DNA interstrand crosslink (ICL)-inducing agents such as cisplatin, melphalan and mitomycin C (MMC) remain a mainstay in the treatment of cancers such as multiple myeloma, ovarian, testis, breast, bladder and head and neck cancers.1–5 However, many cancers become refractory to these drugs by multiple mechanisms during a course of chemotherapy, leading to treatment failure. Re-sensitizing tumor cells to ICL-inducing agents has great potential for treating a wide variety of cancers. One strategy relies on the discovery of small-molecule inhibitors of proteins that control the cellular response to DNA ICLs. Current evidence suggests that the Fanconi Anemia (FA) DNA damage response pathway (which includes breast cancer susceptibility gene products BRCA2, BRIP1/BACH1 and PALB2/FANCN) is involved in the repair of DNA ICLs. As a consequence, acquired ICL-resistance of cancer cells may be associated with re-activation of the FA pathway in tumors initially deficient in this pathway.3, 6, 7 Supporting this idea, epigenetic silencing of FANCF has been shown to occur sporadically in AML7 and in 17, 21 and 30% of breast, ovarian and cervical cancers, respectively.8 Loss of FANCD2 expression has also been linked with sporadic breast cancer.5 Interestingly, a recent report by Kennedy et al. identified a set of DNA repair genes [including ataxia telangectasia-mutated (ATM) and NBS1] that are synthetic lethal with FA genes, demonstrating that inhibition of the FA pathway can selectively kill tumor cells already deficient in another DNA repair pathway, and vice versa.9 Thus, compounds that inhibit the FA pathway are promising candidates to either (i) re-sensitize cancer cells that rely on a functional FA pathway for resistance to chemotherapeutic drugs or (ii) cause synthetic lethality in tumor cells with acquired loss of function in a second, non-FA DNA repair pathway. Yet, a rational approach for discovery of candidate compounds is hampered by the complexity of the pathway and because the functions and interactions of its numerous proteins are not well understood. Moreover, the FA pathway is activated only during S-phase, which renders cell-based assays more difficult to develop.
To tackle these problems, we recently developed a cell-free system to evaluate the function of the FA pathway proteins in the DNA damage response using Xenopus egg extracts.10, 11 These extracts are naturally synchronized in S-phase and are fully capable of supporting DNA replication and DNA damage responses under natural cell cycle control, including the monoubiquitylation of FANCD2.12–15 In both human and Xenopus cells, the FA pathway is controlled by a ‘core complex’ composed of at least 10 proteins (FANCA, B, C, E, F, G, L, M, N, FAAP24) in addition to the FANCD2 binding partner, FANCI. This core complex acts as an E3 ubiquitin ligase that mediates DNA damage-triggered monoubiquitylation of the downstream FA protein, FANCD2.10, 16 Monoubiquitylation of FANCD2 [Fanconi anemia protein D2, long form (FANCD2-L)] during S-phase and following DNA damage is a hallmark of an intact FA pathway and predictive for normal cellular resistance to DNA ICLs.17, 18 We recently demonstrated that in Xenopus egg extracts, FANCD2-L formation is induced by adding DNA molecules mimicking damaged DNA structures, thus bypassing the need for ongoing replication to trigger the FA pathway.10 Using this DNA substrate-inducible FANCD2-L formation as readout for FA pathway activity, we established a cell-free strategy to screen for small molecule inhibitors of the FA pathway. Our approach provides the first example of a rapid and inexpensive in vitro assay for screening compounds that modulate the FA pathway, bridging the gap between biochemical and cell-based assays. Using this system, we identified the menadione analog DDN (2,3-dichloro-5,8-dihydroxy-1,4-naphthoquinone), a known antitumor agent,19, 20 as a novel inhibitor of the FA pathway.
Material and methods
Chemicals and antibodies
Wortmannin (Sigma, St. Louis, MO), curcumin (Sigma), DDN (Calbiochem, San Diego, CA) and cisplatin (Sigma) were resuspended in DMSO. Hydroxyurea (Sigma) was resuspended in H20. Generation of antibodies against xFANCD2 and hFANCD2 was described previously.10 Antibodies against human PCNA and γH2AX were from Santa Cruz (# sc-56, Santa Cruz, CA) and Bethyl Laboratories (# A300-081A, Montgomery, TX), respectively. Antibodies against xMre11 and hFANCD2 (for immunofluorescence) were a kind gift of J. Gautier and K.J. Patel, respectively. The anti-tubulin α antibody was from Sigma.
Xenopus cell-free assay
Low-speed extracts were prepared from Xenopus laevis eggs as described previously.10 Ten microliters of extract were left untreated (negative control) or supplemented with 0.5 μl DNA (3.0 μg/μl) and 1 μl compound. Following incubation for 20 min at room temperature, 100 μl of 1× sample buffer (NuPage, Invitrogen, Carlsbad, CA) were added and the reaction incubated at 95°C for 5 min. Ten microliters of each sample were analyzed by SDS-PAGE and immunoblotting.
Preparation of DNA substrate
Circular supercoiled pBluescript plasmid DNA (pBSKS, 2961 bp) was prepared from E. coli cultures using a QiaFilter Plasmid Maxi Kit (Qiagen, Valencia, CA).
Forty microliters of egg extract (2 mg total protein content) were incubated with plasmid DNA (150 ng/μl) and 500 μM DDN (or DMSO) for 20 min at room temperature before addition of 1 ml lysis buffer (10 mM Tris pH 7.4, 150 mM NaCl, 1% NP40, 0.5% Deoxycholate, 1mM EDTA, 0.5 mg/ml Pefabloc, 1mM DTT). Ten microliters of rabbit polyclonal antibody against xFANCG was added and samples were mixed by rotating overnight at 4°C. One hundred microliters of pre-swelled and washed (50% slurry in PBS) sepharose 4B beads (Amersham, Piscataway, NJ) were added and rotated for 30 min at 4°C. Beads were pelleted by centrifugation, washed 3 times with lysis buffer and proteins were eluted by boiling in protein loading buffer.
Xenopus extracts (20 μl) supplemented with various combinations of 500 μM His-ubiquitin (Boston Biochem, Boston, MA), 150 ng/μl plasmid DNA and 100 μM DDN were incubated for 20 min at 23°C. 1 μl was set aside as input, and 500 μl of buffer A (6 M guanidine-HCl, 0.1 M Na2HPO4-HCl pH 8.0, 0.01 M Tris-HCl pH 8.0, 5 mM imidazole, 10 mM β-mercaptoethanol) were added to the remaining extract volume. 80 μl Ni-NTA agarose beads (50% slurry, Qiagen) equilibrated in buffer A were added and the reaction was incubated for 4 hr at 4°C with constant rotation. Beads were sequentially washed in 1 ml buffer A, 1ml buffer B (8 M Urea, 0.1 M Na2HPO4-HCl pH 8.0, 0.01 M Tris-HCl pH 8.0, 10 mM β-mercaptoethanol) and 1ml buffer C (8 M Urea, 0.1 M Na2HPO4-HCl pH 8.0, 0.01 M Tris-HCl pH 6.3, 10 mM β-mercaptoethanol). Beads were recovered by 2,000 rpm centrifugation for 1 min. After the last centrifugation, buffer C was carefully removed and the beads were resuspended in 40 μl elution buffer (1× SDS loading buffer supplemented with 200 mM imidazole) and boiled for 10 min. Twenty microliters of each eluate was then analyzed by immunoblot along with the respective input extract aliquots.
Protein samples were separated on 3–8% Tris-Acetate or 8–16% Tris-Glycine gels (NuPage, Invitrogen) and transferred to Immobilon P membranes (Millipore, Billerica, MA). After blocking in 5% milk for 30 min, membranes were incubated with anti-xFANCD2 (1:1,000), anti-hFANCD2 (1:500), anti-xMRE11 (1:5,000), anti-tubulin α (1:6,000), anti-γH2AX (1:4,000) or anti-PCNA (1:2,000) and subsequently with HRP-conjugated rabbit or mouse secondary antibody (1:10,000 and 1:2,500, respectively). Immunoblots were developed using the ECL plus system (Amersham). Protein band densitometry was performed using ImageJ. The graphing software used was IGOR Pro (www.wavemetrics.com).
Each compound of the Challenge set (NCI), Natural products set (NCI), Cancer plate (MicroSource Discovery, Gaylordsville, CT) and NINDS II custom collection (MicroSource Discovery, Gaylordsville, CT) was initially tested at 500 μM final concentration in the Xenopus cell-free assay in a 96-well plates format. Immunoblots were probed with antibodies against xFANCD2 and xMre11. Positive compounds identified in the initial screen were further tested by serial dilution to confirm the inhibitory effect in a dose-dependent manner and to establish the IC50.
Cell lines and cell culture
HeLa cells were grown in DMEM medium supplemented with 10% serum in humidified 5% CO2 atmosphere. The FANCF-deficient 2008 human ovarian carcinoma cell line and its FANCF-complemented counterpart (2008+F)6 were grown in RPMI 1640 medium supplemented with 10% serum.
Cell survival assay, FANCD2 ubiquitylation assay and immunofluorescence microscopy
HeLa cells were incubated with 2 mM HU and various concentrations of DDN for 7 hr. Survival assay was performed in triplicate using the CellTiter 96 AQueous One Solution Cell Proliferation Assay (Promega, Madison, WI) following manufacturer's recommendations. For monitoring FANCD2 monoubiquitylation, cells were lysed in RIPA buffer and crude lysates (15 μg total protein) were boiled in SDS-loading buffer for 30 min, resolved on 3–8% SDS-PAGE and immunoblotted. Experiments were repeated twice. For FACS analysis, cells were trypsinized, fixed in 70% EtOH and stained for 16 hr at 4°C in propidium iodide (PI) solution (40 μg/ml PI, 0.2 mg/ml RNAse A, 0.1% triton in PBS). For each point, 10,000 gated cells were counted using a FACSCalibur flow cytometer (BD Biosciences, San Jose, CA). For immunofluorescence microscopy, HeLa cells grown on chamber slides were washed with PBS, fixed with 3.7% paraformaldehyde, permeabilized with PBS containing 0.2% Triton X-100 and blocked with a solution of 7.5% BSA in PBS. Primary and secondary antibodies were diluted in PBS containing 1% BSA. Cells were incubated in primary antibody (1:300 anti-hFANCD2) for 1 hr at 4°C, washed in PBS and incubated in secondary antibody (Alexa Fluor 594-conjugated goat anti-rabbit, 1/1,000; Molecular Probes, Eugene, OR). Nuclei were counterstained with Hoechst 33342 dye (10 ng/ml, Molecular Probes). Cells were mounted with anti-fade mounting solution (Vectashield, Vector Laboratories, Burlingame, CA). For quantification of FANCD2 nuclear foci, wide-field images containing up to 150 cells were taken using a Nikon TE800 microscope with an Optronics DEI750 camera. Each cell was scored for FANCD2 foci using ImageJ software, with cells displaying more than 5 foci considered foci-positive. Two hundred cells were counted per point. Variation significance between points from duplicate experiments was analyzed by standard ANOVA method using the InStat software (GraphPad).
2008 and 2008+F cells were seeded at 4,000 cells/well in 96-well plates and treated with various concentrations of cisplatin (0, 0.05, 0.1, 0.2. 0.4, 0.8, 1.6 μg/ml for 2008 and 0, 0.2, 0.4, 0.8, 1.6, 3.2, 6.4 μg/ml for 2008+F) and DDN (0, 0.34, 0.51, 0.76, 1.14, 1.71, 2.56, 3.84, 5.76, 8.64 μM) alone and in combination. Cell viability was measured in each well after 3 days using the CellTiter 96 assay (Promega). Combinations of cisplatin/DDN concentrations leading to 50% decreased survival (LD50) were determined from survival curves corresponding to one concentration of cisplatin in combination with increasing concentrations of DDN (7 curves) and vice-versa (10 curves). For both compounds, concentrations were expressed as a fraction of the highest concentration in the series (set to 1, corresponding to the LD50 of the compound alone). Each experiment was performed in triplicate and isobolograms were drawn by plotting each of these LD50 pairs onto the same diagram. To graphically highlight the divergence between isobolograms, fit curves were determined using IgorPro, with double exponential curves providing the best fit. Statistical analysis was performed by the Center for Biostatistics, Computing and Informatics shared resource (OHSU). As explained in Tallarida (2001),21 isobologram axes were converted to display the fraction of cisplatin in the total dose on the x-axis vs. total dose (DDN plus cisplatin) on the y-axis, where the line of additivity is y = 1.0. Drug synergism (defined as total dose less than 1.0) was tested using the nonparametric Wilcoxon signed rank test, since sample sizes were small. Overall synergism was tested, as well as synergism at different concentrations of cisplatin on the isobologram. The range of cisplatin values was split into thirds, and each third was examined for synergism for each cell line. Differences between the median total dose of the two cell lines was tested by the Wilcoxon rank sum test.
Calculation of IC50 values
Fit curves were determined from experimental data using IgorPro. The concentration of compound that induces 50% inhibition of the measured endpoint (i.e. FANCD2 ubiquitylation, Mre11 phosphorylation and cell survival) was then plotted on the fit curve.
DNA substrate-stimulated FANCD2-L formation is a readout for FA pathway activation in Xenopus egg extracts
The hallmark of an intact Fanconi anemia pathway is the formation of monoubiquitylated FANCD2 (FANCD2-L) during S-phase and following induction of ICLs or stalled replication forks.17, 18In vitro identification of inhibitors of FANCD2-L formation is currently not possible, because attempts to recapitulate the FANCD2 monoubiquitylation reaction with recombinant proteins have not been successful. To bypass these obstacles we explored a cell-free chemical screening approach using Xenopus egg extracts, a biochemical environment that is naturally synchronized in S-phase and faithfully mirrors DNA damage responses described in human cells. Indeed, we recently found that xFANCD2-L is rapidly induced in egg extracts following addition of several dsDNA substrates including circular double stranded plasmid DNA (Figs. 1a and 1b).11 xFANCD2-L induction reached a plateau at ∼100 ng/μl plasmid DNA. We previously showed that quantitative immunodepletion of the FA core complex protein, xFANCA, abrogates the DNA-stimulated formation of xFANCD2-L in egg extracts, as predicted from observations in human cells where FANCD2-L formation is strictly dependent on the FA core complex.10, 11 We infer that the plasmid DNA-induced formation of xFANCD2-L in egg extracts is dependent on the FA core complex and can therefore be used as readout for activity of the FA pathway.
As an internal positive control, we measured the inducible hyperphosphorylation of xMre11 (xMre11-PPP). As reported previously by Gautier and coworkers, phosphorylation of xMre11 is stimulated by structures containing dsDNA breaks.22 We found that high concentrations of circular plasmid DNA (150 ng/μl) also stimulated xMre11 phosphorylation (Figs. 1a and 1b).11 Thus, our in vitro system is capable of monitoring critical steps in at least two distinct DNA damage response pathways.
A cell-free assay for inhibitors of the FA pathway
On the basis of the robust DNA-inducible formation of xFANCD2-L we hypothesized that the induction of xFANCD2-L could be used as readout in a Xenopus cell-free assay to screen for small molecules that inhibit the FA pathway. Figure 2 outlines the screening process. We performed the assay using saturating plasmid DNA concentrations (150 ng/μl) for induction of xFANCD2-L because it increases the dynamic range of detection of xFANCD2-L modulation, it reduces the variability of xFANCD2-L activation between different extract batches, and it favors the detection of robust FA pathway inhibitors. Moreover, high concentrations of plasmid stimulate Mre11 phosphorylation, which was used to evaluate the specificity of compounds for xFANCD2. To increase the consistency of extracts activation among samples, we decided to pre-mix plasmid DNA and extracts in a large batch before addition of each compound to individual aliquots. This method ensures equal extract activation and xFANCD2 monoubiquitylation among samples, which is crucial to reliably identify inhibitors. On the other hand, this strategy might not detect compounds that are active only when added before DNA stimulation.
Initially, we tested whether compounds that are known to inhibit DNA damage-induced FANCD2-L formation in human cells, such as curcumin,3 could also inhibit plasmid DNA-induced FANCD2-L formation in Xenopus egg extracts. To ensure comparable experimental conditions, egg extracts were aliquoted following preparation and frozen at −80°C. Aliquots of the same extract were used for treatment with the chemical compounds described below, allowing direct comparison of compound activity. As shown in Figures 3a and 3c, curcumin inhibited xFANCD2-L formation in egg extracts in a dose-dependent manner with an IC75 of 596 μM. In comparison, Chirnomas et al. reported an IC50 of 15 μM in HeLa cells.3 The requirement for higher compound concentrations in egg extracts compared to cell-based assays appears to be a general trend,23 possibly due to the high concentration of lipids and soluble proteins in egg extracts (∼60 μg/μl). Phosphorylation of xMre11 was simultaneously monitored to assess whether curcumin affects DNA damage response proteins other than xFANCD2 (Fig. 3a). Curcumin inhibited xMre11-PPP at concentrations similar to those that inhibited xFANCD2-L, suggesting that this inhibitor is not specific for the FA pathway.
We also tested the effect of wortmannin, a specific inhibitor of phosphoinositide 3-kinases (PI-3-Ks) previously shown to inhibit the FA pathway in human cells through suppression of ATR activity.24 In egg extracts, wortmannin did not display a significant activity on xFANCD2-L (IC75 > 1600 μM, Figs. 3a and 3c). In comparison, Andreassen et al. reported that high concentrations of wortmannin (200 μM) significantly suppressed FANCD2-L formation following DNA damage in HeLa cells.24 Interestingly, wortmannin inhibited xMre11-PPP (IC50 = 243 μM, Figs. 3a and 3c), indicating that hyper-phosphorylation of xMre11 in response to plasmid DNA depends on phosphoinositide 3-kinases. Taken together, these data indicated that the Xenopus cell-free system is able to identify inhibitors of the FA pathway that are robustly active in cell-based assays, but does not identify compounds that are weak inhibitors.
Library screen for compounds that inhibit the FA pathway
To identify novel inhibitors of the FA pathway we screened 4 chemical compound libraries: (i) Challenge Set, NCI (57 compounds), (ii) Natural Products Set, NCI (235 compounds), (iii)Cancer Plate, BioSource (80 compounds) and (iv) NINDS II, BioSource (400 compounds) (see Table I). Following a primary screen that identified 54 inhibitors of xFANCD2-L, a secondary screening narrowed the number of candidates down to 29. No activator compound was identified because the screen was performed under conditions of maximum xFANCD2-L induction, leaving no margin for further activation.
Table I. Summary of the Screening of Compounds Modulating xFANCD2 Monoubiquitylation and xMRE11 Phosphorylation
Nb comp. tested
Primary screen corresponds to the initial assay with a single concentration for each compound. The secondary screen was a dose-response experiment. Numbers in brackets indicate the percentage of active compounds in the corresponding library.
Challenge set (NCI)
Natural products set (NCI)
Cancer plate (BioSource)
NINDS II (BioSource)
Challenge set (NCI)
Natural products set (NCI)
Cancer plate (BioSource)
One of the most active inhibitors of xFANCD2-L was DDN (2,3-dichloro-5,8-dihydroxy-1,4-naphthoquinone), an analog of the vitamin K derivative menadione with antitumor activities19, 20 (Fig. 3d). DDN inhibited the DNA-stimulated formation of xFANCD2-L in a dose-dependent manner at much lower concentrations than curcumin, with an IC50 of 30 μM (Figs. 3b and 3c). DDN also inhibited xMre11-PPP at a slightly higher concentration (IC50: 57 μM), suggesting that DDN affects both the FA and the Mre11 DNA damage response pathways.
DDN does not affect the integrity of the FA core complex, H2AX phosphorylation or PCNA monoubiquitylation
The architecture of the FA core complex is conserved in egg extracts.25 To test whether DDN inhibited FANCD2-L formation by disrupting the FA core complex, we treated DNA-containing extracts with DDN and analyzed the integrity of the FA core complex by immunoprecipitation of one of its members, xFANCG (Fig. 4a). Although xFANCD2-L was significantly inhibited in DDN-treated extracts (compare Lanes 1 and 4), core complex proteins xFANCA and xFANCM co-immunoprecipitated with xFANCG regardless of the presence of DDN (Fig. 4a, compare Lanes 3, 6). As expected, xFANCD2 did not co-immunoprecipitate with xFANCG since it is not a core complex protein. We conclude that inhibition of xFANCD2-L formation by DDN is not caused by gross disruption of the FA core complex.
In addition to the FA core complex, the major cell cycle checkpoint kinase ATR appears to be partially required for FANCD2 ubiquitylation in human cells.24 To test whether ATR or related PI-3 kinases like ATM or DNA-PK play a role in DDN activity in extracts, we monitored the DNA-inducible phosphorylation of one of their downstream targets, histone H2AX (γH2AX). The γH2AX signal was strongly induced upon addition of plasmid DNA to egg extracts (Fig. 4b, compare Lanes 1 and 2), similar to a recent study by Dupre et al. describing activation of γH2AX in response to double-stranded plasmid DNA fragments.23 DDN did not affect γH2AX formation at concentrations that completely inhibit xFANCD2-L (Fig. 4b, compare Lanes 3 and 4), suggesting that DDN does not inhibit FANCD2-L through inhibition of PI-3-related kinases. Finally, to test the possibility that DDN could be a general ubiquitylation inhibitor, we monitored the DNA-inducible monoubiquitylation of PCNA in egg extracts (Fig. 4c). Extracts were supplemented with recombinant His-ubiquitin and plasmid DNA, and ubiquitylated proteins were pulled down using Ni-NTA agarose beads under denaturing conditions. Immunoblot analysis revealed that PCNA is monoubiquitylated in DNA-free extracts in the presence of exogenous His-ubiquitin (Fig. 4c, Lanes 3 and 9). Addition of plasmid DNA increased monoubiquitylation of PCNA concurrently with the induction of xFANCD2-L (Lanes 4 and 10). As shown in Figure 4c, addition of DDN to DNA-containing extracts completely abolished xFANCD2-L formation as expected, but did not inhibit PCNA monoubiquitylation (Lanes 6 and 12), suggesting that DDN is not a general ubiquitylation inhibitor.
DDN inhibits the FA pathway in human cells
We next determined if DDN has a similar effect on FANCD2-L formation in human cells. HeLa cells were treated with increasing concentrations of DDN in the presence of 2 mM hydroxyurea (HU), which induces stalling of replication forks and is known to stimulate the FA pathway in human cells.17, 26 Immunoblot analysis showed that DDN inhibits HU-induced FANCD2-L formation in a dose-dependent manner with an IC50 of 7.3 μM (Fig. 5a). Since the FA pathway-dependent activation of FANCD2-L is cell-cycle dependent,27 we tested whether the effect of 15 μM DDN on HU-induced FANCD2-L correlated with differences in cell cycle profile. FACS analysis of propidium iodide (PI)-stained cells indicated that HU treatment strongly decreased the population of G2/M cells (Fig. 5b), consistent with the fact that HU efficiently blocks replication.28 Cells treated with DDN accumulated in S and G2/M phase. However, when DDN was combined with HU, the profile did not significantly change compared to that of cells treated with HU alone. This result suggests that DDN does not inhibit FANCD2-L via perturbation of the cell cycle.
To rule out increased cell death as a possible cause of FANCD2-L inhibition in DDN-treated cells, a survival assay was done using increasing concentrations of DDN in the presence of HU. As shown in Fig. 5c, the concentration of DDN required to inhibit FANCD2-L formation by 50% (IC50, 7.3 μM) had no significant effect on the survival of HeLa cells (DDN LD50: 45 μM).
Following DNA damage, monoubiquitylated FANCD2 is known to re-locate into nuclear DNA repair foci in an FA core complex-dependent manner.17 We therefore tested the effect of DDN on the formation of HU-induced nuclear FANCD2 foci in HeLa cells. DDN reduced the formation of HU-induced FANCD2 foci in a dose-dependent manner (Fig. 5d) within the same range of concentrations as the inhibition of FANCD2-L observed by immunoblot, with complete inhibition of HU-induced foci at 16 μM DDN.
DDN and cisplatin display a greater synergistic effect in FANCF-competent vs. FANCF-deficient cancer cells
Next, we sought to investigate if DDN is a specific inhibitor of the FA pathway by comparing its effect on a FA-deficient cell line and its complemented, wild-type-like counterpart in the presence of cisplatin. If DDN specifically targets the FA pathway, it should sensitize FA-competent cells to DNA crosslinking agents like cisplatin while FA-deficient cells should remain unaffected. To test this hypothesis, we used an ovarian carcinoma FANCF-deficient cell line, 2008, and its complemented counterpart 2008+F.6 As expected, 2008+F cells were 4 times more resistant to cisplatin treatment compared to 2008 cells, with an LD50 of 5.7 μg/ml and 1.5 μg/ml, respectively (Fig. 6a). 2008+F cells were also slightly more resistant (1.5-fold) to DDN treatment than 2008 cells (Fig. 6b), possibly indicating that the FA pathway partially guards cells against DDN toxicity by an unidentified mechanism.
We next treated both cell lines with various combinations of DDN and cisplatin and determined the degree of synergism between these 2 drugs using an isobologram approach.21 An isobologram is the “gold standard” for evaluating drug interactions.29 It is defined as “a graph of equally effective dose pairs (isoboles) for a single effect level. Specifically, a particular effect level is selected, such as 50% of the maximum, and doses of drug A and drug B (each alone) that give this effect are plotted as axial points in a Cartesian plot. The straight line connecting A and B is the locus of points (dose pairs) that will produce this effect in a simply additive combination.”21 Points located below the additivity line denote lesser quantity of each drug necessary to reach the same effect (synergistic effect). Conversely, points above the line denote antagonism between the 2 drugs.
We plotted each pair of DDN/cisplatin concentrations resulting in 50% decrease of cell viability (Fig. 6c, LD50 isobologram). To directly compare cell lines, concentrations were expressed as a fraction of the LD50 values of DDN and cisplatin alone, which were set to a value of 1. Fit curves were calculated for both isoboles to highlight their divergence. As shown in Figure 6c, DDN and cisplatin displayed a synergistic effect in both 2008 and 2008+F cell lines (most data points are below the additivity line, p < 0.0001). The synergistic effect of DDN and cisplatin in the FA-deficient cell line suggests that DDN affects other targets in addition of the FA pathway. More detailed analysis revealed that the synergism is strongest at low cisplatin concentrations (for 0 < cisplatin < 0.33, p = 0.0002 for 2008 and p < 0.0001 for 2008+F). The drugs continued to show a synergistic effect as the concentrations of cisplatin increases (for 0.33 < cisplatin < 0.66, significant association for 2008, p = 0.0156, and marginally significant association for 2008+F, p = 0.0625). At the highest concentrations of cisplatin (0.66–1), synergism was marginally significant for 2008 (p = 0.0742) and not significant for 2008+F (p = 0.6875). Interestingly, the median total dose (DDN plus cisplatin) required to achieve LD50 for 2008+F was significantly less than that for 2008 (p = 0.0461). This result is consistent with the idea that DDN further sensitized the wild type-like 2008+F cells, but not the FA-pathway-deficient 2008 cells, to cisplatin by selectively inhibiting the FA pathway.
In this article, we describe the first cell-free assay that allows the screening of small molecule for inhibitors of the FA pathway. Initial testing of the assay using curcumin, a compound known as a FA pathway inhibitor in human cells,3 confirmed that the activity of cellular inhibitors of the FA pathway are mirrored in the Xenopus cell-free assay. The requirement to use higher compound concentrations in the Xenopus system (Figs. 3a and 3b) is likely due to high protein and lipid concentrations in egg extracts, resulting in nonspecific sequestration of small molecules.23 We were also able to partially assess the specificity of these compounds toward the FA pathway using Mre11-PPP as a control. For example, wortmannin displayed much greater inhibition of Mre11-PPP than xFANCD2-L, suggesting that the low activity of wortmannin toward FANCD2 is the result of a nonspecific effect. Wortmannin is a potent inhibitor of various kinases including phosphoinositide 3-kinases (PI3Ks) such as ATM and ATR, PI3K-related enzymes (mTOR and DNA-PK) and polo-like kinases, which probably explain its robust inhibitory effect on Mre11 phosphorylation.
Our screening of 772 compounds from 4 chemical libraries revealed a total of 29 compounds with inhibitory effect on DNA-stimulated xFANCD2-L formation. Interestingly, we found a larger ratio of inhibitory compounds in the Cancer Plate library (10%) compared to the other 3 compound libraries (1.7, 3.4 and 3%), suggesting that targeted libraries composed of anti-tumor compounds might have a higher chance of identifying new FA pathway inhibitors.
From the Cancer Plate library, we identified DDN, a novel inhibitor of the FA/BRCA pathway. DDN displayed a much stronger inhibitory activity toward FANCD2-L than curcumin and wortmannin (IC75: 14 μM vs. 596 μM and >1600 μM, respectively). In addition, DDN inhibited Mre11-PPP at concentrations similar to FANCD2-L (IC50: 57 μM), suggesting that DDN affects both the FA and Mre11 pathways. Interestingly, the Mre11 complex (MRN) has been shown to be at least partially dependent on a functional FA pathway in human cells.30 Moreover, we recently demonstrated that in response to dsDNA fragments, Mre11-PPP formation is partially dependent on the presence of a functional FA core complex.11 Thus, the inhibition of Mre11-PPP by DDN might be due to its inhibitory effect on the FA core complex.
Further studies showed that DDN has similar inhibitory activity against FANCD2-L in human cells as monitored in HeLa cells by DNA damage-induced FANCD2-L formation and re-localization into nuclear foci. Similar to the effect of curcumin treatment, IC50 were lower in cells than in extracts (7.3 μM vs. 30 μM). Another approach to test whether DDN specifically targets the FA pathway relates to its potential for re-sensitization of cisplatin-resistant tumor cells. Tumors arising from FA-deficient cells are initially exquisitely sensitive to cisplatin but resistance usually occurs within months. Taniguchi et al.6 demonstrated that one mechanism leading to cisplatin resistance involves reactivation of the FA pathway. In this situation, inhibition of the FA pathway is predicted to sensitize FA-competent cells to cisplatin without effect on FA-deficient cells. To test this hypothesis, we treated two isogenic cell lines with various combinations of DDN and cisplatin: 2008 is an ovarian carcinoma cells defective in the FA pathway due to the hypermethylation/silencing of the FANCF promoter whereas 2008+F is its FANCF-complemented, wild type-like counterpart.6 Isobologram analysis showed that DDN synergizes with cisplatin in both cell lines, suggesting that, in addition to FA, DDN has other targets that enhance the effect of cisplatin. The major DNA adducts induced by cisplatin are 1,2- intrastrand crosslink (85–90% of all adducts), which are essentially removed by the NER pathway.31 It is therefore possible that DDN inhibits NER in addition to the FA pathway, a hypothesis that remains to be tested. Thus, DDN might be a candidate lead compound for the generation of a cisplatin sensitizer drug regardless of the activity of the FA pathway in the tumor.
Interestingly, the median total dose of DDN+cisplatin sufficient to kill 50% of the cells was significantly lower in 2008+F cells than in the noncomplemented 2008 cells (Fig. 6c and results), suggesting that by inhibiting the FA pathway, DDN sensitizes the FA competent 2008+F cells to cisplatin to a greater extent than the FA-deficient 2008 cells.
Future studies will address the molecular mechanism of FA pathway inhibition by DDN. As a first step, we showed that the FA core complex is not disrupted in the presence of DDN. Instead, the mode of action of DDN might involve interfering with FA core complex function such as E3 ligase activity, blocking function(s) of non-FA proteins associated with the FA core complex (e.g. members of the BRAFT complex),32 or inhibiting other proteins that act upstream of FANCD2. For example, ATM and ATR have both been demonstrated to be involved in the regulation of the FA pathway in human cells.24, 33 However, several observations suggest that FANCD2 ubiquitylation is not subject to such regulation in our cell-free assay. First, depletion of the obligate ATR partner, ATRIP, inactivates ATR34 but does not affect formation of xFANCD2-L.11 Second, caffeine (a PI-3 kinase inhibitor) does not affect FANCD2 monoubiquitylation in extracts, but inhibits kinase activities of ATM and ATR.10 Third, KU55933, an ATM-specific inhibitor, has no detectable effect on xFANCD2 ubiquitylation in extracts (data not shown). In addition, DDN does not inhibit the plasmid-induced phosphorylation of H2AX, an ATM/ATR/DNA-PK target in the DNA damage response (Fig. 4b). Moreover, general phosphorylation modulators such as tautomycin (phosphatase inhibitor) and SAP (shrimp alkaline phosphatase) have no detectable effect on xFANCD2-L formation in egg extracts (data not shown), suggesting that xFANCD2 monoubiquitylation occurs independently of phosphorylation processes. We infer from these results that DDN does not inhibit xFANCD2-L through a phosphorylation-dependent mechanism. We also investigated whether DDN is a general ubiquitylation inhibitor by monitoring the monoubiquitylation of PCNA in Xenopus egg extracts.35, 36 Contrary to xFANCD2, PCNA monoubiquitylation was not inhibited by DDN, suggesting that DDN activity is restricted to FANCD2 and does not generally suppress ubiquitylation processes. DDN is a polar analog of menadione, a vitamin K derivative that induces growth arrest in G1 and apoptosis in various cancer cell lines. DDN itself has been shown to have anticancer properties in human promyeloid leukemic HL-60 cells and in a Sarcoma 180 tumor-bearing mouse model.20 Known DDN targets include cdc25A phosphatase, Bcl-2 and Akt, none of which have been functionally connected to the FA pathway.19, 20 Interestingly, menadione has been tested in combination with the DNA crosslinking agent MMC in phase II trials for treatment of lung and gastrointestinal cancers.37, 38 The rationale of these studies was that elevated levels of intracellular pools of reduced glutathione in cancer cells correlate with resistance to alkylating agents such as MMC, whereas menadione lowers reduced glutathione levels. Our data further suggest that DDN could be developed as a chemosensitizer of ICL-inducing agents in tumor cells in the presence or absence of a functional FA pathway. DDN, like most small molecule inhibitors, has off-target effects. On the basis of the literature, our expectation is that the screening larger chemical libraries (10,000 or more compounds) using high throughput methods will be required to find compounds with improved specificity for FANCD2-L inhibition.
Interestingly, we found that candidate inhibitors of the FA pathway identified in cell-free Xenopus extracts predict activity in human cells: a) curcumin, an FA inhibitor initially identified in cells, is active in extracts, b) DDN, initially identified using our cell-free assay, displayed a robust inhibitory effect toward the FA pathway in human cells, c) DDN showed off-target effects in both extracts (inhibition of xMRE11 phosphorylation) and in cells (synergy with cisplatin in FA-deficient cells).
Cell-free screening strategies using Xenopus egg extracts have been used previously to identify small molecule inhibitors of the cell cycle machinery.39, 40 Very recently, Dupre et al. developed a similar strategy to successfully identify modulators of histone H2AX phosphorylation, a major marker of the DNA damage response.23 The Xenopus egg extracts approach to identify modulators of the FA pathway, as presented here, provides a significant improvement relative to cell-based screens for several reasons. First, egg extracts are in a replication-competent state that mimics S phase, maximizing the sensitivity of detection of replication-associated DNA damage response events such as FANCD2 monoubiquitylation and Mre11 phosphorylation. The assay takes advantage of the full replication context but the stimulation of FANCD2-L is independent of replication, reducing confounding effects of proliferation that might occur in cell-based assays for inhibitors of FANCD2-L. Second, the simplicity of the system reduces the risks of nonspecific, off-target effects of the compounds screened: as extracts support neither plasmid replication41, 42 nor transcription,43 active compounds are likely to target direct protagonists of the FA pathway, rather than upstream replication or transcription events. Third, screening performed in egg extracts takes place in a fully soluble in vitro context, minimizing cell-based screening problems related to compound toxicity and bioavailability. Fourth, the system is cost-effective, simple to implement and rapid: egg extracts are readily available in large quantities, are easy to handle (e.g. they can be frozen for long periods before use), egg extracts contain abundant and easily detectable FA proteins and are robustly activated within minutes by simple addition of plasmid DNA. Fifth, monitoring the effect of candidate compounds on several DNA repair pathways (as shown here for xMre11, xH2AX and xPCNA) can give indications about its specificity toward the desired target. Sixth, this assay has the potential to identify not only negative, but also positive modulators of DNA repair pathways that might act to prevent de novo tumor formation. For example, our screen identified 2 compounds that induce rather than suppress xMre11-PPP formation (Table I and data not shown).
Using this assay, one can envision a systematic integrated approach in which each compound would be characterized by its effects on key proteins involved in different DNA repair pathways, as illustrated here by PCNA monoubiquitylation and Mre11 and H2AX phosphorylation. Such multidimensional analysis would also be very helpful to dissect the mechanism of action of the compound. Finally, the cell-free screen is easily amendable to high-throughput methods with ELISA or FRET-based detection. Taken together, the characteristics of this assay are particularly well suited for detection of chemical modulators of extended and functionally undefined protein networks that operate in S-phase, such as the FA/BRCA pathway.
The authors thank Dr. J. Gautier and Dr K.J. Patel for generously sharing antibodies for xMre11 and hFANCD2, respectively, Dr. H. Joenje for helpful suggestions and critical reading of the manuscript and Ms. C. Koudelka, MS for statistical analysis of the isobologram data. This work was supported by grants to Dr. M. Hoatlin [NIH, OHSU BioScience Innovation Fund and the Fanconi Anemia Research Fund (FARF)], Dr. A. Sobeck (American Heart Association) and Dr. I. Landais (NRSA postdoctoral training grant).