Low-Dose Radiation and Genotoxic Chemicals Can Protect Against Stochastic Biological Effects

Abstract

A protective apoptosis-mediated (PAM) process that is turned on in mammalian cells by low-dose photon (X and γ) radiation and appears to also be turned on by the genotoxic chemical ethylene oxide is discussed. Because of the PAM process, exposure to low-dose photon radiation (and possibly also some genotoxic chemicals) can lead to a reduction in the risk of stochastic effects such as problematic mutations, neoplastic transformation (an early step in cancer occurrence), and cancer. These findings indicate a need to revise the current low-dose risk assessment paradigm for which risk of cancer is presumed to increase linearly with dose (without a threshold) after exposure to any amount of a genotoxic agent such as ionizing radiation. These findings support a view seldom mentioned in the past, that cancer risk can actually decrease, rather than increase, after exposure to low doses of photon radiation and possibly some other genotoxic agents. The PAM process (a form of natural protection) may contribute substantially to cancer prevention in humans and other mammals. However, new research is needed to improve our understanding of the process. The new research could unlock novel strategies for optimizing cancer prevention and novel protocols for low-dose therapy for cancer. With low-dose cancer therapy, normal tissue could be spared from severe damage while possibly eliminating the cancer.

INTRODUCTION

The shape of the dose-response curve for stochastic effects (mutations, neoplastic transformation, and cancer) of exposure to low doses of ionizing radiation or genotoxic chemicals has been the topic of continuous debate (Pollycove, 1995; Rossi and Zaider, 1997; Calabrese and Baldwin, 1999, 2003a, 2003b; Joiner et al., 1999; Pollycove and Feinendegen, 1999, 2001; Feinendegen and Pollycove, 2001; Schöllnberger et al., 2001a, 2001b, 2001c, 2003). The key discussion relates to whether the linear nonthreshold (LNT) model for low-dose extrapolation of cancer risk is valid (Rowland, 1995; NCRP, 2001). The LNT model is widely used by regulatory agencies and in radiation and chemical protection.

With the LNT hypothesis, any amount of carcinogen exposure increases one’s risk of cancer. Cancer risk increases linearly without a threshold. Based on this hypothesis, tens of thousands of cancer deaths in the United States have been calculated to arise from fallout from nuclear weapons testing (CDC/NCI, 2001) and from nuclear accidents such as that which occurred at Chernobyl.

Now there is growing evidence from epidemiological, experimental, and mathematical modeling studies that does not support use of the LNT model for central estimation of cancer risks at low doses (Hoel et al., 1983; Bond et al., 1987; Feinendegen et al., 1999, 2000; Pollycove and Feinendegen, 1999; Feinendegen and Pollycove, 2001; Schöllnberger et al., 2003; Scott et al., 2003). Instead, the results support the existence of nonlinear dose-response relationships for the frequency of cancer occurrence including hormetic type (Calabrese and Baldwin 2003a, 2003b; Scott et al., 2003).

The U.S. Environmental Protection Agency’s 1996 proposed revision of the carcinogen risk assessment guideline suggests that the most appropriate models for risk extrapolation are those that incorporate mode of action (Wiltse and Dellarco, 1996). The use of mode-of-action information in low-dose risk characterization facilitates reducing uncertainties (Butterworth and Bogdanffy, 1999).

Here we present two biological-based models for low-dose-induced stochastic effects that can lead to hormetic-type, dose-response relationships. Both models incorporate the LNT model as special cases. The first model is called NEOTRANS₂ and relates to low-dose radiation (Scott et al., 2003). The second model is an adaptation of NEOTRANS₂ for application to a prototypic DNA-alkylating agent, ethylene oxide (EO). The adapted model is called NEOTRANS₂-EO (Walker et al., 2003).

Currently, the low-dose, relative-risk (RR) paradigm whereby RR is always greater than or equal to 1 is used in regulating low-dose exposure of humans to radiation and genotoxic chemicals. Using the NEOTRANS₂ and NEOTRANS₂-EO models, we show evidence that this paradigm needs revision to include RR <1 (due to low-dose-induced protective effects).

Low-Dose-Related Stochastic Processes

In developing models for use in low-dose, radiation, and genotoxic chemical risk assessment for humans, one has to consider the related key stochastic processes involved:

• One key stochastic process is associated with the occurrence of genomic instability (Little, 1985, 1999; Little et al., 1990; Marder and Morgan, 1993; Martins et al., 1993; Kadhim et al., 1996, 1998; Kennedy et al., 1996; Morgan et al., 1996; Mothersill and Seymour, 1998b; Wright, 1998).

• A second key stochastic process relates to the occurrence of deleterious bystander effects (Hei et al., 1997; Mothersill and Seymour, 1997, 1998a; Azzam et al., 1998; Prise et al., 1998, 2002; Lyng et al., 2000; Seymour and Mothersill, 2000; Goldberg and Lehnert, 2002; Little et al., 2002; Nagasawa et al., 2002).

• A third key stochastic process relates to the occurrence of a protective bystander effect, thought to be mediated via apoptosis (Jürgensmeier et al., 1994a, 1994b, 1994c; Bauer, 1995, 1996, 2000; Langer et al., 1996; Hipp and Bauer, 1997; Jürgensmeier and Bauer, 1997; Engelmann and Bauer, 2000; Herdener et al., 2000; Barcellos-Hoff, 2001; Barcellos-Hoff and Brooks, 2001; Belyakov et al., 2001a, 2001b, 2002a, 2002b, 2003; Schwieger et al., 2001; Scott et al., 2003), via induced DNA repair (Mitchel et al., 2003; Redpath et al., 2003), or via both processes.

• A fourth key stochastic process relates to the occurrence of problematic mutations (Kent et al., 1994; Kiefer et al., 1999; Tates et al., 1999; Scott et al., 2003).

• A fifth key stochastic process relates to the occurrence of neoplastic transformation (an early step in cancer occurrence) (Little, 1985; Azzam et al., 1994, 1996; Scott, 1997; Redpath and Antoniono, 1998; Mendonca et al., 1999; Redpath et al., 2001; Schöllnberger et al., 2001b; Scott et al., 2003). Complete transformation can occur when oncogenes are activated and tumor suppressor genes are inactivated (Weinberg, 1989; Bauer, 2000; Hanahan and Weinberg, 2000).

• The overall stochastic process of usual interest is cancer and involves the general stages of initiation, promotion, and progression and relates to the other key stochastic processes indicated (Armitage and Doll, 1954; Barrett and Wiseman, 1987; Portier, 1987; Thorslund et al., 1987; Portier et al., 1990; Tan, 1991; Moolgavkar et al., 1993; Portier and Sherman, 1994; Luebeck et al., 1996; Hoel and Li, 1998; Trott and Roseman, 2000; Mebust et al., 2002).

In modeling these key stochastic processes, it is very important to account for repair and misrepair of DNA damage (Friedberg et al., 1995; Stewart, 1999; Thompson and Schild, 1999, 2001; Hanawalt, 2001; Leskov et al., 2001a, 2001b; Jeggo, 2002; Plotkin and Nowak, 2002). Induced adaptation could also be important in vivo (Mitchel, 1995; Mitchel et al., 1997; Stecca and Gerber, 1998).

Deleterious Bystander Effects

In modeling cancer induction by low-dose radiation and genotoxic chemicals, one has to account for both deleterious and protective bystander effects (Scott et al., 2003). Past focus has mainly been on the deleterious bystander effects.

Deleterious bystander effects whereby unirradiated cells are damaged have been examined in two general types of cellular systems. In the first, monolayer cultures have been exposed to very low α-particle fluence rates (particles per unit area per unit time), either from an external source (Azzam et al., 1998; Little et al., 2002; Nagasawa et al., 2002) or a focused microbeam (Hei et al., 1997; Prise et al., 1998). The second technique involves harvesting medium from irradiated cells and incubating it with unirradiated cells (Mothersill and Seymour, 1997; Lyng et al., 2000). Both techniques have demonstrated that cells not being irradiated can still be damaged. Furthermore, the bystander effect does not arise from simply irradiating media. Cell damage and intercellular signaling are essential (Mothersill and Seymour, 1997, 1998a).

Protective Bystander Effects

Evidence is now strong that cell death via apoptosis at low radiation doses can occur via a bystander mechanism (Mothersill and Seymour, 1998a, 1998b; Lyng et al., 2000; Belyakov et al., 2001a, 2001b, 2002a, 2002b, 2003; Prise et al., 2002). Barcellos-Hoff and Brooks (2001) point out that bystander effects in vivo after low doses of radiation occur via extracellular signaling pathways that modulate both cellular repair and death programs. The authors also indicate that transforming growth factor β(TGF-β) is an extracellular sensor of damage. They further indicate that extracellular signaling relevant to carcinogenesis in normal tissue can eliminate abnormal cells or suppress neoplastic behavior.

Yang and colleagues (Yang et al., 2000a, 2000b; Davis et al., 2001) at Case Western University have reported clusterin (CLU, a.k.a. TRPM-2, SGP-2, or radiation-induced protein-8 [XIP8]) to be implicated in selective removal of problematic cells via apoptosis. Their key finding was that enhanced expression and accumulation of nuclear CLU/XIP8-Ku70/Ku80 complexes appear to be an important cell death signal after irradiation. Furthermore, their data suggest that CLU/XIP8 may play an important role in monitoring cells with genomic instability and/or infidelity (e.g., created through translesion DNA synthesis) by facilitating removal of genetically unstable cells as well as severely damaged cells.

A number of well-designed studies that relate to a protective process that acts against existing neoplastically transformed fibroblast cells have revealed the following (Jürgensmeier et al., 1994a, 1994b; Bauer, 1995, 1996, 2000; Azzam et al., 1996; Langer et al., 1996; Dormann and Bauer, 1998; Eckert and Bauer, 1998; Engleman and Bauer, 2000; Herdener et al., 2000; Redpath et al., 2001; Schwieger et al., 2001):

• Low-dose, low-linear-energy-transfer (LET) radiation can turn on the protective process, which can lead to a reduction in the frequency of neoplastic transformation below the spontaneous level (Azzam et al., 1996; Redpath et al., 2001).

• The protective process is rather complex but is largely understood (Bauer, 2000).

• Reactive oxygen and nitrogen species are involved in the protective process (Bauer, 1996, 2000; Engelmann and Bauer, 2000).

• Transformed cells participate in the protective process through signaling (via TGF-β) to nontransformed cells (Engelmann and Bauer, 2000). The release of TGF-β by transformed cells causes normal cells to release a short-lived factor that induces apoptosis in the bystander transformed cells (Jürgensmeier et al., 1994a, 1994b; Bauer, 1996, 2000; Langer et al., 1996; Schwieger et al., 2001). However, individual cell clones of normal cells require different threshold levels of TGF-β to induce apoptosis (Langer et al., 1996). Fibroblast growth factor (FGF) can also trigger the selective apoptosis process (Dormann and Bauer, 1998; Eckert and Bauer, 1998).

• All cells within a population of normal cells can exert induction of apoptosis among transformed cells after being signaled via TGF-β (Langer et al., 1996) and the process does not depend on cell-to-cell contact (Jürgensmeier et al., 1994c; Langer et al., 1996)

• Cells transformed by a variety of carcinogens (including chemical carcinogens, UV light plus TGF-β, oncogenic retroviruses, herpes simplex virus, or viral oncogenes such as ras and src) were found to be equally sensitive to the induction of apoptosis (Jürgensmeier et al., 1994b; Bauer, 1996).

• There is a causal relationship between oncogene expression and sensitivity to intercellular induction of apoptosis of transformed cells in vitro (Herdener et al., 2000; Schwieger et al., 2001). This sensitization seems to be based on a TGF-β–induced decrease in the concentration of endogenous apoptosis inhibitors (i.e., survival factors) (Schwieger et al., 2001). Oncogenic transformation causes a marked decrease of endogenous survival factors (Engelmann and Bauer, 2000). For cancer occurrence, it appears that a necessary genomic change involves the loss of the indicated sensitivity to intercellular induction of apoptosis (Jürgensmeier and Bauer, 1997).

• The intercellular induction of apoptosis in transformed cells does not depend on p53 because p53-negative (i.e., the gene is inactive) transformed cells are sensitive to intercellular induction of apoptosis and p53-negative nontransformed cells are able to induce apoptosis (Hipp and Bauer, 1997; Engelmann and Bauer, 2000).

Figure 1 summarizes the protective process as described by Bauer (2000) and Engelmann and Bauer (2000) for the hypochlorous acid/hydroxyl radical signaling pathway, which is one of two currently identified signaling pathways associated with the protective process. The other (NO/ peroxynitrite) pathway is less complicated (Bauer, 2000) and is not presented here. Here we refer to the overall protective process as the protective apoptosis-mediated (PAM) process. Based on Bauer (2000), during the PAM process, transformed cells release TGF-β and then generate superoxide anions (O₂·^–). The released TGF-β induces nontransformed effector cells to then release peroxidase (P). The O₂·^– anions released from transformed cells spontaneously dismutate and form hydrogen peroxide (H₂O₂), which is used by the peroxidase, along with chloride ions to form hypochlorous acid (HOCl). Hypochlorous acid interacts with superoxide anions to form the highly reactive hydroxyl radical (·OH), which acts as the ultimate apoptosis inducer. Because the range of the superoxide radicals is very short, the interaction with HOCl is limited spatially to the vicinity of the transformed cells, thus allowing for selective apoptosis induction in transformed cells. The superoxide anions derived from the transformed cells are both the basis for HOCl production and its locally restricted activation, which leads to selective induction of apoptosis in transformed cells.

FIGURE 1.

Schematic representation of intercellular signaling associated with the PAM process as specified by researchers at the University of Freiburg (Bauer, 2000; Engelmann and Bauer, 2000). The signaling indicated here is for the HOCl/hydroxyl radical signaling pathway, which is one of two signaling pathways. The other NO/peroxynitrite pathway is less complicated (Bauer, 2000) and is not presented. The figure is explained in the text.

We speculate that nuclear CLU/XIP8-Ku70/Ku80 complexes may play a key role in the initial signaling associated with turning on the PAM process. This is based on the signaling research we have already discussed that was carried out by Yang et al. (2000a, 2000b) and by Davis et al. (2001).

We later show data indicating that the PAM process also selectively eliminates problematic mutants from lung alveolar type II cells in vitro and that the process is turned on by low-dose X-rays. Because low-dose X-rays are far more likely to hit some of the very large number of nonmutant and nontransformed cells in the target population rather than the very small number of mutants and transformed cells present, we think it is likely that the initial signaling event associated with induction of the PAM process by low-dose radiation relates to the nonmutant and nontransformed cells that are hit by radiation sending out signals to the cellular community. The TGF-β signaling from the relatively small number of problematic cells (unlikely to be directly hit by low-dose radiation) would then arise as a secondary event associated with the selective elimination of problematic cells (e.g., mutants and transformed cells) in the manner indicated in Figure 1.

The indicated research findings support the growing view by scientists (e.g., Barcellos-Hoff and Brooks, 2001; Scott et al., 2003; Rothkamm and Löbrich, 2003) that the PAM process associated with low-LET irradiation serves to rid tissue of problematic cells (e.g., mutants, neoplastically transformed cells) and therefore protects from cancer induction. This form of induced protection has been called group (G) adaptation since the cell community acts as a group in eliminating the problematic cells (Scott et al., 2003). In contrast, induced DNA repair in an individual cell has been called individual (I) adaptation.

For split-dose adaptation experiments, where a small adapting radiation dose is delivered then followed after a time delay with a larger test radiation dose, it is clear that induced increased DNA repair capacity (I-adaptation) likely plays a major role in the indicated reduction in the frequency of stochastic effects that has been observed when compared to results obtained with the same test dose alone (Wolff, 1996; Wolff et al., 1988). However, this induced repair adaptation only occurs after a significant time delay. In addition, for brief, very low, single doses of low-LET radiation, there is growing evidence that repair of the very small number of cells that are damaged does not occur (Rothkamm and Löbrich, 2003). However, the damaged cells can be eliminated via apoptosis if the cells are allowed to proliferate after the brief irradiation (Rothkamm and Löbrich, 2003).

We therefore consider the low, brief, single-dose-induced protective process described to be mediated mainly via apoptosis. Hanahan and Weinberg (2000) in their classic paper entitled “The Hallmarks of Cancer” point out the following: “Collectively, the data indicate that the cell’s apoptotic program can be triggered by an overexpressed oncogene. Indeed, elimination of cells bearing activated oncogenes by apoptosis may represent the primary means by which such mutant cells are continually culled from the body’s tissues.”

Assuming that the PAM process indeed protects against both transformants and problematic mutants, one could ask whether the process also protects against cells with DNA repair deficiencies. Kent et al. (1994) published experimental evidence that cells with lowered DNA repair fidelity are hypomutable (rather than hypermutable) at the Hprt locus. The researchers studied mutation induction at the Hprt locus in a human bladder tumor cell line (designated MGH-U1), and in its radiosensitive mutant clone, U1-S40b. There were no significant differences in DNA repair between the two cell types other than a lowered repair fidelity (measured by a plasmid reconstitution assay) for the U1-S40b cells. Our view is that the hypomutability may be related to induction of the PAM process.

Indeed, the induced, selective elimination of problematic cells (transformed cells, cells with activated oncogenes, cells with inactivated tumor suppressor genes, cells with mutations in key repair pathways, cells with mutations in cell cycle regulatory genes, etc.) via the PAM process may be a generalized phenomenon. However, more research related to the PAM process is necessary.

Being able to control and optimize the PAM process would have potentially enormous cancer prevention implications and could lead to novel methods for cancer therapy (e.g., low-dose radiation/chemical therapy). However, some cancer cells may have elevated endogenous survival factors rendering them less likely to undergo apoptosis (Jürgensmeier and Bauer, 1997). If so, therapeutic strategies could possibly be devised that would reduce the level of the endogenous survival factors thereby permitting the PAM process to operate against the cancerous cells.

Nonthreatening mutations may not efficiently sensitize cells for selective elimination via the PAM process. If so, then the efficiency of removal of problematic cells via the PAM process may depend on characteristics of the genomic lesions and possibly also on the external stimulus (and its temporal and spatial distribution) that turns on the process.

The mutation/transformation study described in the next section provides data supporting the possible existence of differential sensitivity of mutant cells to the PAM process.

EVIDENCE FOR DIFFERENTIAL PROTECTION AGAINST MUTANT CELLS

A study that provides data supporting the view that low-dose photon radiation turns on the PAM process that differentially discriminates between the various types of mutant cells was conducted by our research group using mouse alveolar type II cells. The study involved in vitro experiments on X-ray-induced mutations in mouse lung cells (T273; a subclone of C10 cells). The T-cell cloning assay for Hprt mutations developed by Albertini et al. (1982) and modified by Driscoll et al. (1995) was adapted for use with a murine, immortalized lung epithelial cell line. In addition to Hprt mutations, two other types of mutations were also identified and investigated.

The second type of mutations are presently called “assumed problematic mutations of type 1 (Aprob1).” The Aprob1 mutant clones have presumptive mutations in the DNA mismatch repair genes or in genes involved in signaling to apoptosis and are considered to pose a carcinogenic threat to the host if present in vivo. In the case where a mutation occurs in a gene that signals to apoptosis, this does not mean that all signaling pathways to apoptosis are inactivated. Some pathways appear to remain operational.

Our presumption related to the target genes is based on the following evidence:

• The founder T273 cells are highly sensitive to the cytotoxic effects of 6-thioguanine (6-TG). This sensitivity to 6-TG was interpreted to confirm the presence of at least some functional DNA mismatch repair and apoptosis genes (and associated signaling pathways) in the founder cells. T273 cells have a high spontaneous mutation frequency suggesting problems with their DNA repair machinery.

• Radiation seems to induce essentially full removal of the Aprob1 mutants (data shown later) indicating that their endogenous survival factors may be greatly reduced as is the case with transformed cells (Engelmann and Bauer, 2000), suggesting a possible mutation in a key repair gene or gene associated with a key apoptosis signaling pathway.

• The following essential proteins are involved in DNA repair via nonhomologous end-joining of double-strand breaks: DNA-PK, KU80/KU70, DNA-PKcs, DNA ligase IV, and XRCC4. Induced mutations in any of these genes encoding for these proteins lead to heightened radiosensitivity (Daboussi et al., 2002).

• Defects in the ataxia telangiectasia mutated gene lead to radiosensitivity due to a loss in DNA repair capacity (Daboussi et al., 2002).

Thus, induced damage to a variety of genes could lead to increased cell sensitivity for undergoing apoptosis. However, additional studies are needed to better delineate the genes that are mutated in the Aprob1 mutants.

The third type of mutants considered are focus forming cells (transformed cells) which are presumed to be a threat (i.e., are problematic) when they occur in vivo.

The X-ray exposure conducted involved 150-kV X-rays from a Phillips X-ray therapy unit delivered at a dose rate of approximately 100 mGy/min. The T273 cells were exposed to 0, 100, or 1000 mGy of X-rays during log-phase growth.

After exposure, the cells were grown for 2 weeks to allow phenotypic expression of treatment-related changes. The cells were then plated at a specific density and selected with 6-TG to assay for the presence of 6-TG-resistant clones, using a modification of the T-cell Hprt mutation assay. In this experiment, 6-TG resistant clones were subcloned and further characterized by evaluating their ability to survive and proliferate in the presence of 6-TG or hypoxanthine aminopterine thymidine negative (HAT^–) medium. Mutant clones that survived and grew in both 6-TG and HAT^– medium were classified as Aprob1 mutants.

Some mutant clones formed foci, characterized by dense collections of cells that piled up upon each other and displayed a radial arrangement of cells around the periphery of the aggregate. These clones were also resistant to both 6-TG and HAT^– medium. They were classified as focus-forming mutants and are presumed to be transformed (Hurlin et al., 1987). Additional studies are being conducted to characterize some of the genetic alterations present in the indicated mutant clones.

One million cells per treatment group were evaluated. Cloning efficiency was determined using the method described by Wang et al. (1999) and Perera et al. (2002). Mutation frequencies were calculated on the basis of surviving cells (Wang et al., 1999; Perera et al., 2002). The mutation frequency and associated RR for mutation occurrence obtained for the different dose groups are presented in Table 1. The RR is just the risk (frequency of a deleterious effect) after exposure divided by the risk when there is no exposure. The indicated errors in the mutation frequencies are standard deviations. The indicated p-values relate not to the RR values but rather to the corresponding mutation frequencies used to calculate RR. Significant differences (reductions) in mutation frequencies between controls and irradiated cells were evaluated assuming the variance of the difference in the respective mutation frequencies was equal to the sum of their respective variances. A standard normal curve was used to assign p-values. Where no mutants were observed, an upper bound mutation frequency of one per one million cells was assigned in order to assign a p-value associated with no observed mutants (Wang et al., 1999; Perera et al., 2002).

TABLE 1.

Mutation Frequency (MF) per Surviving Cell and Associated Relative Risk (RR) for Mutation Induction among Mouse Lung Cells (T273) Exposed In Vitro to X-Rays

X-ray dose (mGy)	Hprt mutants	Aprob1 mutants	Problematic focus-forming cells (transformed)
0*	MF = 18.9 (± 8.5) × 10–6	MF = 37.9 (±13) × 10–6	MF = 9.6 (±2.6) × 10–6
	RR = 1.0	RR = 1.0	RR = 1.0
100†	MF = 11 (±7) × 10–6	No mutants detected	No mutants detected
	RR = 0.58 ± 0.45	RR≈0	RR≈0
		p < 0.005§	p < 0.002§
1000‡	MF = 9.6 (±2.6) × 10–6	No mutants detected	MF same as controls
	RR = 0.51 ± 0.27	RR≈0	RR = 1.0
		p < 0.005§

*Mean plating efficiency for control groups was 42.4 ± 7.8%.

^†Mean plating efficiency for 100 mGy groups was 36.5 ± 7.5%.

^‡Mean plating efficiency for 1000 mGy groups was 41.5 ± 2.4%.

^§The indicated p-values relate only to mean MF data compared to controls and represent a significant reduction in the MF.

Note the very dramatic protection that appears to be turned on by the 100-mGy photon radiation dose against the spontaneous Aprob1 mutants and against the problematic spontaneous focus-forming mutants. The protection against the Aprob1 mutants was also evident at 1000 mGy (a rather large radiation dose). However, at 1000 mGy, protection against focus-forming cells was not evident. These results suggest that the endogenous survival factors (e.g., Bcl-2) for these cells were greatly reduced at 100 mGy, compared to normal cells. For the Hprt mutations, only modest protection was suggested but was not statistically significant at either dose (p >0.1 in both cases). Although 100% protection was suggested against the spontaneous Aprob1 and focus-forming mutants, the sensitivity of the assay was not sufficient for detecting very low mutation frequencies. However, we can state with confidence that the protection exceeded 99.9% (i.e., most if not all of the mutants were removed) since no mutants were detected among the large number of cells scored.

The indicated results for the three mutant types are consistent with the growing view that an elaborate system of detecting and eliminating problematic cells can be turned on by low-dose radiation. Furthermore, the efficiency for eliminating the problematic cells is expected to be inversely related to the level of endogenous survival factors present (Engelmann and Bauer, 2000). The level of these factors in turn may depend on the nature of the genomic damage present.

Because elimination of transformed cells via the PAM process appears to require the release of TGF-β (or FGF) by transformed cells (Bauer, 2000), it follows that the efficiency of the process against transformations may depend on the spontaneous transformation frequency. The higher the spontaneous transformation frequency, the more transformed cells are available for releasing TGF-β. However, Hprt mutants usually are not transformed cells and so their high spontaneous frequency would not necessarily be associated with an increased efficiency for the PAM process.

We show later that our speculation about a correlation between the spontaneous transformation frequency and the efficiency of the PAM process is consistent with experimental data of Azzam et al. (1996) and Redpath et al. (2001), where greater protection was afforded against neoplastic transformation, for the fibroblast cell population with the higher spontaneous transformation frequency. However, with such limited data (two data sets), no firm conclusions can be drawn.

It should now be clear that in modeling low-dose radiation-induced stochastic effects, such as problematic nonlethal mutations, neoplastic transformation, and cancer, one has to account for both deleterious and protective effects. This was done in developing the NEOTRANS₂ model for low-dose radiation-induced stochastic effects (Scott et al., 2003). The current form of the model is briefly described in the section that follows.

THE NEOTRANS₂ MODEL

Low-Dose Radiation

In our earlier research, we introduced models that relate neoplastic transformation potential to genomic instability status of cells. The models were given the general name “genomic instability state” (GIST) models (Scott, 1997). The expression “genomic instability state” refers to any spontaneous or toxicant-induced instability in the genome, including any initial transient instability, as well as any persistent instability that can be passed to cell progeny.

Our current GIST model for characterizing stochastic effects of low absorbed radiation doses is called NEOTRANS₂ (see Figure 2). We consider absorbed doses up to a few hundred mGy as being low. In addition to a stable genome for resistant cells (not included in Figure 2), NEOTRANS₂ involves three types of genomic instability considered to be important among hypersensitive cells that respond to low radiation doses: (1) normal-minor instability (NMI), which is associated with normal cell function and normal genome status; (2) transient-problematic instability (TPI), which is associated with genomic damage that may sometimes be fully repaired but can be misrepaired; and (3) persistent-problematic instability (PPI), which arises from misrepair that yields nonlethal mutations. Thus, PPI can be passed to progeny, increasing their potential for neoplastic transformation. Neoplastic transformation among progenitors of the PPI cells arise through a stochastic emergence process.

FIGURE 2.

NEOTRANS₂ model transitions for cells that respond to low-dose radiation. Genomic instability states NMI, TPI, and PPI and associated model parameters are explained in the text (Scott et al., 2003).

With NEOTRANS₂, mainly hypersensitive cells respond after very low radiation doses (e.g., 0–100 mGy of γ-rays). Radiation-associated transitions among hypersensitive cells in the NEOTRANS₂ are summarized in Figure 2.

With the NEOTRANS₂ model, a small fraction (T₀ ≪ 1) of the cell population is presumed to have already undergone neoplastic transformation over its life history. Thus, T₀ depends on the life history of the target cell population. For in vivo cell populations, T₀ will depend on the age of the individual. Only hypersensitive cells in the high-vulnerability state PPI (viable mutants) can produce neoplastically transformed progeny.

Potential target genes for the hypersensitive cells include tumor suppressor genes, oncogenes, repair genes, apoptosis genes, and cell-cycle regulator genes (Scott et al., 2003). With NEOTRANS₂, apoptosis is considered the main mode of cell death at very low doses (e.g., 0–100 mGy of photon radiation). Nonlethal mutations are assumed to arise via misrepair. Lethal mutations are assigned to the apoptosis pathway.

The NEOTRANS₂ model presented in Figure 2 applies only to low radiation doses and to the hypersensitive subfraction of cells at risk, f₁, along with the already-existing problematic cells (mutants and transformants). The model parameter α₁, when multiplied by the dose rate c, accounts for low-dose-induced genomic damage among the hypersensitive cells in the population. The parameter α₁ is composed of two parts: (1) one part relates to direct damage to DNA; (2) the other part relates to indirect damage to DNA and includes deleterious bystander effects.

The parameter μ₁ governs the rate of commitment of damaged hyper-sensitive cells to the error-free repair pathway. The corresponding parameter for the misrepair pathway is η₁. Misrepair leads to a variety of viable mutations (PPI cells). The parameter φ₁ governs the rate of commitment of newly damaged cells to the apoptotic pathway.

Typical units for α₁ are mGy^–1. Typical units for μ₁,η₁ and φ₁ are min ^–1 The parameter f₁ is dimensionless. These parameters are stochastic (i.e., have distributions) but currently are not time dependent. The distributions account for variability among cells.

Analytical solutions for the NEOTRANS₂ model that apply to in vitro data for very low radiation doses, ΔD, were developed elsewhere (Scott et al., 2003). The solutions apply to brief exposure to a single type of radiation but not to combinations of different radiations. The steady-state solution for neoplastic transformation frequency per surviving cell (TFSC) after a small dose ΔD is given by the following:

$$\begin{array}{l}\text{TFSC}(\Delta D)=T_0,\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\text{for}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\Delta D=0\\ \text{TFSC}(\Delta D)=(1-f_0)T_0+[(1-T_0)]\hspace{0.17em}{k}_{\text{T}}\Delta D,\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\text{for}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\Delta D>0\end{array}$$ (1)

where

$$k_{\text{T}}=f_1{\alpha }_1{\eta }_1\Omega /({\mu }_1+{\eta }_1+{\phi }_1)$$ (2)

The subscript T is used to indicate the endpoint neoplastic transformation. For evaluating mutations, a corresponding parameter k_M can be used. The parameter Ω is the probability that a cell with induced PPI will produce neoplastically transformed progeny at some point during the follow-up period of interest. For mutations, Ω can be replaced by Λ, with Λ representing the proportion of the induced mutants that is of the type of interest. The parameter f₀ is the fraction of the spontaneous problematic cells (e.g., spontaneous transformants) removed via the radiation-induced PAM process and has been given the special name protection factor (PROFAC; Scott et al., 2003). The PAM process is assumed to start via a bystander cell-signaling mechanism with the initial signals coming from cells damaged by irradiation. Presently, f₀ is assumed to be nonzero only during implementation of the induced PAM process. For brief exposure, newly emergent transformants are assumed to arise in a stochastic manner after signaling related to the PAM process has ceased. The parameter φ₁ accounts for removal via apoptosis of cells with radiation-induced TPI. The threshold dose (stochastic and called StoThresh, Scott et al., 2003) for exceeding the spontaneous transformation frequency is given by

$$D_{\text{Th}}=f_0{T}_0/[(1-T_0)k_{\text{T}}]$$ (3)

For very low doses of low-LET radiation, the predominant term in Eq. (1) is (1 – f₀)T₀ for ΔD > 0 (Scott et al., 2003). This leads to an initial drop in the dose–response relationship from T₀ down to (1 – f₀)T₀. For low-LET γ-ray doses in the range greater than 0–100 mGy, the dose–response curve appears independent of dose remaining at the value (1 – f₀)T₀ (Scott et al., 2003). For this range, the RR for neoplastic transformation is approximately (1- f₀) for all doses. This occurs because γ-rays are not very effective in producing new transformants (e.g., compared to alpha particles). Large doses of γ-rays are needed to produce lots of transformants.

This is shown in Figures 3A and 3B, where RR is presented for γ-ray-induced neoplastic transformation of C3H 10T1/2 cells based on data of Azzam et al. (1996) and for transformation of HeLa x skin fibroblast human hybrid cells based on data of Redpath et al. (2001). For both data sets, there is significant protection against spontaneous transformations associated with the γ-ray exposure. The indicated lines are based on the NEOTRANS₂ model for which RR is approximately equal to 1 – f₀ for the γ-ray dose range 0–100 mGy. The smooth line (in the center) represents the average of the RR data points in the respective figures (zero-dose group excluded). The dashed lines indicate the associated 95% confidence region for the average RR value. The average values for RR in Figures 3A and 3B are 0.29 ± 0.04 and 0.68 ± 0.04, respectively, and are both significantly less than 1 (p < 0.001), assuming RR to be normally distributed. The corresponding averages (and associated standard deviations) for the protection factor (PROFAC = f₀) were 0.71 ± 0.04 for the C3H 10T1/2 cells in Figure 3A and 0.32 ± 0.04 for the HeLa x skin fibroblast human hybrid cells in Figure 3B.

FIGURE 3.

(A) Relative risk (TFSF/T₀) for γ-ray-induced neoplastic transformation of C3H 10T1/2 cells based on data of Azzam et al. (1996); (B) RR for γ-ray-induced neoplastic transformation of HeLa x skin fibroblast human hybrid cells based on data of Redpath et al. (2001); (C) RR for 4.3-MeV, α-particle-induced neoplastic transformation among C3H 10T1/2 cells based on data of Bettega et al. (1992). The dashed curves just above and below the data in (A) and (B) define the 95% confidence region for the mean of the data points indicated (excluding the zero dose group). Error bars associated with data points are ±1 standard deviation. Errors in dose are presumed not to impact on the results presented. A reference horizontal line at RR = 1 is also included.

The PROFAC values we have derived for low-dose photon-radiation-induced stochastic effects (in vitro) are summarized in Table 2.

TABLE 2.

Protection Factors (PROFACs) Associated with Brief Exposure In Vitro to Low Doses of Photon Radiation

Cell type	Genotoxic agent	Stochastic effect	PROFAC*	Data source
T273	X-rays	Hprt mutants	0.42 ± 0.45	This paper†
T273	X-rays	Apop1 mutants	≈1	This paper†
T273	X-rays	Focus-forming cells	≈1	This paper†
C3H 10T1/2	γ-rays	Neoplastic transformations	0.71 ± 0.04	Azzam et al. (1996)
HeLa x skin fibroblast	γ-rays	Neoplastic transformations	0.32 ± 0.04	Redpath et al. (2001)

*Indicated errors are standard deviations of the data used to obtain the average indicated.

^†Study was carried out in the laboratory of D. M. Walker.

The spontaneous transformation frequency for the C3H 10T1/2 fibrob-last cells used by Azzam et al. (1996) was much higher than for the HeLa x skin fibroblast human hybrid cells used by Redpath et al. (2001). Thus, our earlier speculation that the level of induced protection may correlate with the spontaneous transformation frequency is consistent with these data. The higher the spontaneous transformation frequency, the more transformed cells are available for release of TGF-β (or FGF), which is currently considered to be necessary for implementation of the PAM process. However, we have only looked at two data sets and so no firm conclusions can be made as to whether the level of induced protection is positively correlated with the spontaneous transformation frequency.

Although we have explained the low-dose protection as being due to apoptosis (consistent with the NEOTRANS₂ model), others have suggested that DNA repair may also contribute to protection (Redpath et al., 2003) but in a dose-dependent manner. From Eq. (2), the following can be seen: (1) inhibition of error-free repair (i.e., reducing μ₁) would be expected to lead to an increase in the slope factor k_T and a corresponding, dose-dependent increase in the frequency of newly induced neoplastic transformations; (2) inducing an increase in the misrepair (e.g., increasing η₁ via increasing endogenous survival factors that permit survival of cells with DNA repair errors) would be expected to lead to a dose-dependent increase in the frequency of neoplastic transformations.

The information presented in this paper supports the view that apoptosis likely has the major role in the low-dose-induced protective process discussed (related to a reduction in the spontaneous mutation and neoplastic transformation frequency) that has been shown to be associated with brief exposure to very low doses of photon radiation.

For high-LET alpha irradiation, there appears to be essentially no low-dose induced protection (i.e., f₀ = 0) against spontaneous transformants, or that the range of doses over which the protection occurs is too small to be detected from the available data. This is shown in Figure 3C, where RR for neoplastic transformation among C3H 10T1/2 cells appears to increase in accordance with the LNT model, based on data of Bettega et al. (1992). We speculate that for α-irradiation the deleterious bystander effect predominates over the protective bystander effect. It is reasonable to assume that the intercellular signaling efficiency associated with the PAM process depends on the number of cells hit by radiation. However, because of the short range of α-particles in tissue (only a few cell traversals), low-dose α-radiation would be expected to be much less efficient in triggering widespread cell signaling associated with the PAM process. The number of cells hit by low-dose α-radiation would be expected to be much less than that for uniform exposure to the same average absorbed dose of γ- or X-rays. However, for combined exposure to low-dose γ-rays and α-particles, our prediction is that the PAM process may be demonstrable since the γ-rays would be expected to be efficient in turning on widespread cell signaling.

Adapting NEOTRANS₂ Model for Application to a Chemical

We have now adapted the NEOTRANS₂ model to be applicable to mutation induction in vivo (in T lymphocytes in mice) by inhaled EO. The adapted model is called NEOTRANS₂-EO, and the EO concentration C was used as the independent variable (corresponds to variable c in the NEOTRANS₂ model, when the exposure time is fixed). Additional details and analyses are provided in a separate paper (Walker et al., 2003).

EO is an immediate metabolite of ethylene, a normal body constituent (Walker et al., 1990). EO has caused dose-related increases in the incidence of mononuclear cell leukemias, gliomas, and peritoneal mesotheliomas in F344 rats and lymphomas and tumors of the uterus, lung, Harderian gland, and mammary gland of B6C3F1 mice (USEPA, 1990; IARC, 1994; Thier and Bolt, 2000).

Historically, risk assessment for genotoxic chemical carcinogens has been based on the assumption that any exposure carries a cancer risk no matter how small the dose (Butterworth and Bogdanffy, 1999). Using data for EO-induced mutations in T lymphocytes of B6CF31 mice exposed via inhalation (Walker et al., 2003), we present the following evidence against the validity of the LNT model.

Briefly summarizing, two data sets were used to obtain the results presented in Figures 4A and 4B for mutation (Hprt) induction in T cells of B6C3F1 mice exposed via inhalation for 4 weeks (6 h/day, 5 days/week) to EO (Walker et al., 2003). The ethylene-exposed mice had calculated equivalent doses of EO of 0.7, 4.4, and 8.6 ppm. For mice directly exposed to EO, the exposure concentrations were 50, 100, and 200 ppm. Figure 4A features all of the data. Figure 4B features the low-dose data (0.7, 4.4, and 8.6 ppm doses) along with one high dose (50 ppm). Although the lowest three doses are presented as the EO concentration, it should be remembered that the animals were actually exposed to ethylene and their EO exposure was estimated as described by Walker et al. (2000).

FIGURE 4.

(A) Relative risk (MF/MF₀) for ethylene-oxide (metabolite of ethylene) induced Hprt mutations in T lymphocytes of B6C3F1 mice exposed via inhalation to ethylene with associated low doses of EO (0, 0.7, 4.4, or 8.6 ppm) or high doses of EO (50, 100, or 200 ppm) based on application of the NEOTRANS₂-EO model to data; (B) lower portion of dose–response curve in (A). The upper and lower dashed curves in (A) and (B) are 95% (percentile) and 5% values based on the posterior distribution from the Bayesian analysis. Error bars associated with data points represent ±1 standard deviation. Errors in the EO concentration are not included because they are unknown. Errors in the EO concentration cannot explain a decrease in the mutation frequency below the control value (i.e., departure from the LNT model).

As indicated, the EO exposure concentration (C in ppm; replaces the dose rate in the NEOTRANS₂ model) was used as the independent variable. With the present model, variability in C over the target cell population was not considered because existing data do not permit such an analysis. Exposure times were fixed and the model slope parameter k_M (replaces k_T in the NEOTRANS₂ model) has incorporated the constant exposure time. To ade-quately describe the dose-response data using our model, it was necessary to postulate a threshold EO exposure concentration $C_1^{*}$ for turning on the PAM process. Below $C_1^{*}$, DNA misrepair was assumed not to occur (η₁ = 0) This corresponds to a flat dose-response from 0 to the $C_1^{*}$. Above the threshold, misrepair was presumed to occur (η₁ >0) in competition with error-free repair (μ₁ >0), leading to newly induced mutations. In addition, at $C_1^{*}$, the PAM process was modeled as being turned on.

The following equation applies to the mutation frequency (MF) data:

$$\begin{array}{l}\text{MF}=M_0,\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\text{for}\hspace{0.17em}\hspace{0.17em}C<C_1^{*}\\ \text{MF}=(1-f_0)M_0+(1-M_0)k_M(C-C_1^{*}),\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\text{for}\hspace{0.17em}\hspace{0.17em}C\ge C_1^{*}\end{array}$$ (4)

The parameter M₀ represents the spontaneous mutation frequency. Equation (4) was fitted to the data in Figure 4A using Bayesian methods (Carlin and Lewes, 1996; Siva, 1998) and assuming a Poisson distribution (Schollnberger et al., 2001b; Scott et al., 2003) of mutants with means based on Eq. (4). The Bayesian analysis was implemented with WinBUGS (Spiegelhalter et al., 1999) software. WinBUGS implements Gibbs sampling via Markov chain Monte Carlo (MCMC) (Spiegelhalter et al., 1999) and, for each iteration in the MCMC chain, a set of posterior parameter estimates (for f₀, $C_1^{*}$, k_M, and M₀) is generated. Each set corresponds to a different dose–response curve. For some of the posterior parameter sets, both $C_1^{*}$ = 0 and f₀ = 0, which corresponds to the LNT model. We have therefore counted the proportion of the posterior parameter sets generated in the MCMC analysis that yielded zero values for both f₀ and $C_1^{*}$. To facilitate this counting process, all values of f₀ ≤ 0.001 and all values of $C_1^{*}$ ≤0.001 ppm were treated as being essentially zero in our counting.

The frequency of occurrence of both parameters f₀ and $C_1^{*}$ equal to zero represents the posterior probability for the LNT model. A high posterior probability for the LNT-type, dose–response curves would of course favor the LNT model. A low posterior probability would favor a different curve shape.

With Bayesian analyses, prior distribution must be assigned to model parameters. A uniform prior from 0 to 1 was used for f₀. For M₀ a normal prior with mean 2.0 × 10^–6 and precision (inverse variance) of 1.56 × 10¹² was used, which is based on the measured spontaneous mutation frequency. With WinBUGS calculations, precision replaces variance. For $C_1^{*}$ a uniform prior from 0 to 4 was used. If values of zero for both f₀ and $C_1^{*}$ are unlikely, this would be reflected in their posterior distributions.

Starting values for both f₀ and $C_1^{*}$ were set to 0 (i.e., start with the LNT model). Autocorrelation results after the first 5000 iterations were used to assess the additional iterations needed for convergence of the MCMC chains associated with our Bayesian analysis using WinBUGS software. The results indicated that a total of about 20,000 iterations would be adequate for convergence. We, however, conducted 60,000 iterations. The first 40,000 iterations were discarded (the discarded portion is called burn-in and in some instances can contain excessive noise).

The low-dose portion of the dose–response curve in Figure 4A is presented in Figure 4B. The mutation frequency at 4.4 ppm EO equivalence was significantly different from the spontaneous frequency (p = 0.009, Mann-Whitney U-statistics; Walker et al., 2003) and clearly below the spontaneous frequency, indicating presumed protection against spontaneous mutations. The posterior mean and standard deviation for $C_1^{*}$ were 2.1 ± 1.1 ppm, for f₀ they were 0.43 ±0.25, for k_M they were 6.52 ×10^–8 ± 1.45 × 10^–8 per ppm, and forM₀ they were 2.07 ×10^–6 ± 5.93 × 10^–7. The somewhat large uncertainties associated with both $C_1^{*}$ and f₀ was far less than optimal.

As can be seen from Figure 4B, the dose–response relationship does not at all appear to be of the LNT type. In fact, our counting of the proportion of the posterior parameter data sets generated that corresponded to the LNT model yields a posterior probability of 0.002 for that model. Thus, these results strongly favor a dose response other than the LNT type. The dashed curves in Figures 4A and 4B relate to evaluating risk at a given EO concentration over all of the different posterior dose–response curves (associated with parameter sets) generated in the Bayesian analysis. The lower curve shows the 5% (percentile) and the upper curve shows the 95% values for the indicated distribution.

The NEOTRANS₂ and NEOTRANS₂-EO models explain the low-dose- induced protection against spontaneous problematic mutations, neoplastic transformation, and cancer as being mediated via induced apoptosis. Also, for transformation of fibroblast cells, it has been established that equal protection occurs irrespective of whether the transformed state was induced by chemical carcinogens, viruses, oncogenes, or cytokines (Engelmann and Bauer, 2000). This strongly suggests that the PAM process eliminates problematic cells with damage in a variety of genes. Thus, it follows that the PAM process, if turned on, could operate against transformants as well as other cells with problematic genomic lesions caused by a variety of agents. We therefore state the following:

• Low doses of genotoxic agents such as low-LET radiation could possibly protect against spontaneous problematic mutations, neoplastic transformations, and cancer.

• For combined exposure to low doses of different genetoxic agents, the combined exposure could turn on the PAM process, which could lead to a reduction rather than an increase in the risks for problematic mutations, neoplastic transformation, and cancer.

Other Evidence Against the General Applicability of the LNT Model

Rossi and Zaider (1997) critically reviewed the literature on radiogenic lung cancer and concluded that “at radiation doses generally of concern in radiation protection (<2 Gy), protracted exposure to low-LET radiation (x- or γ-rays) does not appear to cause lung cancer. There is in fact, indication of a reduction of the natural incidence.” With such hormetic-type dose–response relationships, there is a low-dose region of beneficial protection against natural cancer incidence before the effects of higher irradiation lead to an excess risk compared to that for an unexposed person. The indicated data are therefore consistent with PROFAC >0 for cancer induction in humans by chronic X-ray or γ-irradiation.

Results of case-control studies of lung cancer among Mayak workers are consistent with possible large thresholds for excess cancers for combined α-and γ-irradiation (Tokarskaya et al., 1995, 1997, 2002). In contrast, in the studies by Kreisheimer et al. (2000, 2003) of lung cancer occurrences among Mayak workers, the LNT model was judged to adequately represent the data when internal controls were used. However, an earlier study by Khokhryakov et al. (1996) using Russian national statistics to obtain baseline cancer rates yielded data consistent with a hormetic-type dose response for lung cancer induction by combined α- and γ-irradiation.

The recent Hanford Thyroid Disease Study did not find evidence of any excess risk for thyroid cancer induction for persons living in the vicinity of the Hanford facility who were exposed to β-radiation from radioactive iodine released from the facility (USDHHS, 2002). For doses in the range 0–100 mGy, risk was not correlated with dose and was less than for the control group based on persons outside what was considered the irradiation zone. In addition, for several health effects, the mean slope of the risk versus dose relationship was negative (indicating a possible hormetic-type dose–response relationship).

Some animal data are also consistent with thresholds for cancer induction by low-dose-rate low-LET radiation (Yamamato et al., 1998; Kondo, 1999; Tanooka, 2000; Yamamato and Seyama, 2000) but the authors do not discuss a beneficial effect. In a study where mice were fed tritiated water (β-radiation source) over their lifetime, lowering the dose rate below 12 mGy per day eliminated all of the lymphomas (Yamamoto et al., 1998; Yamamoto and Seyama, 2000). Higher dose rates produced significant increases in lymphomas. Implied in the reported findings was that spontaneous lymphomas were also eliminated. The data therefore appear consistent with a possible PROFAC = 1 against lymphoma occurrence for dose rates less than 12 mGy per day.

Mitchel et al. (2003) have also demonstrated that low-dose γ-radiation increases the latency for spontaneous lymphoma formation in cancer-prone Trp53 heterozygous mice. They attribute the protection to induced repair. However, the protection could also possibly be related to the elimination of some of the precancerous cells (e.g., problematic mutants, transformants) via radiation induction of the PAM process.

CONCLUSIONS

Evidence was provided to support the view that low-dose radiation and genotoxic chemicals could turn on the PAM process. Doing so could in turn lead to a reduction rather than an increase in the risk of stochastic effects such as problematic mutations, neoplastic transformation, and cancer. Furthermore, for combined exposure to low doses of different genetoxic agents, the combined exposure could turn on the PAM process, which also could lead to a reduction rather than an increase in the risks for problematic mutations, neoplastic transformation, and cancer.

The current cancer risk assessment paradigm is therefore in need of revision to include the possibility for RR <1. Indeed, for exposure to low doses (0–100 mGy) of low-LET radiation, RR <1 may be more the rule than the exception.

New research is needed to better understand the PAM process. The new research could unlock novel strategies for optimizing cancer prevention and novel protocols for low-dose therapy for cancer. With low-dose cancer therapy, normal tissue could be spared from severe damage while possibly eliminating the cancer.