Cell-based screening for small molecules that inhibit the FA pathway

Assembly of DNA damage-induced FANCD2 foci is a widely used indicator of upstream FA pathway integrity [8]. To identify novel small molecules that inhibit the FA pathway, PD20-EGFP-FANCD2 cells [13] were treated with chemical libraries and exposed to IR (15 Gy) to induce FANCD2 foci formation. A significant decrease in the proportion of cells with IR-induced EGFP-FANCD2 foci upon drug treatment was scored as positive (Figure 1A). Using this cell-based assay, we tested more than 16,000 chemicals, and identified 43 compounds (0.27%) that significantly reduced EGFP-FANCD2 foci formation in the initial screen ( Additional file 1: Table S1), including curcumin, wortmannin, alsterpaullone and H-9, as previously described [13]. Fifteen of these 43 compounds were then confirmed to inhibit IR-induced FANCD2 foci formation in multiple cell lines, including PD20-FANCD2, U2OS, HeLa and TOV21G + FANCF ovarian cancer cells, using a wide range of drug concentrations (Table 1, Figures1B and 1C, Additional file 2: Figure S1, and data not shown). Interestingly, some of the drugs independently identified through this screen shared common inhibitory features (Table 1): curcumin [17] and compound 5929407 (see below) are proteasome inhibitors, and curcumin, H-9, and Gö6976 are PKC inhibitors.

Chemicals	Primary known functions	IC50 (μM)
Lactacystin	Proteasome inhibitor	1.7
MG132	Proteasome and calpain inhibitor	0.8
ALLN	Proteasome, calpain, cathepsin L inhibitor	9.2
5929407	Proteasome inhibitor (this study)	6.5
Curcumin	Proteasome, protein kinase C (PKC), EGF-receptor tyrosine kinase, IkappaB kinase and mTOR inhibitor	18.6
H-9	PKA, PKC, PKG inhibitor	34.2
Gö6976	PKC, CHK1 inhibitor	> 50
SB218078	CHK1, Cdc2, PKC inhibitor	1.4
UCN-01	PKC, CHK1, CDK, AKT inhibitor	0.14
Alsterpaullone	CDK, GSK3 inhibitor	1.6
Roscovitine	CDK, ERK1,2 inhibitor	27.5
Geldanamycin	HSP90 inhibitor	0.15
17-AAG	HSP90 inhibitor	0.6
CA-074-Me	CathepsinB inhibitor	16.7
Chloroquine	Lysosome permeabilisation, inhibition of drug efflux pumps (ATP-binding cassette transporters)	67.5
Wortmannin	Casein kinase II, phosphatidylinositol 3-kinase (PI 3-kinase), polo-like kinase 1 (PLK1) inhibitor	84.4
DRB	Casein kinase II, RNA pol II inhibitor	11
HNMPA-(AM)3	Insulin receptor tyrosine kinase inhibitor	4.7
Puromycin	Protein synthesis inhibitor	5.2
TPEN	Heavy metal chelator	5
5656325	Unknown	0.83
5315179	Unknown	5.9
7012246	Unknown	7
5195243	Unknown	2.2
5373662	Unknown	1.3

The identified chemicals are presented with their major known functions and the dose required to inhibit 50% of IR-induced EGFP-FANCD2 foci formation (IC50) in PD20-EGFP-FANCD2 cells. Chemical structures of 5195243 (2-nitro-9,10-phenanthrenedione 10-oxime), 5373662 (2-(2-pyrinylmethylene)quinuclidin-3-one), 5656325 (1-(2-methoxy-1-naphtyl)-3-(3-pyridinyl)-2-propen-1-one), 5929407 (5-[4-(dietylamino)benzylidene]-3-{[(3-methylphenyl)amino]methyl}-2-thioxo-1,3-thiazolidin-4-one), 5315179 (1-(2,4-dimethoxyphenyl)-3-(3-pyridinyl)-2-propen-1-one) and 7012246 (1-(5-bromo-2-hydroxyphenyl)-3-(3-pyridinyl)-2-propen-1-one)) are shown in Additional file 2: Figure S1.

Eleven additional compounds, related to the chemicals identified in our primary screen or identified in unrelated studies [18], were also subjected to secondary screening: two CHK1/PKC inhibitors (UCN-01, SB218078), a CDK inhibitor (roscovitine), an HSP90 inhibitor (17-AAG), four proteasome inhibitors (bortezomib, MG132, ALLN, lactacystin), two compounds structurally related to 5656325 (5315179 and 7012246 ( Additional file 2: Figure S1)), and chloroquine. All of these compounds inhibited DNA damage-induced FANCD2 foci assembly in multiple cell lines, without altering the overall expression of EGFP-FANCD2 or endogenous FANCD2 (Figure 1B Additional file 3: Figure S5 and data not shown). The dose required to inhibit 50% of IR-induced EGFP-FANCD2 foci formation (IC₅₀) in PD20-EGFP-FANCD2 cells was determined for each of these 26 compounds (Table 1). Importantly, 18 of them exhibited IC₅₀ values lower than 10 μM. Although the FA pathway inhibition capacity of these inhibitors may not be due to specific targeting of components of the FA pathway, we will refer to them as FA pathway inhibitors in the remaining text for simplicity.

Identification of a novel proteasome inhibitor among the small molecules that inhibit the FA pathway

All proteasome inhibitors tested (bortezomib, MG132, ALLN, lactacystin, curcumin) inhibited FANCD2 foci formation in multiple cell lines (Table 1 and [18]). Therefore, we hypothesized that some of the newly identified FA pathway inhibitors could also inhibit the proteasome. We first tested proteasome activity using GFPu-1 cells, in which inhibition of proteasome results in increased GFP expression [19]. All proteasome inhibitors as well as the Chembridge compound 5929407 induced a strong increase in GFP expression in GFPu-1 cells (Figure 2A).

We then assessed the effects of these compounds on the three proteases activities associated with the proteasome (trypsin-like, chymotrypsin-like and caspase-like activities), using fluorogenic compounds in HeLa cells extracts. All compounds that increased GFP expression in GFPu-1 cells (bortezomib, MG132, lactacystin, ALLN, curcumin and 5929407) inhibited chymotrypsin- and caspase-like activities of the proteasome, the chymotrypsin-like activity being generally the most affected. In addition, lactacystin and curcumin inhibited trypsin-like activity (Figure 2B). These findings indicate that the compound 5929407 is a novel proteasome inhibitor that preferentially inhibits the chymotrypsin-like activity of the proteasome.

Most chemicals that inhibit the FA pathway inhibit homologous recombination

Since the FA pathway is required for efficient DNA double-strand break repair by HR [4, 20], we tested whether the identified compounds affect HR efficiency in human cells using the DR-GFP reporter system [21]. In this assay, GFP expression reflects the occurrence of an HR repair event; compounds that disrupt HR repair will decrease GFP expression. Twenty-four hours after transfection of a HA-tagged I-Sce1-encoding plasmid, U2OS-DR-GFP cells were exposed to the identified FA pathway inhibitors for 24 hours ( Additional file 4A and B: Figures S2A and B). The concentration used for each drug was optimized to induce minimal decrease in cell viability (more than 90% viable cells after 24 hours, Additional file 4C: figure S2C), and did not significantly affect HA-tagged I-Sce1 expression (averaged 36.3% of the population for all drugs in multiple experiments compared to 39.2% in the non-treated population), with the exception of MG132, UCN-01, compounds 5195423 and 7012246 ( Additional file 4D: Figure S2D). The HA-positive population was analyzed for GFP expression. In the absence of drug, 9.5 ± 0.9% of the HA-positive cells expressed GFP.

With the exception of SB218078, HNMPA-(AM)₃ and TPEN, all FA pathway inhibitors significantly decreased HR (p ≤0.05) (Figure 3 and Additional file 4E and F: Figures S2E and S2F). No significant differences in the cell cycle distribution were observed under these conditions, except for UCN-01, 17-AAG, wortmannin, HNMPA-(AM)₃ and compounds 5929407 and 5315179 ( Additional file 5A: Figure S3A).

BRCA1 and RAD51 are required for efficient HR and are known to interact with FANCD2 [8, 22–24]. We therefore tested whether the FA pathway inhibitors block FANCD2, BRCA1 and RAD51 foci formation upon DNA damage in U2OS-DR-GFP cells. To do so, we used drug treatments identical to those used during the DR-GFP assay, consisting in longer exposure (24 hours) to lower concentrations of chemicals than the initial screen and confirmation experiments. Under such conditions, most of the drugs still significantly inhibited IR-induced foci formation of FANCD2 and RAD51 (Figure 3 and Additional file 6: Figure S4), without significantly modifying cell cycle distribution ( Additional file 5B: Figure S3B). The drugs that failed to significantly inhibit FANCD2 foci formation under these conditions demonstrated significant inhibition at higher dosage (data not shown), consistently with the initial screen. IR-induced monoubiquitination of FANCD2 was in most cases moderately inhibited or unaffected, with the exception of CA-074-Me, which strongly inhibited it ( Additional file 3: Figure S5). IR-induced foci formation of BRCA1 was also mildly affected or unaffected by the compounds (Figure 3 and Additional file 6: Figure S4). By using higher concentrations for a shorter time (8 hours drug exposure), we observed that most drugs significantly inhibited IR-induced FANCD2, RAD51 and BRCA1 foci formation, as well as IR-induced FANCD2 monoubiquitination, in the absence of significant variations of cell cycle distribution (data not shown). In addition, most of the drugs significantly inhibited cisplatin-induced FANCD2 foci formation in 24 hours co-treatment experiments ( Additional file 7: Figure S6). These results demonstrate that most FA pathway inhibitors inhibit HR processes (RAD51 foci formation and HR repair) in addition to FANCD2 foci formation, indicating that the identified chemicals target multiple steps of the DNA damage response pathway and are not specific for FA pathway inhibition. The lack of inhibition of FANCD2 monoubiquitination suggests that the FA pathway inhibitors may inhibit processes involved in the recruitment of proteins at sites of damage, rather than damage signaling upstream of FANCD2 monoubiquitination.

Identification of the compounds that synergize with cisplatin in ovarian cancer cells

Since the integrity of the FA pathway is critical for cellular resistance to ICL-inducing agents such as cisplatin, FA pathway inhibitors may sensitize tumor cells to cisplatin in an FA pathway-dependent manner. To test this hypothesis, we performed isobologram analyses on the ovarian cancer cell line 2008 [25], which is deficient in the FA pathway due to hypermethylation of the FANCF promoter, and on its isogenic, complemented FA pathway-proficient counterpart 2008 + FANCF cell line [11].

First, single agent survival curves were generated, and the dose producing a 50% reduction of cell survival (LD50, lethal dose 50%) was determined ( Additional file 8: Table S2). As previously reported [11], 2008 cells were significantly more sensitive to cisplatin than 2008 + FANCF cells. 2008 and 2008 + FANCF cells were equally sensitive to all FA pathway inhibitors, except for puromycin and geldanamycin. Higher tolerance of 2008 + FANCF cells to puromycin was likely due to the use of puromycin selection to generate the complemented cell line, and therefore, puromycin was omitted from further analysis. The reason for the differential sensitivity of 2008 and 2008 + FANCF cells toward geldanamycin remains unknown.

Next, isobolograms at the LD50 level were generated following the method previously described [26]. Survival assays were performed using the combination of 10 cisplatin concentrations with 6 concentrations of each FA pathway inhibitor. LD50/LD50₀ unit (drug concentration necessary to induce 50% death when combined with cisplatin divided by the drug’s LD50 in the absence of cisplatin) values of each FA pathway inhibitor were plotted against corresponding LD50/LD50₀ unit values of cisplatin. The distribution of points along the line connecting values of 1 corresponds to an additive effect of the two drugs while scattering below or above represents synergism and antagonism, respectively. In addition, combination index (CI) values were calculated according to Chou and Tallay’s method [26]; CI ≤ 0.90 indicates synergism.

Analyses performed at 50% killing revealed that 11 FA pathway inhibitors exhibited synergism with cisplatin in 2008 and/or 2008 + FANCF cells (Figure 4, Additional file 9: Figure S7, summarized in Table 2). Bortezomib, 17-AAG, CA-074-Me, and compounds 7012246 and 5373662 synergized with cisplatin in FA pathway-proficient 2008 + FANCF cells, but not in their isogenic FA pathway-deficient counterpart 2008, consistent with their FA pathway inhibition activity. Geldanamycin, three CHK1 inhibitors (Gö6976, UCN-01, SB218078) and chloroquine synergized with cisplatin in both 2008 and 2008 + FANCF cells. Geldanamycin, UCN-01 and SB218078 exhibited a significantly stronger synergism in 2008 + FANCF cells than in 2008 cells (p ≤ 0.01, unpaired t test), again consistent with their FA pathway inhibition activity. Finally, lactacystin weakly synergized with cisplatin in 2008 cells only. The other compounds had either additive or antagonistic effect with cisplatin.

Analyses performed at 70% killing (Additional file 10: Table S3) showed that bortezomib, 17-AAG, and CA-074-Me, in addition to geldanamycin, Gö6976, and UCN-01, synergized with cisplatin in both 2008 and 2008 + FANCF cells. ALLN, SB218078, and compounds 5656325, 5315179 and 5373662 synergized with cisplatin in 2008 + FANCF only.

Taken together, about half of the FA pathway inhibitors sensitized ovarian cancer cells to cisplatin (11 of 25 at 50% killing (Table 2) and 12 of 25 at 70% killing ( Additional file 10: Table S3)). Most of them exhibit a significantly stronger synergism with cisplatin in FA pathway-proficient 2008 + FANCF cells than in 2008 cells (CI significantly lower for 2008 + FANCF than 2008 (p ≤ 0.05) for 8 of 11 drugs at 50% killing, for 9 of 12 drugs at 70% killing), indicating that their effects are, at least partially, mediated through inhibition of the FA pathway.

We also examined whether the compounds that most significantly synergized with cisplatin would sensitize cells to IR ( Additional file 11: Figure S8 and Additional file 12: Table S4). Geldanamycin and SB218078 synergized with IR in both 2008 and 2008 + FANCF cells, Gö6976 in 2008, and compound 5373662 in 2008 + FANCF cells. Other compounds showed additive effects with IR.

Cell lines and culture conditions

HeLa, U2OS, TOV-21 G and GFPu-1 cells were purchased from the American Type Culture Collections (Manassas, VA). 2008 and FANCF-corrected 2008 + FANCF ovarian cancer cells, TOV-21 G and FANCF-corrected TOV-21 G + FANCF ovarian cancer cells were described previously [11]. FANCD2-deficient fibroblast line (PD20), PD20 corrected with wild-type FANCD2 (PD20-FANCD2) and enhanced green fluorescent protein (EGFP)-FANCD2 (PD20-EGFP-FANCD2 clone 7) were described previously [13]. U2OS-DR-GFP cells were a gift from Drs. Maria Jasin and Koji Nakanishi [49]. Cell lines were grown in DMEM supplemented with 10% fetal calf serum. Gamma irradiation was delivered using a JL Shepherd Mark I Cesium Irradiator (JL Shepherd & Associates).

The present research has been approved by the Institutional Review Board (IRB) Committee at the Fred Hutchinson Cancer Research Center (IRB Protocol file number 6023).

Chemicals

The chemical libraries (ICCB bioactives (489 compounds), Commercial Diversity Set 1 (5,056 compounds), Chembridge DiverSet Library (10,000 compounds) and NINDS II library (1,040 compounds) were used to identify inhibitors of the FA pathway. For subsequent studies, chemicals were purchased from Biomol (lactacystin, penitrem A, splitomicin, (±)-13-HODE), EMD biochemicals (actinomycin D, AG213, AG370, ALLN, alsterpaullone, α-amanitin, BAPTA-AM, bumetanide, CA-074-Me, curcumin, DRB, geldanamycin, Gö6976, HNMPA-(AM)₃, H-9, K-252c, MG132, nifedipine, propidium iodide, puromycin, roscovitine, SB218078, spermine NONOate, TPEN, trichostatin A, wortmannin), Cayman Chemical (leukotriene B₃), Chembridge Corporation (5195243, 5315179, 5373662, 5656325, 5929407, 7012246), Fisher (chloroquine), Millenium Pharmaceutical (bortezomib), MP (3-methyladenine), Sigma (cisplatin, UCN-01), Tocris (ochratoxin A), VWR (17-AAG).

Screen for small molecules that inhibit the FA pathway

The PD20-EGFP-FANCD2 clone 7 was used in the screen [13]. The cell-based screening of ICCB bioactives and Commercial Diversity Set 1 was done at the Institute for Chemistry and Cell Biology (Boston, MA) and a partial result was previously reported [13]. The cell-based screening of Chembridge DiverSet Library and NINDS II library was done at Fred Hutchinson Cancer Research Center. For this screening, duplicate 96-well plates were seeded with PD20-EGFP-FANCD2 clone 7 cells (10,000 cells per well). Chemical compounds from the library were added, five compounds per well, at a single concentration of 7.5 μmol/L. After a 12-hour incubation, cells were irradiated (15 Gy) and fixed for EGFP microscopy 12 hours later. Photomicrographs were obtained for each well and wells with significant (50%) reduction in percentage of EGFP-FANCD2 foci positive cells were identified by visual inspection. The five compounds of each well identified with reduced EGFP-FANCD2 foci were then individually tested.

Immunofluorescence microscopy was performed as described previously [18]. Antibodies against BRCA1 (D-9, Santa Cruz, 1/100), γH2AX (JBW301, Upstate, 1/1000), FANCD2 (NB 100–182, Novus Biologicals, 1/1000) and RAD51 (PC130, Calbiochem, 1/1000 or H-92, Santa Cruz, 1/200) were used. Species-specific fluorescein isothiocyanate (FITC)- or Cy3-conjugated secondary antibodies (Jackson Immunoresearch (West Grove, PA)) diluted in blocking buffer (1/2000) were incubated for 1 hour at room temperature. Images were acquired using a microscope (TE2000, Nikon, Tokyo, Japan) equipped with a 40x immersion objective (1.3NA) and a CCD camera (CoolSnap ES, Photometrics, Tucson, AZ) controlled by MetaVue (Universal Imaging, Downington, PA) and analyzed using MetaVue or ImageJ softwares. At least 100 cells per experimental point were scored for presence of foci, and each experiment repeated at least 3 times independently.

Flow cytometric analyses

Exponentially growing cells were plated in drug-free medium 48 hours before experiment. For proteasome activity assay [19], GFPu-1 cells were exposed to drugs at the indicated concentration for 24 hours, then analyzed for green fluorescent protein (GFP) expression. For cell cycle analyses, cells were exposed to drugs at the indicated concentration for 24 hours, and exposed to IR 8 hours before being pulse-labeled with 30 μM 5-Bromo-2’-deoxy-Uridine (BrdU (Sigma)) for 15 minutes, washed and fixed with 70% ice-cold ethanol. Cells were then stained for DNA content (propidium iodide, PI) and BrdU incorporation with anti-BrdU rat monoclonal antibody (MAS250, Harlan Sera Lab, UK) followed by FITC-conjugated goat anti-rat antibody (Jackson Immunoresearch). For HR assays, cells were transfected with pCBASce (a hemagglutinin (HA)-tagged I-Sce1 expression vector) [50] or the empty pCAGGS vector using TransIT transfection reagent (Mirus) following manufacturer recommendations. 24 hours after transfection, cells were treated with the indicated drugs at the indicated concentration for 24 hours. Cells were then fixed and stained for HA expression with mouse anti-HA antibody (HA.11, Convance, USA, 1/1000) followed by APC-conjugated donkey anti-mouse antibody (Jackson Immunoreserach). To specifically determine the proportion of HR events in I-Sce1 expressing cells, the percentage of GFP-positive cells among the HA-positive cell population was quantified. Flow cytometric analyses were performed on a Becton Dickinson FACScan. Fluorescence data were plotted using FlowJo (Tree Star, Inc., Ashland, OR). At least three independent experiments were carried out for each condition.

Proteasome activity fluorogenic assays were performed as in [51]. Briefly, HeLa cells were treated with the indicated FA pathway inhibitors for 6 hours, scrapped, washed in cold PBS, and lysed by 30 minutes incubation in 5 mM EDTA on ice. Cellular extracts were cleared by centrifugation (10,000 rpm, 5 min, 4°C) and quantified. Fluorogenic peptides specific for the chymotrypsin-like, trypsin-like and caspase-like activities of the proteasome (Suc-LLVY-AMC, Boc-LLR-AMC, Z-LLQ-AMC respectively (Bachem)) were incubated with 5 ug HeLa extracts in specific substrate buffers [51]. Fluorescence emitted by proteasome cleavage of the peptides was monitored every 200 seconds for 1 hour using a fluorometer (Hitachi F-4500 fluorometer) with 380 nm and 440 nm excitation and emission filters, respectively, and maximum linear slopes were measured. Emission of serial dilutions of AMC in extracts was used for fluorometer calibration. Proteasome activity was calculated as concentration of AMC (pM) produced per second per mg of protein. Three independent experiments were performed.

Drug interaction analysis

2008 and 2008 + FANCF cells were plated in 96-well plates at a density of 2000 cells/well. 24 h after plating, cisplatin and FA pathway inhibitors were added concomitantly, or FA pathway inhibitors were added and the cells immediately exposed to IR. Cytotoxicity was measured using the standard crystal violet assay 6 days after drug addition: cells were washed twice with PBS, fixed for 5 minutes at room temperature in 10% (v/v) methanol and 10% (v/v) acetic acid. Adherent colonies were stained for 2 to 10 minutes with 1% (w/v) crystal violet (Sigma) in methanol, rinsed in distilled water, and dried before the adsorbed dye was re-solubilized with methanol containing 0.1% (w/v) SDS by gentle agitation for 1 to 4 hours at room temperature. Dye concentration was quantified using ELx800 Universal Microplate Reader (Bio-Tek Instruments, Inc.) at 595 nm. For quantitation, readings of absorbance at 595 nm were normalized to those obtained from untreated cells, assumed to yield 100% cell survival, and empty wells, considered to be 0% cell survival. Cytotoxicity results were analyzed as described [52]. Briefly, after each experiment, survival curves were generated, for cisplatin and each FA pathway inhibitor alone and for the drug combinations. The LD50s for each drug in combination were determined, and LD50/LD50₀ units were derived as ratio of LD50 for cisplatin or IR and the FA pathway inhibitor relative to LD50 of each drug alone (this value was designated as 1) for each cell line. Isobolograms were generated at LD50 levels. Each plot presents values generated in at least three independent experiments. In addition, combination index (CI) values were calculated by the use of the Chou and Talladay method [26]. An identical analysis was performed at the 70% killing level.

Western blot analysis was done as described [18]. Anti- FANCD2 (NB 100–182, Novus Biologicals, 1/20,000 dilution) and HRP-conjugated ECL anti-rabbit IgG (Amersham, 1/5000) were used. Films were digitalized using a standard scanner and images processed using ImageJ.