High-Throughput Screening Identifies Novel Agents Eliciting Hypersensitivity in Fanconi Pathway–Deficient Cancer Cells

There are at least 12 Fanconi anemia (FA) genes, biallelic mutations of which cause the rare cancer-susceptibility syndrome FA. These FA genes act in a common pathway involved in DNA repair, particularly the repair of DNA interstrand cross-links (ICL) and double-strand breaks (DSB; refs. 1, 2). Inactivation of the FA pathway by either genetic or epigenetic mechanisms also occurs in subgroups of diverse human cancers among the general population (3). In pancreatic cancer, inactivating mutations are identified in BRCA2/FANCD1, FANCC, and FANCG (4–8). Tumor cells having FA pathway inactivation are hypersensitive to ICL agents in vitro and in vivo (5, 9). The extent of hypersensitivity depends on the specific ICL agent and seems to reflect the proportional contribution of ICLs to the overall toxicity of the drug (10). FA pathway–deficient cells are also hypersensitive to some inhibitors of poly(ADP-ribose) polymerase (11–15). Evidence differs as to whether these cells exhibit increased sensitivity to ionizing radiation (10, 16). The identification of novel agents to which FA pathway–deficient cells were hypersensitive could contribute to a better understanding of the biochemical mechanisms that underlie the FA–related chemical hypersensitivity phenotype and could also provide additional therapeutic opportunities. The discovery of agents that confer increased sensitivity via a different mechanism of action than the known ICL agents might also be useful for the development of combination therapies that produce synergistic effects (17).

Using somatic cell gene targeting in the human cancer cell line RKO, we recently engineered several clones that genetically differed only in the presence or absence of single FA genes (FANCC or FANCG). We investigated the effect of these gene defects on chromosomal instability and drug and irradiation sensitivity (10). We report here the application of this model to high-throughput screening (HTS), using four different clones against a panel of 880 active drugs and 40,000 diverse compounds. This model was ideally suited for HTS due to the availability of appropriate control cells (18) and the ability to simultaneously screen multiple clones, facilitating the rapid identification of compounds that specifically affected the growth of cells harboring FA gene defects.

Cell lines, drug, and compound libraries. Parental RKO cells, clones harboring an engineered disruption of FANCC or FANCG, and clones stably transfected with a p53-reporter construct, are described (10, 19, 20). The compound stocks, comprising 880 known active drugs (Prestwick library, Prestwick, Washington, DC) and 40,000 diverse compounds (ChemDiv, San Diego, CA), were maintained in 100% DMSO at −20°C.

Primary (high-throughput) screening and libraries. To validate our HTS approach, the assay was designed to yield a z factor of >0.5 (21). All screening steps were done using a Biomek FX Laboratory Automation Workstation (Beckman Coulter, Fullerton, CA), as previously described (22). Briefly, we used each compound at a final concentration of 2.5 μg/mL in 0.5% DMSO. Two FA pathway–proficient and two FA pathway–deficient lines were employed (parental and FANCG^+/− cells versus FANCG^−/− and FANCC^−/−/− cells). The appropriate number of cells to yield ∼80% confluence after 5 days was determined individually and was similar for all clones (300–350 per well). Cells were plated in 20 μL of complete medium in 384-well tissue culture plates and allowed to settle overnight. Twenty-four hours after seeding, compounds were serially diluted with complete medium to 5 μg/mL; 20 μL of this dilution was added to each well containing 20 μL of medium and cells to yield the final screening concentration. In each 384-well plate, 80 compounds and 16 DMSO (no drug) controls were tested against each of the four cell clones. After 4 days, 40 μL of lysis/detection solution containing 1.2% Igepal CA-630 (Sigma Chemical Co., St. Louis, MO) and SYBR Green I (1:1,000; Molecular Probes, Eugene, OR) were added. Subsequently, fluorescence was measured per well. The data were exported to a custom program that determined growth inhibition by dividing each treated value by the average of the 16 control values. Compounds were scored as “hits,” when they preferentially inhibited the growth of both FA pathway–deficient lines (FANCG^−/− and FANCC^−/−/−) compared with both FA pathway–proficient lines (parental cells and FANCG^+/−). For compounds to be considered for follow-up secondary screening, the average of growth inhibition of the FA pathway–deficient lines divided by the average of growth inhibition of the control lines had to be <0.7 (growth inhibition ratio), and the growth inhibition differences between FA pathway–deficient lines or between FA pathway–proficient lines, respectively, had to be <0.3.

Secondary screening. Compounds found to specifically inhibit the growth of the FA pathway–deficient clones were repeated for confirmation, using standard cell proliferation assays over a broad range of concentrations (covering 100% to 0% survival). Cells (1,500–1,800 per well) were plated in 96-well plates, allowed to adhere, and subsequently treated. After 6 days, the cells were washed and lysed in 100 μL H₂O, and 0.5% Picogreen (Molecular Probes) was added. Fluorescence was measured, and growth inhibition was calculated compared with the untreated samples. At least four independent experiments were done per compound, with each data point reflecting triplicate wells.

Clonogenic survival assays. RKO cells and isogenic derivatives were seeded into six-well plates at multiple concentrations, allowed to attach overnight, and treated in duplicate with 80136342. After continuous exposure for 13 days, the cells were washed and stained with crystal violet. All visible colonies were counted.

Chromosome breakage analysis. FANCC^−/−/− cells were treated with 80136342 at 2.5 μmol/L (representing 80% cell survival) or 10 μmol/L (representing 20% cell survival) for 24 to 72 h. After treatment with colcemid (Sigma, St. Louis, MO) at 0.01 μg/mL for 4 h, the cells were harvested, and chromosome breakage analysis was done by our cytogenetics core facility. Fifty metaphases were assessed per sample for breakage and other structural abnormalities (i.e., rings or radials).

Cell cycle analysis. Cells were treated with 80136342 at multiple doses for 48 h, subsequently washed, fixed in PBS/3.7% formaldehyde/0.5% NP40 (U.S. Biochemical, Cleveland, OH), stained with bisbenzimide (Hoechst 33258, Sigma), and analyzed by flow cytometry. Ten thousand events were acquired per sample. The data were processed using Cell Quest software (Becton Dickinson, Franklin Lakes, NJ).

DNA ICL assays. The compounds were diluted in DMSO to a final concentration of 20 mmol/L. Plasmid DNA (pSP73, 2,464 bp) was linearized with EcoRI, purified, and 3′ end-labeled with α-[³²P]ATP using Klenow DNA polymerase. Labeled plasmid (100 ng) was incubated in the presence of 0.01 mol/L NaClO₄ with the compounds at various concentrations for 24 h at 37°C and subsequently separated using 1% agarose gels. Cisplatin at 3.4 μmol/L served as positive control. As potential ICLs could be labile either at high temperatures or under alkaline conditions, DNA denaturation was done both by gel electrophoresis under denaturing conditions or by heating the samples to 90°C for 5 min before neutral gel electrophoresis. The same experiments were done using conditions in which the tested compounds were incubated with DNA either in 10 mmol/L phosphate buffer (pH 7.5) or in Tris-HCl (pH 7.4). Because mitomycin C (MMC) requires reductive activation to cause the formation of ICLs in isolated DNA, additional experiments were done as previously described (23). Briefly, labeled plasmid was incubated under anaerobic conditions with the test compounds at various concentrations, using Na₂S₂O₄ in 10 mol/L excess as a reducing agent. MMC at 10 to 100 μmol/L was used as a positive control.

H2AX phosphorylation assays. Cells at ∼70% confluency were continuously exposed to 80136342 at 10 μmol/L for 24 h. Ionizing radiation at 10 Gy served as a positive control. Treated cells were subsequently fixed in PBS/4% paraformaldehyde, washed, and fixed in −20°C methanol. Following permeabilization in TBS/0.5% Triton X-100 and blocking in TBS/2% bovine serum albumin/0.5% Triton X-100, the cells were incubated with a primary monoclonal mouse anti-human γ-H2AX antibody (1:200; Upstate, Waltham, MA) for 2 h. The cells were washed and incubated with an Alexa 488 goat anti-mouse IgG secondary antibody (Molecular Probes) for 1.5 h. After washing, nuclei were counterstained with Hoechst 33258 (Sigma) at 10 μg/mL. Slides were then mounted and analyzed. Exposure time and software settings (Metamorph 4.6, Universal Imaging, Downingtown, PA) were kept constant for all samples within each experiment.

Mitotic index assays. FANCC^−/−/− cells were continuously exposed to 80136342 at 20 and 40 μmol/L for 48 h or left untreated. Subsequently, all cells, including dislodged cells, were harvested, fixed in PBS/3.7% formaldehyde/0.5% NP40, and stained with bisbenzimide. The mitotic index was determined as described (24), analyzing at least 200 cells per sample.

p53 reporter assays. The p53 reporter cell line has been described (19, 20). Cells at ∼70% confluence were continuously exposed to 80136342 at 10 and 20 μmol/L for 24 h. Etoposide at 50 μmol/L served as positive control. Subsequently, luciferase substrate (Steady-Glo, Promega, Madison, WI) was added, and light emission was measured. Relative light emission was calculated compared with the untreated samples.

Western blotting. Protein lysates from 200,000 cells, treated for 24 h, were separated on 3% to 8% Tris-acetate gels for 165 min at 150 V and transferred to polyvinylidene difluoride membranes. After blocking, the membranes were incubated overnight with a monoclonal mouse anti-human FANCD2 antibody (1:1,000; Santa Cruz Biotechnology, Santa Cruz, CA). Membranes were washed and probed with a secondary goat anti-mouse antibody (1:10,000; Pierce, Rockford, IL). Detection was done using SuperSignal West Pico Substrate (Pierce).

Two FA pathway–proficient (parental RKO and FANCG^+/−) and two FA pathway–deficient (FANCG^−/− and FANCC^−/−/−) cell lines having similar growth rates were employed to limit clonal variability often associated with HTS and to enhance the specificity of the primary screen. The clones did not systematically differ in growth inhibition when tested with the majority of compounds, ruling out a general hypersensitivity phenotype of the FA pathway–deficient cells. Screening of 880 active drugs from the Prestwick library yielded one hit, carmustine (growth inhibition ratio = 0.29), a known ICL agent. Screening of 40,000 diverse compounds generated 90 primary hits (0.2%) according to the criteria described in Materials and Methods.

The 90 compounds found to specifically inhibit the growth of the FA pathway–deficient clones were repeated for confirmation using standard cell proliferation assays. Compounds were tested over a broad range of concentrations, covering 100% to 0% cell survival. Twenty-one of the 90 compounds (23%) were confirmed in multiple experiments. Except for one compound (termed 80136342), FA pathway–deficient cells were modestly hypersensitive (IC₅₀ ratios ≤2.5) to the identified compounds over a narrow concentration range. Of note, we discovered two groups (termed C499xxxx and 3527xxxx) of structurally related compounds that elicited similar hypersensitivities (Fig. 1). Compared with these compound families, 80136342 conferred a stronger hypersensitivity over a wider concentration range (Fig. 2A). This effect was more pronounced in FANCC-deficient cells than in FANCG-deficient cells (Fig. 2B), as confirmed by employing several independently derived clones (parental RKO cells, one FANCC^+/−/−, one FANCG^+/−, two FANCC^−/−/−, and two FANCG^−/− clones; Fig. 2C). Interestingly, a similar compound from the applied compound library, termed 80136341 and differing from 80136342 only in the absence of one particular methyl group (Fig. 2A, arrow), did not confer hypersensitivity in our model (data not shown).

80136342, when employed in long-term colony formation assays, elicited an increased sensitivity in FA pathway–deficient cells similar to that observed in the short-term assays. However, the growth differences were observed at lower concentrations (Fig. 2B). Similar to the short-term proliferation assays, FANCC-deficient cells were more strongly affected than were FANCG-deficient clones.

Due to its pronounced effects, we further focused our interest on 80136342. To investigate whether this compound elicited increased sensitivity in FA pathway–deficient cells via a similar mechanism as do ICL agents or irradiation, we assessed chromosome breakage, H2AX phosphorylation, p53 activation, ICL formation, cell cycle effects, and FANCD2 monoubiquitination.

Chromosome breakage testing. FA pathway–deficient cells experience increased chromosomal aberrations upon treatment with ICL agents, a phenotype so stringently observed in nonmalignant FA cells that diepoxybutane testing serves as a diagnostic assay (25–27). Using our model, we previously reported drastically increased aberrations in FANCG^−/− and FANCC^−/−/− cells, including the formation of rings and radials, 48 h after treatment with ICL agents compared with parental and heterozygote control cells (10). In contrast, we did not observe increased chromosomal aberrations in parental or FANCC^−/−/− cells 24 or 48 h after treatment with 80136342 at 2.5 or 10 μmol/L, corresponding to 80% and 20% cell survival (≤3 breaks per 50 cells in every sample), respectively. Seventy-two hours after treatment, chromosomal breaks were slightly increased but did not differ significantly between parental cells (0 break in 44 cells, 1 break in 5 cells, and 2 breaks in 1 cell) and FANCC^−/−/− cells (0 break in 42 cells, 1 break in 6 cells, and 2 breaks in 2 cells). The characteristic formation of rings or radials, observed upon treatment with ICL agents, was not seen at any concentration of 80136342 at any time point.

H2AX phosporylation and p53-reporter assays. Phosphorylation of histone H2AX along with activation of p53 can serve as a surrogate marker for the induction of DSBs (19). 80136342 at 10 μmol/L did not induce phosphorylation of H2AX in parental, FANCG^−/−, or FANCC^−/−/− cells, whereas irradiated control cells had a pronounced H2AX phosphorylation (Fig. 3A). Likewise, 80136342 at 10 or 20 μmol/L did not activate p53 in RKO cells stably transfected with a p53-reporter construct, whereas etoposide-treated control cells activated p53 (Fig. 3B). Thus, there was no evidence for the induction of DSBs by 80136342.

DNA ICL assays. In contrast to cisplatin and MMC, neither 80136342 nor compounds in the families C499xxx and 3527xxxx induced ICLs in vitro. To ensure the detection of potential labile ICLs, these compounds were tested under various conditions (alkaline/neutral/reducing conditions during treatment, alkaline/heat denaturation, and alkaline/neutral gel electrophoresis: Fig. 3C; data not shown).

Cell cycle analyses. 80136342 at 10, 20, and 40 μmol/L induced an incremental G₁ arrest in both FA pathway–proficient and FA pathway–deficient cells, as determined by increasing G₁-S-ratios. FA pathway–deficient cells had a more accentuated G₂ arrest at low concentrations of 80136342 than FA pathway–proficient cells. Consistent with the cell survival data, the G₂ arrest was more pronounced in the FANCC^−/−/− cells (i.e., more cells in G₂, observable at concentrations ≥10 μmol/L) than in the FANCG^−/− cells (i.e., fewer cells in G₂, observable at concentrations ≥20 μmol/L). At high concentrations, a late S and G₂ arrest was universally observed in all cells independent of genotype (Fig. 4A and B). We next did mitotic index assays using FANCC^−/−/− cells to distinguish between cells in G₂ and cells in mitotic arrest or cells undergoing mitotic catastrophe. In contrast to the increased fraction of FANCC^−/−/− cells in G₂-M observed with cell cycle analysis, there was a decrease in mitotic cells upon treatment with 80136342 (2.6% mitotic cells without treatment, 0.4% mitotic cells after treatment with 80136342 at 20 μmol/L, and 0.0% mitotic cells after treatment with 80136342 at 40 μmol/L). Therefore, the increased G₂-M population was caused neither by cells in mitotic arrest nor by cells headed for mitotic catastrophe; it thus represents a pre-M phase arrest.

FANCD2 monoubiquitination. FANCD2 monoubiquitination is defective in FA pathway–deficient cells and was therefore assessed in parental and FANCG^+/− cells. 80136342 at 40 μmol/L abrogated FANCD2 monoubiquitination. At 80 μmol/L, some monoubiquitinated FANCD2 was detected (Fig. 4B). These data were consistent with our cell cycle analyses, as FANCD2 monoubiquitination is known to be associated with normal S phase progression (28). Thus, whereas treatment with ICL agents enhances FANCD2 monoubiquitination in parental and FANCG^+/− cells (ref. 10; despite a G₂ arrest), treatment with 80136342 did not.

We tested the combinatorial effects of 80136342 and known ICL agents on cell growth (Fig. 5) by treating the cells simultaneously with both 80136342 and MMC. 80136342 was added either at a fixed concentration to varying MMC concentrations, or the concentration ratio was kept constant for 80136342 and MMC. In the first scenario, a number of fixed concentrations of 80136342 increased the toxicity of MMC. Similar effects were observed when combining the two ICL agents MMC and melphalan in this manner (Fig. 5A; data not shown). In the second scenario, adding 80136342 at a fixed ratio to MMC, we observed a shift of the pharmacogenomic window towards lower concentrations of MMC. Again, similar effects were observed when combining MMC and melphalan in this manner (Fig. 5B). The fixed ratios of MMC and 80136342 or MMC and melphalan, respectively, were chosen from the ratio of the lowest dose of each agent that killed all cells independent of genotype.

FA pathway–deficient cells display a robust hypersensitivity towards certain agents, rendering FA gene defects as promising targets for an individualized, genotype-specific anticancer therapy. We applied our recently developed human cancer cell model of FA gene defects to HTS, using a panel of 880 active drugs and 40,000 diverse compounds, in an effort to identify novel agents to which FA pathway–deficient cells would be hypersensitive.

When treated with 880 active drugs from the Prestwick library, the FA pathway–deficient cells displayed a highly restricted hypersensitivity towards ICL agents, as expected. We only identified one hit, the ICL agent carmustine. The growth-inhibitory effects of this drug on FANCC- and FANCG-deficient cancer cells have been previously characterized in our model (IC₅₀ ratio = 3.5; ref. 10). When treated with 40,000 compounds, the FA pathway–deficient clones did not systematically differ from the FA pathway–proficient clones (hit rate = 0.2%), ruling out a phenotype of general chemical hypersensitivity. Thus, cancer cells harboring an engineered disruption of the proximal FA genes FANCC or FANCG seem not to express a considerable reduction of fitness associated with an increased drug sensitivity as do some BRCA2-deficient cancer cells (17).⁴

We identified two structurally related compound families to which FA pathway–deficient cells were more sensitive than control cells (IC₅₀ ratios ≤ 2.5). The identical structural parts of these compounds suggested the presence of a pharmacophore. Future studies investigating the structure-activity relationships (SAR) will be a critical step in defining the underlying mechanism for the increased sensitivity of FA pathway–deficient cells to these compounds. We also identified a compound, termed 80136342, to which FA pathway–deficient cells were more sensitive than control cells over a wider concentration range, and which conveyed greater toxicity on FANCC-deficient than on FANCG-deficient cells. Initial SAR studies suggested that the increased sensitivity of FA pathway–deficient cells to 80136342 was dependent on a particular methyl group, as no increased sensitivity was observed upon treatment with a similar compound, differing from the former only by the absence of this methyl group.

The mechanism of action of 80136342 was distinct from that of ICL agents or irradiation. In contrast to drastically increased chromosomal aberrations in FANCG^−/− and FANCC^−/−/− cells, including the formation of rings and radials, upon treatment with ICL agents (10), we found only a slight increase in chromosomal breaks, but no rings or radials, upon treatment with 80136342, aberrations that were observed similarly in FA pathway–proficient and FA pathway–deficient cells. Likewise, whereas treatment with ICL agents typically enhances FANCD2 monoubiquitination in FA pathway–proficient cells (10, 29), treatment with 80136342 did not. Treatment with 80136342 also did not induce DSBs, as shown by a lack of H2AX phosphorylation and p53 activation. Finally, we found no evidence that 80136342 would form ICLs in our assays. Taken together, the assessment of chromosomal aberrations, FANCD2 monoubiquitination, H2AX phosphorylation, p53 activation, and ICL induction upon treatment with 80136342 provided no evidence of having induced DNA damage. Thus, the increased sensitivity of FA pathway–deficient cells to 80136342 seemed to be mechanistically distinct from the hypersensitivity to ICL agents or irradiation and could not be attributed to a cellular impairment in ICL or DSB repair. On the other hand, 80136342 mimicked the hypersensitivity phenotype of ICL agents with regard to cell cycle dysregulation (10, 30–32). FANCC^−/−/− cells and, to a lesser extent, FANCG^−/− cells had a more accentuated G₂ (but not M) phase arrest at low concentrations of 80136342 than FA pathway–proficient cells, suggesting that the commonly observed G₂ cell cycle abnormalities of FA pathway–deficient cells may not exclusively reflect the cellular response to DNA damage but might occur upon non-genotoxic influences as well.

Because our findings suggested a unique growth inhibitory mechanism of 80136342, we tested whether this compound would confer additional toxicity when combined with ICL agents. Adding 80136342 either at a fixed concentration or at a fixed ratio to variable concentrations of MMC confirmed an additional toxicity in either setting and excluded possible interferences of 80136342 on ICL agent–induced toxicity. Similar effects were observed when combining two ICL agents that have strong effects on FA pathway–deficient cells (MMC and melphalan). These results suggested that the combinational use of novel non-ICL agents with ICL agents deserves further investigation and indicated a potential application of such therapies to reduce the known harmful side effects of ICL agents when used alone.

Finally, we investigated whether the observed hypersensitivity of FA pathway–deficient cells could be generalized to a human cancer FA model in which defective FA gene function was complemented by exogenous overexpression of the respective gene. We therefore compared the pancreatic cancer cell lines HS766T and PL11, which harbor deleterious mutations in FANCG and FANCC, respectively, with their complemented counterparts overexpressing the respective gene (4, 5). Preliminary experiments showed that the FA–defective cell lines appeared only slightly yet not significantly more sensitive towards treatment with 80136342 than the complemented cells (IC₅₀ ratios ≤ 1.5). Using ICL agents, we reported observing a pharmacogenomic window considerably larger in the knockout model than in the overexpression model (e.g., 14-fold versus 5- to 6-fold for melphalan and 12- to 13-fold versus 8- to 10-fold for MMC; refs. 10, 33). These differences between the two models are likely due to the newer model having avoided the artifacts such as off-target effects and squelching produced by gene overexpression in the older model, illustrating that the compounds identified here using the knockout system could not readily have been found using a less stringent model.

In conclusion, using HTS in a well-controlled isogenic human FA cancer model, we showed a novel approach to identify agents to which FA pathway–deficient cells are hypersensitive. We discovered two groups of compounds eliciting hypersensitivity in a structure-specific manner and a single compound eliciting hypersensitivity via a yet unknown mechanism distinct from that of ICL agents or irradiation. These compounds represent attractive candidates to serve as lead compounds for further development, eventually aiming at an individualized, genotype-based therapy of FA pathway–deficient tumors (17).

Grant support: NIH grant CA62924, Deutsche Forschungsgemeinschaft grant DFG GA762/1-1 (E. Gallmeier), and Bavarian Academy of Sciences (E. Gallmeier).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank L.A. Morsberger (Molecular Pathology Cytogenetics Laboratory at Johns Hopkins) for the cytogenetic analyses and C. Rago, S.C. Cunningham, E.S. Calhoun, J.M. Winter, S.U. Khan, and P.A. Cole for helpful suggestions.