γH2AX Expression in Tumors Exposed to Cisplatin and Fractionated Irradiation

Prediction of tumor response to radiation remains an important goal in radiotherapy (1, 2). However, prediction is complicated by the use of combined modality therapy and the lack of a simple but effective predictor of response in situ. Ionizing radiation and many cancer therapeutics cause phosphorylation of histone H2AX at sites of individual DNA double-strand breaks, providing an exceptionally sensitive method for detection of this critical lesion (3, 4). Production of γH2AX foci can be detected microscopically or by using flow cytometry, and the extent of phosphorylation has been used as a biomarker of both exposure and DNA repair capacity (5–9). The fraction of cells that retain of γH2AX foci, measured 24 h after exposure to either radiation (6) or cisplatin (10), is correlated with the fraction of cells that lose clonogenicity in SiHa and WiDr tumor cells. These in vitro results and the availability of biopsies from patients treated with cisplatin and radiation provided the motivation to examine the response of xenograft tumors to the combination of cisplatin and radiation and then to measure γH2AX expression in clinical biopsies.

The presence of residual γH2AX was examined using three human tumor xenograft models growing subcutaneously in immunodeficient mice given a single dose of cisplatin, 2 Gy × three daily fractions of X-ray, or the combination. To determine the importance of tumor cell location relative to functional blood vessels, tumor cells were sorted based on the fluorescence intensity of Hoechst 33342, a perfusion marker (11). Sorted cells were analyzed for clonogenicity and for expression of γH2AX using flow or image cytometry. A preliminary clinical evaluation was done to determine the feasibility of staining formalin-fixed, paraffin-embedded slides for fluorescence detection of endogenous and treatment-induced γH2AX expression. Pretreatment excision biopsies were obtained from 47 patients, and on-treatment biopsies were obtained from 8 patients undergoing radiochemotherapy for advanced cancer of the cervix.

Cell lines and treatments. SiHa human cervical carcinoma cells and WiDr human colon carcinoma cells were purchased from the American Type Culture Collection. HCT116 human colon carcinoma cells were kindly supplied by Dr. Bert Vogelstein. Cell lines were maintained with twice weekly subculture in MEM containing 10% fetal bovine serum. Multicell spheroids were initiated directly from SiHa or WiDr monolayers and were grown for ∼2 weeks in spinner culture flasks as described previously (12). Spheroids were exposed to 2 Gy X-ray daily for 2 weeks, and samples were taken immediately before the daily irradiation for analysis of clonogenic survival and residual γH2AX expression. Cisplatin (Mayne Pharma) was diluted from the 1 mg/mL stock solution just before use.

Xenograft tumors. Approximately 5 × 10⁵ SiHa, WiDr, or HCT116 cells were injected subcutaneously into the back of NOD/SCID immunodeficient mice. After 4 to 6 weeks when tumors had reached a size of 0.25 to 0.5 g, mice were injected intraperitoneally with 0 to 20 mg/kg cisplatin or mice were exposed to 300 kVp X-ray. For radiation fractionation experiments, exposure to 2 Gy was done daily for 3 days with mice restrained in a Plexiglass jig. Mice were housed in our approved animal holding facility and treated according to the guidelines of the Canadian Council on Animal Care.

Fluorescence-activated cell sorting. To separate tumor cells close to or distant from functional blood vessels, mice were injected intravenously with 0.05 mL Hoechst 33342 (20 mg/mL stock; Sigma) 20 min before sacrifice. Tumors were then excised and a single-cell suspension was prepared by mincing followed by incubation for 30 min in a freshly prepared cocktail of trypsin, collagenase, and DNase (13). Cells were sorted based on Hoechst 33342 fluorescence using a Becton Dickinson FACS440 cell sorter with UV excitation. Gated populations representing the dimmest 1/6 of cells, brightest 1/6 of cells, or all of the tumor cells were obtained under sterile sorting conditions. Sorted cells were analyzed in duplicate for clonogenic cell survival using a conventional colony formation assay (14). A sample of 250,000 sorted cells was also fixed in 70% ethanol and examined for γH2AX antibody binding by flow and image analysis. Four to six tumors per sort group were analyzed in independent sets that included controls, radiation only, cisplatin only, and the combination.

Flow cytometry for analysis of γH2AX. Antibody staining was done as described previously (15). Briefly, 5 × 10⁵ cells were rehydrated for 10 min, centrifuged, and resuspended in mouse monoclonal anti-phospho-Ser¹³⁹ histone H2AX antibody (1:500 dilution; Upstate). After 2 h, cells were rinsed and resuspended in secondary antibody [1:200 dilution of Alexa 488 goat anti-mouse IgG (H + L) F(ab′)₂ fragment conjugate; Molecular Probes]. Cell pellets were finally resuspended in 1 μg/mL 4′,6-diamidino-2-phenylindole (DAPI; Sigma) in TBS. Samples were analyzed using a Coulter Elite flow cytometer with UV and 488 nm lasers for excitation. To account for differences in radiation-induced changes in cell cycle distribution that affect average γH2AX intensity measurements, γH2AX expression per cell was corrected for DNA content and results were expressed as a ratio of the signal intensity for irradiated versus unirradiated cells. Experiments were typically repeated three times.

Microscopic detection of γH2AX foci in individual cells. Cells stained for γH2AX for flow cytometry were also cytospun onto slides for visual analysis of foci (15). Slides were viewed and images were digitized using a Zeiss fluorescent microscope equipped with Neofluor objectives and a Q-Imaging 1350 EX digital camera. For xenograft experiments where single cells were analyzed, cells with fewer than three large foci per cell were scored as negative for γH2AX. However, because foci often cluster together over time after treatment, cells containing one very large focus or cells with foci in micronuclei were also scored as positive. At least 150 nuclei were scored from images obtained using a ×100 objective.

Analysis of γH2AX and carbonic anhydrase 9 in cervical cancer biopsies. Pretreatment cervical cancer biopsies were available from 47 patients with advanced cancer of the cervix. Paraffin-embedded tumor sections from 19 patients were available from a previous analysis of hypoxic fraction at the BCCA (16), 8 additional sections were obtained from a previous analysis of changes in tumor kinetics during treatment (17), and 19 slides were available from patients treated at the National Cancer Center in Korea. Tumors ranged from FIGO stage Ib to IIIb, and the majority of tumors were squamous cell carcinomas.

Sequential incisional biopsies were taken during the course of therapy for advanced cervical cancer as part of a prospective clinical trial, and samples from 8 patients were available for analysis of γH2AX (17). Tumors were FIGO stage Ib to IIIa, and 6 of 8 were squamous cell carcinomas. These patients received weekly cisplatin (40 mg/m²) to a maximum of five cycles and concurrent external beam pelvic radiation of 45 Gy in 25 fractions over 5 weeks. Patients also participated in a separate protocol of 5% carbogen inhalation during external beam radiotherapy (18). Biopsies were obtained before the day's radiation treatment and were immediately fixed in formalin.

To detect regions associated with the presence of tumor hypoxia, slides were costained for carbonic anhydrase 9 (CA9), which is expressed by hypoxic cervical cancer cells (14, 16). Paraffin-embedded tumor tissue (5 μm thick) was prepared, and after dewaxing and rehydrating in graded alcohols, slides were immersed in high pH target retrieval solution (DAKOCytomation) in a 95°C water bath for 30 min. Slides were washed in TBS, blocked for 10 min, and incubated with rabbit polyclonal anti-CA9 (1:400 dilution; Novus Biologicals) for 30 min. After rinsing, slides were incubated with Alexa 488 goat anti-rabbit antibody (1:200 dilution; Molecular Probes) for 30 min. Samples were then fixed in 2% paraformaldehyde, washed, and incubated with mouse anti-γH2AX antibody (1:500 dilution; Upstate Biotechnology) for 1 h. Slides were rinsed and incubated 30 min with Alexa 594 goat anti-mouse secondary antibody (1:200 dilution; Molecular Probes). To stain nuclei, slides were submersed in 0.05 μg/mL DAPI for 5 min, rinsed, and finally mounted with 10 μL Fluorogard. Tumor sections were viewed using a Zeiss Axioplan 2 epifluorescence microscope using ×10, ×40, and ×100 Plan Neofluor objectives. As this was considered a feasibility study, no specific procedures were adopted to ensure that images were obtained randomly throughout the section.

Because sections rather than whole cells were scored, cells with one or more γH2AX foci were counted as positive. Obvious regions of necrosis or areas of stroma were not included in the analysis. Foci results are presented as averages of scores for several high-power images. Digitized images (8-12 images containing 50-100 nuclei each) were scored for the percentage of nuclei lacking γH2AX foci. After thresholding to obtain binary images (16), Northern Eclipse software was used to quantify the number of pixels that were positive for CA9 or γH2AX.

Following antigen retrieval, some slides were incubated for 1 h with FITC-conjugated rabbit anti-caspase-3 antibody (1:200 dilution; BD Biosciences) or rabbit polyclonal anti-cytokeratin (wide-spectrum 20622, 1:500 dilution; DAKOCytomation) followed by staining with Alexa 488-conjugated goat anti-rabbit antibodies. Slides were then incubated with anti-γH2AX antibodies as described above. Selection of images for scoring was based on the presence of caspase-3-positive cells in the field.

Development of γH2AX foci in xenograft tumor cells 24 h after cisplatin treatment. Expression of γH2AX increased over the first 24 h after cisplatin injection in mice bearing WiDr xenografts (Fig. 1A-E). The signal subsequently declined but remained significantly higher than background 72 h after treatment (Fig. 1E). The increase in γH2AX was evident primarily in S/G₂-phase tumor cells, and a minimal increase was seen for normal cell components; cells with 2N DNA content are signified by the arrow in Fig. 1A. A dose-response relationship for γH2AX was confirmed by flow cytometry analysis of average cell fluorescence related to the unexposed control (Fig. 1F) and by analysis of the fraction of antibody stained cells that fell within the control window (Fig. 1G).

SiHa and WiDr tumors were excised 24 h after cisplatin treatment, and single cells were analyzed for clonogenic surviving fraction (Fig. 1H). As expected, cells from SiHa tumors were more sensitive to killing by cisplatin than cells from WiDr xenografts. When cells from these tumors were fixed and analyzed for the percentage of cells that retained fewer than three γH2AX foci 24 h after cisplatin injection, SiHa cells exhibited more cells with residual γH2AX foci than WiDr cells (Fig. 1I). A good correlation was observed between surviving fraction and fraction of cells with fewer than three γH2AX foci (Fig. 1J), and the relationship between residual γH2AX and cell killing was similar for SiHa and WiDr tumors. However, the slopes of the lines for cells treated as xenograft tumors were higher than those previously reported for SiHa and WiDr monolayers (ref. 10; Fig. 1J, dotted line), indicating that the fraction of cells with residual γH2AX underestimated cisplatin toxicity for cells recovered from xenograft tumors. It is known that γH2AX foci only form when cells with cisplatin-induced DNA cross-links move through S phase (10), so it is likely that excision of tumors 24 h after treatment would not allow adequate time for noncycling cells to reenter the cycle.

Response of xenograft tumors to the combination of cisplatin and fractionated radiation. To simulate the clinical protocol for treatment of advanced cervical cancer at the Vancouver Cancer Centre, xenograft tumors from three human tumor cell lines were grown subcutaneously in immunodeficient mice. When tumors reached a size between 0.25 and 0.5 g, mice were injected intraperitoneally with a single dose of cisplatin before three daily fractions of 2 Gy as illustrated in Fig. 2A. SiHa and HCT116 tumors were given 5 mg/kg cisplatin and mice with more resistant WiDr xenografts received 10 mg/kg cisplatin. This treatment is not curative and causes only a minimal growth delay; however, the protocol was designed to provide a way to assess response early into the course of treatment when clonogenic cells are still present. Twenty-four hours after the final fraction of 2 Gy, tumors were excised and single cells were plated to measure clonogenic survival or fixed and analyzed for γH2AX antibody binding. Flow cytometry analysis showed the expected decrease in the fraction of cells with control levels of γH2AX after single or combined treatments (Fig. 2B-E).

For cells obtained from control and cisplatin-treated tumors, the fraction of cells with control levels of γH2AX was similar to the fraction of cells with fewer than three foci per cell as measured by visual scoring, and both were correlated with clonogenic survival (Fig. 3A-C). However, after exposure to X-ray, the fraction of cells with control levels of γH2AX did not correlate well with clonogenic survival. For SiHa tumors, predictive ability was improved by visual scoring of cells with foci because irradiated cells that score as positive for γH2AX foci typically have relatively few residual foci (Fig. 3E). Visual scoring is known to be more sensitive than flow cytometry for detecting low numbers of foci against a background of endogenous γH2AX. Flow cytometry was better able to detect residual γH2AX after cisplatin treatment because cells that retain γH2AX foci after cisplatin treatment typically have more foci (Fig. 3F).

The relation between residual γH2AX expression and cell survival is shown for radiation alone, cisplatin alone, or the combination (Fig. 3G-I). In this study, as in others (19, 20), cisplatin and radiation appeared to act independently, producing an additive response. The slopes of the lines varied for the three tumor types, indicating that the relationship between the fraction of cells that retained foci and the fraction of cells that ultimately died was tumor type dependent.

Relation of tumor cell perfusion to response to cisplatin or X-ray. Hoechst 33342 perfusion followed by fluorescence-activated cell sorting was used to provide populations of tumor cells close to functional blood vessels as well as those distant from vessels, representing well-oxygenated or hypoxic tumor cells (11). When cells were recovered from tumors 24 h after exposure to radiation or injection with cisplatin, cells from poorly perfused regions of SiHa tumors were more resistant to killing by both cisplatin (Fig. 4A) and radiation (Fig. 4B). However, for experiments that employed the protocol in Fig. 2A, there was no significant difference between the responses of cells recovered from well-perfused and poorly perfused regions (Fig. 4C-H). This is not surprising considering that 72 h had elapsed between the start of treatment and tumor cell recovery. Tumor cells perfused at the start of cisplatin or radiation treatment are unlikely to be the same cells perfused at the time of Hoechst 33342 injection and tumor excision. From these results, we expected that retention of γH2AX foci in our clinical samples would not differ in CA9-negative or CA9-positive regions of tumors.

Feasibility of measuring residual γH2AX in cervical tumor biopsies. Dewaxing slides followed by antigen retrieval and antibody staining was found to be effective for fluorescence detection of γH2AX foci together with CA9 staining in 47 archival paraffin-embedded cervical cancer sections. The fraction of cells with >1 γH2AX focus ranged from 0 to 62%, and on average, only 25% of cells scored as γH2AX positive (Fig. 5A). The majority of foci-positive cells expressed one to five γH2AX foci (data not shown). A positive relation between CA9 staining and γH2AX staining was observed by comparing the number of positive pixels in each field (Fig. 5B). However, this relationship is likely to be dominated by cells with pan-nuclear γH2AX staining as described below.

Cells with pan-nuclear γH2AX staining, consisting of hundreds of individual foci, were observed in these tumor sections, typically representing <5% of the cells. Although often randomly located, there was a greater likelihood of observing these cells in regions that stained positive for CA9 (Fig. 5C), possibly indicating that cell death rate is greater in hypoxic regions. Cells with pan-nuclear γH2AX staining also stained positive for the apoptotic marker, activated caspase-3 (Fig. 5D and E). In sections from 3 patient samples costained for γH2AX and activated caspase-3, 56% of cells with pan-nuclear γH2AX staining were positive for activated caspase-3 (Fig. 5D).

Increase in γH2AX expression during radiochemotherapy. Sequential biopsies from 8 patients with advanced cancer of the cervix were taken before and during the course of weekly cisplatin and daily doses of 2 Gy. For 8 pretreatment samples, 15% of the nuclei were scored as γH2AX positive with a range of 8% to 20% (Fig. 6A). At the first sample time during treatment, this fraction increased to 57% (range, 35-71%). However, even at long times after the start of treatment, many nuclei still lacked γH2AX foci. Note that, in Fig. 6A, no information is provided on the number of foci per cell, and this increased in both CA9-negative and CA9-positive cells as treatment time increased. Unlike cell sorting results that showed little difference in the percentage of cells with foci in well-perfused or poorly perfused xenograft tumor cells, CA9-positive cells in all of these clinical biopsies were at least twice as likely to contain γH2AX foci as CA9-negative cells both before treatment and at the first sample time after the start of treatment. Intratumor heterogeneity in the fraction of foci-positive cells before or after treatment was relatively small for some patients (Fig. 6B, patient 2) but larger for others (Fig. 6B, patients 1 and 3). Although no patient developed recurrent local disease, 3 of the 8 patients progressed with pulmonary metastases. There was no relation between foci retention and outcome in this small sample.

Cells lacking foci >2 weeks into treatment could represent tumor cells that have survived treatment, tumor cells that have lost the ability to phosphorylate H2AX, or tumor-infiltrating normal cells. As treatment progressed, the fraction of tumor cells declined, reducing the accuracy of analysis of residual foci in the relevant tumor cell population. To illustrate this problem, loss of cytokeratin-positive cells over the course of treatment is shown for 1 patient in which the tumor sections were costained for cytokeratin and γH2AX (Fig. 6C-H).

No change in radiation-induced phosphorylation of γH2AX after 10 daily fractions of 2 Gy. To examine the possibility that tumor cells may not retain the ability to phosphorylate H2AX at longer times on treatment, SiHa cells were grown as multicellular spheroids and subjected to daily 2 Gy exposures. As expected, residual γH2AX expression measured before each daily treatment increased progressively over the first week, declined over the weekend when radiation was not delivered, and increased again during the second week (Fig. 7A). By the end of the second week, ∼15% of the cells recovered from spheroids still lacked γH2AX foci (Fig. 7B). However, at this point, cell viability was <0.01% (Fig. 7C). After 2 weeks, DAPI-stained nuclei were often enlarged and misshapen, and the samples invariably included nuclear fragments, micronuclei, and apoptotic cells (Fig. 7E and F). This raises questions about the ability of residual γH2AX to accurately predict response in populations of heavily damaged and dying cells. From these images, it was not possible to determine whether the cell membrane of cells that lacked foci was still intact. Loss of membrane integrity would allow entry of extracellular phosphatases and subsequent loss of ability to maintain H2AX phosphorylation. This experiment was repeated using WiDr spheroids. For these radioresistant cells, 15% survived 2 weeks of daily 2 Gy fractions, but 40% of the cells lacked foci (data not shown).

Although cellular debris and micronuclei complicated measurement of the fraction of cells lacking foci, flow cytometry allowed us to gate out debris and improve accuracy (Fig. 7H-K). Challenging spheroids to X-ray after 2 weeks of fractionated treatment indicated that almost all of the cells maintained the ability to phosphorylate H2AX (Fig. 7D and G). Relative to cells from untreated spheroids (Fig. 7H), <2% of spheroid cells given 9 daily fractions of 2 Gy still showed control levels of γH2AX when challenged with 2 Gy and analyzed 1 h later (Fig. 7K). In data not shown here, the rate of loss of γH2AX after exposure to 10 Gy was no different before the start of treatment or 2 weeks after 2 Gy daily exposures, consistent with results showing a similar rate of γH2AX accumulation over 2 weeks (Fig. 7A). Similar results were obtained using cells from WiDr spheroids. Therefore, previously irradiated but nonclonogenic tumor cells retain the same ability to phosphorylate H2AX in response to radiation damage.

Concurrent use of cisplatin with radiotherapy has become the standard of care in the treatment of patients with advanced cancer of the cervix (21, 22). Prediction of response to treatment therefore requires a method that is sensitive to tumor response to both agents. The potential of γH2AX to provide information on tumor response to treatment is only beginning to be explored (3). We previously showed that the fraction of tumor cells that retained γH2AX foci 24 h after cisplatin treatment was correlated with the fraction of cells that lost clonogenicity, and this correlation extended as low as 5% survival (10). Results using cisplatin-treated cells from xenograft tumors also show a correlation between the fraction that retained foci and the fraction that lost clonogenicity (Fig. 1J); however, some cells with fewer than three foci failed to survive so that the relationship between fraction with foci and surviving fraction was no longer close to unity. The stimulus for γH2AX formation after cisplatin treatment is replication fork collapse and subsequent double-strand break formation at sites of interstrand cross-links (10, 23). For cisplatin-treated tumors, 24 h appears to be insufficient time for all tumor cells to transit S phase and develop γH2AX foci. Extending cisplatin recovery time to 3 days (Fig. 2A) should have ensured that surviving cells would have an opportunity to enter the cycle and transit S phase. In fact, residual γH2AX was able to predict clonogenic survival of three xenograft tumors exposed to cisplatin only and analyzed 3 days later (Fig. 3A-C).

Cell loss and repopulation undoubtedly occur during the time between treatment and tumor excision. Cell survival after treatment of SiHa xenografts with 5 mg/kg cisplatin or WiDr xenografts with 10mg/kg cisplatin was only 20% when analyzed 24 h later (Fig. 1H) but 60% when analyzed 72 h later (Fig. 3A and B). This explains the decline in γH2AX at 72 h (Fig. 1E). Cell recovery per gram tumor decreased as much as 50% in some tumors by 3 days after the start of treatment, but this factor was not included in the measure of survival because the goal was to compare γH2AX staining and clonogenicity in the same tumor cell population. Fortunately, daughter cells inherit residual γH2AX foci (6), so that cell division does not eliminate this signal, and γH2AX remaining 24 h after each radiation dose accumulates with each fraction (Fig. 7A). Reduction in γH2AX over the weekend is substantial and would be explained by both cell loss and additional time for repair. Lethal damage that results in nuclei without foci includes damage that occurs during tumor disaggregation as well as elimination of γH2AX foci into micronuclei (24). The remaining nucleus could lack γH2AX foci and be scored as viable, but loss of critical DNA with the micronucleus would lead to cell death. Micronuclei were particularly apparent after exposure of cells to fractionated irradiation (Figs. 3E and 7F). Although the fraction of cells that retained γH2AX foci was a useful indicator for cisplatin killing 72 h after drug treatment, the fraction with residual γH2AX foci underestimated lethal radiation damage in two of three xenograft tumors exposed to fractionated radiation.

Underestimating lethal damage is not a problem if the degree to which damage is underestimated is similar for different tumor types. However, the variation in slopes in Fig. 3G to I indicates significant differences in behavior in these three tumors. Therefore, γH2AX measured 3 days into treatment can provide useful information on the response of a specific tumor type, but comparisons between different tumor types appear problematic. Although nonclonogenic cells remain capable of phosphorylating H2AX (Fig. 7), the efficiency of phosphorylation may be compromised to different extents in different cell lines, cell recruitment into the cell cycle may vary, and cell loss rate may differ. We attempted, unsuccessfully, to improve the ability to detect cells with residual radiation DNA damage by costaining irradiated cells for RAD51 foci. Although almost all cells with γH2AX foci also exhibited RAD51 foci 24 h after a single 2 Gy fraction, after several fractions of 2 Gy, fewer than half of the SiHa cells with residual γH2AX foci also exhibited RAD51 foci. However, like γH2AX foci, RAD51 foci are also sequestered into micronuclei (25).

Analysis of sections from cervical cancer biopsies established the feasibility of fluorescence staining of γH2AX in formalin-fixed tumors. Sections were also successfully costained for an endogenous hypoxia marker (CA9), activated caspase-3, or pan-cytokeratin. There was sufficient resolution to visualize individual foci under high-power magnification. An initial concern was that endogenous γH2AX expression might be too high to allow detection of treatment-induced residual foci (26). Although only half of the pretreatment biopsies exhibited >25% cells with one or more foci, some tumors did show high backgrounds (Fig. 5A). In this study, only cells that lacked foci were scored because our assumption was that only these cells would have the potential to retain clonogenic capacity. However, additional sensitivity could be obtained by counting foci per nucleus because this appeared to increase with time on treatment. A second major problem was intratumor heterogeneity in γH2AX staining made more difficult by the necessity of scoring foci under high magnification. Clusters of cells with foci were often seen, perhaps representing regions undergoing turnover and remodeling, and regions deficient in nutrients that border necrosis. Hypoxia and poor drug penetration should protect regions distant from the blood supply from damage by radiation and cisplatin, as shown in Fig. 4A and B, so these areas might be expected to retain fewer treatment-induced foci. However, there is evidence that DNA repair in hypoxic regions could be compromised (27), which could increase the number of cells with foci. Our results also indicate that cells in hypoxic regions contain more endogenous and treatment-induced foci than cells in well-oxygenated regions (e.g., Figs. 5B and C and 6A). This observation could be important if the absence of γH2AX foci in hypoxic cells can be taken as an indication of their viability. With the exception of the chemical hypoxia markers EF5 and pimonidazole, methods used in the clinic to quantify tumor hypoxia are unable to provide important information on the viability of hypoxic cells.

As discussed above, intratumor heterogeneity is a major limitation in analysis of residual γH2AX foci in clinical biopsies. It is difficult to ensure that a single biopsy is representative of the tumor or even that cells scored for foci are representative of the biopsy. There are several approaches that might help to diminish these problems. It should be possible to use fine-needle aspirate biopsies or disaggregated incisional biopsies to obtain more representative tumor samples. When analyzing sections from a single biopsy, it is recommended that procedures be adopted to ensure that many fields are scored and fields are chosen randomly. In addition, it is important to ensure that the response of tumor cells only is scored perhaps by limiting analysis to tumor cells based on DNA content or cytokeratin expression. Analysis of residual γH2AX within a week of the start of treatment is also recommended to minimize tumor cell loss and avoid problems with tumor-infiltrating normal cells. Our results indicate that perhaps 15% of nuclei may lack foci, although they have lost clonogenicity (Fig. 7B and C). Unless nonviable cells can be excluded by other means, residual γH2AX may only be an accurate predictor of response when cell survival exceeds ∼15%. This would again limit analysis of residual foci to the first few days after the start of treatment. Although there is reason to believe that retention of γH2AX foci will be able to provide an early indication of response to a variety of DNA-damaging agents, application of methods for automated foci quantitation (28), limiting analysis to tumors with low endogenous expression of γH2AX foci, and ensuring that foci analysis is representative of the biopsy will be critical for the success of this approach.

No potential conflicts of interest were disclosed.

We thank Denise McDougal and Darrell Trendall for expert technical assistance and Dr. Jin Zhao for help with tissue staining and foci analysis.