Previous reports have suggested that measuring radiosensitivity of normal and tumor cells would have significant clinical relevance for the practice of radiation oncology. We hypothesized that radiosensitivity might be predicted by analyzing DNA end-binding complexes (DNA-EBCs), which form at DNA double-strand breaks, the most important cytotoxic lesion caused by radiation. To test this hypothesis, the DNA-EBC pattern of 21 primary human fibroblast cultures and 15 tumor cell lines were studied. DNA-EBC patterns were determined using a modified electrophoretic mobility shift assay and were correlated with radiosensitivity, as measured by SF2. DNA-EBC analysis identified a rapidly migrating ATM-containing band (identified as “band-A”) of which the density correlated with SF2 (0.02 ≤ SF2 ≤ 0.41) in primary fibroblasts (r2 = 0.77). The DNA-EBC pattern of peripheral blood lymphocytes was identical to that of fibroblasts. In addition, band-A density correlated with SF2 (0.35 ≤ SF2 ≤ 0.80) in 15 human tumor cell lines (r2 = 0.91). Densitometry of other bands, or total DNA-EBC binding, correlated more poorly with SF2 (r2 < 0.45). These data indicate that DNA-EBC analysis may be a practical, clinically relevant predictor of tumor and primary cell radiosensitivity.

Predicting normal tissue and tumor radiosensitivity has been the subject of intensive investigation, but has yet to be routinely integrated into radiotherapy (reviewed in Ref. 1). Complication risks for an individual irradiated patient can be predicted currently only by the complication rates seen in similar populations. This assessment fails to account for variation in the DNA repair capacity of the individual. Thus, the “standard” dose for a population may be inappropriately low for some patients whose tumors are resistant, whereas this same dose may carry a relatively high risk of complications for others whose normal tissues are sensitive. Modeling data from our group has shown previously that, if the most radiosensitive patients could be identified by a predictive assay, the remaining patients could be safely treated with higher doses (2). Similarly, those predicted to have the greatest risk of complications because of unusual radiosensitivity can have relatively complex/expensive treatment techniques used in an effort to reduce toxicity. In lung cancer radiotherapy, this might involve the use of respiratory gating, using time-consuming but more reliable patient set-up techniques and perhaps the use of radioprotectors such as amifostine. As more resource-intensive, highly conformal therapies become available (e.g., intensity modulated radiotherapy, proton therapy, and so forth), they could be first applied to patients at greatest risk for side effects. Thus, the development of a good predictor of normal-cell radiosensitivity is becoming increasingly important.

Similarly, predicting tumor radiosensitivity has significant clinical applicability. If such prediction could be done accurately, radiation doses could be tailored to the radiocurability of individual tumors. In addition, such an assay could be helpful in determining the optimal doses and schedules of biological and chemotherapeutic radiosensitizers.

Several groups have published modeling data demonstrating the clinical value of predicting normal tissue and tumor radiosensitivity (3, 4, 5, 6, 7, 8, 9, 10). These data indicate that both tumor control probability (TCP) and normal tissue complication probability (NTCP) can potentially be improved by individualizing treatment according to the results of predictive assays. The benefits were dependent on the predictive power of the assay, but clinically meaningful benefit could be demonstrated even if the correlation between test results and TCP or NTCP is within the range of 0.4 and 0.6. Even if the assay only stratifies patients and tumors into only three risk categories (low, medium, and high), the potential gain in TCP was predicted to be between 22% and 33%.

Radiosensitivity of cells in vitro has been shown to be predictive of in vivo radiosensitivity in both normal and tumor tissue. Radiosensitivity has been most often measured by determining the surviving fraction after 2 Gy (SF2). This dose has historically been chosen because curative radiotherapy is typically delivered using a daily fraction size of 1.8–2.0 Gy. However, the quantification of SF2 is time consuming, requiring weeks to grow explants and determine radiosensitivity, and expensive. These problems significantly limit the clinical applicability of the assay. In principal, however, if SF2 could be accurately and reproducibly determined, radiation dose could be individualized to both increase TCP and decrease NTCP. Also by combining tumor and normal radiosensitivity, tumor hypoxia, and proliferative potential measured, a very accurate TCP/NTCP model could be constructed (1).

Radiotherapy kills cells primarily through DNA damage, and DNA double-strand breaks are thought to be the lethal lesion caused by radiation (11, 12, 13, 14, 15). The presence of DNA double-strand breaks activates a variety of signal transduction cascades, such as ataxia telangiectasia mutant (ATM) and DNA-dependent protein kinase (DNA-PK) pathways, that alter cell functions (e.g., cell cycle) to allow time for DNA repair. Both of these proteins can be found at DNA double-strand breaks, and inactivation of either kinase results in profound radiosensitivity (16, 17, 18, 19). Obviously, DNA-repair enzymes must also be present at DNA breaks for repair to occur. Thus, we hypothesized that an analysis of DNA-end-binding complexes (EBCs) would be predict radiosensitivity.

We tested our hypothesis using a modified electrophoretic mobility shift assay. Nuclear extracts from cells of interest were mixed with a radiolabeled oligonucleotide. Electrophoresis of this mixture under nondenaturing conditions, followed by autoradiography, led us to identify 10 DNA-EBC bands in the nuclear extracts of unirradiated human cells. The pattern/density of DNA-EBCs was compared in primary and tumor cells with a variety of known radiosensitivities. This approach does not require that cells grow in vitro, nor does it require radiation. Several practical aspects of DNA-EBC analysis were also explored.

Cell Culture.

The cell lines used in this study and their radiosensitivities are shown in Table 1. Cells were grown as monolayers in DMEM and maintained as logarithmically growing cultures. Primary fibroblast cultures were supplemented with 20% fetal bovine serum, whereas tumor cell cultures were supplemented with 10% fetal bovine serum. The primary human fibroblast cell lines were obtained on protocols approved by our Institutional Review Board. One protocol involved the acquisition of primary fibroblasts from patients with abnormally severe radiation reactions; the other involved the generation of primary fibroblast cultures from patients heterozygous for BRCA mutations (20) but no diagnosis of cancer. Five “BRCA normal” cell lines without known mutations in DNA repair/signaling genes were also included (C42, C74, S23, C37, or C80).

SF2 Determination.

SF2 was determined for each cell line by methods described previously (20). Radiosensitivity was measured by at least three independent clonogenic assays. These were performed within approximately 5–10 population doublings of the population used for DNA-EBC analysis. This was done to minimize possible senescence-related changes in SF2 in primary cells and any possible genetic drift in tumor cell populations. Our data generally agree with the published SF2 values for these cell lines.

Nuclear Extract Preparation.

Cells were rinsed twice with ice-cold PBS, harvested by scraping with a rubber policeman, and centrifuged at 800 rpm for 10 min at 4°C. The cells were resuspended in 400 μl of cold lysis buffer [10 mm HEPES (pH 7.9), 10 mm Kcl, 0.1 mm EDTA, 0.1 mm EGTA, and 1 mm DTT], 2 μg/ml leupeptin, 2 μg/ml aprotinin, and 0.5 mm phenylmethylsulfonyl. After a 10-min incubation on ice, 12.5 μl of 10% NP40 was added, and the mixture was vortexed for 5 s. The lysate was centrifuged for 5 min at 4°C (14,000 rpm). The supernatant was removed and stored as a cytosolic extract. The pellet was resuspended in 30 μl of extraction buffer [20 mm HEPES (pH 7.9), 400 mm NaCl, 1 mm EDTA, 1 mm EGTA, 1 mm DTT, 2 μg/ml leupeptin, 2 μg/ml aprotinin, and 0.5 mm phenylmethylsulfonyl] and extracellular membrane disruption confirmed microscopically. Samples were then vortexed thoroughly and incubated on ice for 30 min. Every 10 min, the preparation was vortexed. After 30 min, the extract was centrifuged for 10 min at maximum speed (14,000 rpm) at 4°C. The supernatant, designated as nuclear extract, was divided into aliquots and stored at −70°C (21). Because SF2 can be dependent on the length of time in culture due to genetic drift of the population, it was considered important to obtain nuclear extracts for cells that were as near as possible in passage number to those used for SF2 determination. Nuclear extracts were made from peripheral blood lymphocytes using a slight modification of the above-described procedure.

Detection of DNA-EBCs.

The DNA-EBC assay, which is a modified electrophoretic mobility shift assay, was performed using the method we described previously for rodent nuclear extracts (22). Briefly, nuclear extracts (0.2–1.0 μg) were incubated with 0.5 ng of 32P-labeled oligonucleotide probe (144-bp fragment of pUC18) for 20 min at room temperature. Plasmid DNA (1 μg of closed circular pUC18 DNA) was used as a nonspecific competitor in a final volume of 20 μl of binding buffer [10 mm Tris-Hcl (pH 8.0), 0.1 mm EDTA, 150 mm NaCl, 1 mm DTT, 1 mm phenylmethylsulfonyl, and 10% (v/v) glycerol]. The protein concentration of each nuclear extract was estimated as we have described previously (22), identical amounts of protein were loaded into each lane, and the electrophoretic mobility of the protein-DNA complexes was analyzed in 5% PAGE gel at 20–25 mA in TBE buffer [45 mm Tris-Hcl (pH 8.0), 45 mm boric acid, and 1 mm EDTA]. The dried gel was subjected to autoradiography. The technique has been widely described by other authors as a way to analyze DNA end binding proteins (23, 24).

pUC18 plasmid was digested with PvuII and EcoRI to generate a 144-bp probe. The probe was purified by 8% PAGE and electroeluted in 10 mm Tris (pH 8.0) and 0.1 mm EDTA. The eluted DNA fragments were concentrated in a Qiaquick gel extraction column (Qiagen, Valencia, CA). The probe was 32P-labeled using the Klenow fragment of DNA polymerase I in the presence of [32P]dATP (Dupont, NEN, Boston, MA), and the unincorporated nucleotides were removed by chromatography on Sephadex G-50 spun columns (25).

DNA-EBC analysis was done at least in triplicate [typically using a phosphoimager (Typhoon 9400; Amersham Biosciences, Piscataway, NJ) and ImageQuant image analysis software (Sunnyvale, CA)], and densitometry was performed. The density of each band was corrected for the density of the corresponding lane in a region below band-A far from a DNA-EBC. It was not possible to run all samples on a single gel. Comparison between gels was accomplished by normalizing all of the band densities to a normal control (usually C80) that was present on all of the gels.

The mixing studies shown in Fig. 5 were performed by first mixing nuclear extracts in the indicated proportion based on nuclear protein concentration. Radiolabeled oligonucleotide was added to the extract mixture, followed by electrophoresis and autoradiography.

Correlations between band-A densities and SF2 were calculated using linear regression, along with their significance expressed as a P. We considered other types of regression analyses, because there was no biological reason to assume that the relationship between band-A density and SF2 should be linear. The r2 value for primary cells was slightly better (0.85) for an exponential fit than for a linear regression (0.77). However, when the data from primary and tumor cells were pooled as shown in Fig. 6, a linear regression was a better fit. Therefore, for consistency, we chose linear regression analysis for all of the results.

DNA-EBC Pattern Predicts SF2 of Primary Fibroblasts.

A representative DNA-EBC analysis is shown in Fig. 1,A. We noted that there were at least 10 bands present in DNA-EBC gels from normal primary human fibroblasts, but the relative intensity of each band could vary significantly from cell line to cell line. It was noted that the relative intensity of the band labeled “band A” decreased as SF2 decreased. Densitometry was performed, and the relative density of band A was plotted versus SF2 (Fig. 1,B). This analysis demonstrated a strong statistical correlation between SF2 and band-A density (r2 = 0.77; P < 0.000005). This analysis included cell lines with marked radiosensitivity such as ATM mutants and BRCA1 homozygous mutants, intermediate radiosensitivity such as BRCA heterozygotes and some cell lines from patients with marked radiation reactions, and normal radiosensitivity in 2 unrelated normal lines (Table 1).

Because AT cells are particularly radiosensitive and because the ATM protein was thought to bind at sites of DNA breaks, it was hypothesized that ATM might be an important component of band A. To test this hypothesis we determined the DNA-EBC pattern of 2 cell lines derived from patients with AT. As can be seen in Fig. 2, band A was essentially undetectable (Fig. 2, Lanes 2 and 3) in AT cells. In fact, two major (Fig. 2, A and B) and four minor (Fig. 2, thin arrows on the left) bands were missing in both ATM mutants compared with a normal control (Fig. 2, Lane 1). One unique band was observed in the ATM cells (Fig. 2, double arrow, right). Also, the relative intensity of bands was different in AT cells compared with controls, with some bands relatively more intense and others less intense. This suggested that mutations in ATM cause widespread changes in the complexes that form at DNA double-strand breaks.

DNA-EBC Analysis of Lymphocytes Is Similar to That of Fibroblasts.

To implement DNA-EBC analysis easily in the clinic, the assay would be better done on more readily available samples than fibroblasts, such as peripheral blood lymphocytes. As a first step in developing a lymphocyte-based DNA-EBC analysis, we compared the DNA-EBC pattern in two primary fibroblast lines (Fig. 3,A, Lanes 1 and 2) with the DNA-EBC patterns from four peripheral blood lymphocytes samples from unrelated individuals (Fig. 3,A, Lanes 3–6). The DNA-EBCs from fibroblasts and lymphocytes were indistinguishable. Fibroblasts and lymphocytes from 2 individuals were also obtained. One patient (C84l Fig. 3,B, left) was heterozygous for an inactivating mutation in ATM, and the other patient heterozygous for a deletion in BRCA1 (C85, Fig. 3,B, right). The DNA-EBC pattern was found to be similar for lymphocytes and fibroblasts derived from these patients. There are also unusual bands (Fig. 3 B, arrows) in each lymphocyte/fibroblast pair that are not seen in the C80 control. Whereas the biological significance of these bands is unclear, these data suggest that DNA-EBC analysis of lymphocyte nuclear extracts can predict SF2 of fibroblasts.

DNA-EBC Pattern Predicts SF2 of Human Tumor Cell Lines.

It might be expected that the genetic variability of human tumors, or their genetic divergence from primary cells, could complicate DNA-EBC analysis. Also, the SF2 of tumor cells is often much higher than that of primary cells, which might exceed the functional range of the assay. Therefore, the DNA-EBC pattern of human tumor cell lines was determined for lines with a wide variation in SF2 (0.35 to 0.80). Fig. 4,A shows representative DNA-EBC analysis. Significant variability was seen in the DNA-EBC pattern of these cells in the bands above band A. However, the band A density continued to correlate strongly with SF2. When the densitometry of band-A was plotted against SF2 (Fig. 4 B), a very strong correlation with band A density was seen (r2 = 0.91; P < 0.000005). These data demonstrate that band A density was a good predictor of SF2 in tumor cell lines as well as normal tissues.

It was possible that either total DNA end binding capacity (i.e., the total density of a lane) or the density of another DNA-EBC might correlate well with SF2 in a manner similar to that found for band-A. To test this hypothesis, densitometry was performed on either the entirety of the lane (below the well to the bottom of band-A) or bands that were easily seen and separated from neighbors (band-B or band-D as shown in Fig. 2) on each of the three DNA-EBC analyses from tumor cell lines. The results of the densitometry correlated poorly with SF2 for any of these parameters, with the correlation coefficients (r2) for linear regression of 0.18, 0.44, and 0.40 for total, band-B, and band-D, respectively (data not shown). We concluded that band-A density correlated better than any other DNA-EBC component with SF2.

DNA-EBC Pattern Is Accurate Despite Contamination with Cells of Different SF2.

Determining DNA-EBC patterns in tumors is likely to be more complex than for cell lines. In vivo, tumors contain both malignant and normal cells. In preclinical animal models, there will also be contamination with normal rodent cells. Therefore, it was important to determine the sensitivity of the assay to contamination with normal cells, because very small changes in DNA-EBC density could lead to a relatively large error in predicted SF2. For this reason, nuclear extracts from normal human fibroblasts (SF2 = 0.4) were mixed with nuclear extracts from cells with a homozygous mutation in ATM and DNA-EBC analysis performed on the mixture (Fig. 5,A). Densitometry demonstrated that the predictive power of DNA-EBC analysis was proportional to the level of contamination. Thus, if both the SF2 and proportion of normal cells within a tumor specimen is known, the SF2 of the tumor cells can be accurately predicted. Similar to the human study, the density of band-A was proportional to the level of contamination with rodent cells (Fig. 5,B) as might be encountered in the study of xenografts. Note that rodent nuclear extracts demonstrate only a single DNA-EBC as we have reported previously (22). Perhaps most interesting was the result of mixing mouse nuclear extracts with those from AT cells (Fig. 5 C). Mouse extracts have no effect on band A at any mixing ratio.

Our data demonstrate that the density of band A is an excellent predictor of primary fibroblast SF2 over a range of radiosensitivities (0.02 ≤ SF2 ≤ 0.41). Whereas the number of samples is small, analysis of Fig. 1,B demonstrates that band-A density analysis can distinguish between three groups, low SF2 (SF2 ≤ 0.1), intermediate SF2 (0.15 ≤ SF2 ≤ 0.33), and high SF2 (SF2 > 0.33). Combining Fig. 1,B and Fig. 4,B (Fig. 6) also clearly demonstrates the predictive power of band-A density. This suggests that band-A density can be used as an intermediate marker for patient intrinsic radiosensitivity. Previous reports have shown that SF2 of patient fibroblasts is predictive of radiation-induced late toxicity. It remains to be determined how well DNA-EBC analysis will predict toxicity; however, these observations and the ease of the assay (particularly when done on lymphocytes) suggests that a large-scale study could be undertaken to determine the predictive power in a much larger sample and correlate this with complication rate.

Using a clonogenic survival assay of dermal fibroblasts, previous reports have correlated SF2 with the degree of skin fibrosis after breast radiation (26). However, acute reactions and skin erythema were not predicted by this assay. Another study (27) also demonstrated that fibroblast SF2 correlated with the maximal toxicity grade for patients irradiated for breast cancer. In selected cases, patients with severe DNA repair deficits such as ataxia telangiectasia can have their treatment tailored to their intrinsic radiosensitivity with good results. In one report, a patient with an inherited defect in DNA repair (ataxia telangiectasia) with medulloblastoma was treated with 0.6 Gy fractions to 15 Gy, based on the measured SF2 of his fibroblasts (28). The sibling of this patient had severe toxicity when treated with standard radiotherapy doses, whereas the patient with individualized therapy had good local control and no severe side effects after 9 months. This demonstrates that, at least in rare patients with severe repair deficits, treatment might possibly be individualized based on SF2.

A study by Stausbol-Gron and Overgaard (29) compared SF2 of tumor cells and local-regional control. In 38 patients, tumors were biopsied, explants cultured in soft agar, and SF2 determined. They found no correlation between SF2 and locoregional control for these patients treated with radiation alone. Interestingly, they also found no correlation between tumor cell SF2 and fibroblast SF2, suggesting that these are independent parameters. One caveat of their observation was that their plating efficiency was extremely low (only 1 of 38 was >1%, and 7 were ∼0.01%), suggesting that they may have measured SF2 only on a small subset of tumor cells. In fact, 5 tumors had SF2 of 1.00, yet 2 of these patients had local control! Therefore, SF2 of small tumor subsets must not be representative of the entire tumor population.

A larger study (30) of 84 curatively treated patients with head and neck cancer did demonstrate a significant correlation between tumor SF2 and local control (P = 0.036), but not survival. These patients were treated with a variety of radiation-containing regimens, and some patients received surgery or chemotherapy. The median follow-up was 25 months in this study, in contrast with the Overgaard study, which had a median follow-up of only 14 months. They did not discuss the plating efficiency for the tumors in their study. Thus, SF2 is predictive of local control in head and neck cancer in the largest study with the longest follow-up.

SF2 may not be predictive of local control for all tumors. For example, SF2 was not predictive of outcome for glioblastomas (31). There are many reasons why SF2 will not predict tumor cure for all patients. Any effects that are due to the tumor environment such as hypoxia (32), angiogenesis (33), tumor immunity (34), or cell cycle distribution (35) will not be predicted by standard SF2 studies. Interestingly, cell cycle effect can be predicted by DNA-EBC analysis.4 However, as we better understand and can predict intrinsic tumor response to therapy, we should be able to better triage patients for aggressive therapy. Certainly, the environment of normal cells should be more similar from patient to patient than the tumor environment, suggesting that DNA-EBC pattern may be a better predictor of toxicity than tumor response.

Band-A density also correlates very well with SF2 of tumor cell lines (r2 = 0.91), and the analysis of Fig. 4,B suggests that band-A density can at least dichotomize between tumors with SF2 above and below 0.60. This is potentially important, because other investigators have shown that tumor cell SF2 can predict outcome, particularly for head and neck and cervical cancers, but not gliomas (29, 30, 31, 36, 37). DNA-EBC analysis has a marked advantage over other assays in that tissue culture is not required. Fig. 6 is a composite of Fig. 1,B and Fig. 4 B, and shows that band-A density is extremely predictive of SF2 over the entire range of SF2 studied (0.02 ≤ SF2 ≤ 0.80; r2 = 0.89; P < 0.000005). Thus, band-A density is predictive of SF2 over the range of radiosensitivities likely to be encountered in clinical practice.

The mixing studies provide additional support for the idea that DNA-EBC analysis of tumors will be practical. Band-A density can be reproducibly and accurately measured when samples are minimally contaminated with cells of different SF2 or from other species. This is in contrast with PCR-based techniques that are extremely contamination sensitive. Interestingly, since rodent ATM and human ATM are highly conserved structurally, mixing nuclear extracts from these cell types might have resulted in the restoration of band A, especially when mixed 50:50. However, our observations that the presence of rodent proteins have no impact on human band-A density suggest that ATM activity (phosphorylation of human proteins) may be required for band-A assembly. ATM probably does not simply play a structural role in band A, because mouse, which is structurally very similar to the human protein, cannot assist with band-A assembly.

The mechanism by which band-A density relates to SF2 is unclear at present. The observation that band density rather than band migration speed (or perhaps broadness of the band) predicts SF2 has potential mechanistic implications, but these cannot be fully explored until all of the band components have been identified. However, we have demonstrated that band A contains at least ATM, Ku70, DNA ligase III, Rpa32, Rpa14, DNA ligase IV, XRCC4, WRN, BLM, RAD51, and p53. However, at least for the BRCA1 heterozygotes, band-A density did not correlate with the nuclear levels of any of these proteins, suggesting that band A assembly may depend on post-translational modification(s) of its components.4 This suggests that DNA-EBC analysis may provide functional information about many proteins simultaneously, in contrast with genomic or proteomic approaches that assess only mRNA or protein levels. This may also allow the detection of post-translational modifications that are important for DNA repair. It will also be important to determine which DNA repair deficits can be predicted by DNA-EBC, and this work is under way.

Several other bands also seemed to decrease with decreasing SF2, in particular those shown by arrows to the left of Fig. 2, all of which are absent from AT cells. We chose to study band A for several reasons. First, the other bands were either very faint or migrated very close to bands that did not have SF2-dependent density. Second, the analysis of band components is easier for more rapidly migrating bands (particularly when using supershift analysis4). Finally, the density of band A correlated better with SF2 than the density of any other easily measured band.

In summary, band-A density was predictive of SF2 over a wide range, although it seemed somewhat more predictive at higher SF2s. Band-A density of lymphocytes may provide an easily accessible source for SF2 prediction and so possibly predict normal-tissue toxicity from radiation. Band-A density was also very predictive of tumor cell SF2, with the potential to stratify individual tumors based on radiosensitivity. Finally, the relative insensitivity of band-A density to contamination may make DNA-EBC analysis more practical than other approaches.

Grant support: Department of Defense Breast Cancer Research Program Career Development Award, BC980154 (to T. A. B.), and NIH Research Grants CA-06294 (to A. M., U. R., R. E. M., M. D. S., and L. M.).

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.

Requests for reprints: Craig W. Stevens, University of Texas M. D. Anderson Cancer Center, 1515 Holcombe Boulevard, Unit 97, Houston, TX 77030. Phone: (713) 792-3400; Fax: (713) 794-5573; E-mail: cstevens@mdanderson.org

4

Unpublished observations.

Fig. 1.

DNA-end-binding complex (EBC) analysis predicts radiosensitivity of normal fibroblasts. A, representative DNA-EBC analysis of primary human fibroblasts with a variety of radiosensitivities. Note that not all primary cell lines described in Table 1 are included in this example. B, densitometry was performed on DNA-EBCs in triplicate, normalized to that of the C29 strain, and plotted versus SF2. The correlation coefficient for linear regression was 0.77. Data points represent the mean of at least three determination of both SF2 and DNA-EBC density.

Fig. 1.

DNA-end-binding complex (EBC) analysis predicts radiosensitivity of normal fibroblasts. A, representative DNA-EBC analysis of primary human fibroblasts with a variety of radiosensitivities. Note that not all primary cell lines described in Table 1 are included in this example. B, densitometry was performed on DNA-EBCs in triplicate, normalized to that of the C29 strain, and plotted versus SF2. The correlation coefficient for linear regression was 0.77. Data points represent the mean of at least three determination of both SF2 and DNA-EBC density.

Close modal
Fig. 2.

Comparison of DNA-end-binding complex (EBC) pattern from normal and AT cells. DNA-EBC pattern of normal human fibroblasts (C29), and fibroblasts from patients with ataxia telangiectasia. Those bands marked at the left were absent from ATM mutants. Bands A and B are major absent bands. The bands marked at the right were not affected by ATM mutation. Band D was the major band that was unaffected by ATM mutation. The double arrow (right top) demonstrates a band present in ATM mutant cells but absent in normal controls.

Fig. 2.

Comparison of DNA-end-binding complex (EBC) pattern from normal and AT cells. DNA-EBC pattern of normal human fibroblasts (C29), and fibroblasts from patients with ataxia telangiectasia. Those bands marked at the left were absent from ATM mutants. Bands A and B are major absent bands. The bands marked at the right were not affected by ATM mutation. Band D was the major band that was unaffected by ATM mutation. The double arrow (right top) demonstrates a band present in ATM mutant cells but absent in normal controls.

Close modal
Fig. 3.

The DNA- end-binding complex (EBC) pattern of lymphocytes is similar to that of fibroblasts. A, comparison of DNA-EBC pattern from fibroblasts (Lanes 1 and 2) and lymphocytes from unrelated individuals (Lanes 3–6). B, DNA-EBC pattern of lymphocytes and fibroblasts from a patient heterozygous for ATM mutation (left), and a patient heterozygous for an inactivating mutation in BRCA1 (right) compared with normal control. An arrow indicates bands seen in both lymphocyte/fibroblast pairs that are not seen in the C80 control.

Fig. 3.

The DNA- end-binding complex (EBC) pattern of lymphocytes is similar to that of fibroblasts. A, comparison of DNA-EBC pattern from fibroblasts (Lanes 1 and 2) and lymphocytes from unrelated individuals (Lanes 3–6). B, DNA-EBC pattern of lymphocytes and fibroblasts from a patient heterozygous for ATM mutation (left), and a patient heterozygous for an inactivating mutation in BRCA1 (right) compared with normal control. An arrow indicates bands seen in both lymphocyte/fibroblast pairs that are not seen in the C80 control.

Close modal
Fig. 4.

DNA end-binding complex (EBC) analysis predicts radiosensitivity of tumor cell lines. A, DNA-EBC pattern of 15 human tumor cell lines, loaded in order of SF2. B, band A density predicts SF2 of tumor cell lines (r2 = 0.91; P < 0.0001).

Fig. 4.

DNA end-binding complex (EBC) analysis predicts radiosensitivity of tumor cell lines. A, DNA-EBC pattern of 15 human tumor cell lines, loaded in order of SF2. B, band A density predicts SF2 of tumor cell lines (r2 = 0.91; P < 0.0001).

Close modal
Fig. 5.

DNA-end-binding complex (EBC) pattern is relatively insensitive to contamination with cells of different SF2. A, band A density of AT cells (which is very low) is stable until it is contaminated by >20% with proteins from normal human cells (C80). B, band A density of C80 normal human fibroblast extracts is stable when <10% contaminated with rodent proteins. C, rodent nuclear extracts do not affect the band A density of AT cell extracts. The difference in quality of A from B and C reflect a departmental change from X-ray film to digital imaging.

Fig. 5.

DNA-end-binding complex (EBC) pattern is relatively insensitive to contamination with cells of different SF2. A, band A density of AT cells (which is very low) is stable until it is contaminated by >20% with proteins from normal human cells (C80). B, band A density of C80 normal human fibroblast extracts is stable when <10% contaminated with rodent proteins. C, rodent nuclear extracts do not affect the band A density of AT cell extracts. The difference in quality of A from B and C reflect a departmental change from X-ray film to digital imaging.

Close modal
Fig. 6.

DNA-end-binding complex (EBC) density versus SF2 for all human cell line. A composite of Fig. 1,B and Fig. 4 demonstrating the correlation between DNA-EBC and SF2 is quite good (r2 = 0.85; P < 0.000005).

Fig. 6.

DNA-end-binding complex (EBC) density versus SF2 for all human cell line. A composite of Fig. 1,B and Fig. 4 demonstrating the correlation between DNA-EBC and SF2 is quite good (r2 = 0.85; P < 0.000005).

Close modal
Table 1

Characteristics of cell lines used in this project

Cell lineCharacteristics/MutationsSF2
GM03395C Fibroblast, ataxia-telangiectasia; 0.02 
AT5BiSV40 Fibroblast, SV40 transformed, ataxia—telangiectasia; (GM05849B) 0.04 
HCC1937 Breast carcinoma, homozygous for 5382C mutation in BRCA1 0.10 
C84 Primary skin fibroblast, ATM heterozygote 0.17 
C42 Primary skin fibroblast, multiple cancers 0.17 
C85 Primary skin fibroblast, BRCA1 mutation 0.18 
C44 Primary skin fibroblast, BRCA1 mutation 0.18 
C74 Primary skin fibroblast acute XRT reaction 0.18 
GM03396A Fibroblast, ATM heterozygote; 0.22 
S23 Primary fibroblast line derived from a patient with breast cancer 0.25 
C75 Primary skin fibroblast, BRCA1 mutation 0.28 
C49 Primary skin fibroblast, BRCA2 mutation 0.28 
C83 Primary skin fibroblast, BRCA1 mutation 0.29 
C51 Primary skin fibroblast, BRCA1 mutation 0.29 
C19 Primary skin fibroblast, patient had breast cancer, no known BRCA mutation 0.30 
C76 Primary skin fibroblast, BRCA1 mutation 0.31 
C63 Primary skin fibroblast, BRCA1 mutation 0.32 
C46 Primary skin fibroblast, BRCA1 mutation 0.32 
C29 Primary skin fibroblast 0.33 
T47D Breast carcinoma 0.35 
PC3 Prostate carcinoma 0.38 
C37 Primary skin fibroblast 0.39 
A375 Malignant melanoma 0.41 
C80 Skin fibroblast, sequence-normal daughter of C75 0.41 
MeWo Malignant melanoma 0.43 
U251 Glioblastoma 0.44 
U87MG Glioblastoma 0.45 
SW620 Colon adenocarcinoma 0.47 
MIA PaCa-2 Pancreatic adenocarcinoma 0.48 
CAPAN-1 Pancreatic adenocarcinoma, BRCA-2 mutation 0.62 
T98G Glioblastoma 0.67 
A549 Lung adenocarcinoma 0.68 
A431 Epidermoid carcinoma 0.74 
NIH/3T3 Mouse embryo fibroblast 0.74 
HN-5 Head and neck, squamous cell carcinoma 0.75 
H1299 Non-small cell lung carcinoma 0.75 
HT29 Colon adenocarcinoma 0.80 
Cell lineCharacteristics/MutationsSF2
GM03395C Fibroblast, ataxia-telangiectasia; 0.02 
AT5BiSV40 Fibroblast, SV40 transformed, ataxia—telangiectasia; (GM05849B) 0.04 
HCC1937 Breast carcinoma, homozygous for 5382C mutation in BRCA1 0.10 
C84 Primary skin fibroblast, ATM heterozygote 0.17 
C42 Primary skin fibroblast, multiple cancers 0.17 
C85 Primary skin fibroblast, BRCA1 mutation 0.18 
C44 Primary skin fibroblast, BRCA1 mutation 0.18 
C74 Primary skin fibroblast acute XRT reaction 0.18 
GM03396A Fibroblast, ATM heterozygote; 0.22 
S23 Primary fibroblast line derived from a patient with breast cancer 0.25 
C75 Primary skin fibroblast, BRCA1 mutation 0.28 
C49 Primary skin fibroblast, BRCA2 mutation 0.28 
C83 Primary skin fibroblast, BRCA1 mutation 0.29 
C51 Primary skin fibroblast, BRCA1 mutation 0.29 
C19 Primary skin fibroblast, patient had breast cancer, no known BRCA mutation 0.30 
C76 Primary skin fibroblast, BRCA1 mutation 0.31 
C63 Primary skin fibroblast, BRCA1 mutation 0.32 
C46 Primary skin fibroblast, BRCA1 mutation 0.32 
C29 Primary skin fibroblast 0.33 
T47D Breast carcinoma 0.35 
PC3 Prostate carcinoma 0.38 
C37 Primary skin fibroblast 0.39 
A375 Malignant melanoma 0.41 
C80 Skin fibroblast, sequence-normal daughter of C75 0.41 
MeWo Malignant melanoma 0.43 
U251 Glioblastoma 0.44 
U87MG Glioblastoma 0.45 
SW620 Colon adenocarcinoma 0.47 
MIA PaCa-2 Pancreatic adenocarcinoma 0.48 
CAPAN-1 Pancreatic adenocarcinoma, BRCA-2 mutation 0.62 
T98G Glioblastoma 0.67 
A549 Lung adenocarcinoma 0.68 
A431 Epidermoid carcinoma 0.74 
NIH/3T3 Mouse embryo fibroblast 0.74 
HN-5 Head and neck, squamous cell carcinoma 0.75 
H1299 Non-small cell lung carcinoma 0.75 
HT29 Colon adenocarcinoma 0.80 

We thank Cora Bartholomew for assistance with the preparation of this manuscript.

1
Peters L., McKay M. Predictive assays: will they ever have a role in the clinic?.
Int. J. Radiat. Oncol. Biol. Phys.
,
49
:
501
-504,  
2001
.
2
Tucker S. L., Geara F. B., Peters L. J., Brock W. A. How much could the radiotherapy dose be altered for individual patients based on a predictive assay of normal-tissue radiosensitivity?.
Radiother. Oncol.
,
38
:
103
-113,  
1996
.
3
Mackay R. I., Hendry J. H. The modeled benefits of individualizing radiotherapy patients’ dose using cellular radiosensitivity assays with inherent variability.
Radiother. Oncol.
,
50
:
67
-75,  
1999
.
4
Burnet N. G., Nyman J., Turesson I., Wurm R., Yarnold J. R., Peacock J. H. The relationship between cellular radiation sensitivity and tissue response may provide the basis for individualizing radiotherapy schedules.
Radiother. Oncol.
,
33
:
228
-238,  
1994
.
5
Burnet N. G., Warm R., Peacock J. H. Low dose-rate fibroblast radiosensitivity and the prediction of patient response to radiotherapy.
Int. J. Radiat. Oncol. Biol. Phys.
,
70
:
289
-300,  
1996
.
6
MacKay R. I., Niemierko A., Goitein M., Hendry J. H. Potential clinical impact of normal tissue intrinsic radiosensitivity testing (letter).
Radiother. Oncol.
,
46
:
215
-216,  
1998
.
7
Norman A., Kagan A. R., Chan S. L. The importance of genetics for the optimization of radiation therapy: a hypothesis.
Am. J. Clin. Oncol.
,
11
:
84
-88,  
1988
.
8
Thames H. D., Schultheiss T. E., Hendry J. H., et al Can modest escalations of dose be detected as increased tumor control?.
Int. J. Radiat. Oncol. Biol. Phys.
,
22
:
241
-246,  
1992
.
9
Brock W. A., Baker F. L., Peters L. J. Radiosensitivity of human head and neck squamous cell carcinomas in primary culture and its potential as a predictive assay of tumor radiocurability.
Int. J. Radiat. Biol.
,
56
:
751
-760,  
1989
.
10
West C. M. L., Hendry J. H. Intrinsic radiosensitivity as a predictor of patient response to radiotherapy.
Br. J. Radiol.
,
24 (Suppl.)
:
146
-152,  
1992
.
11
Thacker J. The nature of mutants induced by ionising radiation in cultured hamster cells.
Mutat. Res.
,
160
:
267
-275,  
1986
.
12
Marder B. A., Morgan W. F. Delayed chromosomal instability by DNA damage.
Mol. Cell. Biol.
,
13
:
6667
-6677,  
1993
.
13
Hartwell L. H., Kasten M. B. Cell cycle control and cancer.
Science (Wash. DC)
,
266
:
182
-188,  
1994
.
14
Karan P. DNA double strand break repair in mammalian cells.
Curr. Opin. Genet. Dev.
,
10
:
144
-150,  
2000
.
15
Nick-McElhinny S. A., Snowden C. M., McCarville J., Ramsden D. A. Ku recruits the XRCC4-ligase IV complex to DNA ends.
Mol. Cell. Biol.
,
20
:
2996
-3003,  
2000
.
16
Peterson S. R., Kurimasa A., Oshimura M., Dynan W. S., Bradbury E. M., Chen D. J. Loss of the catalytic subunit of the DNA-dependent protein kinase in DNA double-strand-break-repair mutant mammalian cells.
Proc. Natl. Acad. Sci. USA
,
92
:
3171
-3174,  
1995
.
17
Kurimasa A., Kumano S., Boubnov N. V., Story M. D., Tung C-S., Peterson S. R., Chen D. J. Requirement for the kinase activity of human DNA-dependent protein kinase catalytic subunit in DNA strand break rejoining.
Mol. Cell. Biol.
,
19
:
3877
-3884,  
1999
.
18
Becker-Catania S. G., Chen G., Hwang M. J., Wang Z., Sun X., Sanal O., Bernatowska-Matuszkiewicz E., Chessa L., Lee E. Y-H. P., Gatti R. A. Ataxia-telangiectasia: phenotype/genotype studies of ATM protein expression, mutations, and radiosensitivity.
Mol. Genet. Metab.
,
70
:
122
-133,  
2000
.
19
Tchirkov A., Bay J-O., Pernin D., Bignon Y-J., Rio P., Grancho M., Kwiatkowski F., Giollant M., Malet P., Verrelle P. Detection of heterozygous carriers of the ataxia-telangiectasia (ATM) gene by G2 phase chromosomal radiosensitivity of peripheral blood lymphocytes.
Hum. Genet.
,
101
:
312
-316,  
1997
.
20
Buchholz T. A., Wu X., Hussain A., et al Evidence of haplotype insufficiency in human cells containing a germline mutation in BRCA1 or BRCA2.
Int. J. Cancer
,
97
:
557
-561,  
2002
.
21
Dignam J. D., Lebovitz R. M., Roeder R. G. Accurate transcription initiation by RNA polymerase II in a soluble extract from isolated mammalian nuclei.
Nucleic Acids Res.
,
11
:
1475
-1489,  
1983
.
22
Stevens C. W., Stamato T. D., Mauldin S. K., Getts R. C., Hahn L. M., Zeng M., Cerniglia G. C. Radiation induced recombination is Ku-80 dependent.
Radiat. Res.
,
151
:
408
-413,  
1999
.
23
Weller G. R., Kysela B., Roy R., Tonkin L. M., Scanlan E., Della M., Devine S. K., Day J. P., Wilkinson A., di Adda di Fagagna F., Devine K. M., Bowater R. P., Jeggo P. A., Jackson S. P., Doherty A. J. Identification of a DNA nonhomologous end-joining complex in bacteria.
Science (Wash. DC)
,
297
:
1686
-1689,  
2002
.
24
Ayene I. S., Stamato T. D., Mauldin S. K., Biaglow J. E., Tuttle S. W., Jenkins S. F., Koch C. J. Mutation in the glucose-6-phosphate dehydrogenase gene leads to inactivation of Ku DNA end binding during oxidative stress.
J. Biol. Chem.
,
277
:
9929
-9935,
25
Sambrook J., Fritsch E. F., Maniatis T. .
Molecular Cloning: A Laboratory Manual
, Cold Spring Harbor Laboratory Press Cold Spring Harbor, NY  
1989
.
26
Johansen J., Bentzen S. M., Overgaard J., Overgaard M. Relationship between the in vitro radiosensitivity of skin fibroblasts and the expression of subcutaneous fibrosis, telangiectasia, and skin erythema after radiotherapy.
Radiother. Oncol.
,
40
:
101
-109,  
1996
.
27
Brock W. A., Tucker S. L., Geara F. B., Wike J., Peters L. J., Turesson I., Nyman J. Fibroblast radiosensitivity versus acute and late normal skin responses in patients treated for breast cancer.
Int. J. Radiat. Oncol. Biol. Phys.
,
32
:
1371
-1379,  
1995
.
28
Hart R. M., Kimler B. F., Evans R. G., Park C. H. Radiotherapeutic management of medulloblastoma in a pediatric patient with ataxia telangiectasia.
Int. J. Radiat. Oncol. Biol. Phys.
,
13
:
1237
-1240,  
1987
.
29
Stausbol-Gron B., Overgaard J. Relationship between tumour cell in vitro radiosensitivity and clinical outcome after curative radiotherapy for squamous cell carcinoma of the head and neck.
Radiother. Oncol.
,
50
:
47
-55,  
1999
.
30
Bjork-Eriksson T., West C., Karlsson E., Mercke C. Tumor radiosensitivity (SF2) is a prognostic factor for local control in head and neck cancers.
Int. J. Radiat. Oncol. Biol. Phys.
,
46
:
13
-19,  
2000
.
31
Taghian A., Ramsay J., Allalunis-Turner J., Budach W., Gioioso D., Pardo F., Okunieff P., Bleehen N., Urtasun R., Suit H. Intrinsic radiation sensitivity may not be the major determinant of the poor clinical outcome of glioblastoma multiforme.
Int. J. Radiat. Oncol. Biol. Phys.
,
25
:
243
-249,  
1993
.
32
Wouters B. G., Weppler S. A., Koritzinsky M., Landuyt W., Nuyts S., Theys J., Chiu R. K., Lambin P. Hypoxia as a target for combined modality treatments.
Euro. J. Cancer
,
38
:
240
-257,  
2002
.
33
Koukourakis M. I. Tumour angiogenesis and response to radiotherapy.
Anticancer Res.
,
21
:
4285
-4300,  
2001
.
34
Pardoll D. Does the immune system see tumors as foreign or self?.
Ann. Rev. Immunol.
,
21
:
807
-839,  
2003
.
35
Antognoni P., Terry N. H. A., Richetti A., Luraghi R., Tordiglione M., Danova M. The predictive role of flow cytometry-derived tumor potential doubling time (Tpot) in radiotherapy: open questions and future perspectives.
Intl. J. Oncol.
,
12
:
245
-256,  
1998
.
36
Buffa F. M., Davidson S. E., Hunter R. D., Nahum A. E., West C. M. L. Incorporating biologic measurements (SF(2), CFE) into a tumor control probability model increases their prognostic significance: a study in cervical carcinoma treated with radiation therapy.
Int. J. Radiat. Oncol. Biol. Phys.
,
50
:
1113
-1122,  
2001
.
37
West C. M., Davidson S. E., Pool C., et al Lack of a relationship between colony-forming efficiency and surviving fraction at 2 Gy.
Radiat. Res.
,
126
:
260
-263,  
1991
.