Fragment-Based Discovery of Novel MUS81 Inhibitors

Abstract

MUS81 is a structure-selective endonuclease that cleaves various branched DNA structures arising from natural physiological processes such as homologous recombination and mitosis. Due to this, MUS81 is able to relieve replication stress, and its function has been reported to be critical to the survival of many cancers, particularly those with dysfunctional DNA-repair machinery. There is therefore interest in MUS81 as a cancer drug target, yet there are currently few small molecule inhibitors of this enzyme reported, and no liganded crystal structures are available to guide hit optimization. Here we report the fragment-based discovery of novel small molecule MUS81 inhibitors with sub-μM biochemical activity. These inhibitors were used to develop a novel crystal system, providing the first structural insight into the inhibition of MUS81 with small molecules.

MUS81 is a structure-selective endonuclease that, when in complex with EME1, cleaves various branched DNA structures (such as Holliday Junctions) arising from natural physiological processes including homologous recombination and genetic recombination.¹⁻⁴ As homologous recombination is one of the major pathways by which cells repair DNA damage, MUS81 contributes to the maintenance of genome stability. As such, there is evidence in support of its function as a tumor suppressor.⁵⁻⁸ Paradoxically, however, several recent studies have now implicated MUS81 activity in playing a role in the progression of several cancers, such as serous ovarian cancer and leukemia.^9,10 Additionally, inhibition of MUS81 activity has been shown to sensitize certain ovarian cancers to chemotherapeutic agents such as cisplatin, as well as to poly-ADP ribose polymerase inhibitors.^9,11−13 There is thus interest in MUS81 as a potential cancer drug target, yet few small molecule inhibitors of this endonuclease have been reported.¹⁴ Here, we report the fragment-based discovery of novel small molecule inhibitors of MUS81-EME1. Hit optimization resulted in compounds with sub-μM biochemical activity and favorable physicochemical and in vitro ADMET properties. We also report a series of crystal structures of MUS81-EME1-inhibitor complexes, providing the first structural insight into small molecule inhibition of the MUS81 endonuclease. Together, these results provide a structural and chemical basis for the further exploration of MUS81 as a potential cancer drug target.

A fragment-based hit finding approach was pursued as a means to find small molecule inhibitors of MUS81. Inspection of the few crystal structures available for human MUS81^15,16 strongly suggested the catalytic site of the enzyme to be the only site on the protein viable for targeting with small molecules. The catalytic site of MUS81 is dominated by a cluster of catalytic acidic residues expected to bind two Mg²⁺ ions.^15,16 Due to this, we designed a small library composed of around 360 fragment-sized molecules (MW < 350 Da) bearing chemical motifs known, or expected, to bind one or two metal (e.g., Mg²⁺) ions. We then screened this library, along with a more standard fragment library composed of around 800 chemically diverse compounds, against a truncated form of the human MUS81-EME1 heterodimer¹⁷ using differential scanning fluorimetry (DSF) (for full details, see the Supporting Information). All fragments were screened against the enzyme at a final concentration of 1 mM, in the presence of either 10 mM MgCl₂ or 10 mM EDTA, or in the absence of additive.¹⁸ The metal-binding library yielded a primary hit rate of around 2.5%, with the chemically diverse library yielding a primary hit rate of only 0.9%. Fragments identified as potential hits by DSF were then analyzed by surface plasmon resonance (SPR) to confirm binding and determine K_d values for MUS81-EME1 (full experimental details can be found in the Supporting Information). From these experiments, around 10 fragments from the metal-binding library were confirmed as true binders of MUS81-EME1, with measured K_d values in the double-digit μM range or weaker, with compound 1 standing out as the most promising hit, with a K_d of 56 μM (Figures 1 and S1). No hits from the diverse library were validated when tested by SPR and were not followed up further.

Figure 1.

(a) Chemical structure of compound 1. (b) Surface plasmon resonance sensorgrams of 1 for binding to MUS81-EME1.

Before attempting to characterize compound 1 structurally or biochemically, we first sought to increase its potency through limited SAR exploration by SPR to assess affinities for MUS81-EME1. Fairly confident that the pyrimidinone motif of 1 would be chelating to the two Mg²⁺ ions expected to be present in the active site of MUS81, we attempted to increase the potency of compound 1 by (1) optimizing Mg²⁺ binding, and (2) exploring substitutions and replacements of the phenyl ring. Replacement of the methyl ester group of the pyrimidinone with primary, secondary or tertiary amides (compounds 2, 3, and 4) all led to a loss or significant reduction in affinity (Table 1). Replacement of the methyl ester with an acidic group (5), however, led to a significant increase in potency (∼20-fold), and gave us further confidence that this series was indeed binding to the Mg²⁺ ions of the catalytic site of MUS81.

Table 1. Activity and Efficiency Metrics of Pyrimidinone Derivatives against MUS81.

^aIC₅₀ and K_d values are the geometric mean ± SEM (for IC₅₀ values) or geometric standard deviation (for K_d values) of at least three determinations.

^bLE (kcal/mol/HA): 1.37 × pIC₅₀/HAC (Heavy Atom Count).

^cLLE: pIC₅₀ – LogD. LogD measured via shake-flask method in octanol and water at pH 7.4.

We next explored the phenyl group of this series, focusing on the acid 5 (Table 1). Replacement of the phenyl ring with a methyl (6) led to a ∼20-fold drop in potency. While not surprising, the reduction in affinity was perhaps not as great as expected, and again reinforced Mg-chelation as the likely mode by which these compounds were binding to MUS81. Replacement of the phenyl ring with the equivalent saturated group (7) resulted in no change in affinity, and we thus pursued the phenyl group for further exploration due to the increased synthetic tractability of the aromatic system over the unsaturated system. Chloro substitution at the ortho or meta positions of the phenyl ring of 5 (8 and 9) slightly reduced binding affinity, while similar substitution at the para position (10) led to a noticeable (∼2-fold), if modest, increase in affinity. Substitution at the para position of the phenyl ring with bromide (11) had a similar effect (∼2-fold increase in affinity for MUS81), as did replacement of the phenyl ring with a β-naphthyl group (12).

With compounds in-hand displaying single-digit μM affinity for MUS81-EME1, we decided to use these compounds as tools to develop a biochemical assay to allow us to assess the ability of these compounds to inhibit the enzymatic function of MUS81. A fluorescence quenching-based assay using branched DNA as a substrate was thus developed (see the Supporting Information for full details) with assessment of compounds yielding IC₅₀ values in excellent agreement with the biophysical K_d values (Table 1).

In order to guide the further improvement of the pyrimidinone series, we wished to obtain structural data concerning MUS81-EME1-ligand interactions. We were able to obtain crystals of MUS81-EME1 in complex with DNA following previously reported procedures,^15,16 however, we were unable to observe electron density for any inhibitors when added either by soaking or cocrystallization. We next turned to hydrogen–deuterium exchange mass spectrometry (HDX-MS) as a means to obtain structural data concerning the site and mode by which the pyrimidinone compounds may be binding to the MUS81-EME1 complex. Comparison of HDX patterns following incubation of MUS81-EME1 with and without compound 10 showed clear protection of solvent-exposed backbone NHs surrounding the catalytic site of MUS81, strongly implying the compound binds to this site as anticipated, presumably via coordination with the Mg²⁺ ions expected to be present (Figure S2). This data gave us confidence in the binding site of compound 10 and compelled us to reassess our crystallographic system.

Inspection of the MUS81-EME1 crystal structures obtained by us, and others,^15,16 suggested that removal of the C-terminal “lobe” of the complex, which does not contain the catalytic site, may lead to a less dynamic protein complex and thereby aid crystallization. We therefore attempted to recombinantly express a protein construct composed of residues 246–462 of MUS81 and residues 246–455 of EME1 (referred to from hereon as “MUS81-EME1^N”) in E. coli.(19) Soluble protein was readily obtained and purified, with apo crystallization experiments using commercially available protein crystallization screens yielding crystals in around 10% of the conditions tested. From these crystals, a 2.15 Å structure was obtained, revealing the truncated MUS81-EME1^N construct to adopt a conformation almost identical to that of the previous crystallization construct,¹⁶ with an RMSD of 0.940 Å (Figure 2a). We next attempted to cocrystallize all compounds that had thus far given single-digit μM IC₅₀ values (Table 1) with this truncated MUS81-EME1^N construct. These efforts yielded good quality electron density maps for MUS81-EME1^N crystals grown in the presence of compounds 10 and 12. Structures determined for these compounds, although at modest resolutions (2.80 and 2.91 Å, respectively), showed both compounds to unambiguously bind to the catalytic site of MUS81, coordinating to two Mg²⁺ ions (themselves coordinated to residues D307, E333, and R334) of the protein) via the pyrimidinone moiety (Figures 2b and S3a). The naphthyl and chloro-phenyl groups occupy very similar positions, and are not buried within a pocket but sit in a shallow groove on the protein’s surface (Figure S3a).

Figure 2.

(a) Overlay of MUS81-EME1^N crystal structure reported here (MUS81, green; EME1, blue) with previously reported MUS81-EME1 crystal structure (PDB entry 4P0P, in light gray (protein) and dark gray (DNA)). RMSD of alignment = 0.940 Å for 327–327 C α atoms. (b–d) Crystal structures of MUS81-EME1^N bound by compounds 10, 16, and 15, respectively. (e) Overlay of MUS81-EME1^N-21 (green carbons) and MUS81-EME1^N-15 (white carbons) complexes. Mg²⁺ ions are shown as orange spheres, polar contacts as black dashes. Crystallographic data collection and refinement statistics can be found in Table S1.

Following detailed inspection of these structures, two potential avenues for further improvement in the activity of the pyrimidinone series were evident: (1) the sterically unhindered (i.e., open) position of the phenyl group suggested significant scope for substitution of all positions of the ring, and (2) it appeared that interactions with the chelated Mg²⁺ ions were key to ligand binding, suggesting further optimization of the Mg²⁺-binding (pyrimidinone) group to be potentially beneficial. To explore the phenyl ring of 5 further, we synthesized compounds with phenyl groups attached to the ortho (13), meta (14), and para (15) positions, with the ortho-substituted biphenyl further modified by the addition of a cyanomethyl group at the para position of the terminal phenyl group (16). While none of these modifications resulted in significantly increased affinity or biochemical activity, all were well-tolerated, giving single-digit μM K_d and IC₅₀ values. We next attempted to cocrystallize these compounds (13–16) with MUS81-EME1^N. These experiments yielded two crystal structures: a 2.02 Å structure for 16 (Figure 2c) and a 2.73 Å structure for 15 (Figure 2d). The crystal structure of MUS81-EME1^N in complex with 16 shows the compound to clearly bind to the catalytic site via Mg²⁺ ion coordination (Figure 2c). Interestingly, the two crystallographically unique MUS81-EME1^N complexes of the asymmetric unit reveal 16 in two quite distinct orientations, 180 degrees relative to one another, while maintaining interactions with the Mg²⁺ ions (Figure S3b). This observation implied strongly that the interactions with the Mg²⁺ ions were driving the affinity between the compound and protein, and that the ortho-phenyl ring was contributing little to binding interactions.

The crystal structure of 15 again revealed the inhibitor to bind to the catalytic site of MUS81 via Mg²⁺ coordination (Figure 2d), with the compound overlaying very closely with that of the chloro-substituted (10) and naphthyl (12) compounds (Figure S3a). Interestingly, the terminal phenyl ring appeared to sit close to an acidic residue of MUS81 (D346), suggesting that a basic group at this position might increase affinity/activity (Figure 2d). Indeed, replacement of the terminal phenyl group of 15 with a piperazine group (17) led to an improvement, albeit modest (∼2-fold), in biochemical activity (Table 1).

Parallel to this work, we sought to further optimize inhibitor interactions with the catalytic Mg²⁺ ions. To achieve this, we synthesized and tested three pyridinone isomer analogues of the pyrimidinone 15 (18–20) (Table 2). Isomer 18 proved the most potent compound of the three, displaying an approximately 3-fold improvement in IC₅₀ compared to the pyrimidinone equivalent (15).

Table 2. Activity and Efficiency Metrics of Compounds 15, 18, 19, 20 against MUS81.

^aIC₅₀ and K_d values are the geometric mean ± SEM (for IC₅₀ values) or geometric standard deviation (for K_d values) of at least three determinations.

^bLE (kcal/mol/HA): 1.37 × pIC₅₀/HAC (Heavy Atom Count).

^cLLE: pIC₅₀ – LogD. LogD measured via shake-flask method in octanol and water at pH 7.4.

With pyridinone 18 identified as our most potent Mg²⁺-binding group, we combined this with the piperazinylphenyl moiety of the most potent pyrimidinone compound (17). This hybrid, 21, resulted in a negligible increase in potency but a significant increase in LLE (from 5.6 (18) to 7.1 (21)) (Table 3). We next sought to optimize the pyridinone series further. Fortunately, we were able to obtain a cocrystal structure of 21 in complex with MUS81-EME1^N (refined to 2.36 Å), which, while indicating the compound to bind to the catalytic site by chelating to the two Mg²⁺ ions, also revealed some differences in the binding mode compared to the pyrimidinone series (Figure 2e). Specifically, the position of 21 appeared shifted relative to that of the pyrimidinones characterized structurally (e.g., 15, see Figure 2e), with movement of the side-chain of R348 creating a new platform which the phenyl ring of 21, which surprisingly is twisted by 90° relative to the pyridinone ring, packs onto. Unfortunately, the piperazine group did not appear to form any meaningful electrostatic contacts with the protein as hoped, however, the new orientation of the compound (relative to the pyrimidinone series, see Figure 2e), plus the twisted conformation of the phenyl ring, offered two further potential routes to optimize the pyridinone series. First, a nitrile group was attached to the ortho position of 21 (giving compound 22) in an attempt to interact with (or displace) a crystallographic water molecule observed close to this position (Figure S3c). However, this modification did not result in any gain in potency for MUS81 (Table 3). Next, noting the position of D346 was still >4 Å from the piperazine group of 21, we grew the compound further by inserting a benzoyl group in between the phenyl and piperazine groups (23), with an ortho nitrile substituted analogue (24) also synthesized. Both 23 and 24 displayed an approximately 2-fold improvement in biochemical activity against MUS81 (compared to 18), with IC₅₀ values of 0.32 and 0.27 μM, respectively (Table 3).

Table 3. Activity and Efficiency Metrics of Pyridinone Derivatives against MUS81.

^aIC₅₀ and K_d values are the geometric mean ± SEM (for IC₅₀ values) or geometric standard deviation (for K_d values) of at least three determinations.

^bLE (kcal/mol/HA): 1.37 × pIC₅₀/HAC (Heavy Atom Count).

^cLLE: pIC₅₀ – LogD. LogD measured via shake-flask method in octanol and water at pH 7.4.

We next assessed the physicochemical and in vitro ADMET properties of top compounds 23 and 24 (Table S2). Both compounds display favorable physiochemical properties, with low logDs (<0) and high solubility (>200 μM), and excellent lipophilic ligand efficiencies (6.9 and 7.6 for 23 and 24, respectively) (Tables 3 and S1). Compounds 23 and 24 were then assessed for metabolic stability in human liver microsomes (HLM) and rat hepatocytes (rHep), with both showing promising stability, with HLM and rHep Clint values of <3 μL/min/mg and 11 μL/min/1 × 10⁶ cells and 7 μL/min/mg and 10 μL/min/1 × 10⁶ cells for 23 and 24, respectively (Table S2). The potential for efflux using the Caco2 assay was also assessed, with both 23 and 24 displaying low efflux ratios (1.5 and 1.1, respectively) (Table S2). 23 and 24 were also assessed for their ability to inhibit the hERG ion channel (in vitro), with both compounds giving IC₅₀ values >40 μM. Both compounds were further screened against a safety panel (Eurofins SafetyScreen panel) of 22 common off-targets, comprising enzymes, receptors and ion channels (e.g., COX2, PDEs, INSR, GABA, NMDA, CaV-L). Pleasingly, neither compound showed any safety concern, with IC₅₀ values >100 μM.

The main synthetic routes used to explore the SAR are shown in Schemes 1–4. Based on the original hit 1, several amide analogues (2–4) were prepared via aminolysis of ester 1 (Scheme 1). Additionally, a number of close analogues to 1 were available commercially including acids 5 and 6.

Scheme 1. Synthesis of Amides 2–4.

Reagents and conditions: (i) RR’NH, MeOH, 80–140 °C.

Scheme 4. Synthesis of Regioisomeric Hydroxy Pyridone Acids 19 and 20.

Reagents and conditions: (i) NaH, MeI, THF, rt to 50 °C; (ii) n-BuLi, THF, −78 to 0 °C then CO₂, −78 °C to rt; (iii) SOCl₂, MeOH, reflux; (iv) [1,1′-biphenyl]-4-ylboronic acid, XPhos Pd G2, Cs₂CO₃, 1,4-dioxane/water, 100 °C; (v) BBr₃, CH₂Cl₂, rt then 30% HBr in AcOH, 120 °C; (vi) BnBr, Cs₂CO₃, DMF, rt; (vii) BnOH, NaH, DMF, 0 °C then 1 M NaOH; (viii) [1,1′-biphenyl]-4-ylboronic acid, Pd(dppf)Cl₂·CH₂Cl₂, Na₂CO₃, DME/water, 80 °C; (ix) 10% Pd/C, H₂, THF/MeOH rt.

The series of hydroxy pyrimidone acids 7–17 were prepared following the method described by Culbertson²⁰ (Scheme 2). Commercial nitriles 7a–17a were reacted with hydroxylamine to afford N-hydroxy amidines 7b–17b, which were in turn treated with DMAD (dimethylacetylene dicarboxylate) to generate intermediates 7c–17c via a Michael-type addition reaction. Intermediates 7c–17c could be isolated but were often used directly without purification in the subsequent cyclization reaction to afford pyrimidone esters 7d–17d which could then undergo base-mediated hydrolysis to hydroxy pyrimidone acids 7–14. In addition, where the aromatic group featured a halide (15d and 16d) a Suzuki–Miyaura cross-coupling reaction could be employed to generate biphenyl derivatives. It was observed that under the aqueous conditions of the Suzuki reaction the ester was found to concomitantly hydrolyze to give the pyrimidone acid directly.

Scheme 2. Synthesis of Hydroxy Pyrimidone Acids 7–17.

Reagents and conditions: (i) 50% Hydroxylamine in water, EtOH, 70 °C; (ii) DMAD, CHCl₃, 60 °C; (iii) xylenes, 140 °C; (iv) LiOH, MeOH; (v) PhB(OH)₂, Pd(dppf)₂Cl₂·CH₂Cl₂, Na₂CO₃, 1,4-dioxane/water, 100 °C; (vi) [4-(cyanomethyl)phenyl]boronic acid, RuPhos Pd G3, RuPhos, K₂CO₃, 1,4-dioxane/water, 120 °C; (vii) TFA, CH₂Cl₂, rt.

The synthesis of hydroxy pyrid-4-one acids 18, 21–24 begins from pyrone Int. A1 which was prepared as previously described in two steps from ethyl 4-(benzyloxy)-3-oxobutanoate.^21,22 Pyrone Int. A1 reacted readily with a range of anilines in the presence of acetic acid which, following acidic debenzylation and ester saponification afforded compounds 18, 21, and 24 (Scheme 3). For 22 and 23, it was first necessary to introduce the piperazine group by amide coupling or a Buchwald–Hartwig palladium catalyzed C–N cross-coupling reaction prior to global deprotection.

Scheme 3. Synthesis of Hydroxy Pyrid-4-one Acids 18, 21–24.

Reagents and conditions: (i) Amine, AcOH, 50–90 °C; (ii) TFA, CH₂Cl₂, rt to reflux; (iii) LiOH, MeOH, rt to reflux; (iv) tert-butyl piperazine-1-carboxylate, Pd₂(dba)₃.CHCl₃, BINAP, Cs₂CO₃, 1,4-dioxane, 100 °C; (v) tert-butyl piperazine-1-carboxylate, HATU, DIPEA, DMF, rt.

Few routes to highly substituted hydroxy pyridone acids have been previously described.²³ We were particularly interested in routes that would generate a protected hydroxy pyridone acid that would allow late-stage variation of the pendant aromatic group to aid SAR exploration (Scheme 4). To this end, 6-chloro 3-hydroxy pyridine 19a was methylated and then the methoxy group was used to direct ortho lithiation and trapping with carbon dioxide to install the acid at C4. Following esterification, 19d could then undergo a late-stage Suzuki cross-coupling reaction to install the biphenyl group prior to global demethylation using boron tribromide to afford 19. In the case of 20, it was convenient to start from dibromopyridine 20a which was first protected as the benzyl ether 20b. The key step was the selective S_NAr displacement of the C4-bromide which was achieved using the sodium alkoxide derived from benzyl alcohol. During this reaction some transesterification was observed so the mixture of methyl and benzyl esters were saponified directly to afford acid 20c. A subsequent Suzuki cross-coupling reaction installed the biphenyl group which then afforded 20 following debenzylation.

We have reported here a fragment-based hit finding campaign targeting MUS81 from which we discovered novel inhibitors (23 and 24) with sub-μM biochemical activity and promising in vitro safety and ADMET properties. Crystal structures determined for several key compounds provide the first structural insight into MUS81 inhibition with small molecules and provide a basis for the further study of this potential cancer drug target.

Glossary

Abbreviations

CyHex

cyclohexyl

dppf

1,1′-ferrocenediyl-bis(diphenylphosphine)

DSF

differential scanning fluorimetry

EME1

essential meiotic structure-specific endonuclease 1

HATU

1-[bis(dimethylamino)methylene]-1H-1,2,3-triazolo[4,5-b]pyridinium 3-oxid hexafluorophosphate

HDX-MS

hydrogen–deuterium exchange mass spectrometry

MUS81

methylmethanesulfonate and ultraviolet sensitive gene clone 81

RuPhos

2-dicyclohexylphosphino-2′,6′-diisopropoxybiphenyl

SPR

surface plasmon resonance

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.3c00453.

• Full experimental details and additional figures and tables (PDF)

Accession Codes

All MUS81-EME1 crystal structures reported here have been deposited in the Protein Data Bank with the following accession codes: 9F98 (apo), 9F99 (10), 9F9A (12), 9F9K (15), 9F9L (16) and 9F9M (21).

Author Present Address

^# MRC Laboratory of Molecular Biology, Francis Crick Avenue, Cambridge, CB2 0QH, U.K

Author Contributions

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

The authors declare no competing financial interest.