Introduction

Radiation-induced genomic instability encompasses a range of measurable end points such as chromosome destabilization, sister chromatid exchanges, gene mutation and amplification, late cell death and aneuploidy, all of which may be causative factors in the development of clinical disease, including carcinoma. The basic experimental research findings underlying the field of genomic instability are reviewed elsewhere in this issue. This article will restrict itself to reviewing clinical and translational research findings, and from this basis will suggest some possible approaches to gaining prospective clinical data on the true incidence and significance of radiation-induced genomic instability in humans.

Clinical implications of genomic instability can be broadly grouped into two main areas: as a marker for increased cancer risk/early detection (i.e. radiation-induced genomic instability as a tool for clinical assessment of tumours or apparently normal tissue), and as a consequence of radiation therapy (IR) that may be causative of, or a strong marker for, the induction of a therapy-induced second malignancy. Clinical implications of genomic instability are a broader area than simply radiation-therapy-induced genomic instability leading to a second malignancy, although clearly this latter area is of significant clinical concern.

The majority of research in genomic instability has used lymphocytes or lymphoid stem cells, despite the fact that these cells are fundamentally different than cells that compose solid organs and therefore present limitations as a surrogate tissue model. Lymphocytes exist for the majority of their lifespan as independent cells in the intravascular compartment. This limitation of cell–cell contacts may allow for a different expression of genomic instability induced by fluid phase mediators and/or a frankly different biology than in solid tissue. Further, lymphocytes are unique in the body for having to undergo VDJ recombination (a process that can be thought of as a very controlled, localized genomic instability) as part of their maturation and this may indicate a different threshold for the induction of genomic instability than in other, solid tissues. Lymphocytes are an attractive model nevertheless because of their accessibility, and do provide a direct model for lymphoma and leukaemia risks. Beyond those cancer types, tissue-specific constraints are likely active in radiation response and so it seems imprudent to assume the generalizability of findings determined in lymphocytes. Experimental confirmation of genomic instability in solid tissue models should be directly obtained.

Radiation-induced genomic instability as a cancer risk marker

Kadhim et al. (1992) identified the persistence of radiation-induced chromosomal instability following -particle irradiation in clonal populations of murine bone marrow cells. The same group followed up this seminal work with an examination of four human bone marrow samples, subjected to ex vivo -particle IR (Kadhim et al., 1994). No history is given on the donors, other than their being hematologically normal. Two of the four samples showed no aberrations after exposure of up to 0.5 Gy of IR, but the other two developed significant aberrations with the majority (60–75%) being of the chromatid type. Further research then demonstrated that the genomic instability phenotype could be transmitted in vivo when murine haemopoietic cells that had been irradiated in vitro were transplanted into mice that had previously had their native bone marrow purged (Watson et al., 1996). Again, this group found the aberrations to be predominantly chromatid type. However, when Whitehouse and Tawn, (2001) examined radiation workers in Sellafield, England who had bone marrow plutonium deposition, evidence for genomic instability was not found. In the largest and longest study of its type, using lymphocyte preparations from men with more than 20% of the maximum permissible plutonium deposition, both stable chromosome aberrations (e.g., translocations) and unstable ones (e.g., dicentrics) were not significantly different from age-matched controls. These authors also matched the samples by smoking history. These data did reveal that there was an overall downward trend in the incidence of chromatid aberrations and gaps over the four time periods studied. The counting of such events was sensitive to the techniques used, with higher frequencies of aberrations scored using G banding over solid Giemsa staining. More recently, Hande et al. published a provocative study using the mBAND technique to evaluate intrachromosomal genomic rearrangements following high or moderate dose plutonium associated with -ray exposure vs -ray exposure alone. All of the patients were also exposed to numerous chemical mutagens simultaneously. The plutonium exposed population showed a markedly increased rate of intrachromosomal rearrangements that seems to suggest that very densely ionizing radiation may produce a unique and durable signature in the genome (Hande et al., 2003). Future work with the mBAND technique may extend upon this demonstration of stable intrachromosomal aberrations to discover evidence of genomic instability.

A similar type of study was conducted by Kryscio et al. (2001) on uranium miners from the former German Democratic Republic. These men were exposed to high levels of radon and its progeny, which resulted in markedly elevated rates of lung cancer development. Lymphocytes from those men who developed lung cancer were compared to those that did not. A control group of 17 healthy donors was also examined. Smoking history was described as 'similar' between the study and control populations. There was also substantial variance within groups for the length of employment in mining, and the radon exposure could not be quantitated. Samples were assessed using a micronucleus assay with scoring restricted to those that did not contain centromeres to enhance the radiation selectivity of the study. While overall micronuclei formation was not different between groups, there were statistically significant differences in centromere negative micronuclei frequency between controls, miners and miners with cancer, with the last group having the highest value. However, when the micronuclei without centromere values were normalized against overall micronuclei frequency per 1000 lymphocytes, the absolute difference in frequency between groups was not significant. Whether this reflects selection bias as the healthy miners had, on average, a longer mining history, or a true biologic similarity between the groups is unclear.

Liu et al. (2002) used a different approach to attempt to elucidate the induction, and significance of, genomic instability in people following -particle irradiation by examining archival tissues sections of human intrahepatic cholangiocarcinoma (ICC). It had been previously shown that ICC can be induced by thorotrast administration, a colloidal suspension of -particle emitting ²³²ThO₂. Using microsatellite instability as the end point, thorotrast-associated ICC showed a 2.75 times increased rate of microsatellite instability, associated with hypermethylation of hMLH1. There were some technical challenges to this work because of the use of archival specimens and the increased MSI was seen in both tumour and adjacent nontumour areas in the thorotrast samples, suggesting that microsatellite instability was inducible by thorotrast in vivo, but not necessarily specific to cancer induction.

Collectively, these data raise several significant issues from the clinical perspective. Kadhim et al.'s data showed that genomic instability could be (a) initiated in vitro, (b) perpetuated in vivo when the process was experimentally initiated in vitro and (c) initiated in apparently normal human hematopoietic cells in vitro. Whitehouse and Tawn's work showed no increase in genomic instability in vivo despite significant ongoing radiation exposure. This suggests that the initiation of persistent genomic instability in humans may have a much higher threshold than seen in isolated cell culture systems. Likewise, Kryscio et al.'s data did not show a correlation between radiation dose and the development of genomic instability with or without an associated cancer diagnosis. The fact that Kryscio's group examined lymphocytes and attempted to correlate their IR response with a lung cancer diagnosis highlights the difficulty in using surrogate tissues for IR response assessment. Lui et al. bypassed this difficulty by examining the target tissue directly and did show an association between -particle irradiation, genomic instability and cancer development. Including control populations of patients exposed to thorotrast, who did not develop ICC, and samples from healthy age- and smoking-history-matched controls would have strengthened the latter study. The hepatocyte regeneration response after cytotoxic stress may be an important factor in explaining the different target response to -particle irradiation-induced genomic instability between these studies, and emphasizes that experimental results may not be generalizable across tissue types.

The downward trend of chromatid exchanges in patients and controls in Sellafield over the 20 years of study, which also corresponded to a decreasing smoking incidence, strongly suggests that other factor(s) besides radiation dose were significant in the induction of the genomic instability. Whether that trend could be explained by smoking habits alone is unclear and smoking history was not included in the reports from Kadhim et al. or Kryscio et al. with which to compare the results. Alternatively, changes in air pollution or some other regional factor may contribute in an additive or synergistic fashion with IR and/or smoking to the observed incidence of chromosomal instability. Dietary micronutrient intake could be a factor in these trends, possibly by altering the metabolism of reactive oxygen species, but this seems less likely as a significant contributing factor to the Sellafield results as it would require postulating a significant dietary change across a population. If changes in cigarette smoking alone accounts for these trends, then studies focused on the mechanistic interaction of these agents should yield clues regarding the relative importance of direct DNA damage, the induction and detoxification of reactive oxygen species, and inflammation and healing responses (i.e., promitogenic stimuli) to the in vivo development of genomic instability. Perhaps in humans it takes at least two different types (i.e., pathways) of genotoxic stress or an overwhelming local radiation dose to induce genomic instability, beyond the levels in the Sellafield or German studies, but perhaps achieved by the thorotrast administration and concentration in the liver. Forced cell division (as would be seen in a regenerating liver) may also be needed to develop genomic instability in vivo. This would be consistent with the in vitro literature (Limoli et al., 2001) and parallel the initiation and propagation steps in carcinogenesis. It is noteworthy that the study populations in Whitehouse and Tawn's work were men who did not have a cancer diagnosis, which should better represent a healthy population than those with an already demonstrated cancer susceptibility.

Once genomic instability following -particle irradiation had been identified, interest in determining whether low LET, -irradiation could also induce these effects was heightened. The original reports from Kadhim et al., (1992) were negative, but further research examining hprt locus mutations convincingly demonstrated that genomic instability was inducible by -irradiation (Kadhim et al., 1995). Clinically, this is a more significant concern as medical irradiation is almost all of the low LET type, as is a significant amount of environmental radiation. Research began to examine radiation-induced genomic instability in lymphocytes in response to ex vivo IR as a predictor of an individual's risk of cancer induction, similar to studies examining susceptibility to other mutagenic agents. These studies were designed to examine the induction of genome damage, not the perpetuation of the instability phenotype, with the latter implied by the existing cancer diagnosis. On a cautionary note, examining genomic instability in cancer patients cannot necessarily be generalized to a 'normal' population. Cancer patients by definition have displayed an enhanced ability to develop, or perhaps a decreased ability to rectify, genomic instability as evidenced by already being able to develop the 6–8 genetic events necessary for cancer induction (Fearon and Vogelstein, 1990).

Buchholz and Wu (2001) examined lymphocytes from patients with bilateral breast cancer and compared IR-induced chromatid breaks to those seen in lymphocytes from healthy gender- and age-matched controls with no cancer history. Smoking history was not documented. Lymphocytes were cultured for 91 h, exposed to 125 cGy of -radiation, and then incubated for a further 4 h to allow the cells to attempt DNA repair. While the sample size was limited, with 26 cases and 18 controls, there was a significantly greater rate of chromatid breaks per cell in the cases than the controls (0.61 vs 0.45). While exciting, there are two important caveats to interpreting these data. Firstly, the cases were highly selected for those with presumed underlying genetic abnormalities giving rise to the otherwise uncommon development of bilateral breast cancers. Secondly, while there were more chromatid breaks/cell with the cancer phenotype, there was significant overlap between the two response patterns, with only a very high level of DNA breakage being specific for cases vs controls (>0.85 chromatid breaks/cell).

Colleu-Durel et al. (2001) also examined genomic instability in breast cancer patients. This group also collected lymphocytes after surgery and before any radiation or chemotherapy was delivered. Lymphocytes were subjected to varying doses of Co-60 -radiation and then evaluated by comet assay (single-cell gel electrophoresis) to measure DNA damage. Score (a summation of tail moment scores from 50 randomly selected cells), mean tail moment and percentage of DNA in the tail were assessed. A total of 19 patients with ipsilateral breast cancer were matched with 19 healthy controls. All patients and controls were nonsmokers. The mean tail moment in patients was markedly higher than in controls, as was percentage of DNA in the tail. When samples were given time to repair damage before the comet assay was performed, patient samples showed lesser repair than normals as measured by the mean tail moment, but when the analysis used percentage of DNA in the tail the patient samples did ultimately repair as fully as controls, although not until the 120 min time point. It is noteworthy in this study that tail moment and percentage of DNA in the tail of patient samples were statistically significantly greater at baseline (i.e., with no IR) than controls.

Another group that also used the comet assay to assess IR-induced damage and cancer incidence focused on lung cancer. Zhang et al. (2000) compared lymphocytes from 31 untreated lung cancer cases vs 39 controls. They found that cancer patients had a significantly increased number of comet cells developed after IR, but that the induced comet tail length was not different.

In separate studies, Scott et al. examined IR-induced chromosomal aberrations in lymphocytes from breast cancer patients (135 patients) and 105 healthy controls following 0.5 Gy X-rays, irradiated ex vivo (Scott et al., 1999). In total, 42% of cancer patient samples vs only 6% of control samples showed increased IR sensitivity, and the overall variability within the values for the patients was much larger than for the controls. Further work by this group using the same assay compared chromosomal aberrations in lymphocytes from family members of sensitive patients vs the index case values (Roberts et al., 1999). These data suggested that as few as two genes may account for the variability of IR-induced chromosomal instability seen within this population.

Taken together, these data suggest that patients with a solid tumour cancer demonstrate greater than expected IR-induced genomic instability in their lymphocytes, but the statistical significance of this is dependent upon how the instability is scored. These experimental approaches use the lymphocytes as biomarkers and the results are consistent with this. Models whereby the lymphocyte behaviour is postulated to reflect the underlying intrinsically decreased DNA repair capacity that exists throughout a patient's tissues could be developed as a prospective cancer risk marker. However, formal linkage of lymphocyte responses and solid tissue responses must be experimentally verified. As noted above, these studies also need to be completed prospectively in noncancer populations, as generalizability from people with cancer to those without is fraught with biases. This is shown by the findings noted above where genomic instability at baseline between cases and controls in the breast cancer studies were significantly different.

Some investigations have focused on strengthening the direct link between target tissue genomic instability, although not IR induced, and cancer development in humans. A detailed examination of prostatectomy specimens focused on identification of chromosomal instability associated with areas of high-grade intraepithelial neoplasia (PIN) either within, or separate from, areas of frankly invasive prostate cancer (Al-Maghrabi et al., 2001). Chromosomal instability was measured by interphase FISH analysis probing for abnormalities in chromosomes 7, 8, X and Y. This group demonstrated that there were more frequent numeric chromosomal abnormalities in invasive prostate cancer (47%) vs PIN associated with cancer (27%), but because of the small sample size of only 35 specimens overall, this did not reach statistical significance (P=0.1). More of the cancers and areas of PIN that were TP53+ than TP53- had chromosomal instability, but again this did not reach statistical significance likely because of the small sample size. When the analysis was restricted to areas of high-grade PIN alone, there was a significant difference in chromosomal instability between TP53+ and TP53- specimens and between areas of PIN associated with invasive cancer (4/15) vs away from any invasive cancer (0/15) (P<0.05). No chromosomal instability was detected in any areas of the prostatectomy specimens where only atrophic, hyperplastic or normal epithelium was seen. This study does not directly address the issue of the significance of chromosomal instability in high-grade PIN as a marker of adjacent invasive cancer, but these results raise the possibility that such an approach may be able to be prospectively validated as a diagnostic aid to stratify PIN risk. For example, in a prostate biopsy obtained because of an elevated screening prostate-specific antigen serum value where only PIN is identified, the presence or absence of chromosomal instability could be useful in assessing the likelihood of invasive cancer being present, but not identified by sampling error alone.

It is worthwhile noting that using radiation-induced genomic instability as a test for the identification of patients at high risk of preneoplastic lesions or higher risk of progression to frankly invasive neoplasia is only of value if there is some intervention that can be offered to alter the natural history at the point of identification. One could always offer patients so identified closer clinical follow-up (e.g., repeat biopsies at shorter intervals), and encourage behaviour modification as appropriate (e.g., aggressive intervention for smoking cessation), but it is important to be aware that such interventions can backfire. Far fewer people get tested for genetic diseases than are offered such screening and compliance with cancer screening varies substantially across populations. The reasons for this are multiple and complex, and include diverse issues such as personal perception of risk, preferring 'not to know', dislike/distrust of the medical field, and/or concerns about future potential discrimination by employers or insurance companies. Further, it also must be recognized that there is a significant difference between increased case-finding of cancers, which may never have been clinically meaningful and a useful screening tool. Similarly, the diagnosis of malignancy at earlier stages without improved treatment can falsely be perceived as an increased survival time, known as lead-time bias. Under these conditions, predictive screening tests provide no benefit and potentially cause harm as the patient has a disease diagnosis for a longer period of time, resulting in an increase in mental and/or emotional suffering.

Secondary genomic instability following medical treatment

A second category of clinical concern is genomic instability that has been induced from medical intervention. Therapy-induced genomic instability can follow chemotherapy or radiation therapy (IR), with IR allowing a more accurate quantification of genotoxic insult (dose actually delivered to the target tissue) and target volume. Radiation therapy is a local treatment modality, in contrast to the systemic nature of chemotherapy, which allows the possibility of examining tissues that have received variable doses of IR. In theory, this would allow a dose–response curve for the induction of genomic instability to be established. With respect to cancer patients, the tissue at risk for the induction of genomic instability is normal tissue that has been exposed to an ionizing radiation insult, but not killed. This population can be further classified into those tissues that received repetitive low-dose exposures (e.g., tissue outside of the radiation portal), variable low or high dose exposures (e.g., tissue that may move into the radiation portal over the course of treatment) or all high-dose exposure by virtue of being within the radiation target area, but not actual malignant tissue. While theoretically radiation delivered with therapeutic intent could also induce genomic instability within the tumour, leading to therapy resistance, it would be difficult if not impossible to convincingly demonstrate that the resultant phenotype was because of the radiation exposure and did not arise simply from the malignant process itself.

Tawn et al. examined peripheral blood lymphocytes in 18 cancer patients over a course of up to 8 years after fractionated therapeutic IR of 35–80 Gy. There was a wide variety of primary tumours and ages at diagnosis. These patients were also workers at the British Nuclear Fuels reprocessing plant in Sellafield, UK. The frequency of aberrant cells following IR was evaluated in 6-month time periods and the rates of aberrancy continued to fall, with the most dramatic decline in the first 6 months after treatment. This is consistent with the expectation that damaged lymphocytes are eliminated from the peripheral blood, and chromosomally unstable bone marrow stem cells will not survive the maturation process. There was no evidence in support of persisting genomic instability. Schmidberger et al. examined lymphocytes in men with the diagnosis of testicular seminoma before and following IR. Using FISH techniques, the group examined the frequency of reciprocal translocations following either para-aortic radiation (including approximately 12% of the active bone marrow, 21 patients, 26 Gy), or para-aortic and ipsilateral iliac nodes (approximately 18% of the active bone marrow, nine patients, 36 Gy). A group of 13 healthy male volunteers served as the control. The cancer group had a statistically significant increase in translocations relative to the controls at baseline, before any radiation therapy was delivered. The range of values of spontaneous translocations was also much greater than in the controls. Post-treatment, translocation rates were elevated, but decreased over time, returning to pretreatment values within 60 months. There was no significant difference in genomic translocation rates between the 2 dose/volume IR schedules. Of significance, the frequency of chromosomal translocations in patient samples remained elevated relative to healthy controls at all study points, highlighting that even successfully treated cancer patients may not ever return to a 'normal' state, or conversely, the 'normal' state for people who develop cancer is fundamentally different than for noncancer patients, consistent with the hypothesis that there is significant interaction between carcinogens and the host for a clinical cancer to develop (Hsu et al., 1991). Tissue biopsies of the areas around the primary tumour that would have received direct radiation exposure would add greatly to our understanding of radiation-induced genomic instability as there would be no 'dilution effect' of only partial bone marrow exposure when measuring lymphocyte translocation rates. Indeed, when Dorr and Herrmann examined the development of second malignancies following radiation therapy, they found that the majority of cancers arose in tissue that received up to 6 Gy. If radiation-induced genomic instability does exist and is of clinical significance in the development of second malignancies, then tissues receiving these doses would be a good study target.

While there are many high-quality studies on radiation-induced second malignancies (e.g., Dores et al., 2002; Dorr and Herrmann, 2002; Matheson et al., 2002; Pickles and Phillips, 2002; Chronowski et al., 2003; Zablotska and Neugut, 2003), none of these were designed to examine genomic instability. One study that did attempt to characterize molecular changes in second primary lung and breast cancers after therapy for Hodgkin's disease found that while they had a similar occurrence of loss of heterozygosity overall at all chromosomal regions, the post-IR cancers had much higher rates of microsatellite alterations than the sporatically arising tumours (Behrens et al., 2000).

Currently, there are far more detailed studies on genomic instability induced by chemotherapy than following radiation therapy. Therapy-induced leukaemia (t-AML), which is induced by certain chemotherapeutic agents such as alkylating agents (e.g., melphalan) or topoisomerase II inhibitors (e.g., etoposide), has a much more complex karyotype than de novo leukaemias. Further, t-AML characteristically has a multilineage dysplasia (Harris et al., 1999; Dann and Rowe, 2001), which is suggestive of a genomic instability component. Chemotherapy-induced AML is resistant to curative treatment and has been shown to have a high prevalence of multidrug resistance (MDR) gene expression, with up to 1/3 of patients having multiple mechanisms of MDR (Leith et al., 1999). The clinical implications of these findings are unclear as Finette et al. (2000) showed that genomic instability as measured by the hprt cloning assay on patient T cells was induced by the treatment of leukaemia in children, but did not correlate with relapse (P=0.2) or inevitably lead to a second malignancy. In this study, nonmalignant T lymphocytes were tested in 103 patients with ALL. In all, 19 patient samples were obtained at the time of initial diagnosis before any chemotherapy, 32 from patients in complete remission, 58 at the time of relapse, prior to retreatment and 49 control samples. Patients in remission or relapse showed equally elevated rates of mutation frequencies, both groups having received alkylating chemotherapeutic drugs. This group suggested that the repeated cycles of cell death followed by proliferation to restore the bone marrow and blood compartment may have been critical to the development of the mutations, especially in those children showing the greatest elevations of mutation frequency. If one attempts to extrapolate these data to the IR situation, the purging/proliferation cycles would be much less and therefore may result in a lesser frequency of genomic instability induction. Mechanistically this could also explain the lack of genomic instability seen in the radiation workers from Sellafield, England or the miners in the former German Democratic Republic, while hepatocyte proliferation for hepatic regeneration could have been a critical difference explaining Lui et al.'s study showing a positive association. It must be emphasized that at present this remains speculative, awaiting experimental data for confirmation or refutation.

While currently not a major concern, ultimately, these issues will also need to be addressed to assess the medical risk to astronauts associated with prolonged space travel. Their risks may vary substantially from that seen in cancer patients both because of differences of the radiation type and exposure times and because astronauts are otherwise young, healthy individuals. Astronauts may therefore have more effective defenses against the development of genomic instability compared to those people who have already developed a cancer.

Future research directions

If we are to move forward in our understanding of radiation-induced genomic instability in people, then prospective trials involving patients and normal control populations must be undertaken. These studies must be designed in as rigorous a manner as possible to minimize confounding variables that may play a role in the expression of individual susceptibility to the induction of genomic instability. These variables include factors that are not usually controlled in clinical trials such as tobacco use, diet, use of dietary supplements and/or previous treatments that may affect genomic stability such as chemotherapy or hormonal therapy. These trials should define the dose exposure precisely so that the shape of the dose–response curve can be characterized. High-precision dosimetry is now available through the use of Monte-Carlo based treatment planning systems, such as PEREGRINE (Hartmann et al., 2001). Solid organs must be biopsied, despite the increased difficulty in obtaining tissue biopsies compared to lymphocytes, as we must examine how well lymphocyte models truly reflect more complex tissues that exist in fixed position in a multicellular admixture where cell–cell dialogue may influence the development or expression of genomic instability. Possible research scenerios include the following: (1) Bladder biopsies post radiation for prostate cancer. Since part of the bladder has been in the high-dose region, part in the low-dose peripheral region, and the very dome of the bladder has likely seen minimal radiation exposure, biopsies at different dose levels could be obtained and compared to biopsies from patients without malignant disease (who have not been exposed to therapeutic IR) and those patients who have a bladder cancer diagnosis, but also have not (yet) been treated with IR. (2) Oral mucosal biopsies after IR for the treatment of a head and neck cancer with samples taken from in-field and out-of-field areas, or (3) Biopsies of tissues following preoperative IR for sarcoma treatment, where large volumes of irradiated tissue may be available for clinical study with the en-bloc resections needed for sarcoma management. In each of these cases, chemotherapy is not a routine in addition to the anticancer treatment and therefore will not be a confounding variable; however, in the head and neck cancers, it must be noted that these patients frequently have a history of tobacco and excessive alcohol use that must be accounted for in any interpretation of the clinical sample results. In any event, designing clinical trials to evaluate genomic instability in solid tissues will bring us that much closer to truly understanding if and how radiation induces this phenomena in vivo, which may lead to new insights into genome-protective cellular defense responses, radiation carcinogenesis, optimization parameters for radiation therapy and radiation risk assessment for health and regulatory purposes.

References

1. Al-Maghrabi J, Vorobyova L, Chapman W, Jewett M, Zielenska M and Squire JA. (2001). Mod. Pathol., 14, 1252–1262. | Article | PubMed |
2. Behrens C, Travis LB, Wistuba II, Davis S, Maitra A, Clarke EA, Lynch CF, Glimelius B, Wiklund T, Tarone R and Gazdar AF. (2000). Cancer Epidemiol. Biomarkers Prev., 9, 1027–1035. | PubMed | ISI | ChemPort |
3. Buchholz TA and Wu X. (2001). Int. J. Radiat. Oncol. Biol. Phys., 49, 533–537. | PubMed |
4. Chronowski GM, Wilder RB, Tucker SL, Ha CS, Younes A, Fayad L, Rodriguez MA, Hagemeister FB, Barista I, Cabanillas F and Cox JD. (2003). Int. J. Radiat. Oncol. Biol. Phys., 55, 36–43. | PubMed |
5. Colleu-Durel S, Guitton N, Nourgalieva K, Leveque J, Danic B and Chenal C. (2001). Oncol. Rep., 8, 1001–1005. | PubMed |
6. Dann EJ and Rowe JM. (2001). Best Pract. Res. Clin. Haematol., 14, 119–137. | Article | PubMed | ChemPort |
7. Dores GM, Metayer C, Curtis RE, Lynch CF, Clarke EA, Glimelius B, Storm H, Pukkala E, van Leeuwen FE, Holowaty EJ, Andersson M, Wiklund T, Joensuu T, Van't Veer MB, Stovall M, Gospodarowicz M and Travis LB. (2002). J. Clin. Oncol., 20, 3484–3494. | Article | PubMed | ISI |
8. Dorr W and Herrmann T. (2002). J. Radiol. Prot., 22, A117–A121. | PubMed |
9. Fearon ER and Vogelstein B. (1990). Cell, 61, 759–767. | Article | PubMed | ISI | ChemPort |
10. Finette BA, Homans AC and Albertini RJ. (2000). Science, 288, 514–517. | Article | PubMed | ISI | ChemPort |
11. Hande MP, Azizova TV, Geard CR, Burak LE, Mitchell CR, Khokhryakov VF, Vasilenko EK and Brenner DJ. (2003). Am. J. Hum. Genet., 72, 1162–1170. | Article | PubMed | ISI | ChemPort |
12. Harris NL, Jaffe ES, Diebold J, Flandrin G, Muller-Hermelink HK, Vardiman J, Lister TA and Bloomfield CD. (1999). J. Clin. Oncol., 17, 3835–3849. | PubMed | ISI | ChemPort |
13. Hartmann Siantar CL, Walling RS, Daly TP, Faddegon B, Albright N, Bergstrom P, Bielajew AF, Chuang C, Garrett D, House RK, Knapp D, Wieczorek DJ and Verhey LJ. (2001). Med. Phys., 28, 1322–1337. | PubMed |
14. Hsu TC, Spitz MR and Schantz SP. (1991). Cancer Epidemiol. Biomarkers Prev., 1, 83–89. | PubMed | ISI | ChemPort |
15. Kadhim MA, Lorimore SA, Hepburn MD, Goodhead DT, Buckle VJ and Wright EG. (1994). Lancet, 344, 987–988. | Article | PubMed | ISI | ChemPort |
16. Kadhim MA, Lorimore SA, Towensend KM, Goodhead DT, Buckle VJ and Wright EG. (1995). Int. J. Radiat. Biol., 67, 287–293. | PubMed |
17. Kadhim MA, Macdonald DA, Goodhead DT, Lorimore SA, Marsden SJ and Wright EG. (1992). Nature, 355, 738–740. | Article | PubMed | ISI | ChemPort |
18. Kryscio A, Ulrich Muller WU, Wojcik A, Kotschy N, Grobelny S and Streffer C. (2001). Int. J. Radiat. Biol., 77, 1087–1093. | PubMed |
19. Leith CP, Kopecky KJ, Chen IM, Eijdems L, Slovak ML, McConnell TS, Head DR, Weick J, Grever MR, Appelbaum FR and Willman CL. (1999). Blood, 94, 1086–1099. | PubMed | ISI | ChemPort |
20. Limoli CL, Kaplan MI, Giedzinski E and Morgan WF. (2001). Free Radic. Biol. Med., 31, 10–19. | PubMed |
21. Liu D, Momoi H, Li L, Ishikawa Y and Fukumoto M. (2002). Int. J. Cancer, 102, 366–371. | PubMed |
22. Matheson JB, Burmeister BH, Smithers BM, Gotley D, Harvey JA and Doyle L. (2002). Radiother. Oncol., 65, 159–163. | PubMed |
23. Pickles T and Phillips N. (2002). Radiother. Oncol., 65, 145–151. | PubMed |
24. Roberts SA, Spreadborough AR, Bulman B, Barber JB, Evans DG and Scott D. (1999). Am. J. Hum. Genet., 65, 784–794. | Article | PubMed | ISI | ChemPort |
25. Schmidberger H, Virsik-Koepp P, Rave-Frank M, Reinosch KR, Pradler O, Munzel U and Hess CF. (2001). Int. J. Radiat. Oncol. Biol. Phys., 50, 857–864. | PubMed |
26. Scott D, Barber JB, Spreadborough AR, Burrill W and Roberts SA. (1999). Int. J. Radiat. Biol., 75, 1–10. | Article | PubMed | ISI | ChemPort |
27. Tawn EJ, Whitehouse CA and Martin FA. (2000). Mutat. Res., 465, 45–51. | PubMed |
28. Watson GE, Lorimore SA and Wright EG. (1996). Int. J. Radiat. Biol., 69, 175–182. | Article | PubMed | ISI | ChemPort |
29. Whitehouse CA and Tawn EJ. (2001). Radiat. Res., 156 (5 Part 1), 467–475, 7587. | PubMed |
30. Zablotska LB and Neugut AI. (2003). Cancer, 97, 1404–1411. | Article | PubMed | ISI |
31. Zhang H, Buchholz TA, Hancock D, Spitz MR and Wu X. (2000). Int. J. Oncol., 17, 399–404. | PubMed |

Acknowledgements

This work was in part supported by the Office of Science (BER), U.S. Department of Energy grant #DE-FG03-01ER63237 and the Office of the President, University of California, CLC Grant number 69895.