DOI:10.1002/chem.201804169 Communication

& Biochemistry Discovery of Small Molecule WWP2Ubiquitin Inhibitors Jessica E. Watt,[a] Gregory R. Hughes,[a, c] Samuel Walpole,[b] Serena Monaco,[b] G. Richard Stephenson,[c] Philip C. BulmanPage,[c] AndrewM.Hemmings,[a, c] Jesus Angulo,*[b] and AndrewChantry*[a]

There have been anumber of recent attempts to develop Abstract: We have screened small molecule libraries spe- HECT ligase inhibitors, and more specifically targeting the cificallyfor inhibitors that target WWP2, an E3 ubiquitin Nedd4sub-family.Heclin wasidentified from abicyclic peptide ligase associated with tumouroutgrowth and spread.Se- libraryscreen and can selectively inhibit Smurf2, Nedd4 and lected hits demonstrateddose-dependentWWP2inhibi- WWP1byinducingaconformational change that causes oxida- tion, low micromolar IC50 values, and inhibition of PTEN tion of the activesite cysteine.[12] Chlomipramineemergedasa substrate-specific ubiquitination. Binding to WWP2 was candidate ITCH ligase inhibitor based on an enzymatic assay- confirmed by ligand-based NMR spectroscopy.Further- basedhigh throughput screen and was shown to block both more, we used acombinationofSTD NMR, the recently ITCH auto-ubiquitination and substrate-specific p73 ubiquitina- developed DEEP-STD NMR approach, and docking calcula- tion, althoughIC50 values are in the low millimolarrange.[13] tions, to propose for the first time an NMR-validated 3D Using acovalent tethering method, anew class of Nedd4 molecular model of aWWP2-inhibitor complex. These first ligase inhibitors were identified and showntoreact with a generation WWP2 inhibitors provide amolecular frame- non-catalytic cysteinetohinderubiquitin binding to this exo- work for informing organic synthetic approaches to im- site and subsequent enzymatic processivity.[14] Additional at- prove activity and selectivity. tempts to discover Nedd4 inhibitors have utilisedinsilico modelling approaches supported by thermal shift binding ex- periments and revealed that derivatives of indole-3-carbinol There are approximately 600 E3 in the , may also bind to the processivity exosite in the Nedd4 catalytic and they can be sub-divided into RING type and HECT domain HECT domain and inhibit substrate-specific ubiquitination ligases.[1] Within the HECT E3s is asmall group (9 in total) re- linkedtooncogenic signalling events.[15,16] ferred to as the Nedd4superfamily,which includes the WWP1, Here, we focused our efforts on developing small molecule WWP2, Nedd4, ITCH and Smurf2Ubligases.[2] The 3D crystal inhibitors of the WWP2 , an enzymethat can structureshave been resolved for the HECT domains within cause ubiquitin-dependent degradation of specific tumour Nedd4,[3] WWP1,[4] and WWP2.[5,6] Accumulating evidence sug- suppressor proteins that are commonly lost in many types of gests that the Nedd4 ligases, and notably WWP1 and WWP2, cancer,namely Smad transcriptionfactors, PTEN, and Oct4. To are frequently mis-expressed in cancer.[7–9] Theiroverexpression that aim, we have used high-throughput screening to identify in animalmodels causesincreased tumour outgrowth,[10] and hits from the NCI DiversitySet V and NCI Approved Oncology Set they also selectively control key oncogenic signaling events V smallmolecule libraries, we have confirmed binding to linked to both cancerinitiation and spread.[11] WWP2byNMR spectroscopy,and used acombination of STD NMR,DiffErential EPitope mapping (DEEP)-STD NMR and dock- [a] J. E. Watt, G. R. Hughes, Dr.A.M.Hemmings,Dr. A. Chantry ing calculations to generate for the first time an NMR-validated SchoolofBiological Sciences 3D molecular model of aWWP2-inhibitor complex. To discover University of East Anglia, Norwich NR4 7TJ (UK) E-mail:[email protected] first generation WWP2ubiquitin ligase inhibitors we used a96- [b] S. Walpole,S.Monaco, Dr.J.Angulo well plate format high-throughput screen (HTS) using recombi- SchoolofPharmacy nant proteinsand auto-ubiquitination as areadout for WWP2 University of East Anglia, Norwich NR4 7TJ (UK) activity (Figure 1a). E-mail:[email protected] Initially,todevelopand optimise the HTS assay full-length [c] G. R. Hughes, Dr.G.R.Stephenson, Prof. P. C. Bulman Page, WWP2(WWP2-FL) expressed as aGST fusionprotein in bacte- Dr.A.M.Hemmings SchoolofChemistry ria was attached onto glutathione-coated plates and then incu- University of East Anglia, Norwich NR4 7TJ (UK) batedwith recombinantE1, E2 and Flag-tagged ubiq- Supporting information and the ORCID identification number(s) for the au- uitin (Supporting InformationFigure S1a). The specificity for thor(s) of this article can be found under: WWP2auto-ubiquitinationwas confirmed by detection of Ub- https://doi.org/10.1002/chem.201804169. Flag conjugated WWP2 in the presence of all assay compo-  2018 The Authors. Published by Wiley-VCH Verlag GmbH&Co. KGaA. nentsand confirmed using acatalytically inactive WWP2-FLC838A This is an open access article under the terms of the Creative Commons At- tributionLicense, whichpermits use, distributionand reproduction in any mutant (Supporting Information FigureS1b). Next,weused medium, provided the original work is properly cited. this assay to screen the NCI Diversity Set V and NCI Approved

Chem. Eur.J.2018, 24,17677 –17680 17677  2018 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim Communication

Table 1. Summary of selected hits from WWP2 inhibitorscreen and IC50 values. Data for STD-NMR andCPMG are included,and (?)indicates in- conclusive data due to poorsolubility and/or low signal/noise ratios.

Cpd./NSC Structure IC50 (mm)STD-NMRCPMG

1/2805 0.38 33

7/228155 0.84 ””

12/44750 0.85 ??

19/228150 1.29 ?? Figure 1. High throughput screenfor WWP2 inhibitors. (a) HTS assays for NCI Diversity Set Vwere carried out using 10 mm compound at 0.1%DMSO in single shot with internal controls for 100 %activity (DMSO) and 0% activi- ty (E3 only) on each individual plate. All assays were normalised using 100 % 20/288387 2.30 33 and 0% controls and had athreshold of 70%asindicated by the yellow line. 0.5 2SDfrom the DMSO control mean) pound effects on E1 activity,weused adifferent gel-based ap- was used to determine which compounds would be selected proach duetodifficulties in developing robustand reproduci- for further testing. Weidentified 24 initial hits from the NCI Di- ble plate assays. RecombinantHis-Uba1 and ubiquitin undergo versity Set Vscreen, but no hits emerged from the NCI Oncolo- E1-mediated transthiolation when incubated together andthe gy set (Figure 1a and Supporting Information Figure S1d). Po- formation of highermolecular weight mono-Ub-E1 conjugates tential hits emerging from the single shot assays were then was clearly apparent when separated by SDS-PAGE and stained tested in triplicate under the same assayconditions and at with Coomassie Blue. In the presence of the WWP2 lead inhibi- 10 mm compound concentration, and 17 of the initial 24 hits tory compounds we found no discernible impact on E1 ubiqui- demonstrated significant decrease in WWP2 activity (Fig- tin conjugation (Supporting Information Figure S3b). Hitcom- ure 1b). These were then further ranked more stringently pounds were next tested against Nedd4 and WWP1 to find out based on <20 %residual activity thresholds, and furtherde-se- whether they were capable of inhibiting otherHECT E3 ligases lected based on dose-dependent WWP2 inhibition (Supporting in the same family as WWP2. GST-tagged Nedd4and His- Information Figure S2). In total, 10 potentialhit inhibitors with tagged WWP1 were immobilisedonto glutathione coated and varied structural compositions were identified andthe majority nickel coated plates, respectively.Auto-ubiquitination levels of these gave IC50 values in the low micromolar range (Sup- were determined using HRP-conjugated anti-Flagantibody and porting Information Table S1), and five of these inhibitors had TMB substrate. Both Nedd4and WWP1 produced asignificant IC50 values below 3 mm (Table 1). increaseinODwhen in the presence of E1, E2, and ubiquitin, We next further analysed all 10 selected hit compounds unlike all other reactions (Supporting Information Figure S3c). based on counter assays against E1, E2, Nedd4 and WWP1 There wasnosignificant differencebetween reactions which ubiquitin ligases. Since the WWP2 auto-ubiquitination is de- lacked one or more components, showingthat Nedd4 and pendentonprior sequential activation of ubiquitin by E1 and WWP1 are capable of auto-ubiquitination under these assay E2 ubiquitin ligases, we addressed whetherany of the lead in- conditions. All inhibitors tested demonstrated some degree of

Chem. Eur.J.2018, 24,17677 –17680 www.chemeurj.org 17678  2018 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim Communication inhibition against all three E3 ubiquitin ligases, although sever- very weak signals both in the binding experimentsand in the al of the inhibitors tested weremore potent against WWP2. control spectra, probablyascribable to solubility issues;while Compounds 1, 2 and 12 showed no significant difference be- compound 7 showednoSTD signals. It is important to high- tween the HECT E3 ligases, implying that these compounds light that fast exchange between the free and the bound state are not selective towards one of the ligases(Supporting Infor- is an essential condition for STD NMR.[21] The STD signal is de- mation Figure S3d). Nevertheless, these three compounds do tected in the ligand free state;ifthe ligand resides within the show asignificant reduction in activity for all three proteins. binding pocket for along period of time, the saturation is not Compounds 7 and 19 showedasignificant difference between sufficiently accumulatedduring the saturation time. The low WWP2 and WWP1, andWWP1 and Nedd4 but 7 and 19 pre- IC50 value determined for compound 7 (0.84 mm)and the ab- sented no differencebetween WWP2and Nedd4. Interestingly, sence of STD signals for this hit could indicate too strong an compounds 16, 17, 18, 20 and 21 showed increased inhibition affinity of binding. This would make compound 7 unsuitable towardsWWP2 overWWP1 and Nedd4 (Supporting Informa- for ligand-based NMR observation, but the resultsdonot nec- tion Figure S3d). We next tested for substrate-specific effects, essarily exclude it as abinder. and determined whether these inhibitors could affect WWP2- STD NMR has also great potential in structurally elucidating dependentPTEN ubiquitination. Reactionscontained E1, E2, the binding mode of aligand in complex with the protein at WWP2, PTEN, ubiquitin andATP and were analysed by western atomicdetail,providing the binding epitope mapping of the blotting. In the presence of selected hit compounds represent- ligand.[21] Among the six identified binders based on STD-NMR ing abroad range of IC50 values based on the HTS assay,itis data, we selected NSC288387 (compound 20)since it showed clear that each can robustly inhibit WWP2-dependent PTEN the clearest STD signals (probably as aconsequence of its mid- ubiquitination (Figure 2). range IC50, of 2.30 mm)todeepen the structuralanalysisby performing STD binding epitope mapping. The STD binding epitope mapping(Figure 3a)suggeststhat the main recogni- tion element in the complexbetween WWP2 HECT and com- pound 20 is the unsubstituted N-phenyl ring (labelled Bin Figure3); ring Aand the aliphatic ether adjacent to it show weaker interactions, suggesting that this part of the molecule would be the most solvent exposed. We then used computational techniques to identify the on the WWP2 HECT domain (based on the PDB ID:4Y075)and to then dock compound 20 into the putative Figure 2. Effects of WWP2inhibitors on substrate-specific PTEN ubiquitina- binding pocket (see Supporting Information TableS2and Fig- tion. In vitro GST-PTENubiquitination reactionsinthe absence and presence ure 3b). The lowest energydocking solution, shown in Fig- of compounds, analysed using western blottingwith anti-GST and anti-His ure 3b (see also Supporting Information Table S2), was in very antibodies. good agreement with the binding epitope experimentally de- Ligand-based NMR wasthen used to confirm binding of hit termined by STD NMR. Furthermore, as astep further forthe compounds to the WWP2 HECT catalytic domain. We used i), Saturation Transfer Difference (STD) NMR and ii), Carr–Purcell– Meiboom–Gill (CPMG) NMR experiments, two widely used techniques for lead screening in drug discovery.[17] STD NMR spectroscopy relies on protein-selective saturation being trans- ferred only to the bound ligands in fast exchange with the re- ceptor.[18,19] Hence, the presence of signals in the STD differ- ence spectra of asample containing one ligand (or more) in the presence of areceptor indicates binding. CPMG NMRex- periments, also called T2-filter experiments, rely on afilter (re- peating spin-echo buildingblocks in the pulse sequence) whose aim is to eliminate the signals comingfrom large mole- [20] cules (usually having ashort transverse relaxation time, T2).

Bound ligands transiently acquireaT2 similartothat of the re- ceptor,and so their signal will be filtered out as well. In the CPMG spectrum, the absence (or strongreduction) of ligand signals in the presence of areceptor indicates binding. Figure 3. Differential epitope mapping of compound20(NSC288387) in The NMR screening provided clear-cutresults for six of the complex with WWP2-HECT domain. a) STD Binding epitope mapping, hit compounds,asshown by positive binding resultsinboth b) lowest energy docking solution of NSC288387 onto WWP2 HECT domain, the STD and CPMG NMR experiments (1, 2, 16, 17, 18 and 20, c) DEEP-STD factors (1.5 ppm/0.5 ppm) and d) details of the binding pocket for NSC288387.Ind), residues within 5 Š from the ligand are shown, protein see Table 1, Supporting Information Table S1 and Figure S4). protonsresonating at 0.5 ppm and 1.5 ppm are enclosed in agreen and Compounds 12, 19 and 21 were excludedasthey showed orangesurface, respectively(see also SupportingInformation Table S5).

Chem. Eur.J.2018, 24,17677 –17680 www.chemeurj.org 17679  2018 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim Communication validation of the 3D model of the complex,weemployedthe Foundation. G.H. was supported by aBiotechnology and Bio- newly developed DEEP-STD NMR approach (relying on protein logical Sciences Research Council (BBSRC) iCASE PhD Student- irradiation at differential frequencies) which allows the identifi- ship awarded to A.C. andP.B.P.incollaboration with Charn- cation of the types of amino-acids lining the binding pocket wood Molecular Ltd. S.M. acknowledges aPhD Studentship that make inter-molecular contactswith specific single protons from the School of Pharmacy at the University of East Anglia of the ligand.[22] The DEEP-STD NMR data obtained by differen- (UEA). J.A. and S.W.received funding from the BBSRC through tial irradiation of the protein at 0.50 and 1.50 ppm (Figure 3c) Research Grant BB/P010660/1 and aDTP-PhD Studentship, re- providedadditional experimentalsupport for the 3D molecular spectively.Wethank Dr.Tharin Blumenscheinfrom the School model of the complex generatedbythe docking approach. of Chemistry at UEA for helpful discussions. Frequencies 0.50 and 1.50 ppm correspond to methyl groups of valine and isoleucine, and the arginine b/g protons, respec- Conflict of interest tively (see SupportingInformation Table S3 and S4). Closer in- spection of the docking solution in the putative binding The authors declare no conflict of interest. pocket reveals the presence of Val806 and Arg803 at close dis- tances from ring Aofthe ligand (Figure 3d and Supporting In- formationTable S5). Keywords: DEEP-STD NMR · drug discovery · ubiquitin ligase · Remarkably,the docking solution structure shows ligand inhibitors · saturation transfer difference NMR protons c, dand epointing towards the methyl group of [1] W. Li,M.H.Bengtson, A. Ulbrich,A.Matsuda, V. A. Reddy,A.Orth, S. K. Val806, and protonsfand g, pointing towards the Arg803, on Chanda, S. Batalov,C.A.P.Joazeiro, PLoS ONE 2008, 3,e1487. the opposite side of the bindingpocket. This is in perfect [2] S. Soond, A. Chantry, Bioessays 2011, 33,749. agreement with the DEEP-STDfactors (1.5/0.5 ppm) observed [3] E. Maspero, S. Mari, E. Valentini, A. Musacchio, A. Fish, S. Pasqualato, S. for these protons (Figure 3c): protons c, dand eshow negative Polo, EMBO Rep. 2011, 12,342. [4] M. A. Verdecia, C. A. Joazeiro, N. J. Wells, J. L. Ferrer,M.E.Bowman,T. DEEP-STD factors, andprotons fand gshow positive DEEP-STD Hunter,J.P.Noel, Mol. Cell 2003, 11,249. factors (suggestingvicinity with protein residues resonating at [5] W. Gong, X. Zhang, W. Zhang,J.Li, Z. Li, Acta Crystallogr.Sect. F 2015, 0.5 and 1.5 ppm, respectively). The excellent match between 71,1251. the docking solution of compound 20 in the putative binding [6] Z. Chen,H.Jiang, W. Xu, X. Li,D.R.Dempsey,X.Zhang, P. Devreotes, C. Wolberger,L.Mario Amzel, S. B. Gabelli,P.A.Cole, Mol. Cell 2017, 66, pocket and the pharmacophore of both the ligand and the 345. binding site (obtained by STD NMR and DEEP-STD NMR, re- [7] E. Cerami, J. Gao, U. Dogrusoz, B. E. Gross, S. O. Sumer,B.A.Aksoy,A. spectively) give us solidground to propose the model as the Jacobsen, C. J. Byrne, M. L. Heuer,E.Larsson, Y. Antipin,B.Reva, A. P. first 3D structure of aWWP2 inhibitor bound to its target Goldberg,C.Sander,N.Schultz, Cancer Discov. 2012, 2,401. [8] S. M. Soond, P. G. Smith,L.Wahl,T.E.Swingler,I.M.Clark, A. M. Hem- (Figure 3and Supporting Information Figure S5). Furthermore, mings, A. Chantry, Biochim. Biophys. Acta Mol. Basis Dis. 2013, 1832, this NMR-validated model is supported by molecular dynamics 2127. simulation studies demonstrating stability of the complex [9] X. Zou, G. Levy-Cohen,M.Blank, Biochim. Biophys. Acta 2015, 1856,91. (Supporting Information Figures S6–S8). [10] S. Maddika,S.Kavela, N. Rani, V. R. Palicharia, J. L. Pokorny,J.N.Sarkaria, J. Chen, Nat. Cell Biol. 2011, 13,728. In summary,the compounds discovered in this study repre- [11] A. Chantry, Cell Cycle 2011, 10,2437. sent the first generation of WWP2 ubiquitin ligase inhibitors. [12] T. Mund, M. J. Lewis, S. Maslen, H. R. Pelham, Proc. Natl. Acad.Sci. USA Additionally,through acombination of HTS, ligand-based NMR, 2014, 111,16736. DEEP-STD NMR and dockingcalculations we have been ableto [13] M. Rossi, B. Rotblat, K. Ansell, I. Amelio, M. Caraglia, G. Misso, F. Bernas- sola, C. N. Cavasotto,R.A.Knight, A. Ciechanover,G.Melino, Cell Death propose for the first time a3Dmolecular modelofaWWP2-in- Dis. 2014, 5,e1203. hibitor complex. Collectively,these data provide aframework [14] S. G. Kathman, I. Span, A. T. Smith, Z. Xu, J. Zhan, A. C. Rosenzweig,A.V. for developing new classes of ubiquitin ligase inhibitors that Statsyuk, J. Am. Chem. Soc. 2015, 137,12442. can progress through further organic synthesis to improve [15] I. Aronchik, A. Kundu, J. G. Quirit, G. L. Firestone, Mol. CancerRes. 2014, 12,1621. compound activity and specificityand have potential togener- [16] J. G. Quirit,S.N.Lavrenov,K.Poindexter,J.Xu, C. Kyauk, K. A. Durkin, I. ate lead therapeutic compounds. We are currently workingto- Aronchik, T. Tomasiak, Y. A. Solomatin,M.N.Preobrazhenskaya, G. L. wards unravelling the mechanism of inhibition by thesenovel Firestone, Biochem Pharmacol. 2017, 127,13. ligandsand further probingthe ligand binding sites using [17] M. Pellecchia,I.Bertini, D. Cowburn, C. Dalvit, E. Giralt, W. Jahnke, T. L. James, S. W. Homans, H. Kessler, C. Luchinat, C. B. Meyer,H.Oschkinat, combined DEEP-STD NMR and crystallographyapproaches. We J. Peng, H. Schwalbe,G.Siegal, Nat. Rev.Drug Discovery 2008, 7,738. anticipate that, buildingupfrom this work, subsequent lead [18] J. Angulo, P. M. Nieto, Eur.Biophys. J. 2011, 40,1357. optimization will provide novel analogues that can progress [19] M. Mayer,B.Meyer, Angew.Chem. Int. Ed. 1999, 38,1784 –1788; Angew. through to potentialtherapeutic compounds. Chem. 1999, 111,1902 –1906. [20] S. Meiboom, D. Gill, Rev.Sci. Instrum. 1958, 29,688. [21] M. Singh, B. Tam, B. Akabayov, Molecules 2018, 23,233. Acknowledgements [22] S. Monaco,L.E.Tailford, N. Juge,J.Angulo, Angew.Chem. Int. Ed. 2017, 56,15289; Angew.Chem. 2017, 129,15491.

The work was supported by funding to A.C. from Prostate Manuscript received:August15, 2018 Cancer UK supporting aPhD Studentship for J.W.through Re- Accepted manuscript online:September12, 2018 search Grant S13-007, and funding to A.C. from James Tudor Version of record online:November 6, 2018

Chem. Eur.J.2018, 24,17677 –17680 www.chemeurj.org 17680  2018 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim