 2). Despite the poor plasma exposures, 7y displayed the highest brain-to-plasma ratio [K p 1.2, based on AUC 0−inf (total drug)] observed to date and was far superior to previous leads 1 (K p ca. 0), 28 3 (K p 0.29), 24 4 (K p 0.16), 26 and 5 (K p ca. 0) 27 where comparable studies were performed. This was viewed to be a significant advantage for this series.
SAR of the C4-substituent of the triazole was studied with retention of the preferred aryl ring of 7y as this group conferred significant activity to the series and was superior to all other substitution patterns explored so far ( Table 2). Substitution at the C5 position was not investigated as previous studies with related 4-Cl-phenyl azoles showed this to be unfavorable. 26 Modification of the −CH 2 OH through addition of the α-Me group (8a), elongation of the carbon chain (8b−8c), or a combination of both these approaches (8d−8e) leads to a dramatic reduction in potency (IC 50 > 1 μM). Oxidation to the  Scheme 5. Preparation of (1,2,3-Triazol-1-yl)pyridine 9c a a Reagents and conditions: carboxylic acid 8f was tolerated with a modest reduction in potency (7y vs 8f), whereas activity was significantly diminished upon conversion to the corresponding primary 8g and secondary amides 8h. Primary amine 8i also had a significant reduction in potency as did "capping" the −CH 2 OH as the −CH 2 OCH 3 ether 8j. Deletion of the −OH from 7y to give the 4-Me triazole 8k had a more modest effect upon the reduction in potency (12-fold) that was consistent with the loss of one Hbonding group from the ligand. So far, all changes to the C4 substituent had been unfavorable, then, somewhat unexpected, potent inhibition of Notum activity was entirely recovered by removal of the substituent as unsubstituted triazole 8l (IC 50 6.7 nM) was the most potent inhibitor prepared in this series. Finally, revisiting the aryl ring substitution with 2,3,4-Cl 3 8m, as a simpler and direct analogue of 8l, was slightly inferior.
The Notum-8l X-ray structure is consistent with the ligand retaining the H-bond between N3 of the triazole and the backbone of Trp128. In addition, 8l has near complete occupancy of the palmitoleate pocket through the 2-Cl-3-CF 3 -4-Cl substituents on the aryl ring. The near optimal filling of the pocket is achieved while still positioning the triazole in a favorable position for a hydrogen bond with the backbone of Trp128 ( Figure 5). This combination of features is likely why such high potency is observed.
1,2,3-Triazole 8l was superior in potency to the isomeric 1,3,4-triazole 9a or the 1,2,3,4-tetrazole 9b (Table 3). These effects seem to be quite subtle as most key inhibitor−Notum interactions are available with these alternative 1-aryl azoles and may be due to a more optimal electronic interaction of 8l with Trp128 or simply that an N4 atom is detrimental. 49 Modification of the phenyl ring of 8l to the corresponding pyridine 9c was equipotent and demonstrated that some modest polarity could be accommodated in the palmitoleate pocket. Dual deuteration of 8l at the C4 and C5 positions of the triazole gave d 2 -8l as a tool to investigate the metabolic fate of the parent molecule.
Triazole 8l has physicochemical and molecular properties consistent with drug-like space. 50 It was noteworthy that 8l had a much lower measured distribution coefficient (LogD 7.4 1.3 ± 0.05) than had been predicted by the calculated partition coefficient (clogP 3.9) despite being non-ionized at physiologically relevant pH. This disconnection could be attributed to the close proximity of the 2-Cl, 3-CF 3 , and 4-Cl substituents on the aromatic ring that results in a reduction in overall hydrophobic surface area when compared to summing their individual lipophilicity fragmental constants (π). These results highlight the importance of generating empirical data to correlate with calculated values. Evaluation of 8l against standard design metrics using Notum inhibition from the biochemical assay (IC 50 6.7 nM, pIC 50 8.2) and measured lipophilicity (LogD 7.4 1.3) calculates 8l to have high LE (0.67), LLE (6.9), and a favorable prediction of brain penetration (CNS MPO 5.4/6.0; BBB Score 5.6/6.0).
Triazole 8l displayed satisfactory aqueous solubility in phosphate-buffered saline (PBS), good metabolic stability in both MLM and HLM, and high permeability in the MDCK-MDR1 cell transit assay with minimal evidence for recognition and efflux by the P-gp transporter (Table 3 and Supporting Information). In contrast, pyridine 9c performed poorly in the permeability assay although these results were confounded by low compound recovery. Comparison of MLM stability of 8l with d 2 -8l (Cl i 3.2 vs 5.4 μL/min/mg protein) showed deuteration of the triazole afforded no improvement suggesting that CYP-mediated metabolism of the triazole C−H bonds was not a major route of clearance in this system.
Based on these data, 8l was selected as our nominated lead from this series for further profiling as it had the best combination of Notum activity and in vitro ADME properties including lipophilicity.
Lead 8l was then evaluated in additional in vitro pharmacology and ADME screens (Table 5). Triazole 8l restored Wnt signaling in the presence of Notum (EC 50 110 nM; n = 4) in a in a cell-based TCF/LEF (Luciferase) reporter assay 23,26,27 and gave standard S-shaped concentration− response curves up to 10 μM ( Figure S9). Performing these experiments in the absence of Notum showed a maximal Wnt response at all concentrations tested (up to 10 μM; n = 4). The activation of Wnt signaling was due to direct on-target inhibition of Notum by 8l and not by assay interference or cell toxicity (up to 10 μM).
The aqueous solubility of 8l could be significantly enhanced to 1 mg/mL with cosolvents bringing added flexibility to study design. Triazole 8l has good stability in liver microsomes and hepatocytes across mouse, rat, dog, and human, which predicts for low clearance across these species. Screening in HLM including eight standard substrates of cytochrome P450 (CYP450) enzymes showed moderate/weak inhibition of 1A2 (IC 50 0.96 μM), 2B6 (IC 50 5.2 μM), 2C9 (IC 50 15 μM), and 2C19 (IC 50 8.1 μM) but not 2C8, 2D6, or 3A4 (IC 50 > 50 μM). Furthermore, there was no evidence for CYP time-dependent inactivation (TDI) as assessed by the preincubation of HLM with 8l in the presence and absence of NADPH, followed by the incubation with the discrete marker substrates (IC 50 shift). There was good cell permeability across a Caco-2 monolayer consistent with good absorption from the gut. The hepatocyte data for 8l was encouraging as this deletion of the −CH 2 OH group from 7y seems to have removed the liability of this metabolic soft spot. Hence, 8l was advanced to rodent PK studies.
PK data for 8l was generated in vivo in mouse, and subsequently in rat, to evaluate plasma and brain exposure (Table 4 and Supporting Information). Following single intravenous administration of 8l to mouse, plasma clearance was low compared to liver blood flow and volume of distribution was moderate resulting in an elimination half-life of 2.4 h ( Figure  S3, Table S11). Following a single oral dose to mouse, 8l was rapidly absorbed and showed good oral bioavailability (66%) ( Figure S4, Table S12). Good brain penetration was confirmed with drug levels in the brain similar to drug levels in the plasma (K p 1.1). The brain-to-plasma ratio was 0.77 (K p,uu ) when free drug concentrations were calculated. This oral dose of 10 mg/kg achieved a concentration in the brain of C max ≈ 300 nM (free drug) that exceeded the Notum EC 50 from the cell-based TCF/ LEF assay, and brain levels were maintained above 100 nM for ∼4 h. Although encouraging, this dose, and the resulting level of brain exposure, may be insufficient to promote a sustained pharmacodynamic (PD) response, where Notum activity was reduced for an extended period of time. This will need to be determined empirically. Hence, we subsequently explored alternative dosing regimes to provide some flexibility in the determination of the required efficacious concentrations (C eff ) in rodent models of disease, and its relationship to Notum pharmacology (EC 50 ), that is, establish the PK−PD relationship (vide infra).
PK evaluation of 8l in rat also showed low plasma clearance and moderate volume of distribution with a half-life of 3.3 h (Table 4). Oral administration to rat showed good absorption, drug levels, and brain penetration (K p 1.4). The oral bioavailability was supramaximal with F o 140%. The plasma (and brain) concentration−time plots clearly show double C max peaks ( Figure S7), which could indicate some contribution from . e Formulation B: 0.1% Tween80 in water. f K p was calculated by the ratio of brain concentration−time profile (AUC (0−inf),brain ) to that of plasma concentration−time profile (AUC (0−inf),plasma ). K p,uu was calculated by multiplying K p with the ratio of free fraction in brain homogenate to plasma (f u,brain /f u,plasma ).

Journal of Medicinal Chemistry
pubs.acs.org/jmc Article excreted drug being reabsorbed from the gastrointestinal tract and/or an effect on gastric emptying times. On balance, these results established that 8l has PK properties compatible with evaluation in rodent models of disease. Triazole 8l emerged as a promising lead with potent inhibition of Notum enzymic activity in both biochemical and cellular assays, along with excellent brain penetration in rodent. The wider pharmacology activity of 8l was then assessed against serine hydrolases, kinases, and representative drug safety targets to identify any off-target activity and to establish its selectivity.
Selectivity against serine hydrolases was assessed by unbiased proteomic analysis in a human colorectal cancer cell line SW620 that has been characterized to have high Notum expression. 51 Quantitative ABPP was performed using FP-biotin to label active serine hydrolases in concentrated conditioned media and cell lysates, combined with tandem mass tagging technology (TMT) for relative quantification. 52 Triazole 8l showed concentration-dependent inhibition of Notum activity (IC 50 170 nM in conditioned media) and demonstrated high selectivity in this system as there was no cross reactivity against any of the other 39 serine hydrolases detected (Figures 6, S11 and S12). 18,52b Compound 8l showed excellent selectivity for Notum over a panel of 485 distinct kinases with <50% inhibition at 10 μM for Journal of Medicinal Chemistry pubs.acs.org/jmc Article all these kinases including casein kinase 1 (CK1) and GSK3 (Table 5). This was of critical importance to studies wishing to assess the function of Notum with 8l, as CK1 and GSK3 phosphorylate β-catenin leading to its degradation and thus reducing Wnt signaling. 53 For some context, 8l showed the highest activity in this kinase panel at CDK14/cyclin Y with 43% inhibition at 10 μM (n = 2) (Table S20).
Compound 8l was screened for off-target pharmacology across multiple drug target classes in the DiscoverX Safety47 panel. This suite of 78 functional assays assesses pharmacological modulation of 47 human off-target liabilities that have been highlighted for the initial safety assessment of lead compounds. 54 There was minimal activity at 75/78 targets having IC 50 /EC 50 > 30 μM with only modest activity at HTR3A (IC 50 4.5 μM), nAChR(a4/b2) (IC 50 10 μM), and MAOA (IC 50 28 μM) ( Table 5). Liability through direct interaction with cardiac ion channels was low as 8l had no significant effect on either the hERG or the cardiac sodium channels (IC 50 > 30 μM) (Table S21).
Having established 8l as a potent and selective inhibitor of Notum carboxylesterase activity, our efforts shifted to developing a multigram synthesis of 8l and identification of a suitable dosing regimen to investigate the role of Notum in mouse models of neurodegenerative disease.
The multigram synthesis of 8l essentially followed the discovery route but with additional precautions to minimize the risk of high energy intermediates on scale such as the azide 12a (Scheme 6). Differential scanning calorimetry (DSC) analysis of azide 12a determined an exotherm onset temperature at 125.8 and so 42°C was established as the maximum safe handling temperature ( Figure S15). Reactions were performed in smaller batches and then pooled for purification to give a single, homogenous product.
Increasing the oral dose of 8l to mouse from 10 mg/kg to twice daily dosing of 30 mg/kg (2 × 30 mg/kg, p.o. at T = 0 and 9 h, formulation A) achieved higher and prolonged drug concentrations in the plasma and brain ( Figure S5, Table S13). The sustained free drug levels of 8l in the brain now exceeded the Notum EC 50 for ca. 20 h. This twice daily oral dosing of 8l was well tolerated in both wild-type and aged APP/PS1 mice for 7 days and was therefore deemed suitable for the in vivo pathfinder experiments. These studies are ongoing and will be reported in due course.

■ CONCLUSIONS
Crystallographic fragment screening of the DSPL for binding to Notum, supported by a biochemical enzyme assay to rank inhibition activity, proved to be successful in identifying chemically enabled hits. This screening strategy identified 58 hits that were shown to bind to the active site of Notum with 20 of these having IC 50 < 100 μM. A pair of outstanding hits were 4-(hydroxymethyl)triazoles 6a (IC 50 12 μM) and 6b (IC 50 0.5 μM), and these were selected for fragment development to improve Notum activity and develop PK properties for oral dosing to achieve good plasma exposure and brain penetration.
Exploration of the substituents on the aryl ring that binds in the palmitoleate pocket of Notum gave potent early lead 7y (IC 50 9.5 nM) and further optimization of the triazole group delivered 8l (IC 50 6.7 nM). It is noteworthy that advanced lead 8l can be characterized as a fragment-sized molecule (MW 282; HAC 17) with high LE and LLE, and that inhibition of Notum activity was increased by over 15,000-fold by optimization of the substituents on the aryl ring (8l vs 7a). These findings highlight the benefits of FBDD when aligned with a suitable drug target.
The Notum-8l X-ray structure clearly showed that 8l has near complete occupancy of the palmitoleate pocket through the 2-Cl-3-CF 3 -4-Cl substituents on the aryl ring. Triazole 8l restored Wnt signaling in the presence of Notum (EC 50 110 nM) in a cellbased TCF/LEF reporter assay confirming functional activity. Assessment in broader off-target and safety pharmacology screens showed 8l to be selective against serine hydrolases, kinases, and representative drug targets. PK studies with 8l in mouse and rat showed good plasma exposure and brain penetration, and was well tolerated.
In summary, triazole 8l (ARUK3001185) is a potent and selective inhibitor of Notum activity with good brain penetration suitable for oral dosing in rodent models of disease.

■ EXPERIMENTAL SECTION
Notum Protein Expression and Purification. Methods have been described in detail elsewhere. 25 Notum OPTS Biochemical Assay. Methods have been described in detail elsewhere. 23−26 Representative concentration−response curves are provided in the Supporting Information.
TCF/LEF Reporter (Luciferase) Cell-Based Assay. Methods have been described in detail elsewhere. 23,26,27 Concentration− response curves are provided in the Supporting Information.
Conditioned Medium Generation. SW620 cells were grown to around 80% confluence, washed three times with PBS, and then placed in serum-free media for 48 h. The medium was then removed, spun at 300G to remove any solids, and then concentrated in 10k MWCO protein concentrators (Pierce) by centrifugation.
Cell Lysate Generation. SW620 cells were grown to around 80% confluence, the medium was removed and cells were washed three times with PBS, and then lysed in the ice cold IP lysis buffer (Pierce). Lysates were clarified by centrifugation at 15k G for 10 min at 4°C, then protein quantification was performed by the BCA assay (Pierce) and lysates were equalized to 1.5 mg/mL. Proteomics Protocol. Either lysates or conditioned medium was treated with the appropriate concentration of each inhibitor or DMSO across a range of concentrations and then equilibrated at RT for 45 min. Following this, 2 μM FP-Biotin (Santa Cruz) was added to each sample for 20 min with agitation. Lysates were then precipitated using chloroform−methanol, and the protein pellet was washed two times with MeOH. Protein pellets were dissolved in 50 mM HEPES pH 8.0 with 2% SDS and 10 mM DTT, then the SDS concentration was adjusted to 0.2% with further 50 mM HEPES pH 8.0. Each solution was then added to an aliquot of equilibrated NeutrAvidin beads, 55 which had been washed two times with 50 mM HEPES pH 8.0 with 0.2% SDS, for 16 h at RT with agitation. Beads were then pelleted by centrifugation, the supernatant was removed and then washed three times with 50 mM HEPES pH 8.0 with 0.2% SDS, followed by three washes with 50 mM HEPES pH 8.0. The beads were then incubated with 10 mM DTT in 50 mM HEPES pH 8.0 for 15 min at RT, followed by 20 mM iodoacetamide, which was incubated for 30 min in the dark. A further aliquot of 10 mM DTT was added to the beads, then they were pelleted by centrifugation, the supernatant was removed, the beads were washed with 50 mM HEPES pH 8.0, and all the supernatants were removed. LysC (Wako) was dissolved in 50 mM HEPES pH 8.0 to give a 1 μg/mL solution, of which 50 μL (50 ng of LysC) was added to the beads and incubated overnight at 37°C. The beads were pelleted, the supernatant was removed, and then washed with additional 50 mM HEPES pH 8.0 to extract as many peptides as possible. 75 ng of trypsin (Pierce) was added from a 20 ng/μL solution in 50 mM HEPES pH 8.0 and incubated at 37°C for 7 h. Each set of conditions were then labeled using eight channels of a TMT10plex (Thermo) following the manufacturer's instructions, then combined. TMT-labeled tryptic peptides were filtered using 1 mm Empore C18 solid-phase extraction discs aka "stage tips" (3M) prior to vacuum centrifugation and resuspension in 0.1% trifluoroacetic acid.
The Orbitrap was operated in "Data Dependent Acquisition" mode, followed by MS/MS in "TopS" mode, with Orbitrap accumulation at R = 50,000 of HCD fragmented parent ions.
Raw files were processed using MaxQuant (maxquant.org) and Perseus (maxquant.net/perseus) with a recent download of the uniprot homo sapiens reference proteome database and a common contaminants database. A decoy database of reversed sequences was used to filter false positives, at a peptide false detection rate of 1%.
Analysis was performed using Perseus and filtered against contaminants, reversed, and proteins identified by the site. Protein groups were then filtered by those which contained valid values in at least six of the eight conditions. Serine hydrolases were identified by filtering for those proteins identified as such in these two reference publications 18,52b with intensity values normalized relative to the DMSO control, and concentration−response curves and IC 50 values were determined using GraphPad Prism 9.1.0.
The identity and quantification of each serine hydrolase identified in SW620 cells by FP-biotin ABPP for 8l (and 1) is provided in the Supporting Information. The MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository 56 with the dataset identifier PXD031338.
Structural Biology. The methods of protein production, crystallization, data collection, and structure determination have been described in detail. 26 Data collection and refinement statistics are presented in the Supporting Information.
To prepare the figures of the crystal structures, the PDB entry was processed using the Protein Preparation Wizard in Schrodinger 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. ADME Assays. Molecular properties (MW, clogP, and TPSA) were calculated with ChemDraw Professional v16.0.1.4(77). Distribution coefficients (LogD 7.4 ) were measured using the shake flask method.
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. Lead 8l was also screened for inhibition of representative CYP450 enzymes, permeability across the Caco-2 cell monolayer, and metabolic stability in liver microsomes and hepatocytes. Assay protocols and additional data are presented in the Supporting Information. ADME studies reported in this work were independently performed by Cyprotex (Macclesfield, U.K), GVK Biosciences (Hyderabad, India), and WuXi AppTec (Shanghai, China).
PK Studies. In vivo PK data in mouse and rat was independently generated at Charles River Laboratories (Groningen, Netherlands), GVK Biosciences (Hyderabad, India), and WuXi AppTec (Shanghai, China). PK was supported by the development of suitable formulations for the route of administration, and the measurement of plasma protein binding (PPB) and brain tissue binding for the calculation of free drug Chemistry. General Information. Methods have been described in detail elsewhere. 27a The purity of compounds 6−9 was evaluated by NMR spectroscopy and LCMS analysis. All compounds had purity ≥95% unless otherwise stated.
General Methods. *Caution* Azides are potentially explosive reagents and therefore all solutions containing azides were concentrated under reduced pressure behind a blast shield at RT. Furthermore, sodium azide reacts with acids to generate gaseous HN 3 , which is a toxic gas. All reactions were conducted behind a blast shield under a flow of nitrogen, and exit gases were passed through an aqueous base prior to their release into a ventilated fume hood.
General Method 1.2. *Caution* Step 1: Sodium nitrite (155 mg, 2.25 mmol, 1.2 equiv) was added portionwise to a solution of aniline 11 (1.87 mmol, 1.0 equiv) in TFA (2.0 mL) at 0°C over 30 min. The reaction mixture was then allowed to warm to RT for 1.5 h before recooling to 0°C. A solution of sodium azide (133 mg, 2.05 mmol, 1.1 equiv) in H 2 O (1.5 mL) was added dropwise over 5 min, stirred for 30 min, and then warmed to RT over 60 min. The mixture was basified to pH 8−9 by dropwise addition of aqueous NaOH (5 M), then diluted with t-BuOH (2.0 mL) to provide a solution of crude azide 12, which was used without further purification.
General Method 1.3. *Caution* Step 1: Aniline 11 (0.65 mmol, 1.0 equiv) was dissolved in H 2 O (0.5 mL) and MeCN (3.0 mL), then aqueous HCl (37%, 7.3 mL) was added at RT and stirred vigorously for 10 min. Sodium nitrite (90 mg, 1.3 mmol, 2.0 equiv) was added portionwise and the solution was stirred for 1 h, cooled to 0°C, and then sodium azide (85 mg, 1.3 mmol, 2.0 equiv) was added portionwise. After 1 h, the solution was diluted with H 2 O (ca. 10 mL) and Et 2 O (ca. 50 mL). The organic phase was separated, diluted with t-BuOH (2.0 mL), and then cautiously concentrated under reduced pressure at 25°C to provide a solution of crude azide 12 in t-BuOH, which was used without further purification.
General Method 1.4. *Caution* Step 1: Sodium nitrite (72 mg, 1.0 mmol, 1.2 equiv) was added portionwise to a solution of aniline 11 (0.9 mmol, 1.0 equiv) in TFA (0.5 mL) at 0°C over 30 min. The reaction mixture was then allowed to warm to RT for 1.5 h before recooling to 0°C . H 2 O (0.1 mL), sodium azide (62 mg, 1.0 mmol, 1.1 equiv) was added portionwise over 30 min, and then warmed to RT over 1 h. The mixture was basified to pH 8−9 by dropwise addition of saturated aqueous NaHCO 3 , then diluted with CH 2 Cl 2 (ca. 50 mL). The organic phase was separated, diluted with t-BuOH (0.5 mL), and then cautiously concentrated under reduced pressure at 25°C to provide a solution of crude azide 12 in t-BuOH, which was used without further purification.
Step 3: MeOH (15 mL) and K 2 CO 3 (6.0 g, 43.5 mmol) were added to crude 17, and the reaction was stirred at RT for 24 h, before H 2 O (ca. 20 mL) was then added until all the solid carbonate dissolved. MeOH was then removed under reduced pressure and the resultant aqueous mixture was diluted with EtOAc (ca. 500 mL). The organic phase was dried, filtered, and concentrated under reduced pressure. The crude product was purified by column chromatography (0−30% EtOAc in cyclohexane) to give 8l and isolated as a white solid (0.83 g, 67% Spectroscopic and analytical data for 8l are presented in the Supporting Information. 1-(2,4-Dichloro-3-(trifluoromethyl)phenyl)-1H-1,2,3-triazole-4,5d 2 (d 2 -8l). *Caution* Step 1: Sodium nitrite (360 mg, 5.22 mmol, 1.2 equiv) was added portionwise to a solution of 2,4-dichloro-3-(trifluoromethyl)aniline 11a (1.0 g, 4.35 mmol, 1.0 equiv) in TFA (10 mL) at 0°C over 30 min. The reaction mixture was then allowed to warm to RT over 90 min. H 2 O (1.0 mL) was added and then the mixture was recooled to 0°C. Sodium azide (310 mg, 4.78 mmol, 1.1 equiv) was added portionwise over 30 min, then the reaction was allowed to warm to RT over 1 h. The mixture was basified to pH 8−9 by dropwise addition of saturated aqueous NaHCO 3 and then extracted with CH 2 Cl 2 (ca. 250 mL). The organic phase was separated, diluted with t-BuOH (10.0 mL), and then cautiously concentrated under reduced pressure at 25°C to provide a solution of crude azide 12a in t-BuOH, which was used without further purification.
Overlay of DSPL screening hits 6; ADME protocols and data; PK studies; Notum OPTS and TCF/LEF screening data concentration−response curves; ABPP optimization; serine hydrolase selectivity of 1; kinase selectivity panel; safety pharmacology studies; DSC for azide 12a; X-ray structure determination including data collection and refinement statistics; ligand electron-density maps; and spectroscopic and analytical data for 8l (PDF) Identity and quantification of each serine hydrolase identified in SW620 cells by FP-biotin ABPP (XLSX) Molecular formula strings and Notum IC 50 data for 6−9 (CSV)

Accession Codes
Coordinates for X-ray structures of Notum crystallized with 6b (7B89) and 8l (7PK3) have been deposited in the Protein Data Bank. Authors will release the atomic coordinates upon article publication.
Stefano Benvegnù− Alzheimer's Research UK UCL Drug