An aza-nucleoside, fragment-like inhibitor of the DNA repair enzyme alkyladenine glycosylase (AAG).

The DNA repair enzyme AAG has been shown in mice to promote tissue necrosis in response to ischaemic reperfusion or treatment with alkylating agents. A chemical probe inhibitor is required for investigations of the biological mechanism causing this phenomenon and as a lead for drugs that are potentially protective against tissue damage from organ failure and transplantation, and alkylative chemotherapy. Herein, we describe the rationale behind the choice of arylmethylpyrrolidines as appropriate aza-nucleoside mimics for an inhibitor followed by their synthesis and the first use of a microplate-based assay for quantification of their inhibition of AAG. We finally report the discovery of an imidazol-4-ylmethylpyrrolidine as a fragment-sized, weak inhibitor of AAG.


Introduction
Alkyladenine glycosylase (AAG; also known as methylpurine DNA glycosylase, MPG, and alkyl-N-purine glycosylase, ANPG, EC: 3.2.2.21) is one of several DNA glycosylases that can initiate the base excision repair (BER) pathway by hydrolysis of a range of alkylated, oxidised or deaminated purine bases from the DNA backbone. 1 AAG's action generates apurinic (AP) sites, which are further processed by downstream BER enzymes. Despite its beneficial role, AAG overactivity on DNA lesions can lead to cellular necrosis and tissue damage. Specifically, AAG has been shown to drive alkylation-induced cytotoxicity in several mouse tissues, namely the cerebellum, spleen, thymus, bone marrow, pancreas and retina, with tissue damage reduced or absent in Aagknockout mice and dramatically increased in Aag-overexpressing transgenic mice. 2,3 Furthermore, in ischemia reperfusion mouse models, Aag-knockout mice display reduced tissue necrosis in liver, kidney and brain when compared to wild type. 4 Ischemia reperfusion models mimic ischaemic stroke, liver and kidney failure, and organ transplantation. In these events the tissues are temporarily starved of their blood supply before sudden reperfusion results in a burst of reactive oxygen and nitrogen species (RONS). RONS can directly oxidise DNA bases but also lead to lipid peroxidation products, which alkylate DNA. 5 Further to this, sterile inflammation is induced and the recruitment of neutrophils and macrophages leads to further production of RONS. 6 Both alkylation-and ischaemia reperfusion-induced AAGdependent tissue necrosis is hypothesised to result from AAG activity on substantial numbers of alkylated and oxidised DNA bases leading to the accumulation of toxic repair intermediates in the DNA, which overwhelm the repair capacity of downstream BER enzymes. 3 Stable, membrane-permeable inhibitors of AAG are required as chemical probes to further investigate these mechanisms and to form leads to potential chemoprotectives for patients on alkylative chemotherapy or to generate rapid treatments for minimising tissue damage from stroke, and organ failure and transplantation.
The only reported specific small molecule inhibitor of AAG is the natural polyphenolic flavonol morin, which was shown to act directly on AAG, and not through binding to DNA, with an IC 50 of 2.6 μM (measured using a gel-based biochemical assay). 7 Although shown through surface plasmon resonance and control experiments to be specific for AAG and effective in cells, a myriad of other bioactivities have been reported for morin resulting from both its antioxidant activity and binding to biomacromolecules. 8,9 These other bioactivities may confound results when used as a probe in biological studies. Therefore, we sought an alternative scaffold for chemical probe inhibitors through ligand-based design. Our specific requirements of a final probe molecule match those of the Structural Genomics https://doi.org/10.1016/j.bmc.2020.115507 Received 10 February 2020; Received in revised form 6 April 2020; Accepted 9 April 2020 Consortium for their Target 2035 program to make chemical or antibody probes available for the entire human proteome. 10,11 Specifically, these are a biochemical assay K d or IC 50 < 100 nM, a cellular assay IC 50 or EC 50 < 1 μM and > 30-fold selectivity against other mammalian BER enzymes and enzymes known to be inhibited by structurally related molecules, such as purine nucleoside phosphorylase (PNP).

AAG small molecule inhibitor design
A duplex DNA 13-nucleotide oligomer containing an ethenodeoxycytidine (εC) nucleotide is reported to inhibit AAG with an IC 50 of 39 nM, and has a K d of 21 nM. 12 This affinity is 2-fold greater than that of a 13-mer substrate ethenodeoxyadenosine (εA)-containing oligomer (K d = 46 nM). Based on this, we first sought to test the lone εC-nucleoside 1 and nucleotide 4 (Scheme 1) for any inhibitory activity with a view to subsequently optimising them into potent, membrane permeable inhibitors if active. However, neither of these compounds, nor the precursor di-or mono-benzylphosphodiesters 2 or 3 showed any inhibition at up to 1 mM concentration.
Another inhibitory duplex DNA oligomer, containing an abasic pyrrolidine (pyrr) nucleotide mimic (pyrr-oligomer, Fig. 1), was designed by the group of Verdine originally as an inhibitor of Escherichia coli 3-methyladenine DNA glycosylase II (AlkA). 13,14 It is proposed that the protonated pyrrolidine mimics the oxocarbenium ion character of the transition state (TS) (or formal oxocarbenium intermediate in an S N 1 mechanism 15 ), which involves cleavage of the N-glycosidic bond between the base and C-1′ of the deoxyribose and attack by the nucleophilic active site water molecule from the opposite side.
In the published co-crystal structure (green, Fig. 1), 13 the protonated pyrrolidine undergoes hydrogen bonding (2.8 Å) with the active site nucleophilic water molecule, which is itself held in place and activated by three hydrogen bonds with Glu125, Arg182 and the backbone carbonyl of Val-262. Hydrogen bonding to this tightly bound water molecule is considered integral to the strong affinity (K d = 23 pM) of this oligomer because the analogous uncharged tetrahydrofuran-containing oligomer showed 7 × weaker binding (K d = 160 pM). 16 It is worthy of note that the pyrrolidine-containing oligomer is an effective inhibitor despite lacking a damaged DNA base mimic to occupy AAG's base-binding pocket. 16,17 Nevertheless, a base mimic (adenine, joined to pyrrolidine via a CH 2 linker for stability) has been incorporated and a K d < 1 pM against AAG reported. 17 To the best of our knowledge, no X-ray crystal structure for this complex has been obtained and the lone nucleotide (or nucleoside) from this oligomer has not been tested against AAG.
To inform the design of small molecule inhibitors, we overlaid the published X-ray co-crystal structures of AAG in complex with the substrate ethenoadenine (εA)-, inhibitory εC-and pyrr-containing duplex oligomers (Fig. 1). 18 [Note that the authors report that the εA-oligomer remained un-hydrolysed in AAG's active site through use of inhibitory MgCl 2 , Mg 2+ ions from which were not seen in the crystal structure.] In the crystal structures, all of the DNA oligomers make several electrostatic/hydrogen bonding interactions between their phosphodiester groups and basic residues on the surface of AAG. Any potent small molecule inhibitor based on a single nucleotide would therefore have to find further binding interactions within the active site to compensate for the loss of the DNA chain (as exemplified by the lack of activity of our εC lone nucleotides). Thus it would be necessary to include a base mimic to find interactions in AAG's base-binding pocket. To this end, we proposed to synthesise hybrid inhibitors based on the immucillin (discussed below) scaffolds containing the pyrrolidine moiety joined to an alkylated DNA-base analogue. This could be achieved in two ways: 1. in the form of inhibitor scaffold 5 where aromatic heterocycles are attached via a C-atom (to deprive them of leaving group ability) to C-1′ position of the pyrrolidine, a design which takes account of the fact that in the overlay shown in Fig. 1, the εC-deoxyribose (pink) overlaps well with the pyrrolidine group (green). Such molecules 5 are the subject of ongoing synthetic efforts in our group. 2. in the form of inhibitor scaffold 6 rationalised as follows: the overlay in Fig. 1 shows that, compared to the εA-oligomer's deoxyribose group, the pyrrolidine is displaced towards the base-binding pocket. This could be to maximize hydrogen bonding with the nucleophilic water molecule or it could be due to the terminus of the εA base pushing back against the end of its binding pocket. The latter is evidenced by the εA-deoxyribose appearing displaced compared to the εC-deoxyribose as well as the pyrrolidine. Either way, this flexibility in the binding mode may permit the N-atom to be moved one position around the ring and itself to bear an alkylated base mimic, which must be linked via a CH 2 group to maintain the basicity (and therefore protonated form) of the pyrrolidine N-atom. Importantly, this positioning of the N-atom has been shown, in Escherichia coli 5′-methylthioadenosine/S-adenosylhomocysteine (MTAN), to better mimic a substrate transition state involving early, S N 1-like departure of the purine where most positive charge accumulates at the anomeric carbon. 19,20 To the best of our knowledge, the "early-" or "late-ness" of the TS of AAG has not been determined and may be substrate specific for this multi-substrate enzyme. However, the comparison of both types of inhibitor in the future may inform on the nature of the TS to some extent.
Both aza-nucleoside inhibitor scaffolds have been reported previously to mimic the transition states in enzymes which attack riboseand deoxyribose-N-glycosidic bonds. 21 For example, they form the first (5) and second (6) generation immucillins, which inhibit PNP by mimicking the TS of phosphorolysis of guanosine and inosine. [22][23][24] Also, compounds based on 6 inhibit the ADP-ribosylating action of the cholera, pertussis and diphtheria toxins, 25 and, as part of RNA-oligomers, the ricin A-chain. 26 Of most relevance to our work, the scaffolds have been previously applied as inhibitors of other DNA glycosylases, either as part of large DNA oligomers or as lone nucleosides. For example, a DNA 11-mer containing the abasic form (lacking any arylmethylene group) of nucleotide 6 showed a K d of 0.11 nM against uracil DNA glycosylase (UDG), 27 and duplex DNA 30-mers containing nucleotides based on 5 and 6 are nano-to picomolar inhibitors of Escherichia coli MutY, bacterial Fpg, human 8-oxoguanine DNA glycosylase 1 (hOGG1) and human Nei-like DNA glycosylase (hNEIL1). 28,29 Herein, we report an investigation of the synthesis of five small molecule, free aza-nucleoside analogues based on scaffold 6 and the finding that one of them is a weak, but ligand efficient, inhibitor of AAG worthy of further optimisation into a chemical probe.

Chemistry
The free nucleoside and nucleotides of εC were synthesised from deoxycytidine as shown in Scheme 1. In our hands, ethenylation of 1 at RT required 7 days at pH 3.5 and complete removal of methanol and water from the purified product required extensive drying (100°C under vacuum for 5 h) which may explain the higher m.p. obtained by us compared to others. 30 Conversion of nucleoside 1 into the phosphate was unsuccessful using POCl 3 with no product being isolable by reverse phase HPLC. 31 To simplify purification, we opted for a dibenzylphosphorylation, which would permit normal phase chromatography of the protected phosphate prior to hydrogenolysis of the benzyl protecting groups. Thus, the tribenzylphosphite-iodine coupling 32,33 was applied to afford a mixture of di-and mono-benzylphosphates 2 and 3 in low yield, along with recovery of starting material. To the best of our knowledge, this is the first application of that coupling to a nucleoside, albeit one bearing a non-nucleophilic base, and in a better-optimised form it may be a useful method for the selective phosphorylation of the primary alcohol of nucleosides without the need for secondary alcohol protection. 34 Subsequently, hydrogenation of dibenzylphosphate 2 afforded phosphate 4 in high yield, which was purified by reverse-phase HPLC prior to biochemical assay.
For the synthesis of azanucleosides 6, several routes to the key dihydroxypyrrolidine 18 (and its enantiomer) have been published, including those starting from sugars, 35,36 those beginning with a 1,3-dipolar cycloaddition and later employing enzymatic chiral resolution, [37][38][39] and those using an asymmetric 1,3-dipolar cycloaddition by means of chiral auxiliaries. 40,41 Despite the latter two being 1-2 steps shorter, we chose to investigate the route based around the enzymatic chiral resolution of hydroxypyrrolidine 13 (Scheme 2), described by Clinch, and we report here some nuances of each reaction which may be of interest to the synthetic chemistry community. 42 The route began with aza-Michael addition of glycine ethyl ester 7 to ethyl acrylate (1 equiv.) to give the corresponding secondary amine 8 in up to 54% yield. 43,44 In several instances the tertiary amine resulting from a second aza-Michael addition of ethyl acrylate, was detected and removed by column chromatography. This has not been reported in the literature and it was not possible to prevent its formation by varying the rate of addition of ethyl acrylate or the reaction time. Subsequent Nbenzylation of purified amine 8 afforded tertiary amine 9 in high yield.
The next step required a Dieckmann condensation to give pyrrolidinone 10. Pinto et al. showed that reaction of amine 9 with KOtBu in THF at −78°C chemoselectively produced pyrrolidinone isomer 10 (74% yield) over the alternative isomer resulting from formation and attack by the enolate of the α-aminoester (13% yield). 45 In our case, formation of 10 was followed by GC-MS (m/z 247 [M + ]) but, during chromatographic purification of the reaction mixture the desired product was oxidised to pyrrole 11 (characterised by GC-MS (m/z 245 [M + ]) and NMR). The oxidation and aromatisation of oxo-and hydroxypyrrolidines to pyrroles in the presence of SiO 2 and air has been  reported by Davis et al. 46 A different purification procedure based on crystallisation of the product at −4°C for 18 h was attempted, giving the desired pyrrolidine 10 in up to 60% yield, albeit with minor impurities. Due to the difficulties in purification of pyrrolidinone 10, alternative, quantitative cyclisation conditions were sought to obviate the need for chromatography. The use of LDA gave a mixture of products and, again, only pyrrole 11 was isolated after chromatography. However, regioselective TiCl 4 -mediated Dieckmann condensation conditions, published by Deshmukh et al. and reported to give pyrrolidinone 10 and related compounds in 40-60% yield after chromatography, 47 proceeded in our hands to give 92% yield of 10 of sufficient purity to be used in the next step without purification, thus avoiding oxidation to the undesired pyrrole. The next step involved borohydride reduction of pyrrolidinone 10, using the conditions reported by Zhang et al., and gave a mixture of both β-hydroxyester diastereoisomers, in a~1:2 ratio in favour of the desired trans form 13, which was obtained in 37% yield after chromatography. 48 This is comparable to the literature yield on the identical substrate at large scale (42%) 49 but a lot lower than that reported for the reduction of the analogous N-Boc-protected (instead of Bn) pyrrolidinone (99%), 48 and could perhaps be improved in the future by use of the milder reducing agent NaBH 3 CN. 50,51 Since fair amounts of the undesired cis diastereoisomer 12 were accumulated during this work, its epimerisation into the desired trans pyrrolidine 13 was studied on small scale (Table 1). Galeazzi et al. achieved quantitative epimerisation of an analogous compound using DBU in toluene at 70°C for 12 h but this was ineffective on pyrrolidine 12 (entry 1) and gave some eliminated alkene 19 along with eliminated and oxidised product pyrrole 20. 53 The latter conversion 19 → 20 has been previously described in dioxane at 90°C. 52 The alternative, weaker base Et 3 N gave no reaction (entry 2) and use of t BuOK gave more of, or mostly, the unsaturated products in toluene or THF (entries 3-5). A balance was found using EtONa in protic solvent EtOH which gave the desired diastereoisomer 13 as the largest component of the mixture (entry 6). Unfortunately, a single attempt at scale-up of this reaction gave none of the desired product, only an intractable mixture of highly polar compounds presumed to include ester hydrolysis products.
Enzymatic resolution of trans-pyrrolidine 13 was achieved according to Clinch et al. by enantioselective acylation catalysed by lipase B of Candida antarctica in tert-butyl methyl ether at 40°C. 42  LiAlH 4 was used for the reduction of the ester moiety of (+)-13 to yield diol (+)-15 in 71% yield. This was followed by N-debenzylation according to Clinch et al., who achieved quantitative yield for hydrogenation over Pd/C in the presence of di-tert-butyl dicarbonate to give the Boc-protected pyrrolidine (+)-17 which is hydrolysed with HCl to give the pyrrolidine hydrochloride (+)-18·HCl. 42 In our hands, this procedure gave unsatisfactory yields of N-Boc pyrrolidine (+)-17 after purification, perhaps due to concurrent O-tert-butoxycarbonylation (O-Bocylation). In one instance, when more equivalents of Boc 2 O were employed N-Boc-Bis-O-Boc-protected pyrrolidine 16 was isolated in 44% yield. Its characteristic features after 1 H NMR analysis were singlets at 1.49 ppm (18H) and 1.45 ppm (9H) corresponding to the three tert-butyl groups. The presence of three Boc groups was consistent with its 13 C NMR spectrum which showed three carbonyl signals at 153.9, 153.0 and 152.4 ppm. O-Bocylation, despite being rare compared to the analogous reaction on amines, has been reported in the literature for protection of alcohols using catalysts such as DMAP, zinc acetate, and perovskites (NaLaTiO 4 ). [54][55][56] As both Boc-protected pyrrolidines could be successfully hydrolysed under the same conditions, a one-pot procedure was applied in later attempts to give (+)-18·HCl in a yield of up to 78%. The specific optical rotation measured for (+)-18·HCl ([ ] D 21 + 16.0) was in broad agreement with that found in the literature ([ ] D 23 + 19.0). 40 Finally, to access the proposed methylaryl inhibitors 6a-e, reductive amination was applied to (+)-18·HCl and five heterocyclic aryl carbaldehydes. Initially, NaBH(OAc) 3 was chosen as reducing agent due to its lower toxicity but this reagent requires the use of aprotic solvents (such as DCE and THF) in which the salt (+)-18·HCl was insoluble. 57 The alternative, but more toxic, reducing agent NaBH 3 CN can be used in MeOH which effectively dissolved (+)-18·HCl and allowed the reaction to proceed. 58 No optimisation was carried out, but each small scale reaction provided sufficient quantity of the proposed inhibitors 6a-e for biochemical testing against AAG.

Biochemical activity against AAG
The biochemical assay used to test inhibitor activity was based on the colorimetric microplate assay, previously reported by two of the authors, for determining base excision DNA repair enzyme activity. 59 Briefly, it involves surface-bound DNA containing a substrate residue for AAG (hypoxanthine) and terminal fluorescein. AAG's action leaves AP sites that render the DNA backbone prone to hydrolysis in the subsequent alkaline hydrolysis step, thus releasing the fluoresceinconjugated part of the DNA. The remaining 'plate-bound' fluorescein is detected using an anti-fluorescein antibody conjugated to horseradish peroxidase (HRP), incubation with a colour-changing substrate of HRP and colorimetric quantification. The assay was shown to be effective by testing published inhibitors εC-oligomer (lit. IC 50   and morin (lit. IC 50 2.6 μM) 7 which gave a comparable IC 50 value of 21 nM for εC-oligomer and a somewhat higher IC 50 of 89 μM for morin. The latter perhaps represents differences in the availability of morin under our conditions, particularly in terms of the buffer components used.
All five azanucleosides 6a-6e were tested in this assay, which revealed that only 6b exhibited detectable inhibitory activity (Fig. 2). Taking the mid-point between the top and bottom points of the curve, the IC 50 of 6b against AAG is 157 μM. While this is weak activity against AAG, 6b has a molecular weight of only 197 g mol −1 and is comprised of only 14 'heavy' (non-hydrogen) atoms, making it akin to a drug discovery fragment and, 60 leaving plenty of scope for increasing potency by the addition of further binding groups. In other words, 6b shows a respectable, approximate (substituting IC 50 for K d ) ligand efficiency of 0.37 kcal mol −1 per heavy atom. 61

Summary and conclusions
We have shown that the small molecule aza-nucleoside mimic 6b possesses weak inhibitory activity against the DNA base excision repair enzyme AAG. Owing to its low molecular weight, 6b is a ligand-efficient starting point for future optimisation into a more potent, selective and membrane-permeable AAG inhibitor. This is in contrast to nucleoside ethenodeoxycytidine 1 (and 5′-phosphates thereof), which we have shown to exhibit no inhibition outside of a DNA oligomer. IC 50 of these small molecules was determined using our microplate-based colorimetric assay as opposed to the more involved but established radiolabel, gel-based assay.
In order to synthesise 6b and its aryl analogues, we made a detailed study of one of the routes, largely reported by Clinch et al., and discovered several nuances of the chemistry including: 1. improved access to pyrrolidinone 10 using a TiCl 4 -mediated cyclisation which negated the need for column chromatography which otherwise led to oxidation to the pyrrole; 2. Isolation of the cis hydroxy-ester 12, as well as the desired trans 13, from reduction of the pyrrolidinone 10, and the possibility (on small scale) of its conversion to the trans using sodium ethoxide; 3. Concurrent O-bocylation (along with desired N-bocylation) of pyrrolidine-diol 15 during hydrogenative removal of the N-benzyl group in the presence of excess Boc 2 O; and, 4. The requirement to use NaBH 3 CN, rather than NaBH(OAc) 3 , for reductive amination of pyrrolidinium salt 18.HCl due to its compatibility with MeOH, the solvent required for dissolution of 18.HCl. In addition, during synthesis of ethenocytidine phosphate, it was found that a tribenzylphosphite-iodine coupling of ethenocytidine, previously never applied to nucleosides, followed by hydrogenation, gave selectively the desired 5′-phosphate.

Epimerisation study 12 → 13 (Table 1, entry 6)
A 0.1 M solution of EtONa was prepared by the addition of sodium (9 mg, 0.4 mmol) to anhydrous ethanol (4 mL) under nitrogen. After complete consumption of the sodium, part of the resulting solution (0.8 mL, 80 μmol) was transferred to a flask containing cis-pyrrolidine 12 (20 mg, 80 μmol) and the resulting solution was stirred at RT for 18 h. DCM (2 mL) was added to produce a precipitate which was removed by filtration through Celite®. The filtrate was evaporated to dryness and analysed by 1 H NMR.

Method 2
Pd/C (10% w/w, 0.091 g) was added to a stirred solution of diol (+)-15 (0.45 g, 2.18 mmol) and di-tert-butyl dicarbonate (0.52 mL, 2.27 mmol) in MeOH (9.1 mL). 42 The atmosphere was replaced with H 2 (balloon) and the reaction mixture was stirred for 24 h. The mixture was filtered through Celite® and the solvent was evaporated to give the crude product (0.44 g) which was dissolved in MeOH (17.3 mL) and aqueous HCl (37%, 8.7 mL) was added at RT. The mixture was stirred for 30 min and the solvent was evaporated to give the title compound in pure form as a yellow oil (0.24 g, 78%); NMR data was identical to that from Method 1, above.

General procedure for reductive amination to give test inhibitors 6a-6e
To a solution of pyrrolidine hydrochloride (+)-18.HCl (50 mg, 0.33 mmol) in anhydrous MeOH (2.2 mL) was added the appropriate aryl aldehyde (1.07 equiv., 0.35 mmol) and the mixture stirred at RT until complete dissolution. Then, NaBH 3 CN was added portionwise and the mixture was left stirring under N 2 for 18 h. General work-up involved filtration of the mixture through Celite® and solvent evaporation.

(3R,4R)-4-(Hydroxymethyl)-1-[(1H-imidazol-2-yl)methyl] pyrrolidin-3-ol (+)-6a
The general procedure for reductive amination was followed. The resulting crude product was adsorbed onto silica and flash column silica chromatography was performed (DCM/MeOH [19:1]  The general procedure for reductive amination was followed. As soon as NaBH 3 CN was added, the mixture, initially a transparent solution, turned cloudy. After stirring at RT for 22 h, LCMS revealed a high concentration of unreacted starting material. For that reason, 1 extra equiv. of 1H-Imidazole-4-carbaldehyde was added (31 mg, 0.33 mmol) and the mixture was stirred for 1 h. It was then filtered through Celite® and loaded onto a 2 g strong cation exchange column.

Procedure
Step 1: Binding of HX02 substrate oligonucleotide to well surface A 0.5 nM solution of oligonucleotide HX02 was prepared by diluting 10 µM HX02 into bicarbonate buffer. It was added to the Nunc® Immobiliser™ Amino plate (100 µL 0.5 nM, 0.05 pmol HX02 per well), which was incubated overnight at 4°C. Then, the liquid was decanted from the wells and the plate was washed with PBST (3 × 150 µL/well) and dried.
Step 2: In situ hybridisation of HX02 and Loop01 A 0.5 nM solution of oligonucleotide Loop01 was prepared by diluting 10 µM Loop01 into hybridisation buffer. It was added to the plate (100 µL 0.5 nM, 0.05 pmol Loop01 per well), which was heated to 95°C for 10 min. Then, it was cooled to 80°C and kept at 80°C for 10 min. After this time, it was cooled to 70°C, 60°C, 50°C, 40°C and 30°C each for 10 min to promote annealing of the DNA strands. It was allowed to cool to RT, the liquid was decanted, the plate was washed with PBST (3 × 150 µL/well) and dried by vigorous tapping onto paper towel.
Step 3: Ligation reaction A 0.04 U/100 µL solution of T4 DNA ligase was prepared by diluting 3 U/µL T4 DNA ligase into ligase buffer. It was added to the plate (100 µL/well), which was incubated at 37°C for 2 h. The liquid was decanted and the plate was washed with PBST (3 × 150 µL/well). The final wash was left in the wells and the plates were frozen overnight. Then, they were warmed to RT, emptied and dried by vigorous tapping onto paper towel.
Step 4: AAG standard preparations with BSA A 100 µg/mL BSA in glycosylase buffer was prepared. This solution was used to prepare increasing concentrations of the different inhibitors which were tested against AAG.
A 0.8 U/100 µL stock solution of AAG was prepared by diluting 10 U/µL purchased AAG into AAG glycosylase buffer/BSA. Increasing concentrations of AAG (0 U/well to 0.4 U/well) were prepared by diluting the stock solution into AAG glycosylase buffer/BSA. 0.05 U/ 100 µL AAG was the concentration selected at which to test the different inhibitors. Each mixture of enzyme and inhibitor was allowed at least 5 min of pre-incubation time.
Step 5: Incubation and work-up The different AAG (+/− inhibitor) dilutions were added to the plate (100 µL/well), each with three replicates, which was incubated at 37°C for 2 h. Then, the liquid was decanted, the plate was washed with PBST (3 × 150 µL) and dried. Alkaline denaturation buffer was added (200 µL/well) and the plate was incubated at 95°C for 15 min. Then, it was allowed to cool to RT, the liquid was decanted, the plate was washed with PBST (3 × 200 µL/well) and dried.
Step 6: Colorimetric detection A solution of BSA in PBST (10 mL, 0.01 g/mL) was prepared. Then, 10 µL of 1:10 diluted goat antibody to fluorescein (goat anti-fluorescein) horseradish peroxidase was added. This solution was added to the plate (100 µL/well), which was incubated at RT for 1 h. Then, the liquid was decanted, the plate was washed with PBST (3 × 200 µL/ well) and dried. A 1:1 mixture of TMB peroxidase substrate solution and peroxidase substrate buffer was prepared. It was added to the plate (100 µL/well). Once sufficient pale blue colour had developed (12 min), phosphoric acid 1 M (100 µL/well) was added, and a colour change to yellow was observed. The absorbance was read at 450 nm.
Step 7: Data processing GraphPad Prism 7.04 was used to process the data. A standard curve of absorbance vs.
[AAG] was plotted and absorbances acquired from wells containing inhibitor were interpolated into this to obtain an apparent [AAG]. The values of apparent [AAG] were used to calculate the % inhibition given that a real [AAG] of 0.05 U/100 μL was used. IC 50 curves of the proposed inhibitors were fitted using equation: "[Inhibitor] vs. response" which also generated IC 50 values from the mid-point between the top and bottom of the curve (not the y = 50% mark).

Declaration of Competing Interest
The authors declare that there are no conflicts of interest.