Background
Several DNA repair pathways have evolved to maintain cell viability after exposure of mammalian cells to DNA damaging agents. Sufficiently high doses of drugs or radiation cause cell killing, and it seems reasonable to expect that those cells that can repair DNA damage will survive while those unable to repair their damage will die. Sensitive detection of residual DNA damage at the level of the individual cell could allow us to identify treatment resistant subpopulations within tumors. This possibility can now be examined by making use of the fact that complex DNA lesions such as DNA double-strand breaks (DSBs) are marked by microscopically visible γH2AX foci [
1].
DSBs rapidly activate kinases that phosphorylate histone H2AX. Resulting γH2AX foci can be used to identify the number and location of DSBs and to follow their fate during recovery [
2,
3]. The fraction of tumor cells that retain γH2AX foci 24 hours after irradiation has been correlated with the fraction of cells that fail to divide and form colonies [
4,
5]. Similar results have been reported for RAD51 recombinase, a key player in DSB repair by homologous recombination [
6]. RAD51 molecules also accumulate slowly as microscopically visible foci that are often co-expressed in cells with γH2AX foci [
7,
8]. Recently, RAD51 foci have been found in association with persistent DSBs [
9]. What is not known for certain is whether the cells that retain γH2AX or RAD51 foci 24 hours after irradiation are actually the cells that die.
γH2AX foci begin to form immediately after irradiation, reaching a maximum size about 30 or 60 min later and disappearing over the next several hours. However, residual foci may remain in some cells for days after exposure and may mark unrepaired or misrepaired sites [
10,
11]. Importantly, residual foci appear to be replicated and retained by daughter cells [
4,
5]. Since rapid loss of γH2AX is contingent upon functional DNA repair, it is not surprising that retention of γH2AX foci has been associated with loss of clonogenic potential. Several studies have reported that repair-deficient cell lines retain more foci and more cells with foci when analyzed 24 hours after irradiation [
12,
13]. The percentage of cells that retained γH2AX foci 24 hours after irradiation was correlated with the percentage of cells that lost clonogenicity, thus making it possible to use the fraction of cells with residual foci as a way to estimate sensitivity to killing by ionizing radiation [
4,
5].
DSBs may be produced either directly or indirectly [
2]. Direct DSBs occur as a result of exposure to ionizing radiation as well as selected drugs including bleomycin and the topoisomerase II inhibitor, etoposide. Indirectly produced DSBs can arise when a single-strand break, crosslinked DNA, or damaged DNA base meets a replication fork [
14]. Phosphorylation of H2AX may also occur indirectly during repair of base damage [
15]. Arguably also an indirect mechanism, extensive H2AX phosphorylation occurs as a result of DNA fragmentation during the process of apoptosis [
16]. Therefore directly or indirectly, the majority of DNA damaging agents are likely to cause H2AX phosphorylation, and cells that subsequently retain γH2AX foci may be more likely to die no matter how the DNA damage was initially produced. To explore this possibility, the fraction of cells that retained γH2AX foci was compared to the fraction of clonogenic surviving cells measured after a short exposure to 8 drugs known to damage DNA and cause H2AX phosphorylation [
2].
A correlation between clonogenicity and fraction of cells lacking foci does not constitute proof that cells that retain γH2AX foci are the cells that will die. Real-time imaging of γH2AX foci is complicated by the necessity of identifying the phosphorylated form of H2AX. However, RAD51 molecules also aggregate as clusters at sites of DNA damage in irradiated cells and are retained by γH2AX [
10]. When labeled with green fluorescent protein (GFP), RAD51-GFP can be used for live cell analysis to determine the fate of an individual cell that retains RAD51 foci [
17,
18]. The ability to follow live cells allowed a direct test of the hypothesis that cells that retain RAD51 foci 24 hours after irradiation are the cells that will eventually die.
Methods
Cell lines and drug treatment
Chinese hamster V79 and CHO cells were maintained by twice weekly sub-cultivation in minimal essential medium (MEM) containing 10% fetal bovine serum (FBS). SiHa human cervical carcinoma cells and HT144 human melanoma cells were obtained from American Type Culture Collection. SKOV3 human ovarian carcinoma cells were obtained from the DCTD tumor repository in Frederick MD. M059J and M059K human glioma cell lines were obtained from Dr. J. Allalunis-Turner, Cross Cancer Center. All tumor cell lines were sub-cultured twice weekly in MEM containing 10% FBS.
To obtain cells that expressed RAD51-GFP, SiHa cells were transfected with a plasmid kindly supplied by Dr. Roland Kanaar [
18]. Transfection was accomplished using Lipofectamine Plus using the protocol supplied by Invitrogen; stably transfected cells were selected by growth in 200 μg/ml G418 (Gibco), and a clone was chosen for further studies.
For drug treatment, 5 × 105 cells/60 mm dish and were exposed as exponentially growing monolayers to selected drugs usually for 30 min or for 60 min (cisplatin, temozolomide) in medium containing 5% FBS. Tirapazamine treatment was conducted using cells in suspension culture incubated for 30 min in drug-containing medium pre-equilibrated for one hour with 95% oxygen and 5% CO2. Cisplatin was obtained from Mayne Pharma and diluted from a stock solution of 1 mg/ml. Temozolomide was prepared in DMSO using a 250 mg capsule from Schering Canada. Etoposide was purchased from Novopharm and diluted from a stock concentration of 20 mg/ml. Doxorubicin, MNNG and hydrogen peroxide (H2O2) were obtained from Sigma and diluted in medium. Camptothecin was purchased from GBiosciences and prepared from a stock solution of 2 mM in DMSO. Tirapazamine was supplied by Dr. J. Martin Brown, and diluted from a stock solution of 2.5 mM in phosphate buffered saline. For experiments using X-rays, cells were exposed using a 300 kV unit at a dose rate of 5.2 Gy/min.
After drug incubation, drug was removed, dishes were rinsed several times, and cells were incubated for 24 hours in fresh complete medium. Trypsin treatment (0.1% for 5 min) was used to produce a single cell suspension. Samples of single cells were plated in duplicate to measure colony formation and resulting colonies were stained and counted two weeks later. The survival of the treated cells was normalized to the plating efficiency of the non-treated cells. Experiments were performed 2-4 times using a range of drug doses. The remaining cells were fixed in 70% ethanol for analysis by flow or image cytometry to measure γH2AX and RAD51 antibody binding.
Flow Cytometry for γH2AX
Antibody staining was performed as previously described using mouse monoclonal anti-phosphoserine-139 H2AX antibody (Abcam #18311; 1:4000 dilution) [
19]. After secondary antibody labelling with Alexa-488 conjugated goat anti-mouse IgG, cells were rinsed and resuspended in 1 μg/mL 4',6-diamidino-2-phenylindole dihydrochloride hydrate (DAPI; Sigma), a UV-excitable DNA stain. Samples were analyzed using a dual-laser Coulter Elite flow cytometer using UV and 488 nm laser excitation. The γH2AX signal was divided by DNA content per cell to account for differences in cell cycle distribution, and results were normalized relative to untreated controls within each experiment. Normalized γH2AX intensity is reported for all of the cells within the population.
Live cell analysis of RAD51-GFP
On average, one SiHa-RAD51-GFP cell was seeded into each well of an 8 well chamber slide with coverslip bottoms (Nalge Nunc International) containing 100 μl fresh complete medium and 100 μl conditioned medium prepared by filtering the complete medium recovered after two days of growth with high density cell cultures. Cells were allowed 4 hours to attach before exposing the chamber slide to 0 or 3 Gy and returning the chambers to the incubator for 24 hours. After 24 hours, each well was examined using a Zeiss inverted microscope with a 63× objective. To ensure high plating efficiency, analysis was restricted to cell doublets since this indicated that cells had attached and divided in the 24 hours period after irradiation. Images of cell doublets were obtained under phase and 488 nm excitation, and the location of the doublet in the well was noted. Almost invariably, both cells of a doublet exhibited RAD51-GFP foci or foci were absent from both cells at 24 hours. After scoring for the presence of foci, chamber slides were returned to the incubator for 2 weeks to allow time for colonies with greater than 50 cells to form. Approximately 40 doublets with foci and 40 doublets without foci were scored for clonogenicity.
Immunohistochemistry for RAD51 and γH2AX
Antibody stained cells prepared for flow cytometry as described above were cytospun onto microscope slides. Alternatively, cells grown on coverslips were fixed for 20 min in 2% freshly prepared paraformaldehyde before incubating with mouse monoclonal antibodies against γH2AX (Upstate, 1:500 dilution) and/or rabbit polyclonal antibodies against RAD51 (Calbiochem, 1:500 dilution or Oncogen Sciences, 1:100 dilution). Cells were viewed using a Zeiss epifluorescence microscope using a 100× Neofluor objective and images were analyzed for foci/nucleus. Analysis of foci/nucleus was always concluded before analysis of clonogenicity so that objectivity of scoring using image analysis was maintained. Experiments were repeated 3 times and independent results for each sample (clonogenic fraction and # cells lacking foci) were plotted.
Comet assays
Alkaline and neutral versions of the comet assay were used to measure MNNG-induced DNA single-strand breaks and DSBs respectively. Exponentially growing V79 hamster cells were exposed to MNNG for 30 min and then embedded in low gelling temperature agarose on a microscope slide. For the alkaline comet assay, slides were placed in a high-salt lysis solution for 1 h at pH 12.3 as previously described [
20]. The neutral comet assay was performed using a 4 h lysis at 50°C, pH 8.3, as previously described [
21]. For each drug dose and time, 150 comet images were analysed for DNA content, tail moment and percentage of DNA in the comet tail. Mean values are shown.
Discussion
We have previously reported that the fraction of SiHa cells that exhibit more than the background level of γH2AX 24 hours after treatment can be correlated with the fraction of cells that will ultimately die after exposure to X-rays and/or cisplatin [
5,
24]. We now show that retention of γH2AX foci can predict clonogenicity after exposure to a variety of DNA damaging drugs, several of which do not produce direct DSBs. DSBs can be formed indirectly by two closely opposing single-strand breaks. This is likely to be the case with hydrogen peroxide since sufficiently high drug doses will produce DSBs [
25]. MNNG-induced double-strand breaks were also detected using the neutral comet assay after exposure to high doses (Fig.
2e), and closely opposed damaged sites, perhaps undergoing base excision repair could be responsible [
26]. However, as a high lysis temperature was used for the neutral comet experiments, opposing heat-labile base damage sites may also be involved in the formation of physical breaks detected using this method [
27].
Several patterns of γH2AX formation and loss have been observed using the drugs listed in Table
1. Foci reach a maximum size within an hour after exposure to X-rays, tirapazamine, doxorubicin or etoposide. However, development of γH2AX is slower when DSB formation requires transit through S phase, for example, after treatment with cisplatin or low doses of MNNG. Significant DSB rejoining was not detected during the 24 hours after treatment with MNNG and γH2AX levels also remained high (Fig.
2c). The inability to repair DSBs caused by MNNG was also suggested by Stojic et al, [
28] based on the persistence of γH2AX foci in MNNG treated cells. Unfortunately, the presence of an unrejoined double-strand break at the site of a residual γH2AX focus cannot be confirmed using a physical method to detect DSBs since these methods lack sufficient sensitivity. A subsequent paper by Stojic et al. indicated that a different mechanism operated to induce γH2AX foci after exposure to high MNNG doses (30 μM) because foci formation became independent of mismatch repair which was associated with replication [
29]. Our flow cytometry results (Fig.
1) support this observation by indicating that cell cycle position is important for γH2AX focus formation after exposure to low but not high (> 20 μg/ml) MNNG doses.
Although our results suggest that residual, not initial γH2AX is the critical factor determining cell fate, initial γH2AX can also be predictive for response if some cells within a population are resistant to the induction of DBSs. Etoposide causes DSBs only in the outer proliferating cells of multicellular spheroids; the non-proliferating inner fraction of cells do not develop γH2AX and therefore survive treatment [
30]. In this case, determining the fraction of cells lacking γH2AX immediately after exposure was also predictive of cell survival [
31]. In the same way, doxorubicin penetrates poorly through the outer cells of spheroids so that only the outer cells developed significant numbers of γH2AX foci. Again the fraction of cells lacking foci immediately after exposure was correlated with the fraction of cells that survived [
31]. However, by counting the fraction of cells lacking foci 24 hours after treatment, the importance of repair capacity as well as susceptibility to a direct-acting genotoxin can be included in the estimate of survival. Moreover, effects of drugs that produce γH2AX only when cells transit S phase can also be evaluated, provided of course that drug-treated cells are given an opportunity to progress through the cell cycle.
A RAD51-GFP construct provided a way to directly address the importance of residual DNA repair foci in determining cell fate. Cells deficient in H2AX also show a deficiency in homologous recombination and RAD51 focus formation [
32,
33], and it is possible that retention of γH2AX may be the signal for retention of repair molecules like RAD51 [
10]. Most but not all cells with RAD51-GFP foci 24 hours after irradiation failed to form colonies. Similarly, most but not all cells lacking foci did form colonies (Fig.
6g). Unfortunately, resolution for detecting foci was reduced under the technical constraints imposed by live cell imaging in multiwells, and 24 hours may not have been an optimum time to score microscopically visible RAD51-GFP foci for all cells. Although the answer was not unequivocal, it does support the idea that residual DNA repair foci mark cells that are likely to die. Since all of the DNA damaging agents we have examined produced residual γH2AX that were predictive of clonogenic survival, it should be possible to use residual foci as a biomarker of response to genotoxic agents.
There are limitations to the application of this approach
in vivo. Tumor heterogeneity is a major consideration especially since both induction of DNA damage and its repair are influenced by the tumor microenvironment. Obtaining a representative biopsy and/or multiple biopsies will be essential [
34]. Second, it is important to consider the endogenous expression of γH2AX since this can be quite variable and will affect the ability to detect residual γH2AX foci [
35]. A pre-treatment biopsy must also be obtained and if endogenous γH2AX is excessive, prediction based on residual foci may not be possible. Third, loss of heavily damaged cells by apoptosis or other mechanisms within the first 24 hours after treatment, or sequestration of foci into micronuclei before scoring will reduce the accuracy of prediction. Although early apoptotic cells exhibit γH2AX foci, necrosis secondary to apoptosis can result in loss of the signal [
16,
36]. Finally, for drugs that produce foci only when DNA replicates, it will be necessary to ensure that all treated cells have the opportunity to transit S phase. In spite of these limitations, the recent application of γH2AX to predict response to cisplatin combined with radiation in xenograft tumors [
24] indicates that this approach has promise for early prediction of tumor response to treatment. Moreover, it should be possible to predict response not only to single drugs but to combinations of DNA damaging agents.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
JPB performed comet assays and the comparisons between γH2AX foci and survival. SHM created the RAD51-GFP transfected cell line which was characterized by DK and CAB. PLO conceived the study and wrote the paper. All authors contributed to study design and data analysis and approved the final manuscript.