Organometallic Inhibitor for the Human Repair Enzyme 7,8-Dihydro-8-Oxoguanosine Triphosphatase

A significant portion of all proteins encoded in the human genome bind or utilize purine-based nucleosides or nucleotides, such as ATP, GTP, NAD(P), FAD, PAPS, and coenzyme A.^[1] These proteins, among which are important classes of enzymes such as kinases, ATPases, ligases, helicases, and chaperones, are involved in all types of cellular processes, are at the origin of many human diseases, and therefore constitute important targets for the design of small molecule drugs. Our group recently introduced a strategy for designing highly potent and selective ATP-mimetic inhibitors of protein kinases based on substitutionally inert transition metal complexes,^[2,3] and we were wondering if –due to certain common features of nucleotide binding sites^[4]– inert transition metal complexes may constitute an attractive class of scaffolds for the development of inhibitors of other nucleotide binding proteins. We here now wish to report how we derived at a novel organoruthenium complex as the first low nanomolar and selective inhibitor of the human repair enzyme 7,8-dihydro-8-oxoguanosine triphosphatase (8-oxo-dGTPase, NUDT1, MTH1), an enzyme that hydrolyzes oxidized purine nucleoside triphosphates and thereby prevents their misincorporation into DNA.^[5]

We started our study with the design of the organoruthenium half-sandwich complex 1, containing a bidentate 8-(pyridin-2-yl)adenine ligand, an η^[5]-coordinated cyclopentadienyl moiety, and a CO ligand (Figure 1). Possessing a ruthenium-coordinated adenine nucleobase in combination with a molecular structure that is devised to mimic the overall shape of nucleotides, we envisioned organometallic 1 to serve as a probe for adenine nucleotide binding proteins. To verify this hypothesis, a probe-based technology using a biotinylated acyl phosphate derivative of ATP and ADP that can irreversibly react with conserved lysine residues in the pocket of ATP binding proteins was employed, thereby allowing to determine direct competition between probe and inhibitor binding within any biological sample.^[6] Accordingly, when organometallic 1 was profiled at a concentration of 100 μM against more than 150 ATP-binding proteins within the lysate of HL60 cells,^[7] 1 was identified to be a binder to a few cellular proteins, particularly the nuclear chaperone midasin and a homolog thereof,^[8] a protein involved in the pyrimidine biosynthesis (CAD protein),^[8] and the repair enzyme MTH1^[5] (Figure 2). We chose MTH1 for further investigations since it plays an important role as a repair enzyme related to oxidative stress in cells and its inhibition might exhibit an interesting way to reduce cancer growth causing RAS-induced oxidative damage, leading to DNA double-strand-breaks, and provoking cells to enter premature senescence.^[5] Furthermore, its co-crystal structure with bound 8‐oxo‐dGMP has been reported, thereby potentially facilitating the identification of important inhibitor-enzyme-interactions for further improvements of affinity.^[10,11]

Figure 1.

Organometallics investigated and developed as inhibitors for human 8-oxo-dGTPase (MTH1, NUDT1). Determined IC₅₀ values are given in brackets. All complexes were synthesized and used as racemates.

Figure 2.

Profiling of complex 1 (100 μM) against ATP binding proteins within the cell lysate of HL60 cells with probe technology (KiNativ™, ActivX Biosciences, La Jolla, CA, USA) using a biotinylated acyl phosphate derivative of ATP and ADP. Shown is the inhibition of probe binding by complex 1 to 207 confirmed sites of 153 ATP binding proteins. Main hits: Midasin (MDN1, 90% inh.), a midasin homolog (90% inh.), CAD protein (87% inh.), and 8-oxo-dGTPase (MTH1/NUDT1, 72% inh.). See Supporting Information for more details.

The concentration of 1 at which the activity of MTH1 is reduced to 50% (IC₅₀ value) was determined with an HPLC assay to be a reasonable starting point with modest 151 μM (Figure 3). It is noteworthy that in addition to its affinity for 8-oxo-dGTP (K_m = 15.2 μM), MTH1 also binds and efficiently hydrolyses oxidized ATP derivatives (e.g. 8-oxo-dATP: K_m = 13.9 μM),^[12] thus rationalizing why the adenine-containing nucleotide probe 1 possesses binding affinity to 8-oxo-dGTPase. Fortuitously, when we replaced the 8-(pyridin-2-yl)adenine ligand of 1 against a cyclometalated 4-amino-6-(pyridin-2-yl)quinazoline, obtaining complex 2, the IC₅₀ value was improved by more than two orders of magnitude to 1.1 μM. A subsequent structure-activity relationship with this metallo-quinazoline lead structure by derivatizing the cyclopentadienyl moiety afforded complex 3, displaying a further decreased IC₅₀ value of 38 nM (see Supporting Information for additional derivatives). Finally, the introduction of a methyl group at position 2 of the quinazoline moiety provided a single digit nanomolar inhibitor for MTH1 (IC₅₀ = 6 nM for complex 4). Compared with the initial probe 1, the organometallic inhibitor 4 displays an improved affinity for MTH1 by a factor of around 26,000.

Figure 3.

Inhibition curves of complexes 1-4 against MTH1: 1 = empty squares, 2 = empty circles, 3 = empty triangles, 4 = filled circles. Assay conditions: Organometallics 1-4 were preincubated with purified MTH1 (10 nM) for 30 min. After addition of 8-oxo-dGTP (25 μM) and incubation for 10 min at 37 °C under shaking, the reaction was quenched and analyzed via HPLC for quantification of the starting material 8‐oxo‐dGTP and the product 8‐oxo‐dGMP.

To understand the binding of the organometallic 4 to MTH1, we co-crystallized racemic 4 with MTH1 and solved the structure to a resolution of 2.70 Å (Table 1).^[13] Figure 4 shows the overall structure. As a member of the Nudix hydrolases superfamily,^[14] MTH1 exhibits the typical Nudix fold with an arrangement of a mixed β-sheet surrounded by two α-helixes, which together with several loops creates the active site. As expected, the organometallic 4 occupies the 8-oxo-dGTP binding site and the electron density distribution, especially of the phenolic ester moiety, allows a clear assignment of the S-configuration at the stereogenic metal center.^[15] As visualized by the semi-transparent surface view in Figure 4, complex 4 is tightly surrounded by residues from the α-helices, β-sheets, and loops to provide an interaction surface of approximately 501 Å^2.16 The quinazoline moiety forms several hydrogen bonds, one between the exocyclic NH₂ group and Asp119, one between the endocyclic N3 and Asp120, in addition to the endocyclic N1 being in hydrogen bonding distance to the backbone NH of Gly34 and the amide side chain of Asn33 (Figure 5a). Further polar contacts at a larger distance can be identified between the ester carbonyl group of 4 and Lys23, and between a methoxy group of 4 and the backbone NH group of Lys38. The inhibitor is in van-der-Waals contact with multiple side chains such as highlighted in Figure 5b. Especially intriguing are the number of aromatic side chains stacking face-to-face (Trp117 and Phe27) or edge-to-face (Trp123, Phe72, Tyr7) to the inhibitor. In order to compare the binding of (S)-4 with 8‐oxo‐dGMP (PDB ID 3ZR0^[10]), both co-crystal structures were superimposed. As displayed in Figure 5c, the pyridylquinazoline moiety of (S)-4 co-locates roughly with the guanine nucleobase of 8‐oxo‐dGMP, the ruthenium center aligns with the ribose unit, whereas the 2,6-dimethoxyphenylester fills the space occupied by the phosphate of 8-oxo-dGMP. Although the position of the heteroaromatic quinazoline moiety is somewhat shifted and tilted compared to the purine nucleobase of 8‐oxo‐dGMP, the same key residues are involved in hydrogen bonding to both ligands, which is possible since the guanine base of 8-oxo-dGMP apparently adopts the enol tautomeric form.^[10]

Table 1.

Crystallographic data and refinement statistics for 4/MTH1 (PDB entry 3WHW).

Data collection and Processing
No. of crystals used	1
Wavelength [Å]	0.91841
Space group	P 21 21 21
Unit cell parameters
a, b, c [Å]	59.76, 67.65, 79.54
α,β,γ [°]	90, 90, 90
Matthews coefficient [Å3/Da]	2.21
Solvent content [%]	44.2
Diffraction data
Resolution range [Å]	25.0 – 2.70 (2.90-2.70)
Unique reflections	9264 (1737)
R(I)sym [%]	16.0 (63.3)
Completeness [%]	99.6 (99.0)
Redundancy	5.8 (5.7)
I/σ(I)	10.62 (3.0)
Refinement
Resolution range [Å]	25.0 – 2.70
Reflections used in refinement (work/free)	8800 / 464
Final R values for all reflections (work/free) [%]	21.5 / 26.6
Protein residues	308
Inhibitor atoms	78
Water molecules	34
Sulfate ions	7
RMSDs
Bonds [Å]	0.0085
Angles [°]	1.280
Ramachandran plot
Residues in most favoured regions [%]	89.6
Residues in additional allowed regions [%]	10.4
Residues in generously allowed regions [%]	-
Mean B factor [Å2]
Protein	51.9
Inhibitor	58.9
Water molecules	33.4
Sulfate ions	99.8

Figure 4.

Co-crystal structure of MTH1 (chain A, semintransparent surface view) with organoruthenium compound (S)-4 bound to the nucleotide binding site. The absolute configuration of the inhibitor was elucidated from its electron-density (shown as blue mesh). The SIGMAA-weighted 2F_obs – F_calc difference electron density map of the inhibitor was contoured at 1 σ.

Figure 5.

Interactions of (S)-4 with the active site residues of MTH1 and superimposed co-crystal structures of (S)-4 (green) and 8‐oxo‐dGMP (grey) with MTH1 (pdb: 3ZR0). Alignment was accomplished with PyMOL Molecular Graphics System, Version 0.99rc6, Schrödinger, LLC.

Overall, the high affinity of (S)-4 to MTH1 can be rationalized by a combination of several hydrogen bonding interactions, a large number of close van-der-Waals contacts, and a shape match between the nucleotide binding site and the organometallic inhibitor. The limited conformational flexibility of 4 does not only provide an entropic advantage, which is reflected in the high binding affinity to MTH1, but also contributes to an impressive specificity. In fact, organometallic 4 does not show any notable protein kinase binding at a concentration of 1 μM as verified in an active-site-directed competition binding assay against a panel of 457 kinases (see Supporting Information).^[17,18] Furthermore, profiling of 4 at 10 μM against 191 protein ATP binding sites confirmed the high selectivity of 4 with only MTH1 and deoxycytidine kinase competing by more than 50% with the reactive acyl phosphate ATP probe (see Supporting Information).

In summary, we here presented the discovery of the first small molecule inhibitor of the repair enzyme 8-oxo-dGTPase (MTH1), which is not a canonical small organic molecule but rather an unconventional cyclometalated ruthenium half sandwich complex. The low nanomolar organometallic inhibitor displays an astonishing specificity as verified in an extended panel of protein kinases and other ATP binding proteins. These results reinforce conclusions drawn from our previous work on metal-based inhibitors of protein kinases, namely that inert metal complexes constitute excellent scaffolds for obtaining highly selective molecular probes, which we attribute to the limited flexibility of the globular metal complexes.^[19] Furthermore, this work provides a blueprint for the discovery and development of organometallic inhibitors of other purine nucleotide binding proteins by using ruthenium coordinated adenine and quinazoline derivatives with different coordination spheres. Finally, the novel organometallic MTH1 inhibitor 4 will be employed as a tool for a closer investigation of the biological function of MTH1 and its merit to serve as a target for anticancer therapy.