CNDAC-Induced DNA Double-Strand Breaks Cause Aberrant Mitosis Prior to Cell Death

The nucleoside analogue 2′-C-cyano-2′-deoxy-1-β-D-arabino-pentofuranosyl-cytosine (CNDAC) and its orally bioavailable prodrug sapacitabine are currently undergoing clinical trials for acute myeloid leukemia (AML) and myelodysplastic syndromes (1–4). CNDAC has a mechanism of action that is distinct from other deoxycytidine analogs (5). After phosphorylation to the triphosphate and its incorporation into DNA, the instability of the analogue induces a β-elimination rearrangement that results in the transformation of the analogue to CNddC, a de facto 3′-terminus DNA chain terminator (6, 7). The resulting single-strand breaks (SSB) have been shown to be converted into double-strand breaks (DSB) during DNA replication in a subsequent S-phase (8). These consequences are supported by evidence of chromosome damage.

The molecular mechanism(s) of action of CNDAC have been investigated by employing cell lines process blocked 3′ ends lead to increased sensitivity to CNDAC. Exposure of cells with defective XPF-ERCC1 nuclease to CNDAC for one cell cycle decreased the IC₅₀ in clonogenic assays by 3- to 5-fold (9). In addition, the viability of cells lacking tyrosyl-DNA phosphodiesterase 1 (TDP1) was similarly sensitized to CNDAC, consistent with the conclusion that it also can remove a CNddC-terminated DNA end (10). In contrast, the viability of cells lacking either base excision repair function or DNA mismatch repair was not changed from those with intact repair pathways.

The nature of the DNA damage initially reported indicated the presence of DSBs. Treatment of cells lacking any of several enzymes (ATM, XRCC3, RAD51D, BRCA2) in the ATM-dependent homologous recombination pathway (HR) with CNDAC for two cell cycle times decreased IC₅₀ for clonogenicity by 30- to 100-fold (8). The finding that the viability of cells lacking nonhomologous end-joining (NHEJ) function was not diminished beyond that of NHEJ-competent cells supported the conclusion that HR repair of CNDAC-induced DSB is the primary mechanism of cell death (8). However, an alternative end-joining pathway, also known as microhomology-mediated end joining, has recently emerged as an important repair mechanism complementary to the classical NHEJ pathway (11). DNA polymerase θ (Pol θ) is a key component involved in alternative end-joining and has been shown to be synthetic lethal with the HR pathway (12, 13). The possible contribution of this pathway to repair of CNDAC-induced damage has not yet been reported.

At the subcellular level, CNDAC exposure gives rise to chromosomal aberrations (14). However, it remains obscure as to how CNDAC-induced DSBs, manifested as chromosomal damage, lead to cell death. In the current study, we focused on CNDAC-caused DNA lesions at the chromosomal level. Chinese hamster ovary (CHO) cell lines deficient in either XPF (a component of the XPF-ERCC1 nuclease) or RAD51D (a component of the HR pathway) were investigated. In addition to chromosomal aberrations, we observed that cells treated with CNDAC formed multiple nuclei, a prominent feature associated with mitotic catastrophe. To better characterize the events before, during and after the aberrant mitosis, we used live cell imaging to track cell fate in real time.

The nucleoside analogue CNDAC was synthesized as described and generously provided by Professor A. Matsuda (University of Sapporo, Hokkaido, Japan; refs. 15, 16). The following reagents were purchased from Thermo Fisher Scientific: Karyomax Giemsa stain stock solution (catalog no. #10092013); Gurr buffer solution tablets (catalog no. #10582013); Histomount mounting solution (catalog no. #008030); OptiMEM I reduced serum medium (catalog no. #31985062); Blasticidin S HCl solution (catalog no. #A1113903). Sources of other reagents are listed as follows: Accutase (catalog no. #07920) from Stemcell Technologies; pBOS-H2BGFP vector (catalog no. #559241) from BD Pharmingen; FuGENE HD transfection reagent (catalog no. #E23110) from Promega; protease inhibitor cocktail tablets (catalog no. #04693159001) and phosphatase inhibitor cocktail tablets (catalog no. #04906837001) from Roche.

mAbs to phospho-Ser1981 ATM (catalog no. #05-740) and phospho-Ser139 H2AX (catalog no. #05-636) were purchased from Upstate/Millipore. Rabbit mAb to DNA-PKcs (catalog no. #ab32566) and polyclonal antibody to ATM (catalog no. #ab17995) were from Abcam. Polyclonal antibody to H2AX (catalog no. #GTX108272) was from GeneTex. α-Tubulin and β-actin mAbs were from Sigma-Aldrich (catalog no. #T5168). mAb to PARP (catalog no. #51-6639GR) was from BD Pharmingen. mAb to caspase-3 (catalog no. #3004) was from BioVision, and polyclonal antibody to cleaved caspase-3 (catalog no. #9661) was from Cell Signaling Technology.

CHO lines AA8 [wild type (WT)] and UV41 (Xpf mutant) were purchased from the ATCC. The AA8-derived Rad51D knockout line, 51D1, and the Rad51D complemented line, 51D1.3 (17), were gifts from Dr. L. Thompson (Lawrence Livermore National Laboratory, Livermore, CA). All CHO lines were grown in αMEM (free of ribonucleosides and deoxyribonucleosides) supplemented with 10% heat-inactivated FBS and GlutaMax. Mouse embryonic fibroblast (MEF) lines Polq^+/+ (55-5 Tag), Polq^+/− (55-7 Tag), and Polq^−/− (55-6 Tag; ref. 11) were cultured in DMEM supplemented with 10% heat-inactivated FBS and GlutaMax. Human osteosarcoma U2OS POLQ^+/+ cell lines were used by RDW and KT to generate two independent POLQ^−/− knockout clones F10 and G6. The U2OS cells were cultured in the same medium as the MEFs. POLQ knockout by frameshift was confirmed by genomic DNA sequencing and reduced POLQ mRNA level was verified by qPCR (Supplementary Fig. S3C). POLQ^+/+, POLQ^−/−, (F10) and POLQ^−/− (G6) were stably transfected with shControl and shDNA-PKcs, respectively. Knockdown of DNA-PKcs was confirmed by immunoblotting. The human lung cancer cell line H460 was a gift from David G. Beer (University of Michigan, Ann Arbor, MI). All human cell lines were authenticated using short tandem repeat DNA fingerprinting by the Characterized Cell Line Core at the MD Anderson Cancer Center (Houston, TX). All rodent and human cells were free of Mycoplasma, as certified by the same core facility using the MycoAlert Kit from Lonza. Each line was used in experiments for no more than 15 passages after recovering from frozen seed stocks that were authenticated prior to freezing.

Adherent cells were detached with Accutase, washed with ice-cold PBS (pH 7.4), and fixed in 70% ethanol for at least 1 hour on ice or overnight at 4°C. Fixed cells were washed with PBS and incubated with 50 μg/mL propidium iodide (PI) and 2.5 μg/mL DNase-free RNase (14). A Becton Dickinson FACSCalibur flow cytometer with BD CellQuest Pro program was used to determine the DNA content in each sample by measuring the fluorescence from at least 20,000 cells.

Detached cells were washed with PBS and resuspended in 100 μL of binding buffer with 5 μL of Cy5-conjugated Annexin V (BD Pharmingen, catalog no. #559934). After 15-minute incubation in the dark at room temperature, 400 μL of binding buffer with 5 μL of 50 μg/mL PI was added to each sample. At least 20,000 cells were analyzed using a BD FACSCalibur flow cytometer with the CellQuest Pro software. Cells stained positive for Annexin V (PI negative and positive) were considered apoptotic cells.

Exponentially growing cells were plated in multiple 100-mm dishes at 10⁵/dish. At indicated times, duplicate dishes of cells were detached using 0.05% trypsin–EDTA solution and subjected to quantitation of total cell numbers using a Coulter Counter (Beckman Coulter Life Sciences). The total number of cells (mean of two dishes) were plotted as a function of time. Population doubling time (PDT) of each cell line was computed using the tool at http://www.doubling-time.com/compute_more.php.

Adherent MEF or human U2OS cells, WT or Polq (or POLQ) knockout, were seeded in 6-well plates one day before they were exposed to a range of concentrations of drugs for 24 hours. Cells were washed into drug-free medium and incubated for an additional 3 days (MEFs) or 6 days (U2OS lines) to form colonies. Colonies in triplicate wells were fixed and stained with crystal violet/ethanol and counted under a dissecting microscope (cutoff ≥ 50 cells) or scored electronically (GelCount, Oxford Optronix). Percent survival is expressed as mean ± SD.

Colcemid (100 ng/mL) was added to cell cultures before harvesting. Mitotic spreads were prepared as described previously (14). A minimum of 50 metaphase spreads were analyzed for each sample by the T.C. Hsu Molecular Cytogenetics Core at the MD Anderson Cancer Center.

CNDAC-treated AA8 cells were detached with Accutase, transferred onto slides using a Shandon Cytospin 3 cytocentrifuge (700 rpm for 5 minutes at room temperature), and air dried briefly. Cells were fixed in 70% ethanol at 4°C for a minimum of 1 hour and air dried again. Giemsa stain solution was freshly diluted in Gurr buffer (1 tablet in 1 L of water, pH 6.8) to 4% (v/v) and used for staining the cells on slides for 20 minutes at room temperature. Cells were rinsed, air dried overnight, and mounted with coverslips using Histomount mounting solution. Images of cells in bright field were captured at randomly selected fields under a Nikon light microscope (20× objective lens) using a Sony 3CCD color video camera (model no. DXC-390, SimplePCI software). For each treatment, nuclear morphology of a minimum of 400 cells was analyzed and categorized according to the samples in Supplementary Fig. S4. Qualitative images were taken using 40× and 100× oil lenses.

Cell lysates, prepared by sonication in RIPA buffer supplemented with protease inhibitor and phosphatase inhibitor cocktail tablets, were loaded on Bis-Tris 4% to 12% polyacrylamide gradient gels (Bio-Rad) and transferred onto nitrocellulose membranes (0.22 μm) as described previously (14). Immunoblots were visualized and target bands were quantitated by Li-Cor Odyssey Version 4.0 software.

DNA strand breaks were detected using single-cell gel electrophoresis under both neutral pH and alkaline conditions using CometAssay Electrophoresis System II (Trevigen, catalog no. #4250-050-ES). After electrophoresis, cells embedded in low-melting agarose (comets) were fixed with 70% ethanol and stained with SYBR Gold (Thermo Fisher Scientific, catalog no. #S-11494). Images of comets in random fields were captured under Olympus IX70 inverted fluorescence microscope using Olympus DP70 CCD camera and 4× objective, and subjected to quantitation using Comet Assay IV software (Instem). At least 200 cells per treatment condition were scored. An increase in Olive tail moment from a neutral comet assay was used as a readout for DSBs, whereas that from an alkaline comet assay for the sum of SSBs and DSBs. Additional images were taken using 10× or 20× objectives to feature more details of comets.

AA8 cells were transfected with pBOS-H2BGFP vector (18) using FuGENE transfection reagent in OptiMEM I. Cells stably expressing H2B-GFP fusion proteins were selected by addition of 5 μg/mL blasticidin in complete medium over a period of 2 to 3 weeks. Cultures were monitored daily under Olympus IX70 inverted fluorescent microscope and imaged as needed using 20× objective at both fluorescent and bright light channels. GFP-positive cells were isolated on BD FACS Aria IIu cell sorter and two populations, GPF⁺ high and GFP⁺ low, were collected and propagated separately. No significant difference in growth rate, based on size of colonies formed, was observed for the two populations. The GFP⁺ high cells were used for live cell imaging. H460 cells were stably transfected with pBOS-H2BGFP vector following the same procedure as AA8.

AA8/H2B-GFP cells were grown in chamber slides overnight and set up in an Olympus live cell imaging station enclosed by a micro cell incubator with humidified air and active controllers for temperature (37°C) and CO₂ (5%). HEPES (10 mmol/L) was supplemented in complete medium to increase pH buffering capacity. A 20× relief phase contrast objective was used on the Olympus microscope. Multiple positions were randomly selected with well isolated cells at a low density and the objective was adjusted to focus for each position. Cells were exposed to CNDAC at 0.5, 1, and 10 μmol/L, respectively, for 6 hours before washout. The microscope and connected computer were programed to capture images of both GFP fluorescence and bright field every 15 minutes after CNDAC washout. The live cell imaging was carried out for 114 hours after the initial 63 hours. Meta Morph Advanced image analysis software was applied to compile images into time-lapse videos and perform subsequent analyses.

For the quantitation of chromosomal aberrations, tail moment of comets, and the live cell imaging, intergroup comparisons of endpoints in response to various treatments were made using a t test with two-tailed distribution and two-sample unequal variance in Excel.

Cellular response to CNDAC is concentration- and time-dependent (14). We postulated that processing or DNA replication across the unrepaired nicks would generate DSBs, which in turn would lead to the death of cells by yet undefined mechanism(s). To address this, cytogenetic endpoints were used to characterize the consequences of deficiencies in specific DNA repair mechanisms. Metaphase chromosomal aberrations (including breaks and gaps) were compared in the WT AA8 line and its mutant UV41, which lacks XPF (19, 20). The cell lines were continuously treated with CNDAC at 1 μmol/L, which is a nonarresting concentration for both lines after one cell cycle (15–16 hours). However, after drug exposure for 2 cell cycles (30 hours), the mutant line was delayed in transiting the second G2, whereas this was not observed for the WT cells (Supplementary Fig. S1). Therefore, the addition of colcemid was scheduled earlier than for WT cells to ensure prolonged incubation time to enrich the number of mitotic cells for analysis. In cells lacking XPF, chromosomal aberrations increased from an average of 0.4/metaphase to 0.9/metaphase after one cell cycle treatment with 1 μmol/L CNDAC, and to 5.6/metaphase after two cell cycles, that is, 30 hours (Fig. 1A, gray bars, n ≥ 60, P < 0.01; Fig. 1B). Chromosomal aberrations in the WT cells did not increase as dramatically, from 0.05/metaphase to 0.2/metaphase after two cell cycles (Fig. 1A, black bars, n ≥ 55, P < 0.05). These data support the interpretation that SSBs induced by CNDAC in the first S-phase can be converted into DSBs during the second S-phase when replication forks encounter unrepaired SSBs and that an XPF-dependent mechanism provides protection against generation of DSBs.

The increase in chromosomal aberrations after CNDAC exposure, regardless of length of time, was greater in UV41 cells compared to the WT. The mean values of chromosomal aberrations increased 6-fold in WT and >6-fold in XPF-mutant cells after two PDTs compared with those treated for only one PDT (Fig. 1A, 30- vs. 15-hour bars). More impressively, the average number of chromosomal aberrations increased 16-fold after two cell cycles in the mutant (5.6 aberrations/metaphase) compared with a 5-fold change in the WT cells (0.22/metaphase, Fig. 1A). The distribution of chromosomal aberration frequencies was also distinct between the two lines. Sixteen percent of WT cells treated with CNDAC for two PDTs had 1 to 2 chromosomal aberrations, and no cells had more than 2 aberrations per metaphase. In contrast, 42%, 30%, and 12% of XPF-mutant cells exhibited 3–5, 6–10, and 10–20 aberrations per metaphase, respectively (Fig. 1C). These results are consistent with a decrease in clonogenic survival in XPF-deficient cells after CNDAC exposure (9), indicating that XPF contributes to the repair of CNDAC-induced damage.

To provide direct evidence for the presence of DNA strand breaks, comet assays were performed in XPF-deficient and WT cells. In XPF-mutant cells, both neutral and alkaline comet assays demonstrated significant elevation in tail moments after CNDAC incubation (Fig. 1D). Under a neutral condition, the tail moment of control cells increased from 4.06 ± 2.72 to 6.12 ± 4.47 after incubation with 1 μmol/L CNDAC for 15 hours, and further to 12.61 ± 6.92 at 30 hours (Fig. 1D; Supplementary Fig. S2A). This result parallels the change in chromosomal aberrations after CNDAC treatment (Fig. 1A). The minor increase in either comets or chromosomal aberrations after 15-hour CNDAC exposure could be interpreted as the conversion of endogenous SSBs into DSBs upon CNDAC incorporation during the first cell cycle. Both comets and chromosomal aberrations soared after 30-hr incubation with CNDAC, indicating an increase in DSBs after a second cell cycle. The comet assay under an alkaline condition revealed a similar change in comets but to a greater extent. The tail moment of untreated UV41 cells was 4.32 ± 4.36, while the CNDAC-treated cells presented a tail moment of 14.63 ± 6.44 at 15 hours and 41.2 ± 9.28 at 30 hours, respectively (Fig. 1D; Supplementary Fig. S2A). Furthermore, the phosphorylation of ATM on Ser1981 and H2AX on Ser139 (γ-H2AX) increased slightly at 15 hours and to a greater extent after 30 hours of CNDAC treatment (Fig. 1E, right). Together, the comet assays and immunoblot results in XPF-deficient cells demonstrated that CNDAC-induced SSBs could be transformed into DSBs after a second S-phase. Although the WT AA8 cells presented increased levels of phosphorylated ATM and H2AX in a similar pattern to that found in the XPF mutants (Fig. 1E, left), comet tails were not discernible under alkaline condition when AA8 cells were exposed to 2 μmol/L CNDAC for 30 hours (2 cycles). However, comet tail legnth greatly increased after 10 μmol/L CNDAC (Supplementary Fig. S2B and S2C). The tail moments of AA8 cells were 7.2 ± 6.4 for the control, 52.9 ± 7.0 after 15-hour treatment and 75.5 ± 10.7 after 30-hour treatment.

HR is the main pathway to repair DSBs induced by CNDAC (8). Here, we compared the cytogenetic changes after CNDAC treatment in cells deficient in RAD51D, a component of the HR mechanism (21), and cells complemented with RAD51D. In the absence of CNDAC exposure, RAD51D-deficient cells had an elevated background of chromosome alterations. Seventy-three percent had no apparent spontaneous chromosomal alteration; 23.8% and 1.6% presented 1–2 and 3–4 breaks per metaphase, respectively; the remaining 1.6% had fusions (Fig. 2A, gray bars, n = 63). In untreated RAD51D complemented cells, no aberrations were detected in 90% of the population, and 1–2 aberrations per metaphase were seen in the remainder of cells (Fig. 2A, black bars, n = 80). These data show a baseline genetic instability that can be attributed to deficiency in RAD51D.

A larger portion of the RAD51D complemented cells (Fig. 2A, 19%, n = 99) bore 1–2 chromosomal breaks, when exposed to 1 μmol/L CNDAC for two PDTs, than those exposed to CNDAC for only one population doubling time (Fig. 2A, 8.3%, n = 72). At both times, a minor portion of cells treated with CNDAC (1%–2%) had more than 2 breaks per metaphase. Notably, after two cell cycles, 99% of all metaphases from CNDAC-treated RAD51D complemented cells could be analyzed, including 78% metaphases free of aberration and only 1% was unscorable (Fig. 2A, n = 99). In contrast, 76% of the metaphases in RAD51D-deficient cells exposed to CNDAC for two PDTs had severe chromosomal damage and therefore were unscorable for individual aberrations (Fig. 2B). Radial formation was apparent in some chromosomes, consistent with a defect in the HR function. Another 9% and 2% presented 1–2 breaks and 3–4 breaks per metaphase, respectively. Only 13% of the population were free of chromosomal aberrations (Fig. 2A, n = 106). As for the RAD51D-deficient cells treated with CNDAC for one population doubling time, 28% and 3% had 1–2 breaks and 3–4 breaks per metaphase, respectively; 1.5% had fusions; and the rest were free of chromosomal aberration (Fig. 2A, n = 68). The RAD51D-deficient cells showed fewer aberrations after one cell cycle than that after two cell cycles. This is consistent with results from the RAD51D complemented cells and the paired XPF cells as well, further supporting the hypothesis that DSBs are formed during a second cell cycle. Importantly, the much greater extent of chromosome damage in cells exposed to CNDAC for a second S-phase relative to that in cells treated for only a single cell cycle is consistent with the much greater clonogenic sensitivity caused by defects in homologous recombination repair than by a defect in XPF-ERCC1 nuclease (8, 9).

The role of the alternative DNA end-joining pathway in the repair of CNDAC-induced damage, if any, has not be defined. We compared the clonogenic survival rate of WT cells with a MEF line in which Pol θ, the product of the Polq gene, had been knocked out (11). The clonogenic survival of MEF cells that were either proficient or deficient in Pol θ was similar (Fig. 3A, P = 0.051). The PDTs of MEFs with different Polq status were 15.9 hours for Polq^+/+, 19.8 hours for Polq^+/−, and 25.5 hours for Polq^−/− (Supplementary Fig. S3A). Consistent with the above observations, clonogenic assays of human U2OS POLQ^+/+ and POLQ^−/− cells transfected with shControl were similarly sensitive to CNDAC (Fig. 3B, left), while POLQ^−/− cells were sensitized about 2-fold to the comparator DNA strand-breaking agent, bleomycin (Fig. 3B, right). Previous investigations demonstrated that cells lacking the classical NHEJ component, DNA-PKcs, were not sensitized to CNDAC (8). Therefore, we investigated the CNDAC sensitivity of cells defective in both classical NHEJ and alternative end-joining. U2OS POLQ^+/+ and POLQ^−/− cells stably transfected with shDNA-PKcs showed similar sensitivity to each agent in clonogenicity assays (Fig. 3C), although the PDT of POLQ^+/+ cells (27.0 hours) was somewhat more rapid relative to that of POLQ^−/− cells (31.8 hours). However, the PDTs were further prolonged in DNA-PKcs knockdown POLQ^+/+ (41.2 hours) and POLQ^−/− (46.1 hours) cells (Supplementary Fig. S3B). Therefore, depletion of Pol θ and/or DNA-PKcs in U2OS cells may hinder cell proliferation. Thus, similar findings in the two cell lines compromised for Pol θ function indicates that the alternative end-joining pathway is dispensable for repair of CNDAC-induced DSBs.

CHO cells manifested metaphases with chromosomal aberrations in response to CNDAC to a greater extent after two PDTs and in the absence of key components of DNA damage repair pathways, such as XPF or RAD51D. To further study the consequences of entering mitosis with chromosomal instability, we analyzed the incidence of multinucleate formation following exposure to AA8 cells to CNDAC 24 hours. Because only WT cells were used in the study, a higher concentration of CNDAC (10 μmol/L, causing G₂ arrest) was tested in addition to two lower and nonarresting concentrations (0.5 and 1 μmol/L). Cells incubated with CNDAC for 24 hours were released into drug-free medium and analyzed for effects on their inability to complete normal mitosis as indicated by multinucleate formation after drug washout. Atypical interphase cells with more than a single nucleus were collectively counted as multinucleate (Supplementary Fig. S4C–S4G). The percentage of such cells increased in a CNDAC concentration-dependent fashion. Notably, among cells exposed to 1 μmol/L CNDAC, multinucleation was identified in approximately 22% and 13% of the population 24 and 48 hours after drug washout, respectively (Fig. 4A and B II). It was more evident that about 24% and 34% of 10 μmol/L CNDAC-treated cells presented multinucleation at 24 and 48 hours, respectively (Fig. 4A and B III-IV). When cells were exposed to CNDAC for 15 hours and then released in drug-free medium, a substantial increase in multinucleation was only observed in cells treated with 10 μmol/L CNDAC, not those with 0.5 or 1 μmol/L CNDAC (Supplementary Table S1). This is likely because longer drug-exposure time allows for more CNDAC incorporation in the DNA of all cells. Nevertheless, in both 24- and 15-hour schedules, multinucleation was not evident until after two PDTs, suggesting that the abnormal nuclear alterations occur after a second mitosis.

Multi-nucleation, a hallmark of mitotic catastrophe, has been found in a significant portion of CNDAC-treated AA8 cells. We used time-lapse video microscopy to test the hypothesis that exposure of S-phase cells CNDAC causes mitotic catastrophe prior to cell death. The fate of AA8 cells stably transfected with a construct expressing H2B-GFP fusion protein after a 6-hour pulse 0.5, 1.0, and 10 μmol/L of CNDAC, respectively, was tracked followed by drug washout. Image capture was started immediately after wash. Cells entering the first mitosis within 5 hours after washout were presumed to have had CNDAC incorporation into DNA and therefore were eligible for analysis (Supplementary Fig. S5A). There was no difference in the duration of first mitosis among all three treatments and control (Fig. 5A–D, orange bars; Fig. 5E, n ≥ 24, P > 0.05). However, the duration of second mitosis was significantly prolonged in a CNDAC concentration-dependent manner in all treatments compared with control (Fig. 5, red bars; Fig. 5E, n ≥ 37, P < 0.05–0.001). The interval from the beginning of the first mitosis to the second was also prolonged markedly in 1 and 10 μmol/L CNDAC-treated samples (Fig. 5E, P < 0.01–0.001). A similar trend was observed in a H2B-GFP transfected H460 lung cancer cells following wash out of CNDAC (0.4 μmol/L, 8-hour exposure). In this case, because of the longer cell cycle time, cells entering first mitosis within 7.5 hours after wash were quantitated (Supplementary Fig. S5B). There was significant increase in the duration of second mitosis and in the interval between the first and second mitoses (Supplementary Fig. S6, n ≥ 38, P < 0.01 and P < 0.001, respectively). Thus, cell-cycle progression was slowed due to CNDAC-induced DNA damage, especially when DSBs are expected to arise.

Notably, large numbers of CNDAC-treated AA8 cells were delayed in completion of the second mitosis, which was abnormal, and associated with the formation of multinucleate daughter cells (Fig. 5, blue bars). For cells exposed to 10 μmol/L CNDAC, multinucleation was found to be coupled to the second mitosis and followed by the onset of apoptosis in 18 out of 25 cells tracked. The snapshots in Fig. 6 and video clip in Supplementary Fig. S7 illustrate how one of those cells (#10) went through a seemingly normal, first mitosis (Fig. 6A), and its two daughter cells progressed through aberrant second mitoses followed by multinucleate formation and cell death (Fig. 6B). Multinucleation was also apparent in cells treated with different schedules of CNDAC exposure and washout (Supplementary Fig. S8). Thus, MM formation was frequently coupled with or followed by onset of apoptosis. Together the results from static nuclei (stained with Giemsa) and live cell imaging indicate that CNDAC-induced DSBs, which, if not repaired, trigger cells to undergo aberrant mitoses resulting in gross nuclear alterations. Multinucleation precedes eventual cell death through apoptosis, which was substantiated by cleavage of caspase-3 and PARP (Fig. 6C), as well as positive staining of Annexin V (Fig. 6D) after CNDAC washout. All the readouts showed concentration- and time-dependency.

This study aimed at defining the mechanism by which CNDAC-induced DSBs lead to cell death. We used cytogenetic approaches to compare CNDAC-caused chromosomal damage in CHO cells proficient and deficient in XPF-ERCC1 nuclease. The increase in chromosomal aberrations after CNDAC exposure for an additional population doubling time (from one cell cycle to two cell cycles), regardless of XPF status, has provided direct cytogenetic evidence for conversion of CNDAC-induced SSBs into DSBs during DNA replication. Spontaneous chromosomal alterations in XPF-deficient cells were 7.6-fold greater than those in WT cells (Fig. 1A), indicating genetic instability conferred by XPF deficiency. The fold increase in chromosomal aberrations after CNDAC treatment was greater in Xpf-mutant than in WT cells, underlining the importance of XPF in repairing initial DNA damage caused by CNDAC. The XPF-ERCC1 endonuclease is multifunctional. It is not only essential for the NER pathway, but also for processing some blocked 3′ DNA ends, and for completing HR initiated by the Fanconi anemia pathway (19, 22, 23). XPF (ERCC4) has been identified as a Fanconi anemia factor, FANCQ (24). XPF-ERCC1 participates in the repair of interstrand crosslinks (ICL) by unhooking the crosslinks in concert with FANCD2 and FANCP/SLX4 (25). The resulting intermediate after resection is further processed by HR. Although both ERCC1 and XPF-mutant cells are sensitized by CNDAC (2.6- and 4.6-fold sensitization, respectively; ref. 9), we focused on XPF deficiency in the cytogenetic analysis and comet assay in the current study. In addition, the XPF-ERCC1 nuclease has been reported to facilitate Ku86-independent end joining of DSBs in yeast and mouse and human fibroblasts (26). Because XPF-ERCC1 can act to remove some blocks from the 3′ ends of DNA (such as 3′ deoxyribose-phosphate and 3′phosphoglycolate; ref. 27), it is likely that the enzyme may perform a similar reaction to excise CNddC, the de facto chain terminating lesion generated by β-elimination after incorporation of a deoxynucleotide following CNDACMP at the 3′ terminus of the DNA (6).

Thus, it is likely that XPF-ERCC1 nuclease contributes to the repair of the SSB caused by incorporation of CNDAC nucleotide into DNA. We have postulated that failure to repair nicked DNA may result in formation of a one-ended DSB upon a subsequent DNA replication cycle. Direct evidence for this was previously obtained by double labeling the DNA of ML-1 cells with a brief pulse of CNDAC and IdUrd (to mark the first S-phase) and followed by washing and addition of CldUrd before the second S-phase (8). Time-course flow cytometry demonstrated that entry of double positive cells into a second S-phase was concordant with DSB formation measured by pulse-field gel electrophoresis. The present studies are consistent with these findings in that they provide additional evidence of chromosome aberrations and DSB formation in the comet assay after two population-doubling times.

Besides the paired XPF cell lines, we chose to investigate cytogenetic alterations in RAD51D-deficient and proficient cells after CNDAC treatment as an indication for HR in the DNA damage repair processes. RAD51D mutations increase the risks for women to develop ovarian cancer (28, 29). Similar to XPF, deficiency in RAD51D renders genetic instability in that spontaneous chromosomal aberrations were more pronounced in AA8-derived cells lacking Rad51D (27% of the cell population) compared to cells complemented with Rad51D (10%, Fig. 2A). Both lines showed minimal cytogenetic changes (<4%) after one cell-cycle exposure to CNDAC, consistent with the timing of DSB induction by CNDAC after a second S-phase. After exposure for two cell cycles, the vast majority of RAD51D repleted cells (99%) could be analyzed, whereas only a minor portion of RAD51D-deficient cells (24%) were scorable, as most cells had extensive chromosome damage (Fig. 2A). This further highlighted the significance of RAD51D and the HR pathway in repair of CNDAC-induced DSBs. The finding that cells lacking HR components such as RAD51D cause significantly more chromosomal aberrations and greater clonogenic sensitization to CNDAC than XPF mutant cells, supports the conclusion that the chromosomal alterations in XPF deficient cells in response to CNDAC are mainly due to an HR-independent function of XPF-ERCC1.

Investigations with the Polq (or POLQ) knockout cell lines indicated that the absence of Pol θ function (30) does not sensitize cells to CNDAC as measured by clonogenic survival (Fig. 3; refs. 6, 31). Thus, it is unlikely that the alternative DNA end-joining pathway has a crucial role in repair of CNDAC-induced DSBs. Considering that deficiency in the classical NHEJ pathway (deficiency in either DNA-PKcs or Ku80) does not confer sensitivity to CNDAC, homologous recombination is indeed the major mechanism for the repair of CNDAC-induced DSBs (8).

Furthermore, we employed a traditional method, Giemsa staining of AA8 cells treated with and then washed out of CNDAC, to demonstrate the presence of multinucleate cells, which is indicative of aberrant mitosis and characteristic of mitotic catastrophe (32). To detect relatively low levels of DNA damage in repair competent cells, we treated cells with lesser concentrations of CNDAC (0.5 and 1 μmol/L) and a greater concentration (10 μmol/L) that delayed cell-cycle progression. The percentage of multi-nucleated cells was dependent upon both CNDAC concentration and length of exposure time. Interestingly, multinucleate cells became discernible only after two PDTs, highlighting that a second mitosis is needed to enable the aberrant nuclear alterations to occur and accumulate.

Finally, we exploited time-lapse video microscopy to better observe cells in real time and verify mitotic catastrophe occurrence after CNDAC exposure and washout. The GFP-tagged histone H2B stable expressing system was used to monitor long-term dynamics of nuclei and chromosomes without compromising their structures (33). Live cell imaging of two cell lines of distinct origins demonstrated that conditions that cause CNDAC-induced DSBs trigger cells to undergo an extended interphase followed by a prolonged second mitosis, coupled with multinucleation and subsequent apoptosis. Such a process characterizes mitotic catastrophe, which is visualized by the multinucleate formation after an aberrant attempt at mitosis, and is regarded as an oncosuppressive process for maintaining genomic stability by leading to apoptosis, senescence, or necrosis (34). Active cancer therapeutics that damage DNA integrity or interfere with mitotic processes are known to induce mitotic catastrophe (35). The demonstration of multinucleate cells following CNDAC treatment suggests the possibility that this may serve as a pharmacodynamic indicator in clinical trials. This possibility may be evaluated in clinical trials of the parenteral formulation of CNDAC (as DFP-10917) or of sapacitabine, the orally bioavailable prodrug of CNDAC, which are undergoing clinical trials in AML and myelodysplastic syndromes with encouraging outcomes (2–4, 36). Furthermore, sapacitabine has also shown antitumor activities in multiple solid tumor types lacking BRCA1 or BRCA2 (37, 38), confirming our findings in hamster and human cell lines (8, 39).

R.D. Wood is a scientific advisory board member for Repare Therapeutics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: X. Liu, W. Plunkett

Development of methodology: X. Liu, Y. Jiang, K. Takata, R.D. Wood, W.N. Hittelman

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): X. Liu, Y. Jiang, K. Takata, B. Nowak, C. Liu, R.D. Wood, W.N. Hittelman

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): X. Liu, Y. Jiang, K. Takata, W.N. Hittelman, W. Plunkett

Writing, review, and/or revision of the manuscript: X. Liu, Y. Jiang, K. Takata, R.D. Wood, W.N. Hittelman, W. Plunkett

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): X. Liu, Y. Jiang, B. Nowak

Study supervision: W. Plunkett

This work was supported by grant R01 CA28596 (to W. Plunkett) and Cancer Center Support grant P30 CA16672 from the NCI, Department of Health and Human Services. R.D. Wood was supported by NIH P01 grant CA193124 and the Grady F. Saunders, PhD Distinguished Research Professorship. K. Takata was supported by an Institutional Research Grant (IRG) from The University of Texas MD Anderson Cancer Center and by the Center for Radiation Oncology Research (CROR). We thank Drs. Simona Colla and Andrea Santoni for their assistance with Comet Assay IV software.

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.