The mutagenic and carcinogenic effects of sparsely ionizing

Transcription

1 Radiation dose-rate effects, endogenous DNA damage, and signaling resonance Michael M. Vilenchik* and Alfred G. Knudson *The Sally Balin Medical Center, 110 Chesley Drive, Media, PA 19063; and Division of Population Science, Fox Chase Cancer Center, 333 Cottman Avenue, Philadelphia, PA Contributed by Alfred G. Knudson, October 2, 2006 (sent for review June 15, 2006) We previously concluded, from our analysis of the published data of other investigators, that the yield of germ-line and somatic mutations after exposure to ionizing radiation is parabolically related to the logarithm of the dose-rate at which a given dose is administered. Here we show that other data reveal a similarly parabolic relationship for other ionizing radiation-associated phenomena, namely, genetic recombination, chromosomal translocation, cell inactivation and lethality, and human leukemogenesis. Furthermore, the minima for all effects fall in a relatively narrow range of the dose-rate logarithms. Because the only mechanism common to all of these phenomena is the double-strand break (DSB) in DNA, we refer to our previous analysis of the endogenous production of DSBs, from which we concluded that 50 endogenous DSBs occur per cell cycle, although most are repaired without error. Comparison then reveals that their rate of production falls within the range of minima for the several end points pursuant to radiation-induced DSBs. We conclude that the results reflect a physiological principle whereby signals originating from induced DSBs elicit responses of maximal effectiveness when they are produced at a rate near that of the production of endogenous DSBs. We refer to this principle as signaling resonance. DNA repair mutilation repair leukemogenesis parabolic minimum The mutagenic and carcinogenic effects of sparsely ionizing radiation (IR), such as x-rays or -rays, decrease for a given dose as the dose-rate (DR) is decreased in the range of centisieverts (csv) min, a phenomenon known as the direct DR effect (DRE) and attributed to increasingly effective repair of DNA damage. On the basis of observations on locus-specific germ-line mutations in mouse spermatogonial stem cells, Lyon concluded that there could be an increase in mutation rates below this DR, with a minimum rate of damage near 0.1 csv min (reviewed in ref. 1), but this conclusion was not generally accepted. Our analysis of available data for both germ-line and somatic mutations demonstrated that this inverse DRE is characteristic and that the overall DREs are parabolic functions of the logarithms of DRs over a range of approximately six logs, with minimal mutation rates per unit of dose in the range of csv min (1). This finding then raises the question of why the minimum occurs in this range, far above a common rate of background radiation on the order of 10 7 csv min. Here we analyze published data on both mutation and other radiation genetic endpoints not only for the existence of DREs, but also for clues to a mechanistic explanation. Such studies have been published for somatic recombination, for chromosomal translocations, for cell inactivation and lethality, and for human leukemogenesis. Because all of these phenomena can involve double-strand breaks (DSBs) in DNA, we compare the results of all of the DREs with those of our previous analysis of the spontaneous production of endogenous DSBs (EDSBs) (2) in an attempt to ascertain an underlying principle. Results and Discussion Mutagenesis. Germ-line mutations. Data from the irradiation of spermatogonia in mice have been published by two groups led, Fig. 1. Parabolic plot of published DR data on seven specific loci mutations induced by low-linear energy transfer IR in mouse spermatogonial stem cells. These data from the Oak Ridge and Harwell laboratories were summarized by us elsewhere (1). Here we plot them as the increase over the background (spontaneous) frequency (in % per 10 csv) versus the log 10 of the DR. The data include an ultra-high DR, i.e., 1,000 csv min. These latter DR data were not shown previously but are shown here because we compare these effects on mutation rate with those on IR-induced leukemia rate in the ultra-high DR region, estimated to be 500 csv min or higher. The mutation data are fitted to y x x (R 0.92). respectively, by Lyon (3) and Russell (4, 5). From their published results we have calculated the DREs and shown that their sets of data are in close agreement (1). Both show a direct DRE at DRs above 1.0 csv min; i.e., the frequency of mutations increases as the rate of delivery of a given dose increases. Conversely, both sets of data demonstrate an inverse DRE at DRs below 0.03 csv min; i.e., the frequency of mutations increases as the rate of delivery of a given dose decreases below this DR. A minimum for the DRE is observed at csv min, and the overall pattern of DREs is parabolic (Fig. 1). The increase over the background mutation frequency (y) is plotted (in % per 10 csv) versus (x), the log 10 DR (in csv min), to facilitate comparison with our later analysis of data on leukemogenesis. As indicated in Fig. 1, y x x R 0.92 [1] Author contributions: M.M.V. and A.G.K. designed research, performed research, analyzed data, and wrote the paper. The authors declare no conflict of interest. Abbreviations: IR, ionizing radiation; DSB, double-strand break; EDSB, endogenous DSB; DR, dose-rate; DRE, DR effect; DD, doubling dose; ERR, excess relative risk; Sv, sievert; csv, centisievert. To whom correspondence should be addressed. alfred.knudson@fccc.edu by The National Academy of Sciences of the USA PNAS November 21, 2006 vol. 103 no cgi doi pnas

2 Fig. 2. DRE of -rays on recombination in a diploid strain of S. cerevisiae (data from ref. 7). Data are plotted as slopes of the recombination induction curves [a measure of the yield or induction efficiency per given dose (10 csv) as a function of DR]. The data for very high DRs are not shown here. for specific-locus germ-line mutations. Somatic mutations. We have previously demonstrated (1), using published data on IR-induced HPRT mutations, that rodent and human somatic cell lines show both direct and inverse types of DRE for HPRT mutations, with an overall parabolic pattern of DREs similar to that seen in Fig. 1 for specific-locus mutations in spermatogonia. We analyzed data on DRE for HPRT mutations in mouse lymphoma L5178 cells (6) on the same scales used for the germ-line data in Fig. 1 and found a minimum effect in the region of DRs 1.0 csv min, a higher value than that for the in vivo studies of germ-line mutation noted above (data not shown; see figure 1 in ref. 1). The difference between the two becomes smaller if we include all somatic mutation data on different mammalian cells, with the values for somatic cells falling in the interval csv min, with a mean near 0.5 csv min (1), as discussed in Results and Discussion. Mitotic Recombination and Chromosomal Translocation. Mitotic recombination, another genetic phenomenon, has also been studied for DREs. In the diploid form of the yeast Saccharomyces cerevisiae, resistance to the drug canavanine was concluded by the authors (7) to be induced by mitotic recombination in heterozygotes for mutation at the can locus. The rates of mitotic recombination were minimal at a DR of 1.7 csv min (Fig. 2). Interestingly, although there is a direct DRE in the range of 1 20 csv min, in the higher DR region, up to several thousand csv min, the cells show again an inverse DRE, indicating inhibition of recombination in the ultra-high DR region. There are several plausible mechanisms for such inhibition that are not discussed here [for further consideration of this problem see Kiefer et al. (7)]. A minimum for radiosensitivity with respect to recombination is clearly seen in the DR region of 1.0 csv min, near the minimum observed for mutations in somatic cells. Several studies have demonstrated the existence of direct DREs on chromosomal translocations in somatic mammalian cells exposed to IR either in vitro or in vivo (reviewed in ref. 8). In these studies, however, only DRs in the high and intermediate DR region were usually used. For germ-line translocations one study reported both direct and inverse DRE in mouse spermatogonia stem cells (9), with a transition from the direct DRE to the inverse DRE in a region of DRs 0.05 csv min. The DRE data for both specific-locus mutation and translocation in mouse Fig. 3. The pattern of the DRE on clonogenic cell death of human PC3 prostate cancer cells exposed to 5 Sv of low-linear energy transfer IR in vitro. Asynchronously growing cells were exposed to 5 Sv of low-dr -rays or by 5 Sv of x-rays in the case of the highest DR. The data are from refs. 15 and 16, where the cell-survival data are calculated as the relative fraction of irradiated cells that show clonogenic survival (S). To compare DREs on the clonogenic inactivation of the mammalian cells with those on genetic effects and leukemogenesis (Fig. 1), we plotted the value of the fraction (in %) of the cells whose clonogenic capacity was inactivated by radiation, i.e., 1-S, or 100% (control cells) % of cells remaining clonogenic. spermatogonia suggest a minimum, csv min, that is lower than that for mutations in somatic cells (see Somatic mutations). Clonogenic Cell Inactivation and DSBs. The direct DRE on cell survival has usually been well studied for high and intermediate DRs ( csv min) (reviewed in refs. 10 and 11), but Mitchell et al. (12) had reported that cell lethality from exposure of HeLa cells at a DR of 37 csv h (0.62 csv min) was increased in comparison with irradiation at 154 csv h (2.57 csv min) (12). More recently, inverse DREs have also been reported for glioblastoma cells and prostate cancer cells (13). From these and other recent data (14 16), we conclude that different mammalian cell lines show a parabolic dependence of the loss of clonogenic survival [per given dose of 5 sieverts (Sv)] on DR. The reproductive cell death (lethality) for human PC3 prostate cancer cells was decreased as the DR was decreased from 55 csv min to 1.8 csv min, but there is an inverse DRE with increase in lethality as the DR was decreased further from 1 to 0.03 csv min (Fig. 3) (data from ref. 15). Our analysis of data on glioblastoma cell line T986, again at a dose of 5 Sv, also revealed a pattern of DRE with a minimum in radiosensitivity with respect to cell inactivation at 1.8 csv min (15, 16) (data not shown here). These minima are close to the values in the range of csv min shown for somatic mutations in human and rodent cell lines (1) (see Somatic mutations and Fig. 1). It is established that DSBs and errors in their repair are responsible for most, if not all, IR-induced lethality (14, 17 19) and are major factors in mutagenesis (8, 20). The close relationship between the above parabolic curves for lethality inactivation and for somatic mutation suggests to us that DSBs, both endogenous and IR-induced, may be limiting factors for both DR processes. We shall return to this idea in Optimal Sensing of the Frequency of Induced DSBs. These data clearly show a parabolic dependence on the log of the DR, just as in the cases of the genetic phenomena discussed MEDICAL SCIENCES Vilenchik and Knudson PNAS November 21, 2006 vol. 103 no

3 Table 1. Estimation of the ERR of radiation-induced leukemia at 10 csv Observed expected deaths Leukemia 5 10 csv dose group 10 csv dose group Combined data ( 10 csv) Estimated ERR, % All subtypes Excluding CLL The table is based on data on leukemia mortality published from one of the largest and most recent epidemiological studies of radiation carcinogenesis: a cohort of radiation workers of the United Kingdom Atomic Energy Authority workforce (compared with nonradiation workers) with cumulative radiation doses in the range of 5 10 csv and 10 csv, adjusted for age, calendar year, and social class (24). ERR in % [(observed risk expected risk) expected risk] 100. CLL, chronic lymphocytic leukemia. Fig. 4. ERR for leukemia mortality among the most studied cohorts of individuals exposed to whole-body -radiation at DRs that are very high, very low, or intermediate. (A) The risk is estimated by using a linear quadratic model for Japanese AB survivors (point 1) (ref. 22 and references therein), Chernobyl accident operation workers (point 2) (21), nuclear industry workers (Three-Country Study) (point 3) (22), the U.S. nuclear power workers study (point 4) (23), and the United Kingdom Atomic Energy Authority study (point 5) (Table 1). (B) ERR for leukemia mortality as in A, but point 1 is calculated by using a linear rather than linear quadratic model as shown in ref. 22. above. Furthermore, the minimum for cell killing inactivation approximates those for these same genetic effects. Relative Risk of Leukemogenesis over a Wide Range of DRs. Although we have a great interest in cancer generally, we selected leukemogenesis because human leukemia is the only type of cancer for which there are available data on the risk associated with exposure to IR delivered over a broad range of DRs. Leukemia is the most radiogenic neoplasm, with a relatively short latent period. Furthermore, translocations and other events that involve DNA DSBs are major contributors to leukemogenesis. Data are available from cohort studies of Japanese atomic bomb survivors (reviewed in ref. 21), nuclear plant (and other nuclear industries) workers (22 24), and Chernobyl s accident operation ( clean-up ) workers (21). These data provide estimates for the leukemia risk (per unit of dose) induced in three very different DR regions, calculated here as percent increase of background risk per 10 csv (Fig. 4). What is shown is the excess relative risk (ERR), expressed in percentages, for each of the four cohorts of exposed persons compared with the expected background risk. The first very large group of exposed individuals is the cohort of atomic bomb survivors, who experienced exposure to wholebody irradiation at an average dose of 25 csv. The DR for their exposure has been estimated at 500 csv min. The ERR has been estimated at 2.15 per 1 Sv and 4.55 per 1 Sv based on a linear quadratic or linear model, respectively (22, 23). The leukemia mortality risk was reported to be 21.5% per 10 csv for the first study (22) and 455% per 100 csv, or 45.5% per 10 csv with their use of a linear model, for the second study (23) (Fig. 4). This result was then compared with that of nuclear industry workers, including nuclear plant personnel, for whom the estimate varies, depending on the cohort of individuals studied. The leukemia risk varies between 20% [which is statistically significant for the cohorts investigated in the multinational study organized by the International Agency for Research of Cancer, Lyon, France (22), known also as the Three-Country Study ] and 56.7% in the study of nuclear plant workers only, in the U.S. (23). We used the data of Atkinson et al. (24) to estimate the ERR for leukemia mortality risk for the two groups of U.K. nuclear industry workers, who experienced exposure to two dose levels for which 10 csv is expected to be close to an average dose for the two groups. For a duration of 5 years of exposure, 10 csv would yield a DR of csv min. We directly estimated the standardized mortality ratio for each of the groups and then estimated the average for the two groups (Table 1). This analysis provides an estimated induced risk at 10 csv (in percentage of the background risk, as we use for all other estimations of the risk) of 20%. This value is close to the values of leukemia risk among the cohorts of both nuclear industry workers and Japanese atomic bomb survivors in the Three-Country Study and Lifespan Study, respectively. Thus, it could be concluded that there is no DRE for human leukemogenesis. This conclusion is in strong contradiction to reports on the pattern of DREs for germ-line and somatic mutation and other genetic outcomes of IR that we considered above. Furthermore, comparison of data on leukemia mortality and incidence among the above cohorts, with the data reviewed by the United Nations Scientific Committee on the Effects of Atomic Radiation in 2000 (21) on the cohort of the Chernobyl accident operation ( cleanup ) workers, indicates the existence of an inverse DRE for leukemia too (Fig. 4). Although this latter cohort experienced exposure to whole-body irradiation by -rays at different doses, delivered with widely different DRs, many individuals were exposed to a dose of 10 csv delivered at a DR in the range of csv min, and all of the studied individuals experienced exposure at a DR that is clearly intermediate between the DRs for atomic bomb survivors and nuclear industry personnel. Moreover, a group of experts with the United Nations Scientific Committee on the Effects of Atomic Radiation concluded that there was no evidence for radiation-associated excess risk for leukemia mortality or incidence among these clean-up workers. The calculations of the cgi doi pnas Vilenchik and Knudson

4 Fig. 5. Excess leukemia mortality risk plotted as in Fig. 4, but the very low DR-induced risk (point 3) is represented solely by data from the new, and largest, 15-country International Agency for Research on Cancer study (26). (A) Point 2 is the same as in Fig. 4, but point 1 was calculated by the authors by using only males years for a better comparison with data in points 2 and 3. (B)AsforA, except that B used a linear rather than linear quadratic model for point 1. ERR for the three cohorts (Figs. 4 and 5) fit well to the parabolic DRE relationship for mutation (Fig. 1). A recent publication lends confirmatory support to this conclusion (25). This multinational, collaborative, retrospective study of workers in the nuclear industry (in 15 nations, thus referred to as the 15-Country Study) was coordinated by the International Agency for Research on Cancer and greatly extends the previous Three-Country Study (22), the data from which are plotted in Fig. 4. The principal cohort of this new study comprised 407,391 individuals, most of whom were men who received 95% of the collective dose and who were monitored for at least 1 year, with a total follow-up of 5,192,710 personyears. Individuals with significant ( 10% of whole-body dose) exposure to neutrons or radiation other than -rays or x-rays were excluded. For comparison, only men aged years were included from the Japanese atomic bomb cohort. Thus, all data in Fig. 5 concern very comparable cohorts, because, except for the very high DR group, most of the individuals were males. For the Japanese cohort, both linear and linear quadratic models for risk per unit of dose were used for comparison of the nuclear industry cohort. The results of this comparison are plotted in Fig. 5. Both linear and linear quadratic analyses are shown for the Japanese cohorts (Fig. 5). We include the available data on the Chernobyl accident clean-up workers. The relevant equations are shown in the figures. Parabolic Relationship of Biological Effects to DR. We have analyzed the DREs upon various genetic phenomena, showing both inverse and direct DREs in a parabolic relationship. Here we employ the principle of doubling dose (DD) espoused by James Neel for mutation (reviewed in ref. 8) and extend it to permit quantitative comparisons for the different effects. The DD is the dose that doubles spontaneous (background) effects. In general, DDs for mutation in mammals, including humans, are in the range of 1 2 Sv for germ-line genetic effects. However, such high doses are rarely observed in epidemiologic studies of cancer risk, so we used the data obtained in previous reports (21 24) to estimate the increases (in percent of background level) induced by 10 csv, a level close to those encountered in acute and chronic exposures (21). We let y represent the quantitated biological effect under consideration and x represent the log 10 of the DR in csv min. There is a very good fit of male germ-line mutations in mice related to DRs by Eq. 1 and Fig. 1. We then compared Eq. 1 with those obtained for radiationinduced human leukemia (Eqs. 2 and 3 and Fig. 5). y x x [1] y x x [2] y x x [3] The minimal values of y occur when dy dx 0, namely, at x , , and , respectively, for Eqs The corresponding values in csv min are 0.028, , and , respectively, reflecting lower values for germ-line mutations and leukemia than for somatic mutations in cultured cells. We calculated the increased risk for the outcomes (mutations or leukemia) imparted by 10 csv of IR delivered at a DR of csv min. The values for germ-line mutations and leukemia risk are increases of 19.5% and 31.8% respectively, suggesting a similarity in the mutagenic and oncogenic responses to IR such as -rays in germ-line stem cells and the precursors of leukemia, hematopoietic stem cells, respectively. We also compared the leukemia data with data on chromosomal translocations, which constitute a major mechanism of leukemogenesis, and which arise after DNA DSBs. The relevant study for our analysis (9) reported the yield of germ-line translocations over a wide range of DRs. We found a parabolic effect with a minimum at log 1.08 (0.08 csv min), which is nearer the minimal DRs for germ-line mutations, csv min, and leukemia risk, or csv min, than to those for somatic mutations. The parabolic pattern of the genetic DREs of IR was originally demonstrated for mutation rates in mammalian and human somatic and germinal cells but have now been extended to mitotic recombination, chromosomal translocations, cell inactivation and death, and leukemogenesis. We conclude that all of these effects are parabolic when the y axis is the measured effect expressed linearly and the x axis is the log 10 of the DR. The minima for different effects and different cells cluster in a range of x values of 1.55 to 0.118, corresponding to DRs of csv min. What then is a plausible explanation for the observed range of minimal effects and their relationship to leukemogenesis? (For reviews of leukemogenesis see refs ) MEDICAL SCIENCES Vilenchik and Knudson PNAS November 21, 2006 vol. 103 no

5 Optimal Sensing of the Frequency of Induced DSBs. Both endogenous and induced damage to DNA can involve single-strand lesions or DSBs. Our analysis of published data indicated that single-strand lesions occur endogenously at a rate on the order of 5,000 per cell during a mammalian cell cycle, that most of these are repaired promptly and correctly, but that 1%, or 50, are converted to DSBs (1, 2). Genetic effects (including mutation, recombination, and translocation), cell inactivation, and leukemogenesis all show DREs (Figs. 1 5). The repair of, and response to, IRinduced DNA damage are presumably and predominantly determined by signaling from DSBs (2, 14, 17, 30, 31). Now we show that the DREs are related to the rate of production of EDSBs. We noted earlier that the DD of IR for genetic effects is considered to be in the range of 1 2 Sv for humans. IR produces DSBs per Sv beyond the background rate (17, 19), so a dose of Sv would produce 50 DSBs and constitute a DD. Again there is a close relationship between genetic effects and DSBs. The 50 or so EDSBs are predominantly produced during the S phase of the cell cycle, a period of 6 h for cultured human cells such as fibroblasts. Accordingly, the rate of their production and repair is , or 0.14 DSB per minute. On the other hand, IR induces additional DSBs with a yield of DSBs per Sv, or DSB per csv, through the cell cycle, including S phase. At a rate in the mid-range of the DR for minimal somatic genetic effects, namely, 0.5 csv min, DSB would be produced per minute, a rate that is close to that for the production of EDSBs, during the S phase of the cell cycle. Thus, in somatic cells, such as diploid fibroblasts, a dose of IR of 0.5 csv in 1 min during S phase of the cell cycle approximately doubles the endogenous production of DSBs, from 0.14 DSB min to , or 0.30 min. We then introduce the term doubling DR for 0.5 csv min of IR such as -rays. We interpret this to mean that induced DSBs are optimally recognized and signal a need for error-free repair when they are produced at a rate close to the endogenous rate. We propose therefore that, at DRs below those that are optimal for repair, the frequencies of signals are too far below the background signaling ( noise ) associated with repair of EDSB to be detected with maximal efficiency, whereas at frequencies much above the optimal point the signaling and repair systems are less effective or even fail. Sensing of the induced DNA damage of replicating mammalian cells depends on the rate at which this damage is delivered, as has been directly demonstrated by DeWeese and colleagues (14). This latter group of investigators (14) studied several cell lines for cell survival in response to 2 Gy (2 Sv) of IR at DRs of 4,500, 25, 9.4, and 2 cgy h, or 75, 0.417, 0.157, and cgy min. Cell killing inactivation showed an inverse DRE. The investigators then tested the idea that signaling from DSBs was defective at the lowest DRs. They found that one protein, ATM, known to be critical in signaling from DSBs, showed relatively less efficient activation at low DR levels, and another protein, H2AX, showed relatively less phosphorylation at those levels. The repair of EDSBs is accomplished mainly by homologous recombination repair, whereas repair of radiation-induced DSBs is primarily accomplished by nonhomologous end joining. Why then are induced DSBs optimally repaired when their rate is similar to that of EDSBs? As noted above, the proteins ATM and H2AX are important for recognition of both induced DSBs and EDSBs, as are some other components in the DSB repair process, e.g., the complex of RAD50, MRE11, and NBS1 proteins, which are shared by homologous recombination repair and nonhomologous end joining. It is interesting that Collis et al. (14) found that interactions between NBS1 and ATM are compromised at very low DRs of IR, with considerable reduction of levels of -H2AX-marked DSBs. This conclusion is also consistent with the report of Rothkamm and Lobrich (32) that failure of DSB repair can occur at a very low dose of IR (1 mgy). Conclusions We previously reported that the genetic effects (mutations) observed for exposure of both germ-line and somatic cells to IR are parabolically related to the logarithm of the DRs, for a given dose, with minimal effects at DRs in the range of csv min. Here we report that a similar result is obtained for other effects, namely, mitotic recombination, chromosomal translocation, and cell death or inactivation, and note that all of these effects can be related to the production of DSBs in DNA from both endogenous and exogenous sources. Because radiation-induced leukemias, especially myeloid leukemias, frequently reveal chromosomal translocations, we estimated the incidence of induced leukemia as a function of DR. Data on DRs for three different exposure groups (atomic bomb survivors, nuclear plant workers, and Chernobyl accident cleanup workers) enabled us to show again a parabolic relationship to DR for a given dose, with a probable minimum in the range of that for specific-locus mutations and translocations in spermatogonial stem cells. Using previously published data on the occurrence of endogenous DNA single-strand lesions (1) and DSBs (2) in normal somatic cells, wherein we estimated 50 DSBs per S-phase of the cell cycle, we compared the rate of induced DSBs at an intermediate point in the DR range for minimal somatic genetic effects, namely, 0.5 csv min, with the endogenous rate. The former result was 0.16 DSBs per minute, and the latter result was 0.14 DSBs per minute. We conclude that maximal repair efficiency is obtained when the induced DSB rate is similar to the endogenous rate and that signaling for a response to induced damage involves mechanisms that are related to those used in responding to endogenous damage. As radiation DRs increase above the level of maximal DNA repair efficiency, cellular responses to radiation are not so efficient; as induced DRs decrease below this level, the ratio of signal from induced DSBs to the noise of EDSBs becomes less favorable for repair with high fidelity. We previously concluded that cells are able to measure the frequency of DNA damage and to compare that frequency with a spontaneous, or background, frequency (1). The mechanism would therefore resemble that of stochastic resonance that is observed for signal-to-noise relationships in physics and physiology. Here we propose that the noise is the frequency of EDSBs ( hits ) and that the signal is the frequency of induced DSBs. We refer to this process of the comparison of frequencies as signaling resonance. Methods of Analysis We analyzed published data on mutation, recombination, leukemogenesis, and cell inactivation killing, phenomena that are not usually analyzed with the same ordinates. Data on leukemogenesis are usually reported as ERR for leukemia above background per unit of dose, which we report as a percentage increase. In many cancers ERR is usually expressed per Sv, but for leukemia, in the cohorts we analyzed, exposure was far below 1 Sv. Because the Chernobyl accident clean-up workers were exposed to an average dose of 10 csv, and because many nuclear industry workers and atomic bomb survivors on average, received, respectively, several csv and 20 csv, we calculated the effects per 10 csv. To facilitate comparison with genetic effects (mutation and recombination), we also calculated the latter as ERR per 10 csv above background. Cell inactivation has been reported in terms of survival, but to compare the results with those on genetic and leukemogenic effects we converted the data to percentages of inactivation killing in response to 5 Sv cgi doi pnas Vilenchik and Knudson

6 M.M.V. thanks Prof. Arthur K. Balin and Dr. Loretta Pratt of The Sally Balin Medical Center for their support. We thank Drs. Anna Marie Skalka and Theodore DeWeese for their constructive comments on the manuscript. This research was supported by an appropriation from the Commonwealth of Pennsylvania and National Cancer Institute Core Grant CA06927 to the Fox Chase Cancer Center. 1. Vilenchik MM, Knudson AG (2000) Proc Natl Acad Sci USA 97: Vilenchik MM, Knudson AG (2003) Proc Natl Acad Sci USA 100: Lyon MF, Papworth DG, Phillips RJ (1972) Nat New Biol 238: Russell WL, Kelly EM (1982) Proc Natl Acad Sci USA 79: Russell WL, Kelly EM (1982) Proc Natl Acad Sci USA 79: Furuno-Fukushi I, Ueno AM, Matsudaira H (1988) Radiat Res 115: Kiefer J, Muller J, Gotzen J (1988) Radiat Res 113: United Nations Scientific Committee on the Effects of Atomic Radiation (1993) Sources and Effects of Ionizing Radiation (United Nations, New York). 9. Pomerantseva MD, Goloshchapov PV, Vilkina GA, Shevchenko VA (1984) Mutat Res 141: Steel GG (1996) Phys Med Biol 41: Hall EJ (2000) Radiobiology for the Radiologist (Lippincott Williams & Wilkins, Philadelphia). 12. Mitchell JB, Bedford JS, Bailey SM (1979) Radiat Res 79: DeWeese TL, Shipman JM, Dillehay LE, Nelson WG (1998) J Urol 159: Collis SJ, Schwaninger JM, Ntambi AJ, Keller TW, Nelson WG, Dillehay LE, DeWeese TL (2004) J Biol Chem 279: Mitchell CR, Folkard M, Joiner MC (2002) Radiat Res 158: Mitchell CR, Joiner MC (2002) Int J Radiat Biol 78: Iliakis G (1991) BioEssays 13: Baverstock KF (1991) Int J Radiat Biol 60: Nevaldine B, Longo JA, Vilenchik M, King GA, Hahn PJ (1994) Radiat Res 140: Little JB (2000) Carcinogenesis 21: United Nations Scientific Committee on the Effects of Atomic Radiation (2000) Sources and Effects of Ionizing Radiation (United Nations, New York). 22. Cardis E, Gilbert ES, Carpenter L, Howe G, Kato I, Armstrong BK, Beral V, Cowper G, Douglas A, Fix J, et al. (1995) Radiat Res 142: Howe GR, Zablotska LB, Fix JJ, Egel J, Buchanan J (2004) Radiat Res 162: Atkinson WD, Law DV, Bromley KJ, Inskip HM (2004) Occup Environ Med 61: Cardis E, Vrijheid M, Blettner M, Gilbert E, Hakama M, Hill C, Howe G, Kaldor J, Muirhead CR, Schubauer-Berigan M, et al. (2005) Br Med J 331: Felix CA (2001) Med Pediatr Oncol 36: Greaves MF, Wiemels J (2003) Nat Rev Cancer 3: Janz S, Potter M, Rabkin CS (2003) Genes Chromosomes Cancer 36: McHale CM, Smith MT (2004) Am J Hematol 75: Rothkamm K, Kuhne M, Jeggo PA, Lobrich M (2001) Cancer Res 61: Thacker J (1992) Adv Radiat Biol 16: Rothkamm K, Lobrich M (2003) Proc Natl Acad Sci USA 100: MEDICAL SCIENCES Vilenchik and Knudson PNAS November 21, 2006 vol. 103 no