Identification of phosphotyrosine mimetic inhibitors of human tyrosyl-DNA phosphodiesterase I by a novel AlphaScreen high-throughput assay

Human tyrosyl-DNA phosphodiesterase I (Tdp1) is a newly discovered enzyme involved in the repair of DNA lesions created by the trapping of human topoisomerase I (Top1) on DNA. Top1 can be trapped by abasic sites, oxidative and methylation base damage, carcinogenic adducts, and strand breaks (1) or following treatment by anticancer agents such as camptothecins and indenoisoquinolines (for review, see refs. 2, 3). Tdp1 belongs to the phospholipase D superfamily (4) and was discovered by Pouliot et al. (5) as the enzyme capable of hydrolyzing the covalent bond between the Top1 catalytic tyrosine and the 3′-end of the DNA (6). The hydrolysis leads to a 3′-phosphate DNA end, which is further processed by a 3′-phosphatase called polynucleotide kinase phosphatase (PNKP; Fig. 1). In humans, Tdp1 and PNKP form a multiprotein complex with XRCC1, poly(ADP-ribose) polymerase, β-polymerase, and ligase III (refs. 7–9; Fig. 1, bottom). This complex contains the critical elements for base excision repair.

Tdp1 is ubiquitous in eukaryotes and physiologically important because the homozygous mutation H493R in its catalytic pocket causes spinocerebellar ataxia with axonal neuropathy (SCAN1; ref. 10). This mutation inactivates Tdp1 by trapping Tdp1-DNA intermediates (11). SCAN1 cells are hypersensitive to camptothecin (8, 11–14) and ionizing radiation (15) but not to etoposide or bleomycin (11). The budding yeast TDP1 knockout is viable (5) and hypersensitive to camptothecin only when the checkpoint gene Rad9 is simultaneously inactivated (16) or when some endonuclease repair pathways (Rad1/Rad10 and Slx1/Slx4) are defective (17–19). Tdp1 function is probably not limited to the repair of Top1 cleavage complexes as it could also be involved in the repair of DNA lesion created by the trapping of topoisomerase II (12, 20). Tdp1 can also remove 3′-phosphoglycolate generated by oxidative DNA damage (15, 21), suggesting a broader role in the maintenance of genomic stability (22) and making it a rational anticancer target (2).

Aminoglycoside antibiotics and ribosome inhibitors inhibit Tdp1 at millimolar concentrations (23). Vanadate and tungstate act as phosphate mimetics in cocrystal structures and also block Tdp1 activity at millimolar concentrations (24). Furamidine inhibits Tdp1 at micromolar concentrations but may have additional targets due to its DNA-binding activities (25). It is therefore rational to develop Tdp1 inhibitors for cancer treatment in combination with camptothecins and indenoisoquinolines. The anticancer activity of Tdp1 inhibitors may prove to be dependent on the presence of cancer-related genetic abnormalities, because hypersensitivity to camptothecin in Tdp1-defective yeast is conditional for deficiencies in the Rad9 checkpoint (see above; refs. 5, 17, 18), leading one to speculate that Tdp1 is primarily required when checkpoints are deficient.

There is an obvious need to identify new Tdp1-inhibiting chemotypes, but simple homogeneous assays amenable to high-throughput screening (HTS) have been lacking. Standard activity assay involves radiolabeled DNA-phosphotyrosine substrates with polyacrylamide gel analysis (6). Although this separation-based approach is thorough, in that both substrate and product are accounted for, it is not suitable for HTS. Screening-friendly schemes have included chromogenic (p-nitrophenyl based; refs. 6, 26) and fluorogenic (4-methylumbelliferone based; refs. 27) substrates. However, these assays were either relatively insensitive, requiring high enzyme and substrate levels (to develop the color of the p-nitrophenyl reporter), or used an incomplete substrate (the DNA-phospho-4-methylumbelliferyl substrate is missing the tyrosine moiety). Additionally, the fluorogenic assay operated in the blue-shifted region of light detection where the most interference from compound autofluorescence has been shown to occur (28) and the released 4-methylumbelliferone was fluorescent in strictly basic pH environment.

We recently reported an enhanced chemiluminescence assay for the discovery of Tdp1 inhibitors (25). Due to its high cost (>60 cents per well), its applicability strictly to 96-well format, and its requirement for washing steps and the use of specialized detector, the enhanced chemiluminescence assay was unsuitable for application in highly miniaturized automated systems designed to screen millions of samples. Our present approach is based on the use of the AlphaScreen (Amplified Luminescence Proximity Homogenous Assay) bead system, originally designed as an ultrasensitive, low-background immunoassay (29) but nowadays a method that is finding an increased use in HTS. AlphaScreen beads are true colloidal size, hydrophilic, and stable and are processed by liquid dispensers in the same manner as ordinary homogeneous solutions. On illumination with 680 nm centered light (see Fig. 2), a phthalocyanine photosensitizer contained within the donor bead converts ambient oxygen molecules in its vicinity to singlet oxygen. The latter can diffuse ∼200 nm in solution within its 4 μs half-life. If an acceptor bead is within that proximity, for example, as a ternary complex with an analyte and a donor bead, the singlet oxygen can react with and trigger light release from the thioxene, anthracene, and rubene derivatives contained within the acceptor bead, ultimately resulting in a light emission in the 520 to 620 nm range. Due to this unique signal generation mechanism, the AlphaScreen technology offers very high sensitivity in combination with very low background and has been used to configure a wide range of assays, including single nucleotide polymorphism genotyping and HTS for proteases, cyclic AMP, kinases, inositol phosphatases, and protein-protein interactions. Herein, we describe the development of a simple homogeneous AlphaScreen assay for Tdp1 and show its suitability for large-scale HTS. Further, we report the identification and initial characterization of four inhibitors of the enzyme out of a pilot screen of the Library of Pharmacologically Active Compounds LOPAC¹²⁸⁰ collection.

Tween 20, KCl, FITC, and PBS were procured from Sigma-Aldrich. DMSO-certified ACS grade was obtained from Fisher. The AlphaScreen FITC/streptavidin detection kit was from Perkin-Elmer Life and Analytical Sciences. AlphaScreen Assay buffer consisted of PBS (pH 7.4), 80 mmol/L KCl, and 0.05% Tween 20. Recombinant Tdp1 was expressed and purified as described previously (25). The Sigma-Aldrich LOPAC¹²⁸⁰ library of 1280 known bioactives was received as DMSO solutions at initial concentration of 10 mmol/L and plated as described previously (30, 31).

The biotinylated phosphotyrosine deoxyoligonucleotide (5′-biotin-GATCTAAAAGACTT-pY-3′) was synthesized, purified, and quality control-tested by Midland Certified Reagent. Standard overnight 4°C FITC coupling was done in-house. The FITC-labeled substrate was purified on Biospin-6 desalting columns (Bio-Rad) and its concentration was determined by UV-visible spectrophotometry.

Assay reaction mixtures and substrate/bead titrations were initially done at 20 μL final volume in 384-well plates. For the subsequent 1,536-well-based experiments, Flying Reagent Dispenser (Aurora Discovery, presently Beckman-Coulter; ref. 32) was used to dispense reagents into the assay plates, whereas a pintool was used to deliver DMSO solutions of the test inhibitors. Plates were read on Envision plate reader (Perkin-Elmer) equipped with AlphaScreen optical detection module. For testing the effect of enzyme reaction conditions on the IC₅₀ values for select inhibitors, the respective compounds were prepared as 24-point 2-fold intraplate dilution series and assayed as described.

Three microliters of reagents (buffer in columns 3 and 4 as negative control and 1.33 nmol/L Tdp1 in columns 1, 2, and 5-48) were dispensed into 1,536-well black solid-bottomed plate. Compounds (final concentrations in the range of 0.7 nmol/L to 57 μmol/L) were transferred via Kalypsys pintool equipped with 1,536-pin array (33). The plate was incubated for 15 min at room temperature, and 1 μL substrate (15 nmol/L final concentration) was added to start the reaction. After 5 min incubation at room temperature, 1 μL bead mix (15 μg/mL final concentration) was added and the plate was further incubated for 10 min at room temperature before signal measurement. Substrate and beads were prepared and kept in amber bottles to prevent photodegradation. Library plates were screened starting from the lowest and proceeding to the highest concentration. Vehicle-only plates, with DMSO being pin-transferred to the entire columns 5 to 48 compound area, were included at the beginning and the end of the screen to record any systematic shifts in assay signal. Screening data were corrected, normalized, and curve-fitted (34) by using in-house developed algorithms.

5-³²P-labeled substrate (1 nmol/L; N14Y; 5′-GATCTAAAAGACTT-pY-3′) was incubated with 0.1 nmol/L recombinant Tdp1 in the absence or presence of inhibitor for 20 min at 25°C in AlphaScreen assay buffer. Reactions were terminated by the addition of 1 volume of gel loading buffer [96% (v/v) formamide, 10 mmol/L EDTA, 1% (w/v) xylene cyanol, and 1% (w/v) bromophenol blue]. The samples were subsequently heated to 95°C for 5 min and subjected to 20% denaturing PAGE. When the Tdp1 mutant H493R (100 nmol/L) was employed, the reactions were stopped with 1 volume of SDS loading dye and analyzed by 4% to 20% SDS-PAGE. All gels were dried and exposed on a PhosphorImager screen. Imaging and quantification were done using a Typhoon 8600 and ImageQuant software (GE Healthcare).

To configure a new assay for Tdp1, we used the ability of the AlphaScreen system to report on the integrity of the phosphotyrosine-based substrate, viewed here as a multisite analyte. The AlphaScreen signal is a direct consequence of the close proximity of donor and acceptor beads, which is achieved by the formation of a ternary donor-analyte-acceptor recognition complex. To realize the AlphaScreen assay using the DNA-pY substrate for Tdp1, 5′-GATCTAAAAGACTT-pY-3′ (7, 23, 25), recognition elements for both bead types had to be introduced. To this end, we coupled a FITC to the amino group of tyrosine in a phosphotyrosine-containing single-stranded DNA substrate bearing a biotin at its 5′-end (Fig. 2). Thus, intact substrate, when mixed with streptavidin-donor and anti-FITC antibody-acceptor beads, was expected to yield a high AlphaScreen signal, whereas the Tdp1-catalyzed substrate hydrolysis would result in a decreased signal. Conversely, if the Tdp1 catalysis was inhibited, an elevated signal would be observed for that sample.

In a 384-well plate, the addition of an equimolar mix of anti-FITC-acceptor and streptavidin-donor AlphaScreen beads to FITC-DNA substrate generated a strong signal, whereas omission of substrate resulted in background readings. The concentration-response curve of the FITC-DNA substrate titrated against constant bead concentration (Fig. 3A) exhibited a maximum ∼15 nmol/L followed by a signal decrease to background levels as the substrate concentration was further increased; this trend was reproduced with two different enzyme batches and two independently synthesized substrate lots. The biphasic behavior is characteristic of such polyvalent interactions and in this case was due to the saturation of all available binding sites on the beads and the buildup of bipartite populations of beads saturated exclusively by substrate molecules at the expense of tripartite donor-substrate-acceptor assemblies. The maximum signal concentration of 15 nmol/L falls below the previously reported K_m value of Tdp1 of 80 ± 20 nmol/L (35), thus ensuring good assay sensitivity with respect to potential inhibitors.

When the same-sequence construct devoid of FITC was tested, a weak and considerably right-shifted signal evolution was noted (Fig. 3A). Although this result points to some nonspecific binding, the contribution of the latter to the total signal at the 15 nmol/L point appears to be minimal. Moreover, the Tdp1-induced cleavage of the FITC-DNA substrate, when driven to completion, yielded background signal (Fig. 3B), thus indicating that the small signal increase observed here was likely due to the interaction between the acceptor bead and the tyrosine moiety. In the presence of increasing concentrations of Tdp1, the FITC-coupled tyrosine group was released, preventing anti-FITC acceptor bead positioning close to the donor bead and subsequently leading to an enzyme-dependent loss of signal (Fig. 3B). DMSO had no effect on the enzymatic reaction (data not shown). Neomycin and vanadate, two previously described weak Tdp1 inhibitors, both inhibited Tdp1 in this assay (Fig. 3C). The test method and reaction conditions of the original studies of neomycin and vanadate (23, 24) differ from those used in the present assay; nevertheless, the potencies of the two compounds determined here were in agreement with these prior publications, with neomycin being a weak low millimolar inhibitor and vanadate exhibiting its inhibition in the submillimolar range.

The assay was further miniaturized to a final volume of 5 μL in 1,536-well format by direct volume reduction. The inclusion of Tween 20 in the assay buffer helped prevent enzyme absorption to the polystyrene wells and minimized the interfering effect of promiscuous inhibitors acting via colloidal aggregate formation (36). To verify assay integrity over the intended timeline for automated robotic screening, reagent stability over time was assessed by running the assay at multiple time points while storing the working reagents at 4°C. Reagents remained stable for ∼2 days of storage (Supplementary Fig. S1),³

To further validate the present assay in a real 1,536-well-based HTS context, the LOPAC¹²⁸⁰ was screened in quantitative HTS mode as eight-point concentration series (30, 31). The pilot screen performed robustly, yielding average Z′ factor (37) of 0.71. After data analysis, four active compounds were selected for further studies. These included the highly potent aurintricarboxylic acid, the known protein tyrosine phosphatase inhibitor methyl-3,4-dephostatin, and the Gα-specific G-protein antagonists suramin and NF449 (Fig. 4A). In enzyme turnover assays, the extent of substrate conversion plays a role in determining the assay sensitivity toward inhibitors. Figure 4B shows the IC₅₀ potencies of the four inhibitors identified from the above pilot screen as a function of changing reaction conditions designed to achieve a gradual decrease in the fraction of substrate consumed. For all four inhibitors, a significant downward shift in IC₅₀ was observed on progression to the assay with the lowest substrate conversion (IC₅₀ trends in Fig. 4B and concentration-response curves derived from condition C in Fig. 4C), in agreement with published analyses (38). As expected, a second pilot screen, run at the more sensitive assay conditions, resulted in the appearance of additional actives (Fig. 4D, right versus left), all of which were less potent than the originally identified four compounds.

The four inhibitors were purchased separately and tested in secondary radiolabel gel-based assays (22, 24). All were found to be active against recombinant Tdp1 enzyme. Methyl-3,4-dephostatin, a stable analogue of the nitrosoaniline dephostatin (Fig. 5A), inhibited Tdp1 in our gel-based assay with an IC₅₀ value of 0.36 ± 0.20 μmol/L (Fig. 5B and C). In contrast, dephostatin, which differs from methyl-3,4-dephostatin by the position of one of its hydroxyl groups (Fig. 5A), displayed only a trace level of inhibition within the same concentration range (Fig. 5B and C), suggesting that specific structure modification on the nitrosoaniline motif is sufficient to determine Tdp1 inhibition. The inhibition curves obtained for methyl-3,4-dephostatin in both the primary and the secondary assays had a similar shape and were almost superimposable (Fig. 5C). These results show the high sensitivity of the Tdp1 AlphaScreen assay and further validate it biochemically, with its output being comparable with that of conventional separation-based assays.

To gain additional insight into the mechanism of action of methyl-3,4-dephostatin, experiments were carried out using the SCAN1 mutant of Tdp1 (10). This mutant bears a H493R substitution affecting one of the two catalytic histidine residues (for review, see Fig. 3 in ref. 2). The two catalytic histidines, H263 and H493, participate sequentially in the Tdp1 cleavage reaction. H263 is first responsible for the nucleophilic attack on the phosphotyrosyl bond, whereas H493 donates a proton to the tyrosyl-peptide-leaving group (Fig. 6A). The resulting transient Tdp1-DNA covalent complex (Fig. 6B) is then hydrolyzed by a water molecule activated by H493 (Fig. 6C). It has been shown that this transient Tdp1-DNA complex accumulates in SCAN1 cells (Fig. 6D) and suggested that this Tdp1 trapping could be responsible for the SCAN1 phenotype (11, 39). We took advantage of this characteristic of the SCAN1 Tdp1 mutant to investigate whether an inhibitor acts before or after Tdp1 forms a transient covalent intermediate with DNA (Fig. 6B). When incubated with the H493R Tdp1 mutant, methyl-3,4-dephostatin completely prevented the accumulation of the covalent Tdp1-DNA intermediate at 10 μmol/L concentration (estimated IC₅₀, 3 μmol/L), whereas dephostatin had a much weaker effect at this concentration (Fig. 6E). This result suggests that methyl-3,4-dephostatin likely interacts with Tdp1 before the enzyme forms a transient covalent intermediate with DNA and therefore interferes with the binding of the DNA substrate inside the Tdp1 catalytic site.

Tdp1 catalyzes the hydrolysis of a variety of substrates, including 3′-phosphotyrosyl oligonucleotides, protein-DNA 3′-phosphoamide linkages, and 3′-phosphoglycolate adducts (15, 21). A major biological function ascribed to Tdp1 is the liberation of genomic DNA from the covalently trapped Top1 to allow further repair of the strand break by repair enzymes such as PNKP in the base excision repair and single-strand break repair pathways (40). A point mutation in the Tdp1 gene is responsible for the SCAN1 disorder where cells exhibit hypersensitivity to Top1 inhibitors (8, 11, 14), suggesting that inhibitors of Tdp1 could be used in association with Top1 inhibitors in a combined anticancer therapeutic regimen. Tdp1 has been shown to preferentially process short oligonucleotide DNA substrates with an exposed 3′-phosphotyrosyl bond (Top1-DNA junction; refs. 3, 6, 41). Thus, to design the most relevant assay for Tdp1, the DNA-pY-Top1 consensus sequence 5′-GATCTAAAAGACTT-pY-3′ was used (7). The need to use the substrate at nanomolar concentrations [close to the 50-100 nmol/L K_m range reported for Tdp1 (35) to maintain assay sensitivity (42)] meant that the sensor component(s) to be used in this assay had to both possess high affinity and afford high signal amplification.

The AlphaScreen system fulfilled the above requirements and appeared to be superior to traditional dual-label fluorescence resonance energy transfer or donor-quencher-based approaches. For a successful fluorescence resonance energy transfer assay, fluorophore and acceptor need to be matched in terms of excitation and emission; additionally, the strong dependence of fluorescence resonance energy transfer signal on the precise distance between donor and acceptor labels necessitates investigation of multiple candidate substrates. Lastly, due to the nature of the substrate, the labeling of the latter with two tags at varying locations to explore the best fluorescence resonance energy transfer pair is not synthetically feasible and might result in constructs not accepted by Tdp1 as substrates.⁴

The Tdp1 AlphaScreen substrate had to contain recognition elements for the donor and acceptor beads such that the Tdp1-dependent event of phosphate bond hydrolysis could be reported on. The previously used 5′-biotin was the natural choice for the donor bead attachment due to both the ease of 5′-biotin incorporation during oligo synthesis and the availability of streptavidin-coated donor beads as a standard reagent. On the phosphotyrosine end of the substrate, we surmised that the tyrosine free amine was both most distal to the phosphodiester bond and offered the opportunity for standard conjugation chemistries. In choosing a recognition pair for that site, we were similarly guided by the desire to use as many standard reagents as possible. We thus selected a fluorescein label in combination with anti-fluorescein antibody-coated acceptor bead due to the availability of the latter as a standard catalogue item. The excellent reproducibility and reagent stability associated with the assay make it especially suitable for HTS. Furthermore, the AlphaScreen format did appear to be less susceptible to light interference as the large number of colored and fluorescent compounds known to be present in the LOPAC¹²⁸⁰ collection (28) failed to produce activity in this assay.

The present studies led to the identification of several potent new inhibitors of Tdp1. Aurintricarboxylic acid (IC₅₀, 12 nmol/L) has been noted in the literature as a potent inhibitor of a variety of DNA-interacting enzymes. As such, its identification was not unexpected but rather served as an additional biochemical validation of the assay. The other three inhibitors vary in potency, but all bear resemblance to the phosphotyrosine-oligo substrate. Although the exact mechanism of Tdp1 inhibition by suramin, NF449, and methyl-3,4-dephostatin has yet to be established, it appears unlikely that these compounds act as single-stranded DNA binders. Suramin blocks the binding of various growth factors, including insulin-like growth factor-I, epidermal growth factor, platelet-derived growth factor, and tumor growth factor-β, to their receptors, thereby inhibiting endothelial cell proliferation and migration (43). Both suramin and its analogue NF449 have been recognized as G_sα-selective G protein (44) and P2 receptor antagonists (45). Herein, they may be acting as decoys, as Tdp1 may perceive their extended sulfone group network as multiple phosphotyrosine sites.

The nitrosoaniline methyl-3,4-dephostatin is a stable analogue of dephostatin and is the smallest of the four active compounds (Figs. 4A and 5A); of note, it is considerably more potent than vanadate and neomycin (23). Dephostatin and its analog are known protein tyrosine phosphatase inhibitors (46, 47) that also probably act as phosphotyrosine mimetics. Methyl-3,4-dephostatin, in contrast to dephostatin, does not inhibit CD45-associated protein tyrosine phosphatase, pointing to the exclusivity of the respective binding site. We found that methyl-3,4-dephostatin inhibited Tdp1 at submicromolar concentration in both the primary and the secondary assays (Fig. 5B and C) and therefore represents the most potent Tdp1 inhibitor reported to date. Dephostatin itself was not identified as a hit in our pilot screen and showed only trace inhibition when procured subsequently and tested in both the AlphaScreen HTS assay and the secondary assay (Fig. 5C). Therefore, dephostatin could be of value as a negative control counterpart of methyl-3,4-dephostatin when these compounds are used as small-molecule probes of Tdp1 mechanism and function.

To probe which step of the catalytic reaction was inhibited by methyl-3,4-dephostatin [cleavage of the phosphotyrosyl group (Fig. 6A and B) or release of the DNA from the transient covalent complex (Fig. 6B and C)], DNA trapping experiments were done using the SCAN1 Tdp1 mutant (11). The results indicate that methyl-3,4-dephostatin interferes with the early catalytic step, the binding of the Tdp1 DNA substrate, whereas dephostatin has almost no effect on this activity. These marked differences in potency for the two isomers (Fig. 5A) suggest that the position of the two hydroxyls on the nitrosoaniline motif is critical for Tdp1 inhibition.

In summary, we have developed a robust 1,536-well-based assay for the identification of inhibitors of Tdp1. The assay uses commercially available common reagents and what we believe to be the most active site relevant Tdp1 substrate reported to date. The general assay scheme presented here should be applicable to the testing of other DNA repair enzymes where a cleavage step is involved. Furthermore, the assay is tunable and responsive to a wide range of inhibitor potencies, and its low cost of ∼3 cents per well makes it attractive for both large-scale automated HTS and secondary screening. Methyl-3,4-dephostatin was identified from our pilot quantitative HTS and represents the most potent Tdp1 inhibitor known to date. Extended SAR studies on dephostatin analogues are warranted to further understand the mechanism of action of methyl-3,4-dephostatin and determine its value as a Tdp1 lead inhibitor.

No potential conflicts of interest were disclosed.

The SCAN1 mutant Tdp1 construct was generous gift from Drs. J.J. Champoux (University of Washington) and H. Interthal (University of Edinburgh). We thank Carl Apgar for expert advice during the initial adoption of the AlphaScreen assay.