Health Effects of Exposure to Low Levels of Ionizing Radiations: Time for Reassessment? (1998)

Estimation of radiation-related risk for radioepidemiologic tables is a useful check for the modeler and it relates to actual applications of the model such as claims for compensation in individual contested cases. Each possible estimate is to be treated as the current scientific consensus judgment in a particular case. For example, a model that produces a sharp change in estimated probability of causation, depending on whether exposure occurred at age 39 or 40 yr, would lack credibility even though it might fit the data better than a model with a smooth exponential decline in ERR with increasing age at exposure. It can be argued that a model that agrees well with scientific observations while avoiding anomalies that would seem unreasonable and capricious as a basis of real-life decisions would suit the interests of the sponsors of BEIR VII.

There is a legal requirement for periodic revision of the 1985 radioepidemiologic tables, which now are out of date in view of changes in understanding of radiation-related risk over the last decade or so. Any such revision presumably will depend heavily on the algorithms developed in the most recent BEIR reports, modified as necessary to meet the requirements for plausibly equitable decisions in individual cases. It would be appropriate for the BEIR VII phase-2 committee to produce its risk estimates in a format that would enable scientists revising the radioepidemiologic tables to incorporate the latest BEIR estimates.

The genetic material of cells is DNA, which is distributed among the chromosomes of eukaryotic cells and is bound to structural and other proteins. Because the two strands of DNA are complementary—a purine base (adenine or guanine) on one strand pairing with a pyrimidine base (thymine or cytosine) on the other strand—the information contained in the sequences of bases is redundant. In a haploid human cell, there are about 3 × 10⁹ base pairs of DNA, which includes about 100,000 genes whose sequences specify all the structures and reactions that make up the cell and the entire human being, including the control of DNA replication and cell division. If there are DNA sequence changes in germ cells that affect offspring, the changes that result are hereditary changes or mutations. Changes in the normal sequences of bases in somatic-cell DNA as a result of endogenous reactions or exogenous agents might alter the normally well controlled cellular processes and result in loss of homeostatic regulatory mechanisms, loss of inhibition or stimulation of cell growth and division, or cell death. The uncontrolled and metastatic growth of tumor cells, derived from previously normal cells, is associated with changes in DNA sequence in somatic cells.

DNA in humans is a large molecule subjected to hydrolytic attack and to endogenous oxidative and other damage at 37ºC. For example, it has been estimated that 2-10 × 10³ DNA purines (of a total of about 3 × 10⁹) turn over in each human cell each day (Lindahl 1993). Over a 70-yr lifetime, depurination could affect 10% of a person's DNA. Furthermore, DNA alterations caused by the deamination (removal of an amine group) of cytosine and 5-methyl-cytosine (and to a lesser extent adenine and guanine) lead to coding changes that must be rectified.

In addition to the damage that results from its normal chemical bond breakage and reunion errors, DNA is assaulted by reactive oxygen species generated by "leakage" from mitochondria, flavin-catalyzed reactions, and many other sources, including phagocytosis and inflammation (Beckman and Ames 1997). The superoxide radical (O₂^-), formed by one-electron reduction of molecular oxygen, is generated in all aerobic cells. Chemical or enzymatic dismutation of (O₂^-) produces hydrogen peroxide, H₂O₂. The toxicity of these species has been attributed to the highly reactive hydroxyl radical (OH^.), which can be formed by reactions of (O₂^-) and H₂O₂. Floyd (1995) has estimated that about 1% of the oxygen consumed by human cells is diverted to oxidizing cellular protein and that 0.001% of the oxygen molecules damage DNA and RNA; these numbers undoubtedly increase under conditions of oxidative stress, such as during chronic inflammation. Although protein and small molecules, such as glutathione, serve as scavengers for reactive oxygen and thus protect the nucleic acids, there is a considerable amount of oxidative DNA-base damage per cell per day (Saul and Ames 1986). However, the steady-state level of DNA damage is low, so most of the spontaneous and metabolically-generated damage is apparently repaired efficiently and correctly. Poor repair would allow the accumulation of excessive DNA damage that could interfere with DNA replication and transcription and ultimately threaten survival. Thus, although DNA in cells is frequently damaged, the damage is counteracted by DNA-repair processes.

Added to the sources of spontaneous damage and metabolically produced oxidative DNA damage is natural background radiation. The principal sources of external exposure from natural sources are cosmic radiation and naturally occurring radionuclides in the earth/soil. The primary sources of internal exposure are radionuclides, such as potassium-40, deposited within tissue. Collectively, these two sources deliver effective (whole body) close rates to members of the US public that range from 1 to 2 mSv per year. One sievert represents an amount of absorbed energy equivalent to 1 J/kg, adjusted to take into account the quality factor of the radiation. Artificial radiation sources, such as x-rays used in medical diagnosis and radiopharmaceuticals used in nuclear medicine, add an additional effective dose rate to the average member of the US public of about 0.50 mSv per year—0.40 mSv from medical x-rays, and about 0.14 mSv from nuclear medicine (NCRP 1989). The total effective dose rate from these two artificial sources is thus about half that from the natural background sources cited above. In addition, naturally occurring indoor radon and its

airborne radioactive decay products, whose concentrations vary widely from one geographic location to another, add an estimated effective dose rate of 2 mSv per year to the average member of the US public (NCRP 1987). Ionizing radiation produces OH^. and other radicals ( e_aq^- and H atoms) by interacting with cellular water and exerts the bulk of its biologic effects in cells through these free radicals, in particular OH^.. Ionizing radiation produces several classes of damage to DNA, including single-strand breaks (SSBs) and double-strand breaks (DSBs) in the DNA chains, DNA-DNA covalent cross-links, DNA-protein covalent cross-links, and a large variety of oxidative changes in the nucleotide bases (Hutchinson 1985; Ward 1988). The identified oxidative base products of ionizing radiation are chemically identical with those produced by other oxidizing agents, such as H₂O₂ in the presence of iron or copper ions, and those resulting from the normal metabolic production of free radicals that are byproducts of the transport of electrons to oxygen in mitochondria (Dizdaroglu and others 1987, 1991a,b; Gajewski and others 1990; Nackerdien and others 1991; Dizdaroglu 1992; Beckman and Ames 1997). Ionizing radiation damages DNA both through direct deposition of energy in the DNA (which is considered to include the first layer of tightly bound water) and indirectly through the generation of OH^. radicals in the water within the immediate vicinity of the DNA. Early experiments demonstrated that about 70% of the DNA damage can be prevented by the addition of OH^. scavengers (Roots and Okada 1972). Because OH^. is so highly reactive, it has been estimated that only the radicals formed within about 3 nm of the DNA can react with it (Ward 1994).

It has been argued in both the scientific and lay press that the quantity of spontaneous and metabolically generated DNA damage is many orders of magnitude larger than that resulting from low, protracted doses of radiation from environmental sources implying that the contribution from low doses of ionizing radiation is trivial (Billen 1990; Beckman and Ames 1997)—in other words, that the DNA damage produced by background radiation and the even higher doses to which some workers are exposed does not add appreciably to the extensive spontaneous and metabolic damage and can be ignored.

A counterargument is based on unique aspects of ionizing radiation damage to DNA. Accumulated evidence shows that the products of ionizing radiation differ from chemically generated oxidation products in the microdistribution of the damage rather than in the chemistry of the individual lesions (Ward 1981, 1988, 1994). A portion of the energy of ionizing radiation, primarily that from secondary electrons, is deposited in large-enough packets to produce clusters of OH^. radicals. Clusters of ionization were first observed in a cloud chamber (Wilson 1923), then extended to liquid water (Samuel and Magee 1953), and later shown to result from properties of the radiation-track structure (Goodhead 1989, 1994; Pimblott and Mozumder 1991). Because OH^. has a very short range owing to its high reactivity, it can produce a cluster of damage within a

few base pairs along the DNA if the cluster is generated within 3 nm of the DNA. Ward and others (1985) have referred to such lesions as multiply damaged sites (MDSs). The probability of clustered damage increases with dose and linear energy transfer (LET) but is independent of dose rate because it results from the passage of a single-particle track (Prise and others 1994; Holley and Chatterjee 1996; Rydberg 1996). A DSB resulting from a single energy deposition is the most obvious example of a MDS, but other combinations of strand breaks, cross-links, and base or sugar products can also occur (Ward 1994). Furthermore, both direct interactions of radiation with DNA and reactions of OH^. contribute to the complexity of MDSs (Nikjoo and others 1997).

A second property of ionizing radiation that might distinguish it from chemical radicals is the extensive production of peroxyl radicals due to initial damage to molecules other than DNA (Floyd 1995; Milligan and others 1996). Peroxyl radicals produce oxidized bases but not DNA strand breaks and might account for the greater-than-expected yield of base damage, as opposed to strand breaks, observed in irradiated cells (Nackerdien and others 1992), as well as the production of double base lesions by single radicals that have been observed in irradiated oligonucleotides (Box and others 1995).

Ward has calculated that 5 μM H₂O₂ can produce 15 Gy-equivalents of SSBs in mammalian-cell DNA in 30 min through OH^. generation catalyzed by iron ions bound to DNA; on the basis of these SSB yields, 1000 Gy-equivalents are required to kill cells (Ward and others 1985). Similarly, on the basis of the mount of Oxidative base damage excreted by rats each clay (4,600 molecules of thymine glycol, an amount equivalent to that produced by 4.7 Gy of ionizing radiation per day), or the measured mount of 8-oxoguanine generated daily in each rat liver cell (80,000 molecules or 40 Gy-equivalents), base damage cannot be of consequence in the killing of cells by ionizing radiation (Ward 1988; Beckman and Ames 1997). In fact, at the D₃₇ dose for cell-killing, it has been calculated that each cell will have sustained 2.5 million SSBs for H₂O₂ and 400,000 pyrimidine dimers for UVC radiation. In contrast, the D₃₇ dose for low-LET ionizing radiation produces only 1,000 SSBs but 40 DSBs, a type of damage that is not characteristic of lethal doses of H₂O₂ or UVC radiation. Such data suggest DSBs are the critical lethal phenomena. DSBs and other MDSs are peculiar to ionizing radiation and a few radiomimetic agents, such as bleomycin and neocarzinostatin.

The mount of energy deposited that can yield MDS increases with LET, and MDSs are generally thought to explain the increased biologic effectiveness of high-LET radiation and the poor repairability of the induced DNA damage. At the least, clustering will create complex DSBs within up to 10 bp or so (Ward and others 1985; Holley and Chatterjee 1996). Because of the wrapping of DNA around nucleosomes and the organization of the chromatin fiber, some clusters might include DSBs at two or more sites that are several kilobase pairs apart or even removed from each other by the distance of a chromosomal loop of about 100 kbp (Lobrich and others 1996; Rydberg 1996). A relationship between the protein composition of the nuclear matrices of cells deficient in

the repair of DSBs and chromosomal-loop dynamics is consistent with the idea that chromosome structure affects both DNA and cellular radiosensitivity (Roti Roti and others 1993; Malyapa and others 1994, 1996). Chromatin proteins and condensation can also directly affect access of OH^. to DNA and thereby protect DNA from damage (Ljungman 1991; Ljungman and others 1991; Warters and Lyons 1992; Elia and Bradley 1992; Chiu and others 1992; Xue and others 1994). A more open structure can make active chromatin domains more sensitive than the bulk condensed chromatin to radiation damage (Chiu and others 1982; Bunch and others 1992). The nuclear matrix and its associated DNA can also suffer excess damage, both DSBs and DNA-protein covalent cross-links, became of its more open structure relative to the bulk chromatin and because of the binding of metal ions capable of catalyzing the formation of additional OH^. (Chiu and others 1986, 1993, 1995; Balasubramaniam and Oleinick 1995).

For cells to survive without mutations, DNA damage must be faithfully repaired. Whereas spontaneous damage is readily repaired in repair-competent cells, the DSBs and clustered lesions produced by even low-LET radiation are likely to be repaired with difficulty or incorrectly, if at all (Ward 1988, 1994). However, conventional assays for monitoring the yield and repair of DSBs would not detect the majority of MDSs (ones that contain one or no initial strand breaks) and would treat complex DSBs as simple ones. One approach to study the repair of clustered damage is to synthesize oligonucleotides that contain defined sets of damage and to monitor the ability of specific repair enzymes to act on those sequences, as opposed to sequences that contain simple types of damage. One study has demonstrated impairment of repair if two base damages lie within 5 bp of each other (Chaudhry and Weinfeld 1995). Given the very large number of possible combinations of lesions within MDSs and the several types of enzyme systems that might attempt repair, considerably more work is needed on this subject. If two or more DSBs occur within a single chromosomal loop, the fragment between the two breaks is theoretically no longer bound to the nuclear matrix and might be more difficult to repair.

Within the limits of detection of standard assays of DNA damage, induction of DSBs and other lesions in cellular DNA is generally found to depend linearly on radiation dose (Iliakis and others 1992; Lange and others 1993). Assays for the measurement of removal of base damage or the rejoining of SSBs or DSBs reveal that repair begins in cells as soon as radiation damage occurs. DSB rejoining proceeds rapidly with apparently biphasic kinetics; the half-time for the first (rapid) repair phase has been estimated at about 10-20 min and that for the second phase about 0.5-2.0 h (for example see Metzger and Iliakis 1991). The initial rate of damage removal decreases modestly with increasing dose, and the extent of residual unrejoined DSB might increase with dose. The data suggest that the enzyme systems for DSB rejoining are constitutively present in repair-competent cells. However, such measurements are made at supralethal radiation doses and cannot detect the removal of all the lesions; furthermore, some components of repair of the measured DSBs might be inducible.

The steady-state level of DNA damage is low, so extensive spontaneous damage must be rapidly and effectively repaired. Errors in DNA replication, such as the placing of a thymine opposite a guanine, create mismatches that are corrected by "proofreading" activities of the DNA polymerase complex and removal of the newly incorporated incorrect base before the next base is added. Alternatively, if the mismatch is not at the growing end of a DNA chain, mismatch-repair enzymes remove the wrong nucleotide, and the resulting gap in one strand is filled in properly by a DNA polymerase. Defects in the mismatch-repair enzymes have been associated with genetic instability and the human familial syndrome hereditary nonpolyposis colon cancer (Modrich 1994; Fishel and Kolodner 1995; Marta and Boland 1995). Most oxidative base damage and SSBs, including those derived from ionizing radiation and from metabolic sources, are efficiently repaired by the base-excision repair pathway, which initiates the removal of damaged bases via the generation, by one of several specific N-glycosylases, of an apurinic-apyrimidinic (AP) site, which is then a substrate for associated AP endonucleases (Demple and Harrison 1994; Wallace 1994). Some kinds of oxidative base damage are also repaired via the nucleotide-excision repair pathway that is thought to be the primary repair mechanism for UVC-induced pyrimidine dimers and bulky adducts (Sancar and Tang 1993; Sancar 1995). In spite of the importance of those repair systems, none of the human syndromes that are characterized by a sensitivity to ionizing radiation have been attributed to defects in the repair of oxidative damage, with the possible exception of Cockayne syndrome, discussed below.

The repair of DSBs in human cells is effected primarily by nonhomologous end joining (NHEJ) and less by homologous recombination and single-strand annealing (Thompson 1996). NHEJ requires the participation of DNA-dependent protein kinase (DNA-PK), the enzyme system that carries out the end-rejoining component of V(D)J recombination in developing immune cells. DNA-PK is composed of a dimer of DNA-end-binding proteins Ku-70 and Ku-86 (the Ku autoantigen), which serve as a nucleus for the binding of the 470-kDa catalytic subunit (DNA-PKcs) (Jeggo and others 1995). Cells deficient in DNA-PK subunits are defective in DSB repair and V(D)J recombination and highly sensitive to ionizing radiation (Biedermann and others 1991; Taccioli and others 1993). DNA-PKcs is a serine-threonine kinase that is a member of the phosphatidyl inositol-3-kinase family (Hartley and others 1995). Another member of this kinase family is ATM, the protein mutated in ataxia telangiectasia (AT), a disease that is also characterized by immune deficiencies and radiosensitivity (Jorgensen and Shiloh 1996). In spite of limited homology between DNA-PKcs and ATM in the kinase domain, the substrates for the two enzymes are different, and cells from AT patients, although highly sensitive to the lethal effects of ionizing radiation and defective in normal radiation-induced cell-cycle progression delays by radiation damage, are not defective in the repair of DSBs (Meyn 1995). Evidence of the existence of other rejoining mechanisms, such as one operating in late S/G2 phase of the cell cycle, has also been obtained (Whitmore and others 1989). The fidelity of the rejoining process is largely accurate to within ±100 nucleotides (Lobrich and others 1995), but the nucleotide-sequence accuracy of the process remains to be determined.

Interestingly, the enzymes of DNA repair in normal mammalian cells are constitutively present and do not require induction to repair DNA damage. Thus, if ionizing-radiation damage is produced as a linear function of dose, if a component of that damage is unique (such as MDSs) and not found as a segment of the background DNA damage, if the requisite repair enzymes do not require induction, and if the repair rate is not markedly altered as a function of dose, one could conclude that even the lowest dose of radiation can be biologically significant. In contrast, radiation damage can trigger a plethora of inducible processes, some of which can affect damage-recognition processes, repair, or the cellular responses to initial or unrepaired damage.

As elaborated in the previous section, the cell contains a variety of mechanisms for repairing or tolerating damage deposited in DNA by spontaneous or endogenous events, as well as by environmental radiation and chemical agents. The capacity of these repair systems is set at some constitutive level such that a steady-state concentration of DNA lesions remains in the genome of normally growing cells. It is interesting and important that this steady-state concentration is nonzero and that some of the lesions that persist are known to contribute to mutagenesis and other potentially deleterious biologic end points. In principle, the frequency of spontaneous mutations would be lower if the constitutive efficiency of DNA-repair were higher. We do not yet understand how the efficiency of the DNA repair systems is regulated so as to maintain the steady-state concentration of lesions, but we do appreciate that the constitutive repair systems provide only a limited capacity to deal with additional damage that might be inflicted by external threats, such as ionizing radiation. We have also learned in recent years that not all lesions are equally accessible to recognition by repair enzymes. Intragenomic DNA repair is heterogenous; lesions in some domains of the genome are poorly repaired; whereas those in others are repaired with relatively high efficiency (Hanawalt 1991). Thus, some bulky DNA adducts in the highly repetitive DNA sequences found near centromeres are poorly repaired, in comparison with the overall genome, and lesions that block transcription appear to be preferentially repaired. In particular, the lesions in the transcribed DNA strand that arrest the progression of polymerase II are preferentially repaired. Perhaps this so-called transcription-coupled repair mode has evolved because the stalled RNA polymerase otherwise encumbers recognition and repair of the arresting lesion (Hanawalt 1994). The existence of intragenomic DNA-repair heterogeneity means that we need to understand the fine structure of lesion processing in relation to the particular genes or genomic domains responsible for the biologic end point of interest, such as cancer. In addition to the heterogeneity of DNA repair at the level of the gene, there is heterogeneity at the level of the nucleotide. Thus, a given type of lesion might be much more efficiently repaired in one nucleotide-sequence context than in another. In some cases (for example, in the p53 tumor suppresser gene) the sites of slow repair have been shown

to correlate with the sites that are most frequently found to be mutated in tumors (Tornaletti and Pfeifer 1994).

The cell is not passive in its response to environmental genotoxic threats. A wide variety of genes are known to be activated by such agents as ultraviolet light (UV) or ionizing radiation, although only a few of them have been directly implicated in DNA repair in mammalian cells. Genes inducible by x-rays include the p53 tumor-suppresser gene, the proliferating-cell nuclear antigen (PCNA), and the DNA polymerase ß. Polymerase ß is used in the primary pathway of base-excision repair, so it is important for the repair of some of the principal types of base damage produced by radiation. PCNA is the "sliding clamp" that ensures processivity of DNA polymerase δ/ε for both chromosomal replication and repair replication in the process of nucleotide-excision repair. Additional DNA-repair genes will probably be shown to be inducible in mammalian cells. In model bacterial systems, several inducible systems are now well understood at the biochemical level and clearly involve up-regulation of DNA-repair gene expression. Thus, the SOS system, controlled by the recA-lexA regulatory circuit, results in the induction of the uvrA, uvrB, and uvrD genes (and others), with consequent enhanced efficiency of nucleotide-excision repair. In the adaptive response to alkylation damage, glycosylases specific to DNA lesions and a 6-alkyl-guanine transferase are induced, thereby leading to greatly enhanced tolerance of agents that produce alkylation damage.

We know much less about the induction of specific DNA-repair pathways in mammalian cells, but some reported phenomena are consistent with the existence of such pathways. Thus, the basic phenomenon of "Weigle reactivation" that originally led to the discovery of the SOS response in bacteria, has also been confirmed in mammalian cellular systems. In brief, that phenomenon involves the enhanced survival of UV-damaged virus when the host cells have been preconditioned by exposure to low doses of UV or other DNA-damaging agents. In bacteria, the enhanced survival is now known to be due primarily to the up-regulation of genes involved in the damage-recognition step of nucleotide-excision repair, as noted above. Similarly, the UV-induced and p53-dependent up-regulation of global excision repair in human cells might be due to enhanced expression of DNA-repair genes (Ford and Hanawalt 1995, 1997).

A number of recently reported provocative phenomena in mammalian cell systems deserve careful study to determine their biochemical mechanisms and possible relevance to the low-dose response to radiation and the question of linearity of that response. It has been shown, that in cultured human lymphocytes, low doses of radiation result in the "protection" of the cells from the chromosomal aberrations or mutations that would otherwise result from later exposure to radiation (Olivieri and others 1984; Sanderson and Morley 1986; Wiencke and others 1986; Kelsey and others 1991; Wolff 1992; Shadley 1994) or to some chemical agents that produce DSBs. In human fibroblasts, a low chronic exposure to radiation was found to reduce the frequency of micronuclei derived from later acute radiation exposure; this finding is evidently correlated with an increased

rate of DSB repair. Furthermore, it has been shown in C3H10T_1/2 cells that the frequency of spontaneous neoplastic transformation can be reduced by a factor of 3 or 4 by a single exposure of the quiescent cells to radiation doses as low as 0.1 cGy. However, it is important to note that the subclone of C3H10T_1/2 cells used in these experiments exhibited an unusually high level of spontaneous transformation and that the basis for that phenotype is not understood (Azzam and others 1996).

Some recent studies suggest that important biologic effects, including the induction of sister chromatid exchanges and changes in gene expression, can occur in an irradiated population in cells that have received no direct radiation exposure. These so-called bystander effects might be a consequence of communication among cells in the population, although in some cases the results might be explained by indirect effects, such as radiation action on components of the culture medium. Recent evidence has implicated the up-regulation of oxidative metabolism and the production of active oxygen species as mediators of the effects. It is important to appreciate that the intercellular communication that exists in the normal tissue environment of cells in an intact organism (such as a human) is a complication that ultimately limits the utility of model cultured-cell systems in vitro. Another interesting phenomenon is the induction by ionizing radiation of a type of genomic instability whereby important biologic effects occur in the progeny of irradiated cells after many generations of cell replication. The occurrence of this effect has now been confirmed in a number of laboratories for end points that include mutagenesis, cytogenetic changes, and reproductive failure (Kadhim and others 1992; Grosovsky and others 1996; Morgan and others 1996; Little and others 1997). The mechanisms by which this instability is induced and maintained over a long period remain to be elucidated.

The p53-regulated pathways are important and have received much recent attention because mutations in the p53 gene are found in a large percentage of human tumors. p53 is regulated primarily at the level of translation and the stability of the protein, and it is involved in cell-cycle checkpoints, in apoptosis, and in nucleotide-excision repair. In the cancer-prone Li-Fraumeni syndrome, fibroblasts expressing only mutant p53 exhibit little apoptosis and are therefore radiation-resistant. Interestingly, they are deficient in global nucleotide-excision repair but proficient in transcription-coupled repair (TCR) (Ford and Hanawalt 1995). The loss of p53 function can lead to genomic instability by reducing the efficiency of genomic repair whereas cellular resistance is ensured through the operation of TCR and the elimination of apoptosis. Recent reports from several laboratories suggest that an important inducing signal for p53 stabilization and consequent apoptosis is the arrest of transcription at lesions in the DNA. In the case of Cockayne syndrome, characterized by deficiency in TCR, p53 and apoptosis are induced by much lower doses of radiation than in normal cells or in xeroderma pigmentosum complementation group C cells, which are proficient in TCR but deficient specifically in global genomic repair (Ljungman and Zhang 1996). The p53-induction pathway might be of particular relevance to low-dose radiation effects because of the demonstration that some base damage (such as thymine glycol damage) is subject to TCR and that people with Cockayne syndrome are defective in the TCR of this type of damage (Leadon and Cooper

1993). Cells from people with Cockayne syndrome have been shown to be sensitive to radiation and to UV radiation; this leads to the suggestion that the characteristic developmental problems in this hereditary human disease are be caused by cell (particularly neuron) loss due to enhanced apoptosis (Leadon and Cooper 1993; Hanawalt 1994).

A large number of radiation-induced gene products have been identified by comparing 2D gels after electrophoresis of extracts from irradiated and untreated control cells. Differential screening of cDNA libraries has been used to identify radiation-induced genes. Although most of these remain to be characterized, some of the early radiation-induced genes have been identified, including AP-1 and NF-KB, in addition to p53. In fact, AP-1 and NF-KB sites have been found in many UV-induced and radiation-induced genes, and these factors have also been shown to contribute to the induction of HIV-LTR after UV exposure. Cytokines have been shown to be induced by radiation, including IL-1d, TNF, interferons, IL6, TGFB, and bFGF. An important caution, however, is the finding that pathways of gene induction after radiation exposure might be different in endothelial cells from such pathways in other cells. One must therefore be cautious about generalizing from an inducible response in one type of cell to that in another—or for that matter from cultured cells to cells in a tissue. It is also important to understand the interaction between different repair pathways in that the results of knocking out or up-regulating a particular pathway are often unpredictable. The disruption of mismatch repair has been shown in a number of studies to enhance tolerance to DNA damage, including that produced by reactive oxygen species. Possibly the mismatch-repair system normally interferes with the processing, by nucleotide-excision repair, of some lesions produced by reactive oxygen species.

Comparisons between the widely varied genes induced by UV should be informative. For some proteins, the induction occurs within minutes and can be observed with x-ray doses as low as 10 mGy. The p53 response to x-rays reaches a peak several hours after irradiation but it is transitory and smaller than that after UV exposure (Lu and Lane 1993). Enhanced expression of p53 has also been reported in bystander cells in cultures exposed to alpha rays (Hickman and others 1994). The implication is that substantial communication occurs among the cells in culture and that the biologic effects in cell populations might not be restricted to the responses of the individual damaged cells, as noted earlier. Again, the complication of intercellular communication, when one considers cells in a tissue, is surely important.

The existence of inducible repair systems that improve the efficiency of DNA repair has fueled speculative proposals that low levels of ionizing radiation actually have beneficial, rather than deleterious, effects. These suggestions of hormesis in the radiation response must be considered seriously but critically. Some of the arguments do not take into account the important differences between the DNA-damage spectrum produced by radiation and that produced by endogenous reactive oxygen species—differences that influence the spontaneous mutagenesis level substantially. Thus, an argument for low-dose radiation hormesis goes as follows: If the low-LET background of 1 mGy/yr were

increased to 10 mGy/yr, and that stimulated a 10% increase in DNA-repair efficiency, then mutations due to background radiation would increase from 10⁹/day to 9 × 10⁹/day (not 10×10⁹/day), and the overall level of background mutations due to endogenous damage would decrease from 10¹³/day to 9 × 10¹²/day (for a net decrease of 1×10¹²) (Myron Polycove, personal communication).

Some epidemiologic data have been cited as consistent with the existence of hormetic mechanisms. However, there have been no carefully controlled studies that negate the conservative view that even very low doses of radiation simply add to the burden of cellular damage and thereby increase the likelihood of deleterious mutagenesis. We need to obtain much more mechanistic information in the general area of inducible responses to DNA-damaging agents such as radiation.

Since the acceptance of the unineme structure of chromosomes, it has been generally agreed that the DNA DSB, equivalent to a chromosomal backbone break, is the critical radiation-induced damage that leads to chromosomal aberrations. Experiments with different restriction enzymes that induce specific types of DSBs provide good evidence that both blunt-ended and staggered-ended DSBs can cause chromosomal aberrations (Bryant 1984; Natarajan and Obe 1984; Obe and Winkel 1985; Winegar and Preston 1985). The prevailing concept of the formation of chromosomal aberrations proposes that radiation induces two DSBs that interact with each other to produce aberration configurations—such as dicentrics, reciprocal translocations, and rings—through incorrect rejoining of the broken ends. Two mutually exclusive models are used to describe the formation of aberrations: the ''classical" or "breakage-reunion" hypothesis and the "exchange" hypothesis (Lea 1946; Revell 1974; Bender and others 1974; Savage 1989). More recently, it has been proposed that aberrations might be derived from one DNA DSB via nonhomologous, or illegitimate, recombination (Chadwick and Leenhouts 1981; Szostak and others 1983; Goodhead and others 1993), although this hypothesis is not widely held.

Mutations are generally classified as point mutations, which are intragenic and thought of as small changes in a DNA gene sequence, or chromosomal mutations, which are intergenic and thought to result from major alterations in chromosomal structure. Mutations of both types are considered to arise from DNA DSBs, as is supported by evidence from restriction-enzyme cutting experiments (Singh and Bryant 1991), and the second type is clearly associated with large deletions of DNA and possibly chromosomal aberrations. Radiation is generally considered to induce chromosomal mutations rather than point mutations, although point mutations are found after radiation exposure (Vrieling and others 1985; Thacker 1986; Breimer and others 1986; Liber and others 1986; Kraemer and others 1994; Little 1994; Meuth and Bhattacharyya 1994; Simpson and others 1994).

Dose-effect relationships for chromosomal aberrations and mutations induced by sparsely ionizing radiation are invariably interpreted to have linear-quadratic dose kinetics, moving to linear dose kinetics for more densely ionizing radiation, which is more effective per unit of dose than sparsely ionizing radiation (Lloyd and others 1975, 1976; Cox and Masson 1979; Thacker and others 1979; Liber and others 1983; Hei and others 1988; Metting and others 1992; Jostes and others 1994). Dose-rate and dose-fractionation effects are observed for chromosomal aberrations, including translocations (Lloyd and others 1975; Schmid and others 1976) and for mutations (Asquith 1977; Thacker and Stretch 1983; Evans and others 1990; Thacker 1992; Elkind and others 1994). Correlations have been made between chromosomal aberrations and mutations and the cell killing that follows radiation exposure; they suggest that some types of radiation-reduced damage are common to the different biologic end points (Dewey and others 1970, 1971; Thacker and Cox 1975; Cox and others 1977; Bryant 1985).

Recent technologic developments and advances in molecular biology have led to new approaches to the investigation of the mechanisms underlying the induction of both aberrations and mutations. Those developments include the use of prematurely-condensed-chromosome techniques (Hittelman and Rao 1974; Pantelias and Maille 1985; Hittelman and others 1994) to investigate the early response to radiation damage in interphase chromosomes and, more important, the use of fluorescent in sire hybridization (FISH) or "chromosome painting," centromere and telomere staining (Pinkel and others 1986; Gray and others 1991; Natarajan and others 1992; Bauchinger and others 1993; Straume and Lucas 1993; Savage and Simpson 1994; Savage and Tucker 1996), and PCR methods and DNA sequencing (Meuth and Bhattacharyya 1994; Okinaka and others 1994; Singleton and others 1994) to identify the different types of chromosomal aberrations and the DNA-sequence changes in mutations formed after radiation exposure.

Inasmuch as specific types of chromosomal changes are associated with specific cancers and mutations in oncogenes and tumor-suppressor genes are involved in cancer development, it is important to understand the mechanisms underlying the radiation induction of aberrations and mutations. The dependence of the response of these cellular end points on dose, dose rate, and radiation quality needs to be defined because of the relevance of this knowledge to the assessment of low-dose radiation risk.

Research on radiation-induced chromosomal aberrations has been most fertile and has provided the data leading to the development of theoretical approaches to describe the action of radiation at the cellular level (Lea 1946; Revell 1955). It continues to do so, probably because the aberrations visualized with a microscope represent an early indication of the radiation damage. The increasingly refined methods of staining chromosomes, either by using prematurely condensed chromosomes (PCCs) or by chromosome painting, and the use of specific types of radiation, such as carbon K-characteristic x-rays, continue to provide newer insights into the mechanisms of formation of chromosomal aberrations.

There is no consensus on those mechanisms even though their understanding is of utmost relevance to the assessment of radiation risk at low doses. However, there is a relatively good consensus on the shape of the dose-effect relationship for aberration induction suggesting that it is curvilinear in general and linear at very low doses. Attempts by groups of collaborating cytology laboratories to measure the dose-effect relationship for dicentrics at low doses have shown that linearity can be demonstrated down to 20 mGy but that at lower doses statistical variations mask any effect of radiation; measurements at doses below 20 mGy produced yields of dicentrics that were less than background but not significantly so (Pohl-Rohling and others 1983; Lloyd and others 1988). Those experiments did not reveal any evidence of a supersensitive response at the low doses; the researchers concluded that in view of the very large number of cells scored, it would be "very unlikely that the true response at doses less than 20 mGy will ever be measured directly with these techniques" (Lloyd and others 1992). The dose-effect relationship at doses lower than 20 mGy will have to be inferred from an understanding of the mechanisms of aberration formation.

More recent work with fluorescence in situ hybridization (FISH), or chromosome painting, to study the dose-effect relationship for the induction of translocations shows the same curvilinearity as found for dicentrics and a linear response at low doses. Some workers found that the proportion of dicentrics to translocations induced is 1:1 (Nakano and others 1993; Straume and Lucas 1993; Finnon and others 1995; Lucas and others 1995), while others found that there are relatively more translocations, with the proportion varying from 1:2 to 1:1.5 (Natarajan and others 1992; Schmid and others 1992; Bauchinger and others 1993; Tucker and others 1993), although it is expected that the probabilities of formation of symmetrical exchanges (such as reciprocal translocations) and asymmetrical exchanges (such as dicentrics) would be the same. More detailed measurements have revealed, moreover, that some chromosomes seem to be over represented in aberration formtion on the basis of the DNA content of the different chromosomes (Knehr and others 1994, 1996; Slijepcevic and Natarajan, 1994a,b; Finnon and others 1995). Chromosome 4 in humans seems to be particularly over represented. This work suggests that either the induction of chromosomal damage by radiation is nonrandom, or the rejoining of the breaks is nonrandom, but the implications for the mechanism of aberration formation are unclear.

Investigations into the mechanisms of aberration formation have become more sophisticated with the use of the newer PCC and chromosome-painting detection methods and the ultrasoft x-rays. The PCC technique permits the visualisation of interphase chromosomes and the scoring of fragments some 30 min after irradiation. The dose-effect relationship for the fragments is usually found to be linear with dose although the numbers of fragments found are considerably lower than the numbers of DSBs from which they are assumed to be derived. However, the yield of fragments measured in PCC studies has been found to depend on the mitotic cells that are used to induce chromosomal condensation in the irradiated cells (Cheng and others 1993), so the comparison of PCC experiments is compromised, and some care is needed in their interpretation.

Experiments with carbon K-characteristic ultrasoft x-rays, which create very short electron tracks (less than 7 nm), demonstrated an efficient induction of exchange aberrations with a strong linear component in the dose-effect relationship (Virsik and others 1980; Thacker and others 1986). Those results have been interpreted as implying "either that the participating DNA helices must be lying extremely close together at the time of radiation damage, so that one track can effectively damage both helices, or that only one radiation-damaged chromosome is needed to promote an exchange event" (Thacker and others 1986). The results appear, at first sight, to contradict the breakage-reunion concept of aberration formation, although explanations of the ultrasoft x-ray results have been sought in the application of "proximity" concepts to the breakage-reunion theory (Sachs and others 1997) and in the kinetics of formation of aberrations (Brenner 1990; Greinert and others 1995, 1996). However, a chromosome painting study of the induction of complex chromosomal-exchange aberrations, with more than two breaks, by ultrasoft aluminium K x-rays that have a track length of 70 nm has concluded that "for the classical breakage-and-rejoining theory to hold, very large interaction distances are needed . . . unless many sites pre-exist where several different chromosomes come very close together" (Griffin and others 1996). The authors of the study suggest as an alternative that damaged DNA interacts with undamaged DNA to produce an exchange aberration. In a study of the nature of chromatid breaks, using bromodeoxyuridine (BrdU)-differentiated sister chromatids to estimate the proportion of breaks associated with a "color jump" and thus arising from incomplete intra-arm intrachanges rather than from simple breaks, Harvey and Savage (1997) found that the proportion of the color-jump breaks is substantial and almost constant irrespective of radiation dose, radiation quality, BrdU concentration, and cell origin and is similar to the proportion after restriction-enzyme cutting. They concluded that the results are ''very difficult to reconcile with the expectations of breakage-and-reunion theory," although they are in line with both the exchange hypothesis and the idea that a single damaged chromosome can interact with an undamaged chromosome to yield an exchange.

The concept of the interaction of a damaged and an undamaged chromosome in exchange formation was not supported by experiments in which irradiated cells were fused with unirradiated cells to determine whether exchange aberrations were formed between the irradiated and unirradiated chromosomes (Cornforth 1990). The low frequency of intergenomic exchanges found suggested that the "majority of radiation-induced exchanges do require damage to both chromosomes." That result casts doubt on the concept of interaction between a damaged and an undamaged chromosome in the formation of aberrations. However, one of the attractive features of the concept is that it is based on a proposal for the recombination repair of DNA DSBs (Resnick 1976), which makes use of known enzymatic processes such as exonuclease degradation, endonuclease nicking, topoisomerase unwinding, and polymerase and ligase sealing—as well as suggesting a role for homology in DSB repair, of which there is increasing evidence in yeast (Resnick and others 1996). The concept also provides a potential link between DNA-repair studies and chromosomal-aberration formation.

There is no clear consensus on the mechanisms Evolved in the formation of chromosomal aberrations. The "classical" theory has been modified by "proximity" factors to take account of the results of the ultrasoft x-ray experiments but remains in doubt as a consequence of the "color-jump" experiments on chromatid breaks, which clearly favor the "exchange'' hypothesis or the "recombination" hypothesis. Whichever hypothesis is finally shown to be correct, DNA DSBs and complex damage are currently the implied relevant radiation-induced lesions, and the dose-effect relationship at low doses is assumed by many to be linear.

The type of mutation most often associated with radiation exposure is a large chromosomal deletion that can lead to the loss of the target gene and loss of additional DNA extending on both sides of the gene. The mount of DNA lost in a mutational deletion present in a viable mutant depends on whether the adjacent genes are essential for cell viability; and studies of deletions at different target genes in different cell systems reveal considerable variation in the mount of material lost (Thacker and others 1979; Evans and others 1986; Waldren and others 1986; Thacker 1990). In some cases, the amount of material lost is so large that it can be detected cytologically (Simpson and others 1993, 1994). The target genes most often used in mutation studies are present in the cell in only one functional copy, that is, monosomic, such as the HPRT gene on the X chromosome in male cells whereas most genes are present in the cell in two copies—that is disomic. In the case of disomy, a recessive mutation in one (allele) of the target genes would not be revealed in the phenotype, because of the presence, and activity, of the other copy (allele). Therefore, experimentally, one copy of the target gene normally carries a small inactivating point mutation, and the radiation effect is studied in the other copy. In such studies, it has often been found that the frequency of induced mutants is higher than in the monosomic case (Evans and others 1986; Moore and others 1986; Yandell and others 1986; Bradley and others 1988), most probably because the allele with the point mutation is in a maintained genetic region, so that large deletions in the other target copy can be tolerated. Here, again, the mutant frequency depends on whether essential and active genes are in the neighborhood of the target gene. Differences have been found in the mutant frequencies that result when the target gene is switched from one TK allele to the other in lymphoblastoid cells (Amundson and Liber 1991, 1992). Thus, depending on the local genomic situation of the target gene, wide variations in the mutant frequency induced by radiation in different target genes can be expected hacker 1996).

The loss of large mounts of DNA in radiation-reduced mutations makes it difficult to sequence them, but in the few that have been sequenced it has been found that short, direct or inverted repeat sequences are associated with the break points (Miles and others 1990; Nicklas and others 1991; Morris and Thacker 1993; Morris and others 1993). A comparison of the spectra of mutations induced by sparsely and densely ionizing radiation has revealed conflicting data; some results indicate differences between the spectra of different radiation types, and other results indicate very little difference (Thacker 1986;

Gibbs and others 1987; Kronenberg and Little 1989; Lutze and others 1990; Whaley and Little 1990; Aghamohammadi and others 1992; Lutze and others 1992, 1994; Jostes and others 1994; Bao and others 1995; Jin and others 1995; Kronenberg and others 1995; Chaudhry and others 1996). There is some indication that after high-LET radiation, more-complex mutational rearrangements are involved, in addition to the short repeat sequences (Meuth 1990; Simpson and others 1993; Thacker 1996).

The findings of short direct-repeat DNA sequences at sites of large-deletion rejoining, as well as the more complex rearrangements, suggest that a form of illegitimate recombination initiated by a break in DNA is involved in the mutational process. That idea is supported by the results of experiments that reconstructed the process of illegitimate recombination in cell-free conditions by using a DNA substrate with a site-specific DSB and showed that misrejoining is associated with short direct—repeat sequences on either side of the break (North and others 1990; Ganesh and others 1993; Thacker 1994). Research with heterozygotes, in addition to indicating tolerance of large deletions, has also indicated the possibility of mitotic recombination or nondisjunction with a suggestion that recombination is more common than deletion in spontaneous mutants (Fujimori and others 1992; Li and others 1992). There is also an interesting result of a comparison of two cell lines derived from the same original cell but differing in p53 status, DSB rejoining, and recombination ability. The cell line resistant to radiation-reduced killing had a higher radiation-induced mutant frequency, which suggests that it has a higher rate of recombination and can then survive with a concomitant higher rate of mutation (Amundson and others 1993; Xia and others 1994). Delayed apoptosis might well be the reason for this cell's resistance to cell killing (Xia and others 1995; Zhen and others 1995).

In addition to the large deletions induced by radiation, a study of the HPRT and APRT genes has revealed that all types of small mutations occur in response to radiation—such as base-pair substitutions, frameshifts, and small deletions—and that they occur at sites distributed throughout the target genes (Grosovsky and others 1988; Miles and Meuth 1989; Nelson and others 1994). In contrast, spontaneous point mutations tend to occur preferentially at particular sites in the genes. Radiation leads to more transversion and frameshift mutations than are found spontaneously, but large intergenic mutants occur spontaneously at a substantial frequency.

The determination of quantitative dose-effect relationships is more difficult in the ease of mutation than in the ease of chromosomal aberrations but the measurements that have been made indicate a curvilinear relationship for sparsely ionizing radiation, in general, and a linear relationship at low doses (Cox and others 1977). Recent molecular-biology techniques are providing more insight into the mechanisms that lead to mutations after radiation. Although DNA DSBs and complex damage are clearly implicated with recombinational repair processes, the precise mechanisms remain to be elucidated.