Synthetic Routes to N-9 Alkylated 8-Oxoguanines; Weak Inhibitors of the Human DNA Glycosylase OGG1

The human 8-oxoguanine DNA glycosylase OGG1 is involved in base excision repair (BER), one of several DNA repair mechanisms that may counteract the effects of chemo- and radiation therapy for the treatment of cancer. We envisage that potent inhibitors of OGG1 may be found among the 9-alkyl-8-oxoguanines. Thus we explored synthetic routes to 8-oxoguanines and examined these as OGG1 inhibitors. The best reaction sequence started from 6-chloroguanine and involved N-9 alkylation, C-8 bromination, and finally simultaneous hydrolysis of both halides. Bromination before N-alkylation should only be considered when the N-substituent is not compatible with bromination conditions. The 8-oxoguanines were found to be weak inhibitors of OGG1. 6-Chloro-8-oxopurines, byproducts in the hydrolysis of 2,6-halopurines, turned out to be slightly better inhibitors than the corresponding 8-oxoguanines.


Introduction
Chemo-and radiotherapy are, in addition to surgery for removal of solid tumors, the two main treatment protocols currently available to improve the outcome of cancer patients in general, but treatment-related toxicity, the risk of secondary cancers, and the emergence of resistance limit their effectiveness [1]. Some chemotherapeutic drugs and radiotherapy work partly by imposing high concentrations of DNA damage on the genome of cancer cells, beyond the repair capacity of those cells. The drug-exposed cancer cells are heavily dependent on efficient DNA repair to survive. Consequently, inhibitors that reduce DNA repair activities should sensitize cancer cells to chemo-and/or radiotherapy [2][3][4][5].
Several DNA repair mechanisms counteract exogenous and endogenous processes that destabilize or directly damage genomes. The processes include, among others, base excision repair (BER), a mechanism that depends on enzymes that recognize small modifications in the native bases in DNA, resulting from alkylation, oxidation, deamination, or hydrolysis of the DNA bases. The pathway is initiated by a damage-specific DNA glycosylase that removes the altered base [6]. Some of these enzymes mainly remove oxidized bases, such as the human 8-oxoguanine DNA glycosylase (OGG1) that removes guanines that have been oxidized at the C8-position. The 8-oxoguanine base in the DNA is flipped into a lesion recognition pocket on the enzyme surface, exposing the Watson-Crick signature of guanine and the oxidized C8 position (Figure 1).  [7]). The protein backbone is shown as a blue ribbon/helix. Selected amino acid side chains and the 8oxoG base are shown as ball-and-stick. Hydrogen bonds between the protein and 8oxoG are shown as dashed lines. Asp268 is the catalytic residue in OGG1. Symbols 5′ and 3′ indicate the position of the 5′ and 3′ phosphodiester links in the DNA.

Chemistry
We found it most convenient to start the synthesis of 9-alkyl-8-oxopurines from commercially available purines, and in our opinion the best way to introduce the 8-oxo group would be by hydrolysis of an 8-halopurine. However, there still was the question of whether the halogen or the N-9 substituent should be introduced first and which protection/activation groups should be employed in the synthesis. Ideally, such groups should also be removed in the final hydrolysis step. Regioselectivity in N-alkylation of guanine derivatives was also an issue [20][21][22][23][24][25][26][27]. We chose to start from two guanine precursors, commercially available 2-amino-6-chloropurine (1a) and the O-carbamoylguanine 1b, easily available from guanine [28,29]. The synthetic routes explored are all summarized in Scheme 1. Table 1; (b) See Table 2; (c) 1. Ac2O, NaOAc, AcOH, 2. NaOH(aq), ∆; (d) 1. LDA, 2. (CCl2Br)2, THF, −78 °C; (e) Br2, CHCl3; (f) See [30]; (g) HCl(aq), EtOH. Scheme 1. Synthetic routes to 8-oxoguanines 5.  First we chose to N-alkylate the substrates 1 before C-8 halogenation and hydrolysis. Alkylations were conducted by various methodologies in order to find the conditions that gave the desired N-9 alkylated isomer 2 with high selectivity and in a good isolated yield (Scheme 1, Table 1). Relatively simple alkylating agents were chosen for the model reactions and we focused on alkylation with alkyl halides in the presence of base, Mitsunobu reactions, and Pd-catalyzed allylic alkylation.

Reagents and conditions: (a) See
The cyclohexylmethyl substituent could be introduced at N-9 either by reaction with alkyl bromide in the presence of a base [31,32] (Table 1, Entries 1 and 3) or with cyclohexylmethanol under Mitsunobu conditions (Table 1, Entries 2 and 4). The latter is often claimed to be more N-9 selective compared to classical alkylations of purines [33][34][35]. In all cases a mixture of the N-9 and N-7 alkylated isomers (2 and 3) was formed with good selectivity for the desired isomer 2. The isomers were identified from HMQC and HMBC-NMR, as described before [31].
The guanine precursor 1b, carrying a bulky substituent at C-6 that may sterically block N-7, is reported to react with high N-9 selectivity in other N-functionalization reactions [28,29,[36][37][38][39][40][41]. Nevertheless, we found the regioselectivity in N-alkylation of purine 1b equal or slightly poorer compared to 6-chloroguanine 1a in all reactions performed in this study. In the alkylation of compound 1b, minor amounts of other relatively polar products were formed under both reaction conditions. These often made purification of the N-7 alkylated isomer 3 difficult. The identity of the byproducts could not be determined, but they may be formed as a result of cleavage of the O 6 -protecting group. Alkylation of N 2 , as observed by others [41], was not seen.
Introduction of the cyclohexyl group at N-9 turned out to be quite difficult (Table 1, Entries 5-12). Both starting materials (1a and 1b) did not react with cyclohexyl bromide (data not shown) and reacted slowly with cyclohexyl iodide or the corresponding tosylate, but compounds 2b and 2f could be isolated in modest yields (Table 1, Entries 5, 6, 10 and 11). It is, however, well known that cyclohexyl halides or pseudo halides may react sluggishly in substitution reactions [42]. The results were not significantly improved when the Mitsunobu reaction was employed ( The cyclopentyl group could easily be installed at N-9 on both starting materials 1a and 1b by reaction with cyclopentyl bromide and base (  Entries 14 and 16). The selectivity for N-9 was higher in the Mitsunobu reactions, but the isolated yields were comparable due to more tedious purification when Mitsunobu conditions, also producing phosphine oxides and reduced azodicarboxylates, were employed.
Finally we introduced the cyclopent-2-enyl group at N-9 (Table 1, . These reactions were only conducted at the guanine precursor 1a, since we so far had not observed any significant improvement in regioselectivity when compound 1b was employed and we had observed problems with compounds derived from purine 1b later in the planned synthetic sequence. In addition to alkylation with the halide and Mitsunobu reaction with the alcohol, we also attempted palladium catalyzed alkylation with the allylic acetate [43]. 3-Bromocyclopentene could only be generated as a 15% solution in CCl4 and the reagent had a limited stability, probably partly due to traces of the radical initiator used in the synthesis left in the solution [44], which may explain the low yield of product 2d (Table 1, Entry 17). The Mitsunobu reaction between purine 1a and cyclopenten-2-ol was surprisingly slow, and full conversion was not achieved even after several days. Furthermore, the N-9/N-7 selectivity was only ca. 4:1 ( Table 1, Entry 18). Pd-catalyzed allylic alkylation of purine 1a went to completion and gave the isomers 2d and 3d in a 4:1 ratio (Table 1, Entry 19).
Attempts to brominate the O-carbamoylguanine 2e failed (Scheme 1). Treatment with bromine or lithiation followed by trapping with CCl2BrCCl2Br only resulted in cleavage of the carbamoyl protecting group to give the guanine derivative 7. When compound 2e was treated with NBS, no reaction took place at all. Thus, no attempts were made to brominate the carbamoyl protected guanines 2f and 2g.
Since bromination of the cyclopentenylpurine 2d turned out to be a challenge, we also examined the possibility for introducing the 8-halo substituent before the N-9 alkyl group (Scheme 1). We chose to brominate the THP protected compound 8 [30] and removed the protection group under mild acidic condition, but direct bromination of purine 1a in a moderate yield has also been reported [48].
Alkylation of 8-bromo-6-chloropurin-2-amine (10) by bromomethylcyclohexane in the presence of K2CO3/DMF (Table 2, Entry 2) occurred slowly compared to alkylation of 2-amino-6-chloropurine (1a) under the same set of reaction conditions (for alkylation of compound 1a see Table 1). NMR analysis showed that approximately 50% of the starting material was intact even after 96 h reaction time and the desired product was isolated in a low yield. Also, ca. 4% of N-7 alkylated isomer was formed, as judged by NMR. When compound 10 was reacted under Mitsunobu (Table 2, Entry 3) conditions, high conversion (ca. 95%) and almost full selectivity towards the desired N-9 alkylated isomer 4a was achieved, as judged by 1 H-NMR. However, the product 4a was isolated only in 56% due to tedious separation from reduced DIAD. Since compound 10 reacted slower (conventional alkylation) or comparably (Mitsunobu alkylation) to compound 1a, it was concluded that there were no benefits associated with introducing the bromide before the N-alkyl group for the synthesis of 8-bromopurines 4a-c.
Also, synthesis of the 9-cyclopentenylpurine 4d by N-alkylation of compound 10 was examined (Table 2, Entries 7 and 8) since bromination of 2-amino-6-chloro-9-cyclopentenylpurine 2d turned out to give only a low yield of the desired product. Again, isolation of the desired product from alkylation under Mitsunobu conditions turned out to be troublesome. We tried this Mitsunobu alkylation using the water-soluble azodicarboxylate DMEAD (di-2-methoxyethyl azodicarboxylate) as well as DIAD [49]. Purification of the product was less complicated, but the conversion was low and ca. 40% of starting material 10 was recovered. Also, Pd-catalyzed allylation turned out to be a very slow reaction and even after six days only 29% of the desired compound 4d could be isolated, together with 32% unconverted starting material 10.

Biology
As previously mentioned, our hypothesis was that N-alkyl-8-oxoguanines may inhibit the human 8-oxoguanine DNA glycosylase (OGG1). Other substituents in the purine 8-position are probably not tolerated, for instance 8-bromo-and 8-aminoguanines are reported to be enhancers for OGG1 activity [50]. Thus, the 8-oxoguanines 5 as well as the partly hydrolyzed 6-chloro-8-oxopurines 6 were tested against human DNA glycosylases OGG1 and NTH1. A general structure of the tested compounds is shown in Figure 2 and the results are presented in Tables 3 and 4 Table 3. Table 3. % Activity of OGG1 in the presence of compounds 5 or 6 at 0.2 mM concentration.
Cl c-pent-2-enyl 84 ± 3 a The predominant 6-oxo tautomer of compounds 5 is shown in Scheme 1. Table 4. % Activity of NTH1 in the presence of compounds 5 or 6 at 0.5 mM concentration.
Compounds 6b and 6c inhibit the OGG1 enzyme by ca. 30%, followed by compounds 5a, 5b, and 6d at ca. 10%-15%, all at 0.2 mM ligand concentration. Interestingly, the halogenated compounds seem in general to be better inhibitors than their 6-oxo derivatives. To check enzyme specificity, we tested the same seven compounds at the higher concentration of 0.5 mM against NTH1, a structural but not functional homolog of OGG1. Both enzymes have a deep pocket for binding of oxidized bases; in general, OGG1 repairs oxidized purines while NTH1 is involved in repair of oxidized pyrimidines. Compound 6b reduced the NTH1 activity by around 25% at 0.5 mM ligand concentration. An effect of varying the N-9 substituent is not so evident from the few compounds examined.

General Information
1 H-NMR spectra were recorded at 300 MHz with a Bruker DPX 300, at 400 MHz with a Bruker DPX 400 or at 600 with a Bruker AVI 600 instrument (Bruker BioSpin AG, Fällanden, Switzerland). The 13 C-NMR spectra were recorded at 75, 100, or 150 MHz with the Bruker instruments listed above. Assignments of 1 H and 13 C resonances are inferred from 1D 1 H-NMR, 1D 13 C-NMR, DEPT, or APT, and 2D NMR (HMQC, HMBC) spectroscopical data. 1 H-and 13 C-NMR spectra of all novel compounds can be found in the Supplementary Material (Figures S1-S18). HRMS (EI) was performed with a double-focusing magnetic sector VG Prospec Q instrument (Waters, Manchester, UK) and HRMS (ESI) with a TOF quadrupole Micromass QTOF 2 W instrument (Waters). Melting points were determined with a Büchi Melting point B-545 apparatus (Büchi Labortechnik AG, Flawil, Switzerland) and are uncorrected. Dry DMF and THF were obtained from a solvent purification system, MB SPS-800 (MBraun, Garching, Germany). Acetic anhydride and diisopropylamine were distilled over CaH2. DMSO was dried over activated 3 Å molecular sieves for four days. Potassium carbonate was oven dried at 150 °C under high vacuum for 12 h. A saturated aqueous solution of Br2 was prepared by stirring water (20 mL) with Br2 (0.200 mL) in a closed container for 15 min at ambient temperature. Sodium hydride (ca. 60% in mineral oil) was washed with dry pentane under inert atmosphere prior to use. All other reagents were commercially available and used as received. The following compounds were available by literature methods: Cyclohexyl tosylate [51], cyclopentenyl bromide [44], cyclopent-2-enol [52], cyclopentenyl acetate [53], 1b [29], 8 [30].
Method C: A solution of compound 1a (100 mg, 0.590 mmol) and NaH (18 mg, 0.77 mmol) in dry DMSO (5 mL) was stirred at room temperature for 20 min under Ar atmosphere. The mixture was added to a solution of cyclopent-2-en-1-yl acetate (0.070 mL, 0.77 mmol) and Pd(PPh3)4 (103 mg, 0.0890 mmol) in dry DMSO (5 mL) and the resulting mixture was stirred at 50 °C for 48 h under Ar and evaporated in vacuo. The product was purified by flash chromatography as described in Method B to yield 2d (73 mg, 53%) and 3d (25 mg, 18%).

DNA Glycosylase Activity Assay
The enzyme OGG1 (residues12-327) was diluted to the desired concentration (60 pM) using a protein dilution buffer (15% glycerol, 1 mM EDTA, 25 mM HEPES pH 7.9, 1 mM DTT, 0.1 μg/μL BSA). Enzyme, compound 5 or 6 (0.2 mM), and 5′-32 P end-labeled duplex DNA containing an 8-oxo-G/C base pair were mixed in a 10 μL reaction volume of 50 mM MOPS pH 7.5, 1 mM EDTA, 5% glycerol, and 1 mM DTT. The sequence of the damaged strand in the DNA substrate used is 5′-GCATGCCTGCA CGG-8oxoG-CATGGCCAGATCCCCGGGTACCGAG-3′, which was annealed with a complementary strand containing a C opposite 8oxoG. The reactions were incubated for 10 min at 37 °C, followed by addition of 2.5 μL 0.5 M NaOH and incubation for 20 min at 70 °C, in order to stop the reaction and ensure complete strand cleavage. Then 0.5 M HCl/0.25 M MOPS pH 7.5 (2.5 μL) was added to each sample to neutralize the pH. Formamide DNA loading buffer (15 μL) was added to the reaction mixtures and the samples were incubated at 95 °C for 5 min to denature the DNA. The reaction products were analyzed on 20% denaturing urea gels. The gels were transferred to 3M paper and dried at 80 °C for 45 min. The dry gels were placed in a storage phosphor screen overnight, and subsequently scanned on a Typhoon 9410 Variable Mode Image. ImageQuant TL Version 2003.02 (Amersham Biosciences, Piscataway, NJ, USA) was used to analyze the results. For human NTH1, the same procedure was followed, except that the DNA substrate contained a 5-hydroxyuracil/G base pair instead of the 8oxoG/C pair in the OGG1 substrate. The concentration of NTH1 was 3 nM to make sure the activity in the assay was within the linear range. Compounds were screened at 0.5 mM concentration.

Conclusions
Synthetic routes to 8-oxoguanines have been examined. The best reaction sequence from chloroguanine 1a to the target compounds was found to be N-9 alkylation, C-8 bromination, and finally simultaneous hydrolysis of both halides. Bromination before N-alkylation should only be considered in cases where the N-substituent is not compatible with bromination conditions, since a bromide in the purine 8-position lowers the reactivity in N-alkylations. In most cases, alkylation with an alkyl halide in the presence of a base compared favorably to reactions under Mitsunobu conditions. 2-Amino-6-chloropurine (1a) turned out to be a superior guanine precursor compared to the O-carbamoylguanine 1b. The latter did not result in improved N-9 selectivity in the alkylation and was not compatible with standard reaction conditions for C-8 bromination.
Enzymatic assays show that partly hydrolyzed 6-chloro-8-oxopurines 6 are somewhat better OGG1 inhibitors than the 8-oxoguanines 5. However, an inhibitory effect was only observed when using at least 0.2 mM concentration of the compounds, suggesting that the R-group should be extended even further to make more interactions with the enzyme's substrate recognition pocket. Further, testing of the same compounds at a 2.5-fold higher concentration against human NTH1, which is a structural homolog of OGG1, showed that the synthesized compounds do not inhibit NTH1 at 0.5 mM, except possibly for a weak effect for compound 6b. To develop these compounds into more potent inhibitors of OGG1, one possibility is to try compounds with more ribose-like R-groups. In the present study, the R-group contains a cyclic hydrocarbon only, and it would also be interesting to replace this with carbocyclic 2′-deoxyribose derivatives, as in antiviral drugs like abacavir and entecavir. In these nucleoside analog drugs, the R-group is not particularly larger than the R-group in our study, but it contains 5′ and/or 3′ hydroxyl groups. Since the structure of the OGG1 enzyme is known [7], molecular modeling will be included in the search for more potent OGG1 inhibitors in the future.