Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 65
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2 3 Radiation-Induced Cancer: Mechanisms, Quantitative Experimental Studies, and the Role of Genetic Factors INTRODUCTION The process of cancer development (tumorigenesis) is recognized to involve multiple changes in genes involved in cell signaling and growth regulation, cell cycle checkpoint control, apoptosis, differentiation, angiogenesis, and DNA damage response or repair. Changes in these genes can involve (1) gene mutations or DNA rearrangements, which result in a gain of function as in the case of the conversion of proto-oncogenes to oncogenes; (2) mutations or DNA deletions or rearrangements, which result in loss of gene function as in the case of tumor-suppressor genes (Kinzler and Vogelstein 1998). The long latent period between radiation exposure and cancer development together with the multistage nature of tumorigenesis make it difficult to distinguish radiation-induced changes from those alterations that occur once the process has been initiated. Radiation-induced cancers do not appear to be unique or specifically identifiable (UNSCEAR 2000b). The mutations in tumors and their growth characteristics are not readily distinguishable from those in spontaneously occurring tumors of the same site or from tumors at the same site induced by other carcinogenic agents. Attempts to identify radiation-specific changes in human tumors have not been particularly successful despite fairly extensive investigation (UNSCEAR 1993, 2000b). There are, however, clues to possible underlying mechanisms of radiation-induced cancer that emerge from epidemiologic and experimental investigations. Based mainly on experimental studies, it is generally believed that complex forms of DNA double-strand breaks are the most biologically important type of lesions induced by ionizing radiation, and these complex forms are likely responsible for subsequent molecular and cellular effects (see Chapters 1 and 2). Attempts to repair complex DNA double-strand lesions are judged to be error prone, and there is evidence that this error-prone repair process can lead to gross chromosomal effects and mutagenesis. Molecular analyses of radiation-induced mutations have found a full range of mutations including base-pair substitutions, frameshift mutations, and deletions. Importantly, the most common radiation-induced mutations are deletions rather than base-pair changes in genes (point mutations; Chapters 1 and 2). Therefore, theories of radiation-induced cancer have generally centered on postirradiation tumor-suppressor gene inactivation that would be expected to occur through DNA deletion rather through the induction of point mutations. Oncogene activation through specific forms of induced chromosome translocation is also a candidate radiation-associated event, particularly for leukemia and lymphoma (UNSCEAR 2000b). Thus, mechanisms involving gene and/or chromosome rearrangements and loss of heterozygosity (signaling specific regions of DNA loss) are considered the most likely radiation-induced events that contribute to cancer development (UNSCEAR 2000b). More recently, experimental studies have questioned whether the initiating events produced by radiation are indeed direct effects on specific genes (e.g., Little 2000). Rather, it has been proposed that the gene or chromosomal mutations involved in radiation tumorigenesis arise indirectly as a consequence of persistent genomic instability (Chapter 2) induced by the radiation exposure. This chapter focuses first on studies relevant to mechanisms of radiation-induced tumorigenesis, with particular emphasis on the potential implications for low-dose risks. Subsequently, experimental studies addressing the quantitative relationship between radiation dose and cancer development are reviewed with particular regard to their consistency with proposed underlying mechanisms and the overall implications for cancer risk at low doses. Advances in human and animal genetics have also highlighted the contribution made to cancer risk by heritable factors (Ponder 2001). Much of the available information concerns germline genes that influence the risk of spontaneous cancer and the mechanisms through which they act. However, evidence is also emerging on the impact of such genes
OCR for page 66
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2 on radiation cancer risk (ICRP 1998). Relevant data on genetic susceptibility to cancer are reviewed in the final section of this chapter, and some interim judgments are developed about their implications for radiation cancer risk in the population. MECHANISMS OF TUMORIGENESIS Gene and Chromosomal Mutations in Spontaneously Arising Human Tumors Studies on the cellular and molecular mechanisms of tumorigenesis have in recent years cast much light on the complex multistep processes of tumorigenesis and its variation among tumor types. There is a vast literature on tumor biology and genetics (Bishop 1991; Loeb 1991, 1994; Hartwell 1992; Levine 1993; Vogelstein and Kinzler 1993; Hinds and Weinberg 1994; Weinberg 1994; Boland and others 1995; Karp and Broder 1995; Levine and Broach 1995; Skuse and Ludlow 1995; Kinzler and Vogelstein 1998; Rabes and others 2000; Khanna and Jackson 2001; Balmain and others 2003), and it is sufficient to highlight the principal points of current fundamental knowledge that may serve to guide judgments on the impact of ionizing radiation on cancer risk. Tumor development is generally viewed as a multistep clonal process of cellular evolution that may be conveniently but imprecisely divided into a number of overlapping phases: (1) tumor initiation, which represents the entry via mutation of a given normal somatic cell into a potentially neoplastic pathway of aberrant development; cellular targets for this process are generally held to have stem cell-like properties; (2) tumor promotion, which may now be viewed as the early clonal development of an initiated cell; cell-cell communication, mitogenic stimulation, cellular differentiating factors, and mutational and nonmutational (epigenetic) processes may all play a role in this early pre-neoplastic growth phase; (3) malignant conversion, which represents the tumorigenic phase where the evolving clonal population of cells becomes increasingly committed to malignant development; mutation of genes that control genomic stability is believed to be particularly important; and (4) malignant progression, which is itself multifaceted, is a relatively late tumorigenic phase during which neoplastic cells become increasingly autonomous and gain a capacity for invasion of surrounding normal tissue and spread to distant sites (metastasis); the development of tumor vasculature is important for the development of solid cancers (Folkman 1995). In addition, there is evidence that inflammatory processes and the microenvironment in which tumors develop are important cofactors for malignant progression (Coussens and Werb 2002). Overall, it is clear that only a small fraction of cells that enter tumorigenic pathways complete the above sequence that results in overt malignancy (Rabes and others 2000), and that the whole process can take many years. The balance of evidence suggests that sequential gene and chromosomal mutations act as the principal driving force for tumorigenic development, with phase transitions being dependent on the selection and overgrowth of clonal neoplastic variants best fitted for the prevailing in vivo conditions. Although there are exceptions, the consensus view is that tumor initiation or promotion is a monoclonal process having its origin in the appearance of a single aberrant cell (Levy and others 1994; Rabes and others 2000). The tumor initiation phase is most difficult to study directly, but in recent years it has become evident that a relatively tissue-specific set of so-called gatekeeper genes (Kinzler and Vogelstein 1997; Lengauer and others 1998) may be critical mutational targets for cellular entry into neoplastic pathways. Table 3-1 provides examples of such genes and their principal associated neoplasms. These gatekeepers are frequently involved in intracellular biochemical signaling pathways, often via transcriptional control, and are subject primarily to productive loss-of-function mutations. They fall into the tumor-suppressor gene category consistent with the germline role of many of these genes in autosomal dominant familial cancer (see “Genetic Susceptibility to Radiation-Induced Cancer,” later in this chapter). The somatic loss of function associated with gatekeeper gene inactivation can arise by point mutation (often of the chain-terminating type), intragenic deletion, or gross chromosomal loss events (Sidransky 1996; Kinzler and Vogelstein 1997, 1998). For some genes, epigenetic silencing events may also be important (Jones and others 1992; Feinberg 1993, 2004; Ranier and others 1993; Merlo and others 1995; Issa and Baylin 1996; Roth 1996). It is evident from Table 3-1 that the gatekeeper gene hypothesis applies principally to the genesis of solid tumors. For lymphomas and leukemia a somewhat different mechanism appears to apply. In these neoplasms, the early productive events often involve chromosomally mediated gain-of-function mutations in tissue-specific proto-oncogenes (i.e., gene activation or intragenic fusion involving juxtaposition of DNA sequences by specific chromosomal exchange; Rabbitts 1994; Greaves and Wiemels 2003). In many instances, these leukemia- or lymphoma-associated chromosomal events involve the DNA sequences (TCR [T cell receptor] and IG [immunoglobin]) involved in immunological TABLE 3-1 Examples of Human Tumor-Suppressor Genes of the Gatekeeper Type Gene Principal Cancer Type Mode of Action APC Colon carcinoma Transcriptonal regulator NF1 Neurofibromas GTPase-activator VHL Kidney carcinoma Transcriptional regulator WT-1 Nephroblastoma Transcription factor PTCH Skin (basal cell) Signaling protein NOTE: GTPase = guanosine 5′-triphosphatase.
OCR for page 67
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2 response (Rabbitts 1994). Tumorigenic chromosomal exchange events are less well characterized in solid tumors but do occur in certain sarcomas and in thyroid tumors (Rabbitts 1994; Mitelman and others 1997). However, in accord with data from solid tumors, gene deletion and other loss-of-function mutations are not uncommon in lymphohemopoietic tumors (Rabbitts 1994; Mitelman and others 1997). In relation to tumorigenesis in general, a second broad category of so-called caretaker genes has also been identified, although it is important to stress that the distinction between gatekeeper and caretaker genes is somewhat artificial—there are examples of genes that fulfill both criteria. Caretaker genes are those that play roles in the maintenance of genomic integrity (Kinzler and Vogelstein 1997, 1998). Table 3-2 provides examples of such tumor genes and their associated neoplasms. In such cases, mutational loss of function can lead to deficiency in DNA damage response and repair, repair or recombination, chromosomal segregation, cell cycle control, and/or apoptotic response (Loeb 1991; Hartwell and others 1994; Fishel and Kolodner 1995; Kinzler and Vogelstein 1996, 1998). Almost irrespective of the specific nature of the tumor gene in question, the net result of caretaker gene mutation is to elevate the frequency of gene or chromosomal mutations in the evolving neoplastic clone, and there is evidence that in some tumors this phenotype can arise at a relatively early point in neoplastic growth (Schmutte and Fishel 1999). This increased mutation frequency can be seen to provide the high level of dynamic clonal heterogeneity characteristic of tumorigenesis, thereby facilitating the selection of cellular variants that have gained TABLE 3-2 Examples of Human Tumor Genes of the Caretaker Type Gene Principal Cancer Type Mode of Action TP53 Multiple types Transcription factor (DNA damage response) ATM Lymphocytic leukemia PI-3 kinase (DNA damage response) MSH2, MLH1, PMS Colon or endometrial carcinoma DNA mismatch repair BRCA1/2 Breast or ovarian carcinoma Transcription factor (DNA damage response) XPA-G Squamous, basal cell carcinoma, melanoma Nucleotide excision repair MYH Familial adenomatous polyposis in families that lack the inherited mutation in the APC gene Removes adenines misincorporated opposite the mutagenic lesion 8-oxoguanine the capacity to evade or tolerate antitumorigenic defenses (Tomlinson and Bodmer 1999). These defenses would include cell-cell communication, apoptosis, terminal differentiation, cell senescence, and immune recognition (Rabes and others 2000). Gene and chromosomal mutations conferring enhanced tumor cell survival or growth characteristics have been identified in a range of malignancies (Greenblatt and others 1994; Branch and others 1995; Kinzler and Vogelstein 1998; Greider 1996; Orkin 1996). In summary, gene and chromosomal mutations of the general types induced by ionizing radiation are known to play a role throughout the multistep development of tumors. Loss of function of gatekeeper genes may be of particular importance in the initiation of common solid tumors, while gain-of-function chromosomal exchanges and gene loss events can arise early in lymphoma and leukemia. The relatively early spontaneous development of genomic instability via specific mutation of caretaker genes is believed to be important for tumorigenesis in many tissues, but epigenetic gene silencing or activation events have also been characterized. The emphasis placed here on early events in tumorigenesis derives from the prevailing view from epidemiologic and animal studies that ionizing radiation acts pri.cipally as a tumor-initiating agent. Mechanisms of Radiation Tumorigenesis Data from quantitative animal tumorigenesis (UNSCEAR 1988; Rabes and others 2000) and human epidemiologic studies (UNSCEAR 1994) imply that low-LET (linear energy transfer) ionizing radiation acts principally as a tumor-initiating agent. Specifically, in humans and animals, single acute doses of low-LET radiation produce a dose-dependent increase in cancer risk with evidence that chronic and fractionated exposures usually decrease that risk. Also, experimental animal data show that radiation only weakly promotes the development of chemically initiated tumors, and the generally greater tumorigenic sensitivity of humans to acute irradiation at young ages is more consistent with effects on tumor initiation than with promotional effects that accelerate the development of preexisting neoplasms. In this section, molecular and cytogenetic data on radiation-associated human and animal tumors are summarized in the context of the mutagenic and tumorigenic mechanisms discussed previously. Particular attention is given to the proposition, based on somatic mutagenesis data, that early arising, radiation-associated events in tumors will tend to take the form of specific gene or chromosomal deletions or rearrangements. Gene and Chromosomal Mutations in Radiation-Associated Human Tumors The acquisition of data on TP53 tumor-suppressor gene mutational spectra in human tumors associated with ultra-
OCR for page 68
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2 violet radiation (UVR) and chemical exposures was followed by searches for potential TP53 mutational signatures in excess lung tumors arising in Japanese A-bomb survivors and radon-exposed uranium miners (Vahakangas and others 1992; UNSCEAR 1993; Taylor and others 1994b; Venitt and Biggs 1994; Bartsch and others 1995; Lo and others 1995; Rabes and others 2000). Subsequently, attention was also given to TP53 mutations in liver tumors arising in excess in patients receiving the alpha-emitting radiographic contrast agent Thorotrast (Iwamoto and others 1999). Interpretation of these data are problematical, and although one study of lung tumors from uranium miners was suggestive of a possible codon-specific mutational signature of radiation (Taylor and others 1994b), this finding was not confirmed by others (Venitt and Biggs 1994; Bartsch and others 1995; Lo and others 1995). The studies on liver tumors from Thorotrast patients provide some comment on secondary TP53 mutation and possible instability effects but, overall, the studies cited above do not give consistent evidence that TP53 is a primary target for ionizing radiation. A cytogenetic-molecular data set is available on papillary thyroid cancer (PTC) (Bongarzone and others 1997) arising in excess in 131I-exposed children in areas contaminated by the Chernobyl accident (UNSCEAR 2000a). These mechanistic studies were guided by the knowledge that chromosomally mediated rearrangement and activation of the ret proto-oncogene is a frequently early arising feature of PTC (Richter and others 1999). Three different forms of ret gene rearrangement have been characterized at the cytogenetic and molecular levels (i.e., ret/PTC1, ret/PTC2, and ret/PTC3), and the prevalence of these events has been investigated in post-Chernobyl childhood PTC (Klugbauer and others 1995; Bongarzone and others 1997; Williams 1997; Smida and others 1999a, 1999b). As expected, ret activation events were found to be recurrent in Chernobyl-associated childhood PTC, and a similarly high frequency has been reported in adult thyroid cancer of patients with a history of radiation (Bounacer and others 1997). These studies suggest that the spectra of ret mutations differ between tumors of adults and children. Some investigations suggest that ret/PTC3 events in post-Chernobyl childhood cases are more frequent than expected. However this view is questioned by the study of 191 cases by Rabes and colleagues (2000), which provides evidence that the spectrum of ret rearrangements may be dependent on postirradiation latency, degree of tumor aggression, and possibly, dose to the thyroid. At present, causal relationships between ret gene rearrangement, childhood PTC, and radiation remain somewhat uncertain. However, a possible clue to radiation causation is the finding that breakpoints in the majority of ret rearrangements carry microhomologies and short direct or inverted repeats characteristic of the involvement of nonhomologous endjoining (NHEJ) mediated misrepair (Klugbauer and others 2001). Other investigations have reported that TP53 gene mutation does not play a significant role in the development of post-Chernobyl PTC (Nikiforov and others 1996; Smida and others 1997). Some informative molecular data are also available for basal cell skin carcinomas (BCCs) arising in X-irradiated tinea capitis patients (Burns and others 2002). In five out of five tumors analyzed there was evidence of DNA loss events which encompassed the Ptch gene (the gatekeeper for BCC development) plus the closely linked XPA gene. Overall, the studies summarized above, together with reports on the cytogenetic characterization of acute myeloid leukemias in A-bomb survivors (Nakanishi and others 1999) and radiotherapy-associated solid tumors (Chauveinc and others 1997) do not provide clear evidence on the causal gene-specific mechanisms of radiation tumorigenesis. In general however, they do support a monoclonal basis for postirradiation tumor development and suggest that the characteristics of induced tumors are similar to those of spontaneously arising neoplasms of the same type. A possible exception to this is that an excess of complex chromosomal events and microsatellite sequence instability was observed in late-expressing myeloid leukemias arising in A-bomb survivors exposed to high radiation doses (Nakanishi and others 1999); these data are discussed later in this chapter. Gene and Chromosomal Mutations in Animal Tumors Although radiation-induced tumors from experimental animals have been available for study for many years, it is only through advances in cytogenetics, molecular biology, and mouse genetics that it has become possible to investigate early events in the tumorigenic process. The most informative data on such early events derives from studies of tumors induced in F1 hybrid mice in which specific DNA loss events may be analyzed by loss of heterozygosity for genomically mapped polymorphic microsatellites. Mouse Lymphoma and Leukemia Early studies with radiation-induced thymic lymphoma provided evidence of recurrent RAS gene activation and some indication that the RAS gene mutational spectra differs between X-ray and neutron-induced lymphoma (Sloan and others 1990). Other molecular studies include the finding of recurrent chromosome (chr) 4 deletions in thymic and nonthymic lymphomas (Melendez and others 1999; Kominami and others 2002) and T-cell receptor (Tcr) gene rearrangements and chromosomal events in thymic lymphoma. However, the above and other somatic mutations in mouse lymphoma have yet to be specifically associated with initial radiation damage. The situation in mouse acute myeloid leukemia (AML; Silver and others 1999) is clearer. AML-associated, region-specific deletion of chr2 has been shown by cytogenetic analysis of in vivo irradiated bone marrow cell populations to be a direct consequence of radiation damage; clonal pre-
OCR for page 69
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2 neoplastic growth of carrier cells has also been reported (Hayata and others 1983; Breckon and others 1991; Bouffler and others 1997). These deletions, which are characteristic of ~90% of AML induced by various radiation qualities, have been analyzed in detail, and a putative myeloid suppressor gene target was identified within a chr2 interval of ~1 centimorgan (cM; 1 centimorgan equals about 1 million base pairs; Clark and others 1996; Silver and others 1999). Site-specific breakage of chr2 is characteristic of early radiation-induced events in AML, and there are cytogenetic and molecular data that support the involvement of telomere-like repeat (TLR) sequence arrays in chr2 breakage and rearrangement at fragile sites (Finnon and others 2002). Initial hypotheses on this form of postirradiation chromosomal fragility centered on increased recombinational activity of such TLR sequence arrays (Bouffler and others 1997). However, the data of Finnon and colleagues (2002) are more consistent with a mechanism of domain-specific chromosomal rearrangement involving chromatin remodeling that is mediated by TLR-associated matrix attachment sequences. With regard to radiation-induced osteosarcoma, Nathrath and colleagues (2002) have provided evidence for the involvement of two tumor-suppressor gene loci, but whether these loci are direct targets for radiation remains to be determined. Mouse genetic models of tumorigenesis have also proved to be instructive about the nature of radiation-associated early events in tumor induction. In these models, the germline of the host mouse carries an autosomal deficiency in a given tumor-suppressor or gatekeeper gene, thus exposing the remaining functional (wild-type) copy to spontaneous or induced mutation and thereby tumor initiation (see “Genetic Susceptibility to Radiation-Induced Cancer”). The nature of these tumor gene-inactivating events has been studied in models of different tumor types. In mice deficient in the Trp53 tumor suppressor gene (Trp53+/− and Trp53−/−), quantitative tumorigenesis studies implied that loss of the wild-type (wt) gene of Trp53+/− heterozygotes was a critical early event for the radiation induction of lymphoma and sarcoma (Kemp and others 1994). Molecular analysis confirmed the loss of wt Trp53 from tumors but also showed a high frequency of concomitant duplication of mutant (m) Trp53—such duplication was much less frequent in spontaneous tumors (Kemp and others 1994). Subsequent cytogenetic studies showed that Trp53+/− mice were highly prone to radiation-induced whole chromosome loss and gain (aneuploidy), and that the molecular data on tumorigenesis could be explained by radiation-induced loss of the whole chromosome (chr11) bearing wt Trp53, with duplication of the copy bearing mTrp53 being necessary to regain cellular genetic balance (Bouffler and others 1995). Thus, in this genetic context, Trp53 loss and tumorigenesis were relatively high-frequency events dependent upon the cellular tolerance of aneuploidy. However a recent study poses questions about whether Trp53 is indeed a direct target for radiation tumorigenesis in these knockout mice (Mao and others 2004). This study has raised the hypothesis that after radiation, the wt Trp53 gene in +/− mice activates the Fbxw7 gene, leading to genome instability, aneuploidy, and thereby increased Trp53 loss. Radiation-induced intestinal tumorigenesis has been studied in F1 hybrid mice of the Apc+/− genotype (Luongo and Dove 1996; van der Houven van Oordt and others 1999; Haines and others 2000). In this mouse model, DNA may be sampled from very small, early arising adenomas, thus focusing attention on early clonal events in tumor development (Levy and others 1994). Loss of wt Apc with the whole of the encoding chr18 is a relatively common early event in spontaneous intestinal tumorigenesis in Apc+/− mice. However, in tumors arising in low-LET-irradiated mice, the spectrum of wt Apc loss events was dominated by interstitial chromosome deletions. One study (Haines and others 2000) implicated a second chr18 locus in these early radiation-associated losses and also identified loss of the Dpc4 gene as a common secondary event in spontaneous and induced tumors. In some genetic backgrounds, mammary, ovarian, and skin tumors also arise in excess in Apc+/− mice (van der Houven van Oordt and others 1999). The same molecular genetic approach to experimental radiation tumorigenesis has been used in tumor-prone rodents that are heterozygous for the Ptch and Tsc-2 tumor-suppressor genes. Mice deficient in the patched gene (Ptch+/−) are susceptible to both spontaneous and radiation-induced BCC and medulloblastoma (Hahn and others 1998; Aszterbaum and others 1999; Pazzaglia and others 2002). Of particular note are the recent data of Pazzaglia and colleagues (2002) showing that neonatal mice are highly susceptible to X-ray-induced medulloblastoma and that the predominant mutational event in these tumors is loss of Ptch+. Loss of Tsc-2+ was similarly observed in many X-ray-induced renal carcinomas of Tsc-2+/− rats (Hino and others 2002), although intragenic deletions and point mutations were also observed. Importantly, the data available in this rodent genetic model (Hino and others 2002) reveal different spectra of tumor-associated Tsc-2+ mutations in spontaneous, X-ray, and ethylnitrosourea (ENU) induced renal carcinomas, which strongly suggests that the wt gene in target kidney cells is a direct target for carcinogens. As predicted from in vitro studies on somatic mutagenesis (Thacker 1986), tumors induced by the powerful point mutagen ENU were not characterized by Tsc-2+ gene loss events. Studies with gene knockout mice are providing further evidence on the role of DNA damage response genes in determining the in vivo radiosensitivity of cells and tissue, together with the impact on growth or development and spontaneous tumorigenesis (Deng and Brodie 2001; Kang and others 2002; Spring and others 2002; Worgul and others 2002). It is expected that such animal genetic models will, in due course, yield more detailed information on the in vivo mechanisms of radiation tumorigenesis.
OCR for page 70
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2 In summary, although studies with radiation-associated human tumors have yet to yield unambiguous data on the nature of causal gene and chromosomal mutations, animal studies are providing valuable guidance on the issue. Three principal points may be made. First, mechanistic studies on murine1 AML, lymphoma or sarcoma in Tp53+/− mice, intestinal adenoma in Apc+/− mice, medulloblastoma in Ptch+/− mice, and renal carcinoma in Tsc-2+/− rats all argue that the induction of critical cellular events by radiation occurs early in the tumorigenic process—a conclusion that is consistent with previous judgments on the issue. Second, the cytogenetic and molecular data cited for AML and intestinal tumors provide evidence for early monoclonal development of characteristic radiation-induced pre-neoplastic changes implying an initial, single-cell target. Third, for induction of AML and intestinal, medulloblastoma, and renal tumors, the radiation-associated events are predominantly DNA losses targeting specific genomic regions harboring critical genes. This in vivo DNA deletion mechanism is consistent with that understood in greater detail from in vitro somatic mutation systems. Also, many of the radiation-associated DNA loss events recorded in tumors are of cytogenetic dimensions. It is therefore possible to draw parallels with in vitro data on chromosome aberration induction where the predominant importance of DNA DSB induction and postirradiation error-prone NHEJ repair has been used in this report to argue against the proposition of a low-dose threshold in the dose-response. Evidence on the single-cell origin of radiogenic animal tumors, the in vivo gene or chromosomal loss mechanism for tumor initiation that appears to apply, and the close parallels that may be drawn with mechanisms and dose-response for in vitro induction of gene or chromosomal mutations argue in favor of a no-threshold relationship between radiation dose and in vivo tumor risk. In the examples cited, there is generally concordance between gene loss or mutational events recorded in spontaneous and radiation-associated tumors of a given type; although the data are more limited, such concordance tends to apply to other tumorigenic agents. A degree of gene specificity for different tumor types is also evident. An obvious caveat to this conclusion is the degree to which these limited mechanistic data provide support for broad judgments about radiation risk at low doses. For example, the data cited on the tolerance of aneuploidy in the bone marrow of irradiated Trp53-deficient mice can explain the high-frequency development of lymphoma but may not be wholly relevant to other tissues and/or other genetic settings. Data discussed in the following section on the potentially powerful effects of genetic background on tumorigenic risk in irradiated mice also caution against a dogmatic approach to judgments about low-dose risk that are based on current mechanistic knowledge. In this respect, the following section summarizes data concerning novel aspects of radiation response that may have relevance to unconventional mechanisms of radiation tumorigenesis. RADIATION-INDUCED GENOMIC INSTABILITY IN RADIATION TUMORIGENESIS As noted earlier in this chapter, the spontaneous development of tumors is frequently accompanied by the acquisition of genomic instability phenotypes that serve to promote the mutational evolution of more aggressive neoplastic clones. This form of genomic instability is increasingly well understood, and many of the responsible tumor gene mutations have been identified. Also noted in Chapter 2 is the large body of data showing that initial radiation-induced lesions are processed rapidly and expressed as chromosome aberrations at first postirradiation mitoses. However, during the last decade, evidence has accumulated that under certain experimental conditions, the progeny of cells surviving radiation appear to express an excess of new chromosomal and gene mutations over many postirradiation cell generations. This feature of cellular response (reviewed in Chapter 2) is generically termed radiation-induced persistent genomic instability. There are a variety of different manifestations of this phenomenon, and the developing field has been the subject of a number of recent reviews (Morgan and others 1996; Mothersill and Seymour 1998b; Wright 2000). The available data do not allow for generalizations on the onset and duration of such phenomena. On the basis of these data and previous reports of high-frequency neoplastic cell transformation (Clifton 1996), it has been suggested that epigenetic changes affecting a substantial fraction of irradiated cells can serve to destabilize their genomes and that the elevated postirradiation mutation rates in cell progeny, rather than gene-specific initial mutations, act to drive radiation tumorigenesis (Little 2000; Wright 2000). This section of the chapter focuses attention on in vivo studies of induced genomic instability that address the relevance of the phenomenon to radiation tumorigenesis. Chromatid Instability in Hematopoietic Cells Radiation-induced genomic instability in hematopoietic cells was first revealed by studies showing a persistent excess of chromatid-type aberrations in the progeny of mouse bone marrow cells irradiated in vitro with α-particles and subsequently grown in culture (Kadhim and others 1992). Alpha particles were considered to be substantially more effective than low-LET radiation in inducing this form of genomic instability (Wright 2000), which has also been reported in the progeny of cells that had not been traversed by an α-particle track (i.e., a bystander effect for instability; Lorimore and others 1998). Posttransplantation growth in vivo of in vitro irradiated bone marrow cells was also re- 1 Genus mus. A rat or mouse.
OCR for page 71
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2 ported to result in excess chromatid instability (Watson and others 1996). However, on the basis of the data summarized below, the consequences of postirradiation chromatid instability of bone marrow cells for hematopoietic neoplasia remains somewhat doubtful. Cytogenetic characterization of myeloid leukemia induced in the same mouse strain by α-particles, neutrons, and X-rays did not reveal evidence of the LET-dependent cytogenetic footprint of induced chromatid instability that might be expected from in vitro cellular studies with bone marrow cells (Bouffler and others 1996). In addition, the very high α-particle relative biological effectiveness (RBE) for induced genomic instability in bone marrow cells in culture (Kadhim and others 1992) is somewhat inconsistent with the low α-particle RBE suggested to apply to leukemogenic risk in vivo (Breckon and Cox 1990; UNSCEAR 2000b). Early studies of this form of induced instability depended on in vitro irradiation. Studies with humans exposed in vivo to low- and high-LET radiation (Tawn and others 2000b; Whitehouse and Tawn 2001) have found no evidence of induced chromatid instability in hemopoietic cells. The same negative result was obtained experimentally in the CBA/H mouse strain (Bouffler and others 2001). However Watson and colleagues (2001) provided data that suggested variable expression of in vivo induced chromatid instability in the CBH/H mouse strain. Since CBH/H is a highly inbred strain, such variable expression of chromatid instability cannot be ascribed to genetic variation. Experimental factors may therefore be of considerable importance, and relevant to this are the data of Bouffler and colleagues (2001), which indicate the existence of confounding stress factors that may account for in vitro and in vivo differences in the apparent expression of such instability. These in vivo observations cast considerable doubt on the relevance of radiation-induced chromatid instability for risk of lymphohematopoietic tumors. This view is strengthened by studies showing that the genetic determinants of induced chromatid instability in mouse bone marrow cells differ from those of susceptibility to induced lymphohematopoietic neoplasia (Boulton and others 2001). A similar degree of doubt has been expressed following reanalysis of genomic instability data (Nakanishi and others 1999, 2001) relating to myeloid leukemia arising in A-bomb survivors (Cox and Edwards 2002; Little 2002). Chromatid Instability in Mouse Mammary Epithelial Cells Differences in radiosensitivity and susceptibility to radiation induction of specific tumors among inbred mouse strains are well recognized, and there is good evidence that the BALB/c mouse is unusually sensitive to the induction of tissue injury and mammary tumors (Roderick 1963; Storer and others 1988); on these criteria the C57BL/6 mouse falls into the radioresistant category. Initial cytogenetic studies showed that mammary epithelial cells cultured from irradiated BALB/c mice persistently expressed substantially more chromatid aberrations during passage than those derived from irradiated C57BL/6 animals (Ponnaiya and others 1997; Ullrich and Ponnaiya 1998). In follow-up investigations, the chromatid instability phenotype of BALB/c was shown to be associated with a partial deficiency in the NHEJ repair protein DNA-dependent protein kinase catalytic subunit (DNA PKcs) together with compromised postirradiation DNA DSB repair (Okayasu and others 2000). This study, which included an intercomparison of inbred mouse strains, showed deficiency of DNA-PKcs and DNA DSB repair to be restricted to BALB/c suggesting genetic associations with persistent genomic instability and mammary tumor susceptibility. In accord with this, molecular genetic analyses showed BALB/c to carry a rare variant form of the gene (Prkdc) encoding DNA-PKcs, and subsequent analysis of recombinant mice provided strong evidence that variant Prkdc directly determined DNA-PKcs deficiency and postirradiation chromatid instability in mammary epithelial cells (Yu and others 2001). On the basis of these data it was proposed that induced genomic instability and mammary tumor susceptibility were genetically codetermined. Importantly, these investigations provide genetic evidence that deficiencies in the repair of DNA DSB, rather than as-yet-undefined epigenetic phenomena, are likely to determine persistent chromatid instability in this mouse. The question as to whether such instability is a primary causal element in mammary tumorigenesis or a secondary in vitro consequence of DNA repair deficiency and clonal growth selection remains to be resolved. Recent studies have also suggested a linkage between DNA-PKcs and maintenance of functional telomeres (Bailey and others 2004a, 2004b). As noted elsewhere in this report, the products of telomere dysfunction are dicentric chromosomes created by end-to-end fusion and sister-chromatid fusions, both of which can be associated with breakage-fusion-bridge cycles. More recently, a second product of telomere dysfunction, fusions between telomeres and the ends of broken DNA strands (i.e., DNA DSBs), have been described. Since telomere-DSB fusions have properties that differ from both chromosomal end fusions and ordinary chromosome aberrations, such fusions offer a potentially important new mechanism for induction of instability. These fusions appear to occur only under conditions of telomere dysfunction resulting from defects in the NHEJ pathway (Bailey and others 1999; Mills and others 2004). This suggests that genomic instability as a mechanism in radiation-induced cancer may be limited to specific circumstances in which individuals harbor specific DNA-repair deficiencies. Telomere-Associated Persistent Chromosomal Instability Telomeric repeat sequences (Bertoni and others 1994) cap the ends of mammalian chromosomes and serve to protect against replicative erosion and chromosomal fusion; in nor-
OCR for page 72
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2 mal human cells in culture, telomere shortening and instability is a natural feature of replicative cell senescence (Harley and Villeponteau 1995; Bacchetti 1996). In often degenerate forms, telomeric repeats are also found in subtelomeric and interstitial chromosomal locations, and there is some evidence that these loci may act as sites at which radiation-induced and other forms of genomic damage are preferentially resolved (Bouffler 1998). Early studies of the postirradiation development of chromosomal instability in in vitro passaged human diploid fibroblasts were suggestive of instability effects in a high proportion of irradiated cells (Sabatier and others 1992). However, subsequent detailed cytogenetic analyses suggested that passage-dependent instability in cultured human fibroblasts primarily takes the form of telomeric events expressed in cell clones naturally selected by growth rate during passage (Ducray and others 1999). Overall, the data obtained may be interpreted as initial radiation exposure bringing forward in time the natural process of clonal telomeric sequence instability associated with cell senescence and telomere shortening. A different form of postirradiation telomere-associated instability is expressed in a hamster-human hybrid cell system (Marder and Morgan 1993) where, in some clones, chromosomal instability is persistently expressed at translocations that have telomeric sequences at their junction (Day and others 1998). Similarly, unstable structures have been observed in unirradiated hamster cells undergoing gene amplification (Bertoni and others 1994), and again it may be that radiation is inducing genomic structures that enhance the natural expression of instability. There is good evidence that telomeric sequence instability is a recurrent feature of tumorigenic development (Bacchetti 1996; Chang and others 2001; Murnane and Sabatier 2004). Of particular relevance to the question of unstable translocation junctions are the so-called segmental jumping translocations that have been well characterized in spontaneously arising human leukemias (Shippey and others 1990). In respect of radiation tumorigenesis, detailed cytogenetic analyses suggest an excess of complex aberrations and segmental jumping translocations in myeloid leukemias arising at old ages in high-dose-exposed atomic bomb survivors (Nakanishi and others 1999). These and other data on excess microsatellite instability in A-bomb myeloid leukemias (Nakanishi and others 2001) have been reanalyzed in respect of dose and probability of tumor causation (Cox and Edwards 2002; Little 2002). These reanalyses largely uncouple the expression of leukemia-associated jumping translocations and microsatellite instability from radiation causation and argue that the potential contribution of induced instability to leukemogenic risk is likely to be small. Telomeric sequence instability at radiation-associated deletion or translocation breakpoints in mouse myeloid leukemia has also been recorded; this is not a general characteristic of these tumor-associated events, and recent studies argue against the direct involvement of telomeric sequence instability in these events (Bouffler and others 1996; Finnon and others 2002). In conclusion, although the position regarding radiation-induced persistent genomic instability and its causal association with tumorigenesis is not well understood, a few specific points can be made: In the case of radiation-associated persistent telomeric rearrangement and unstable chromosome translocation junctions, a coherent case can be made that a certain fraction of misrepaired genomic damage after radiation may be prone to ongoing secondary change in clonal progeny. There is evidence that such secondary genomic rearrangement can be a normal component of tumor development, in which case it is reasonable to assume that excess instability of this type could be a feature of some radiation-associated tumors, particularly those arising after high-dose irradiation where multiple or complex rearrangements may be expected. The genetic evidence from mouse studies that postirradiation chromatid instability can be associated with mammary tumor development is also persuasive, although it leaves unanswered questions on the causal role of the excess chromatid damage observed in vitro. Thus, in certain genetic settings of DNA repair deficiency, a role for postirradiation chromatid instability in tumorigenesis appears reasonable, and the potential linkage with telomere dysfunction could also be important. Based on the negative or inconsistent data on in vivo induced genomic instability in bone marrow cells, the nonsharing of genetic determinants, and the contention on data regarding A-bomb leukemias, induced genomic instability is judged unlikely to impact appreciably on the risk of lymphohematopoietic tumors after low-dose radiation. There are very few data on radiation-associated human solid tumors from which to assess the potential contribution of induced genomic instability. The central problem is the inherent difficulty in distinguishing this specific radiation-induced phenotype from spontaneously developing genomic instability as a natural consequence of clonal selection during tumor development. Stated simply, does tumor instability correlate with initial radiation damage or with neoplastic phenotype? This problem is well evidenced by molecular studies on post-Chernobyl (Belarus) childhood thyroid cancer. Initial studies showed evidence of excess microsatellite alterations in these radiation-associated tumors when compared with a reference group of adult thyroid cancers (Richter and others 1999). However, more detailed follow-up studies showed that the principal correlation was between microsatellite alterations and the aggression of early arising tumors. When this factor was taken into account, microsatellite loss or mutation in the early Belarus tumors was shown to be similar to that of the adult reference cases (Lohrer and others 2001).
OCR for page 73
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2 Based on consideration of the available in vivo data it is concluded that, at present, only a weak scientific case can be made for a discernible impact of induced genomic instability on radiation cancer risk. This conclusion is strengthened when account is also taken of the uncertainties noted in Chapter 2 regarding the biological basis and generality of the expression of induced genomic instability in cultured mammalian cells. QUANTITATIVE STUDIES IN EXPERIMENTAL TUMORIGENESIS General Aspects of Dose-Response The preceding discussion of potential mechanisms for radiation-induced cancer has indicated an important role for radiation-induced DNA DSBs, damage response pathways, and gene or chromosomal mutations in the initial events leading to cancer development. On this basis it would be predicted that the form of the dose-response for radiation-induced cancer and the effects of fractionation or reduced dose rate on this dose-response would be compatible with such underlying mechanisms unless factors involved in the expression of initiated cells are limiting in neoplastic development. Such a mechanistic model provides specific predictions with respect to dose-response and time-dose relationships for initial events and provides a framework for prediction of dose-response and time-dose effects for radiation-induced cancer (Ullrich and others 1987). Animal studies can be used to test these predictions. This framework is based on the αD+βD2 dose-response model for chromosome aberration induction described in Chapter 2. For single acute exposures the dose-response would be predicted to follow this model such that at low doses the relationship between cancer incidence and dose would be linear, while at higher doses this relationship would follow a function more closely related to the square of the dose. It is unlikely from a statistical standpoint alone that such a function could be proven to hold to the exclusion of all other dose-response models for any set of experimental data. Because of this, time-dose studies using both fractionated and low-dose-rate exposure regimens are important components in testing mechanistic predictions. On the basis of this model, it would be predicted that the dose-response following low-dose-rate exposures would be linear, with the same slope as the linear portion of the acute dose response model. In other words, at low doses the risk of radiation-induced cancer is independent of the time over which exposure occurs and is a cumulative function of dose. Fractionated exposures can further test these time-dose relationships and also provide information on the kinetics of processes involved. Such kinetic information, while limited, can provide insight into the nature of cellular versus tissue effects as major components in cancer risks in the specific experimental model under study. Any critical analysis of quantitative data on radiation-induced cancer requires informed selection of data sets. First, the adequacy of a study with respect to statistical power and use of appropriate analytical methodology must be considered. Second, biological factors involved in the pathogenesis of specific neoplasms must be considered with respect to the applicability of the experimental model to carcinogenesis in general and to cancer risk in humans in particular. Given these caveats, there are relatively few studies on animal carcinogenesis where the data are sufficient to address the issue of dose-response relationships or the issue of dose-rate effects and/or fractionation effects. Those studies in which such analyses are possible are limited mainly to rodent studies, principally mice. Biological factors in neoplastic development must also be noted. As discussed later in this chapter genetic background has a major role in determining neoplastic development at the level of sensitivity to both initiating events and events involved in expression. Therefore even in mouse studies in which there is sufficient statistical power to address questions of low-dose effects and time-dose relationships, the data are limited to mouse strains that are highly susceptible to specific forms of neoplasias. While variations in susceptibility must be considered potential confounding factors in applying animal data to human risks, careful analyses of human and animal data suggest that animal data do in fact have predictive value—for example, they can guide judgments on the choice of cancer risk models (Carnes and others 1998; Storer and others 1988). On the other hand, there are specific murine neoplasms whose pathogenesis appears to be unique to the mouse. In these specific instances it is unlikely that data derived using these systems would be applicable to human risks. These neoplasms are identified in sections below. Specific Murine Neoplasms Leukemia and Lymphoma The induction of leukemia and lymphoma has been examined in a number of murine systems, but the most extensive quantitative data on both dose-effects and time-dose relationships are for myeloid leukemia and thymic lymphoma. The most comprehensive data for myeloid leukemia with respect to dose-response relationships, and fractionation and dose-rate effects are in CBA male mice and RFM male mice (Upton and others 1970; Mole and Major 1983; Mole and others 1983). Interestingly, susceptibility in female mice of the same strains is markedly lower. The CBA mouse has also been used as an important model to dissect underlying radiation-induced molecular events described earlier (Bouffler and others 1991; Clark and others 1996; Silver and others 1999). For both strains, studies have been conducted over the dose range 250–3000 mGy (Upton and others 1970; Mole and Major 1983; Mole and others 1983). Analyses of
OCR for page 74
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2 data sets from both strains have yielded similar conclusions. Briefly, a number of dose-response models were seen to describe the data sets adequately. Data on incidence as a function of dose for both strains could be described by quadratic, linear-quadratic, and simple linear dose-responses with insufficient statistical power to exclude any of these three models on the basis of acute exposure data alone. Fractionation of the dose or low-dose-rate exposures resulted in a linear dose-response consistent with expectations of radiobiological theory in which the dose-response is linear quadratic for acute exposures and linear for low-dose-rate exposures, with the linear slope of the linear quadratic predicting the low-dose-rate and fractionation responses. These results are compatible with the apparent role of alterations in chromosome 2 in initial events for murine myeloid leukemogenesis and consistent with mechanistic predictions of dose and time-dose relationships described previously. This is not the case for studies on thymic lymphoma. In contrast to myelogenous leukemia, for which male mice are the most sensitive, female RFM mice are significantly more sensitive to the induction of thymic lymphoma following radiation exposures (Ullrich and Storer 1979a). For radiation-induced thymic lymphoma in female RFM mice, the data suggest a more complex relationship between radiation exposure and neoplastic development. Following single acute exposures over the 100–3000 mGy dose range, no simple dose-response model was found to describe the data (Ullrich and Storer 1979a). Low-dose-rate exposures, although significantly less effective with respect to induction of thymic lymphoma than single acute exposures, still resulted in a complex dose-response with a clear suggestion of a large threshold (Ullrich and Storer 1979c). These results should not be unexpected since the development of thymic lymphoma in mice following irradiation is an extremely complex process. The target cells for induction of thymic lymphoma are thought to be in the bone marrow rather than the thymus, and the pathogenesis of the disease appears to be largely mediated through indirect mechanisms with cell killing playing a major role (Kaplan 1964, 1967; Haran-ghera 1976). For example, the expression of thymic lymphoma can be substantially reduced or eliminated by protection of bone marrow stem cells from radiation-induced cell killing. The complex nature of the pathogenesis of this disease and the lack of a comparable counterpart in humans argues against thymic lymphoma as an appropriate model for understanding dose-response and time-dose relationships in humans. Solid Tumors Data from experimental studies examining dose-response and time-dose relationships are also available for a limited number of solid cancers in female RFM and BALB/c mice, including pituitary, Harderian gland, lung, and breast cancers (Ullrich and Storer 1979b, 1979c; Ullrich 1983). In a large study examining dose and dose-rate effects in female RFM mice, increased incidences of pituitary and Harderian gland tumors were reported. In spite of the large numbers of animals used, analyses of the data with respect to dose-response models could not distinguish between linear and linear-quadratic models (Ullrich and Storer 1979b). However, when the data for low-dose-rate exposures were considered as well, they were most compatible with a linear-quadratic model (Ullrich and Storer 1979c). Importantly, with respect to low-dose effects, these data support a linear response at low doses that is independent of exposure time. Such a response is consistent with predictions of the mechanistic model outlined earlier in this chapter. Although the number of animals used was smaller, a study examining radiation-induced lung and mammary adenocarcinomas in female Balb/c mice reached similar conclusions with respect to dose-response functions and low-dose risks (Ullrich and Storer 1979c; Ullrich 1983). This model was tested further in a series of experiments comparing the effectiveness of single acute exposures, acute fractionated exposures, and low-dose-rate exposures on the induction of lung and mammary tumors in the Balb/c mouse (Ullrich and others 1987). Importantly, in this study the hypothesis of time independence of effects at low doses was critically tested and found to hold. Specifically, similar effects were observed whether the same total dose was delivered as acute low-dose fractions or as low-dose-rate exposures. While the data for solid tumors described above are compatible with mechanistic models detailed earlier in this chapter, there are data sets that do not support a linear-quadratic dose-response model. Extensive data for mammary cancer induction in the Sprague-Dawley rat appear more consistent with a linear model over a wide range of doses and with linear, time-independent effects at low doses, low-dose fractions, and low dose rates (Shellabarger and others 1980). Although questions have been raised about the applicability of this model system to radiation-induced breast cancer in humans, much of the data from this rat model, from the mouse model in Balb/c mice, and from epidemiologic studies in exposed human populations appear to be consistent with respect to low-dose risk functions (Preston and others 2002b). In contrast to the data for leukemia and for pituitary, Harderian gland, lung, and mammary cancer described above, data from studies examining radiation-induced ovarian cancer in mice and bone and skin cancer in various animal species are more compatible with threshold dose-response models. In each instance it appears that an important role for cell killing in the process of neoplastic development and progression may explain these observations. Analysis of the dose-response for radiation-induced ovarian tumors following single acute or low-dose-rate exposures in RFM female mice indicated a marked sensitivity to induction at relatively low radiation doses, but equally importantly the analysis of the data strongly supported a threshold dose-response model (Ullrich and Storer 1979b, 1979c).
OCR for page 75
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2 In fact, this is one of the few instances for which a linear relationship could be rejected statistically. Studies in other mouse strains, while having less statistical power, also suggest a high sensitivity to induction of ovarian tumors at relatively low doses but with an apparent threshold (Lorenz and others 1947; Ullrich and Storer 1979c). This relatively unusual dose-response combining a threshold with high sensitivity to induction is unique to the mouse. Ovarian cancer in the mouse appears to involve an indirect mechanism for induction involving oocyte cell killing and subsequent alterations in the pituitary ovarian hormonal interactions (Kaplan 1950; Foulds 1975; Bonser and Jull 1977). The hormonal alterations are the proximate cause of tumor formation, with the role of radiation being relatively indirect as a result of its cell-killing effects. Because mouse oocytes are uniquely sensitive to the killing effects of radiation (the LD50 [lethal dose—50%] is ~50 mGy), ovarian tumors occur at very high frequencies following relatively low doses of ionizing radiation (Ullrich and Storer 1979c). A threshold appears to exist because a certain level of oocyte killing is required to cause the hormonal alterations that result in ovarian tumor formation. The principal effect of lowering the dose rate is to increase the threshold. In the RFM mouse, estimates of thresholds were reported as 110 mGy for acute exposures and 700 mGy for low-dose-rate exposures (Ullrich and Storer 1979b, 1979c). In contrast to the mouse, oocytes in humans are relatively resistant, with an LD50 of several grays. This difference in sensitivity is apparently because mouse and human oocytes are at different stages of differentiation in the ovary (Brewen and others 1976). The unique sensitivity of the mouse ovary to radiation makes it unlikely that results using this model system would have general applicability to risks in humans. Radiation-induced skin cancer has been studied in both mice and rats, although the majority of such studies have focused on the rat model because the rat is significantly more sensitive to skin tumor induction than the mouse (Burns and others 1973, 1975, 1989a, 1989b). In both rats and mice, relatively high total doses are required to induce skin cancer, and there is a clear threshold below which no tumors are seen. Multiple repeated radiation exposures are generally required for tumors to develop in mouse skin, while a single high dose (>10 Gy) is capable of inducing tumors in rat skin. It was for skin tumorigenesis that many of the concepts of multistage carcinogenesis were developed, including concepts related to initiation, promotion, and progression, and it is within this framework that the data for radiation-induced skin tumors are best considered (Jaffe and Bowden 1986; Burns and others 1989b). It appears from a variety of studies that single doses of ionizing radiation are capable of initiating cells with neoplastic potential, but that these cells require subsequent promotion in order to develop into tumors (Hoshino and Tanooka 1975; Yokoro and others 1977; Jaffe and Bowden 1986). Without this promotion these latent initiated cells will not express their neoplastic potential. Several lines of evidence support this view. Hoshino and Tanooka have demonstrated that small doses of beta irradiation are capable of inducing initiating alterations in mouse skin that required subsequent promotion with 4-nitroquinoline N-oxide (4NQO) for tumors to develop. Jaffe and Bowden (1986) have demonstrated the initiating potential of single doses of electrons when followed by multiple exposures to the tumor-promoting agent TPA (12-O-tetradecanoylphorbol-13-acetate). Fry and his coworkers (1986) have shown that X-ray-initiated cells can be promoted to develop skin tumors by exposure to ultraviolet light. This group has demonstrated further that the apparent threshold dose-response for skin tumorigenesis can be converted to a linear UVR dose-response when promotion is used to maximize the expression of latent initiated cells. Based on such observations it is logical to speculate that the multiple high-dose fractions of radiation that are generally required to induce skin tumors in mouse skin are acting not only to initiate cells but also to induce tissue damage via cell killing, which in turn acts as a promoting stimulus to facilitate the progression of these initiated cells into skin tumors. Likewise in the rat, the high doses required to produce tumors are likely to produce both transformation of cells and sufficient cell killing to promote the transformed cells. This phenomenon does not appear to be unique to these animal systems. Most evidence suggests that relatively high doses of radiation are necessary to induce skin tumors in humans and that these effects can be enhanced by exposure to UV light from the Sun (Shore 2001). It is also important to note studies by Jaffe and Bowden demonstrating that multiple low doses of radiation to the skin that did not produce tissue damage were not effective in promoting skin tumors initiated by chemical agents (Jaffe and Bowden 1986). These data support the view that the predominant role for low-dose radiation is tumorigenic initiation. Studies of bone cancer also suggest a threshold response and a requirement for prolonged exposure for tumor development from exposure to low-LET radiation (NCRP 1990). Unfortunately most of the available data have focused on observations of effects rather than dissecting potential underlying mechanisms. Attempts have been made to model bone tumorigenesis however, and these models have again focused on an important role for a mechanism involved in the expression of initiated cells in controlling tumor development (Marshall and Groer 1977). Although speculative, it is likely that mechanisms similar to those proposed for skin tumorigenesis involving the cell-killing effects of radiation are likely involved in producing a threshold response for bone tumors. Fractionation Kinetics Studies using fractionation regimens have been useful in addressing issues of time-dose relationships in radiation carcinogenesis. In a few instances, investigators have also used
OCR for page 80
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2 TABLE 3-3 Examples of Autosomal Recessive Disorders of DNA Damage Response Disorder Genes or Locus Defect Proposed Major Clinical Features Cancer Approximate Prevalence (per live births) Xeroderma pigmentosum XP-A to XP-G and XPV Excision or postreplication repair Photosensitivity and cancer of UVR-exposed skin Squamous cell skin carcinoma, basal cell carcinoma, and melanoma 1 in 250,000 Cockaynes syndrome CS-A, CS-B Transcribed strand repair Photosensitivity, dwarfism No excess a Trichothiodystrophy XP-D Excision repair Photosensitivity, abnormal sulfur-deficient hair Variable excess (skin) a Ataxia-telangiectasia ATM Kinase activity Radiosensitivity, neuro-and immunodeficiency Lymphoma 1 in 100,000 Nijmegen breakage syndrome NBS NHEJ factor (Mrell/RAD50/nbs) Radiosensitivity, microencephaly, immunodeficiency Lymphoma a Fanconi’s anemia FA-A to FA-C DNA cross-link repair Bone marrow deficiency, skeletal abnormalities Leukemia 1 in 300,000 aLess than 1 in 100,000. TABLE 3-4 Examples of Autosomal Dominant Disorders of Tumor Suppressor Genes, Proto-oncogenes, and DNA Damage Response or Repair Genes Disorder Genes or Locus Defect Proposed Cancer Approximate Prevalence (per live births) Tumor-Suppressor Disorders Familial adenomatous polyposis APC Transcriptional regulation Colorectal cancer (multiple polyps) 1 in 8000 Von Hippel-Lindau disease VHL Transcriptional regulation Renal cancer 1 in 30,000 Denys Drash syndrome WT1 Transcriptional regulation Nephroblastoma (+ others) ? Neurofibromatosis type 1 NF-1 GTPase regulation Neurofibroma Schwannoma 1 in 3000 Neurofibromatosis type 2 NF-2 Cytoskeletal linkage Meningioma Neurofibroma 1 in 30,000 Nevoid basal cell carcinoma syndrome PTC Cellular signaling Basal cell skin cancer Medulloblastoma 1 in 50,000 Tuberous sclerosis TSC1 TSC2 Cellular signaling Cellular signaling Benign lesions of skin, nervous tissue, heart, and kidneys 1 in 20,000 Retinoblastoma RB1 Transcriptional regulation Retinal tumors, bone or soft-tissue sarcoma, brain cancer, and melanoma 1 in 25,000 Proto-oncogene Disorders Multiple endocrine neoplasia (2A and 2B) and familial medullary thyroid cancer RET Cellular signaling Thyroid or parathyroid neoplasms ? DNA Damage Response or Repair Disorders Hereditary nonpolyposis colon cancer MLH1, MSH2, PMS1, PMS2 DNA mismatch repair, apoptosis Colon cancer, endometrial cancer 1 in 2000 Li-Fraumeni syndrome TP53 (others?) DNA damage recognition Various 1 in 50,000 Heritable breast or ovarian cancer BRCA-1 BRCA-2 Transcriptional regulation, DNA repair Breast or ovarian cancer Breast cancer (also male) 1 in 1000
OCR for page 81
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2 degree of spontaneous tumor risk that is imposed must be sufficient to distinguish that family from others that are noncarriers. Given that, on average, spontaneous cancer incidence in the general population is around 30%, the information currently available is restricted largely to mutations where the cancer in question is expressed at a high relative frequency in gene carriers (i.e., so-called high-penetrance mutations). Other features of importance are (1) the organ specificity of many cancer-predisposing mutations, (2) the age of onset of given neoplasms in gene carriers that usually occurs at younger ages than in noncarriers, (3) the frequent occurrence of multiple tumors in gene carriers, and (4) the substantial variation for cancer risk between carriers of a given gene mutation, suggestive of major influences from the genetic background and/or life-style of the host. These issues of heritable cancer risk have been summarized by the International Commission on Radiological Proterction (ICRP 1998) and more recently by Ponder (2001). The crucial point, to be developed later, is that current knowledge of heritable cancer susceptibility in humans is restricted largely to relatively rare mutations of high penetrance. Cancer may be regarded as a multifactorial disorder (see Chapter 4), and genetic views developed from the study of other multifactorial conditions, such as coronary heart disease, suggest strongly that there will be many more variant cancer genes having lower penetrance than those listed in Tables 3-1 and 3-2. The current lack of knowledge about the nature, frequency, and impact of such genes imposes fundamental limitations in respect of the objectives stated earlier. Mechanistic Aspects of Genetically Determined Radiation Response In making judgments on the radiation response of cancer-prone individuals it is valuable to consider first the theoretical expectations that follow from current knowledge of the cellular mechanisms that are likely to be involved in cancer susceptibility. Germline mutations in DNA damage response or repair genes, tumor-suppressor genes, and proto-oncogenes are considered in turn. DNA Damage Response-Repair Genes As outlined in Chapters 1 and 2, different forms of DNA damage are recognized and processed in mammalian cells by different biochemical pathways, which share few genetic determinants. Accordingly, there is no expectation of a global association between DNA damage response or repair deficiency and sensitivity to the tumorigenic effects of radiation. Rather, the expectation is that a deficiency of genes associated with recognition or repair of the form of damage that is critical for cellular response to radiation (i.e., DNA DSB) will be of greatest significance for radiation cancer risk. On this basis the autosomal recessive disorders AT and NBS in Table 3-3 might be judged to exhibit increased cancer risk after ionizing radiation, whereas XP would not. Stated simply, germline deficiency in the recognition and/or repair of induced DNA damage of specific forms is expected to increase the abundance of genome-wide damage in the somatic cells of body tissues. This increased mutational load will tend to increase cancer risk, albeit with differing degrees of expression among tissues. It is important to recognize, however, that a number of autosomal dominant conditions, particularly Li-Fraumeni syndrome (TP53+/−), are determined by genes that play more general roles in the control of stress responses, apoptosis, and/or coordination of the cell reproductive cycle (Chapter 2). Abnormal cellular response or cancer risk in such disorders might be expected for a range of DNA-damaging agents including ionizing radiation. Tumor-Suppressor Genes For tumor-suppressor genes such as VHL and NF1 in Table 3-4 there is no specific association with DNA damage response or repair. Accordingly there is no expectation of increased genome-wide sensitivity to the mutagenic effects of radiation. In these instances increased radiation cancer risk may be anticipated on the basis of the now well-supported hypothesis of Knudson (1986). In brief, there is good evidence that many tumor-suppressor type genes act as tissue-specific gatekeepers to neoplastic pathways (Kinzler and Vogelstein 1997). Since loss or mutation of both autosomal copies of such genes from single cells is believed to be rate limiting for the initiation of neoplastic development, tumor initiation in normal individuals is expected to be a rare cellular event. A carrier of a germline mutation in a given tumor-suppressor gene will however show loss of function of one such gene copy, thus “unshielding” the second copy in all target somatic cells. The lifetime risk of spontaneous loss or mutation of that second copy from any given population of target cells will be relatively high—hence the often dramatic increase in organ-specific cancer risk. There is also a clear expectation that exposure of the carrier individual to ionizing radiation or indeed other genotoxic carcinogens would, via the same genetic-somatic mechanism, result in a greater-than-normal risk of organ-specific cancer. Stated simply, the enhanced radiation cancer risk in the carrier individual would be driven by a reduction in the target gene number from two to one; in a given disorder the organs at increased risk would tend to be the same as those involved in spontaneous neoplasia. Proto-oncogenes There are few well-characterized germline, gain-of-function mutations in proto-oncogenes that have unambiguous associations with cancer risk; a series of characterized ret gene mutations are however known to increase the risk of
OCR for page 82
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2 thyroid neoplasia (Table 3-4). As in the case of tumor-suppressor gene loss, germline ret mutation may be viewed as removing one early rate-limiting step in multistage thyroid tumorigenesis such that the carrier individual is at increased risk of neoplastic development via the accumulation of further mutations in other genes. Again, greater-than-normal radiation risk to the target organ should be anticipated. In the following sections, the above propositions are examined on the basis of available cellular, animal, and epidemiologic data. Cellular Data on Heritable Human Radiosensitivity Cellular data on heritable radiosensitivity in respect of cell inactivation have been reviewed recently (ICRP 1998). In brief, although there are isolated instances of cancer and/or radiotherapy patients showing clear evidence of radiosensitivity, it is only for AT and NBS that there is unambiguous evidence of profoundly increased radiosensitivity to cell killing associated with known human disorders of DNA damage response or repair and cancer. Claims for increased radiosensitivity in other cancer-prone disorders remain controversial and do not provide clear guidance on radiation cancer risk. Although sensitivity to cell killing after radiation may at present not be a particularly useful surrogate for cancer risk, there are closer parallels between the induction of chromosome damage and cancer. Although not without some uncertainty, the data accumulating on the patterns of chromosomal radiosensitivity in human cancer-prone disorders are worthy of some attention. These data, considered by Scott and colleagues (1998) and reviewed by the National Radiological Protection Board (NRPB 1999) show that, compared with healthy controls, cells cultured from AT and NBS patients typically exhibit two- to threefold greater chromosomal radiosensitivity, but in some cytogenetic assays, the increased sensitivity can be up to tenfold (Taalman and others 1983; Taylor 1983). The NRPB has summarized a large body of cytogenetic data on which claims of associations between chromosomal radiosensitivity and human cancer susceptibility have been based. As in the case of cell killing, some of these claims remain controversial. More recent studies on the possible radiosensitivity of cells from breast cancer-susceptible BRCA1 and BRCA2 patients have also provided conflicting evidence (Buchholz and others 2002; Trenz and others 2002; Powell and Kachnic 2003). Of additional interest are the data on G2 cell cycle radiosensitivity, which among other findings suggest that AT heterozygotes are indeed radiosensitive and that up to 40% of unselected breast cancer cases also exhibit modestly elevated radiation-induced chromosome damage (Scott and others 1994; Parshad and others 1996). There is also some evidence of elevated chromosomal radiosensitivity in cells from patients with malignant gliomas (Bondy and others 1996) and colorectal cancer (Baria and others 2001). In summary the evidence available on human chromosomal radiosensitivity suggests that AT and NBS may be up to tenfold more sensitive than normal; some uncertainty surrounds the chromosomal radiosensitivity of other cancer-prone disorders, but any such increase in sensitivity appears to be modest—not more than two- to threefold. Although critical data are lacking, it is a reasonable assumption that, in general, a heritable increase in chromosomal radiosensitivity would be associated with increased radiation cancer risk, albeit with possible differences in the response of different tissues. Data from G2 chromosomal radiosensitivity assays are generally supportive of this association, but some data remain controversial. Animal Data on Radiosensitivity and Tumorigenesis The experimental data available about the impact of heritable factors on radiosensitivity and tumorigenesis derive principally from studies on the genetic homologues of some of the human disorders listed in Tables 3-3 and 3-4. These studies are summarized in Table 3-5 with references. Although there are some differences in the patterns of phenotypic expression, in the main the rodent genetic homologues of AT, Li-Fraumeni syndrome (LFS), familial adenomatous polyposes, neroid basal cell carcinoma syndrome (NBCCS), and tuberous sclerosis recapitulate many of the features of their human counterparts. In respect of early responses, Atm−/− mice show extreme radiosensitivity; there is also evidence of moderate in vivo radiosensitivity in Atm+/− mice. Studies with Atm+/− knockout mice (Barlow and others 1999) provided evidence of increased in vivo radiosensitivity but failed to demonstrate differences in radiation induced tumorigenesis between +/− and +/+ genotypes. However, more recent data on spontaneous tumorigenesis (Spring and others 2002) imply that such studies are best conducted with Atm knock-in mice, which recapitulate known human mutations. Data on BRCA1- and BRCA2-deficient mice have yet to provide clear evidence on the role of these genes in radiation tumorigenesis. The principal benefit of the referenced studies noted in Table 3-5 is the provision of a growing association between the Brca genes, Rad51, cell cycle perturbation, and DNA damage response. The most valuable animal genetic data on radiation tumorigenesis have been developed from studies on mice heterozygously deficient in the tumor-suppressor genes Tp53, Apc, and Ptch and in a rat strain (Eker) heterozygously deficient in Tsc2 (see Table 3-5 for references). In all instances, the germline mutational loss of one copy of the respective tumor-suppressor gene leads not only to an increase in the rate of spontaneous tumorigenesis but also to increased sensitivity to the induction of the same tumor types by whole-body low-LET radiation with doses up to around 5 Gy. These data provide strong support for the contention, discussed earlier, that the unshielding of tumor-suppressor
OCR for page 83
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2 TABLE 3-5 Radiation Response and Tumorigenesis in Rodent Homologues of Cancer-Prone Human Genetic Disorders Genotype Human Homologue Radiation Response Comment Key References Early response Tumorigenesis Mouse Atm−/− Ataxia-telangiectasia (AT) Radiosensitive in vivo or in vitro May be dependent upon Atm genotype (see text) Defects in meiosis, immunity, and behavior Barlow and others (1996) Elson and others (1996) Xu and others (1996) Mouse Brcal−/− BRCA+/− Heritable breast cancer in heterozygotes Cellular and embryonic radiosensitivity No published study identified Embryolethal; association with Rad51−/− phenotype Gowen and others (1998) Sharan and others (1997) Mizuta and others(1997) Connor and others (1997) Mouse Tp53+/− Li-Fraumeni syndrome (LFS) Excess aneuploidy and G2/M checkpoint defect in bone marrow cells Highly sensitive to induction of lymphoma or sarcoma Tumorigenesis associated with loss of Tp53+ Kemp and others (1994) Bouffler and others (1995) Mouse Apc+/− Familial adenomatous polyposis None reported Highly sensitive to induction of intestinal adenoma (breast and other cancers in some genetic backgrounds) Tumorigenesis associated with loss of Apc+ and other loci Luongo and Dove (1996) Ellender and others (1997) van der Houven van Oordt and others (1999) Haines and others (2000) Mouse Ptch+/− Nevoid basal cell carcinoma syndrome Some evidence of cellular radiosensitivity Sensitive to induction of medulloblastoma Tumorigenesis associated with loss of Ptch+ Hahn and others (1998) Pazzaglia and others (2002) Rat Tsc2+/− Tuberous sclerosis None reported Sensitive to induction of renal neoplasia Tumorigenesis associated with loss of Tsc2+ Hino and others (1993, 2002) genes by germline mutation will lead to a significant increase in individual susceptibility to radiation tumorigenesis. Critical mechanistic support for this hypothesis has been provided by molecular analysis of tumors arising in irradiated Tp53+/−, Apc+/−, and Ptch+/− mice and Tsc-2+/− rats; as predicted, such analyses strongly suggest that radiation acts by inactivating the wild-type tumor-suppressor gene copy in target somatic cells. These wild-type genes appear to be mutated by radiation through mechanisms principally involving substantial DNA loss events, although there are examples of whole chromosome losses as well as intragenic deletions and point mutations. Although the above studies provide proof-of-principle experimental evidence of strong genetic effects on radiation tumorigenesis in mammalian species, quantification of the genetically imposed radiation risk is most problematical. An ICRP (1998) Task Group, in reviewing much of the data of Table 3-5, suggested that radiation tumor risk in such suppressor-suppressor gene-deficient mice might be elevated by up to a hundredfold or more but cautioned against firm judgments because of (1) problems associated with experimental design and (2) preliminary evidence that natural variation in the genetic background of host animals can have major modifying effects on tumor yield. During the last few years the impact of such modifier genes on the expression of tumorigenesis in mice has been demonstrated more clearly (Balmain and Nagase 1998). The principal message from this experimental work is that because of the strongly modifying effects of genetic background, rodent homologues are unlikely to provide a quantitatively reliable representation of radiation tumorigenesis in cancer-prone human genetic disorders. Such genetic modification is to be expected in humans, but the specific nature and impact of the modifier genes are likely to differ among species. The issue of genetic modification of radiation response is considered further in the section of this chapter that deals with cancer-predisposing mutations of low penetrance. Human Data on Radiosensitivity and Tumorigenesis As noted earlier in this chapter unambiguous evidence of human genetic disorders showing hypersensitivity to tissue injury after radiation is confined to AT and NBS, where conventional radiotherapy procedures have proved disastrous to patients. Adverse, but less profound, reactions to radiotherapy are however reported to occur in around 5% of cancer patients (Burnet and others 1998). Studies on in vitro
OCR for page 84
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2 cellular radiosensitivity in such radiotherapy patients have, so far, failed to reveal evidence of strong correlations between in vivo and in vitro responses although subsets of these patients do show statistically significant increases in cellular radiosensitivity under some assay conditions (Burnet and others 1998). Similarly limited molecular studies show no correlation between adverse reactions to radiotherapy and heterozygous ATM gene mutation (Appleby and others 1997; Burnet and others 1998). The question as to whether adverse tissue reaction to radiotherapy signals potentially increased risk of therapy-related second tumors has yet to be addressed in epidemiologic studies. Postradiotherapy observations on specific sets of cancer patients have, however, revealed valuable information on genetic associations with risk of second tumors (Meadows 2001). These data are summarized and referenced in Table 3-6. In brief, there is evidence of an excess of radiotherapy (RT)-related tumors in the human cancer-prone conditions heritable retinoblastoma, NBCCS, and LFS plus related conditions, as well as in children from families with a history of early onset cancer. In addition there are reports suggesting that neurofibromatosis is a positive factor for RT-related tumorigenesis (Robison and Mertens 1993). By contrast, a variety of studies discussed by Mark and colleagues (1993) provide no clear evidence that genetic factors are important for RT-related breast cancer. Recent studies provide no evidence that the status of BRCA genes influences post-radiotherapy outcomes at 5 years (Pierce and others 2000). In Table 3-6 the data suggesting that NBCCS and LFS patients have substantial increases in tumorigenic radiosensitivity are in accord with data obtained experimentally with their rodent genetic homologues. For retinoblastoma (RB), the large size of the U.S.-based epidemiologic studies of Eng and colleagues (1993) and Wong and coworkers (1997a) allows some judgments to be developed on the degree to which this suppressor gene disorder predisposes to (second) radiogenic soft-tissue sarcoma and bone cancer. Although there is a clear dose-response for radiation tumorigenesis, these data imply that excess relative risk (ERR) in heritable RB patients may be lower than in the nonheritable controls. The background rate of tumorigenesis in RB is, as expected, rather high, and for the purposes of this report, excess absolute risk (EAR) may be a more useful measure of tumorigenic radiosensitivity than ERR. In considering this issue, the ICRP (1998) and NRPB (2000) suggest that the EAR in heritable RB is around fivefold higher than in the nonheritable group. It is notable that low values of ERR for radiogenic cancer in such cancer-prone conditions are consistent with other epidemiologic data on radiation tumorigenesis where high background cancer rates also tend to be accompanied by lower ERRs. Abramson and colleagues (2001) have also reported on third tumors in RB patients after radiotherapy. As might be expected, the sites of these additional tumors generally accorded with the irradiated volume of normal tissue. In summary, although clinical and epidemiologic data on RT patients are limited, they are sufficient to confirm the view developed from mechanistic knowledge and experimental studies that human genetic susceptibility to spontaneous tumorigenesis is often accompanied by an increase in absolute cancer risk after ionizing radiation. Quantifying that risk is problematical, but the single study on RB patients that has this capacity is suggestive of relatively modest (about fivefold) increases over that of normal individuals. In the TABLE 3-6 Postradiotherapy Observations on Risk of Second Tumors in Humans Genetic Disorder or Study Group First Tumor Observations Key References Retinoblastoma Retinoblastoma Excess bone tumors and soft-tissue sarcomas, large cohorts; some dose, dose-response, and risk estimates possible Tucker and others (1987a) Eng and others (1993) Wong and others (1997a) Abramson and others (2001) NBCCS Medulloblastoma Excess basal cell skin neoplasms and ovarian fibromas, short latency; case reports only Strong (1977) Southwick and Schwartz (1979) LFS and related conditions Various Follow-up of children developing posttherapy soft-tissue sarcoma, bone tumors, and acute leukemia—linkage with family histories of cancer Strong and Williams (1987) Heyn and others (1993) Robison and Mertens (1993) Malkin (1993) Case-control study of therapy-related second tumors Various Excess posttherapy tumors in children from non-LFS families with a history of early onset cancer Kony and others (1997)
OCR for page 85
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2 future, the growing capacity of molecular screening techniques to detect cancer-susceptible genotypes in the general population will, in principle, allow the radiation risk of such genotypes to be assessed in a number of suitable human cohorts. A summary of such molecular epidemiologic approaches to spontaneous cancer risk is given later in this chapter. Population Modeling of Radiation Cancer Risk: Impact of Strongly Expressing Genetic Disorders In conjunction with the work of an ICRP (1998) Task Group, Chakraborty and colleagues (1997, 1998a) have constructed and illustrated the use of a population-based computational model that serves to describe the impact of cancer-susceptible genotypes on radiation cancer risk in the population. For reasons of data sufficiency, breast cancer risk in typical Western populations was considered and illustrated. This approach, which is based on established Mendelian principles, employed best estimates of the prevalence of known, high-penetrance breast cancer-predisposing genes (BRCA1 and BRCA2), the relative risk of spontaneous breast cancer in such genotypes, and a range of factors that describe in a hypothetical fashion the increase in radiation risk imposed by the given gene mutations; the risk of radiogenic breast cancer in normal individuals was based on data from Japanese atomic bomb survivors. Other issues that were considered included increased gene frequency in certain genetically isolated populations (Ashkenazi Jews) and the influence of reduced penetrance on population risk. The following points summarize the outcome of these modeling exercises. Using best estimates of breast cancer gene frequencies, the genetic impact on excess breast cancer in an irradiated Western population would be small even if these mutations were to impose a radiation risk that was as much as a hundredfold greater than that of normal genotypes. Using estimates of the higher gene frequencies in Ashkenazi Jewish populations, the genetic impact on radiation-associated breast cancer can become significant but only if the genetically imposed radiation risk is very high. The genetic impact of such mutations will be diluted in proportion to decreasing penetrance. This model and its predictions have been used by the ICRP (1998) and NRPB (1999) to provide interim judgments on the implications of genetic susceptibility to cancer for radiological protection. Since the overall prevalence of highly penetrant cancer-predisposing mutations in typical human populations is judged to be 1% or less (ICRP 1998) and since available data tend to argue against extreme increases in genetically imposed radiation cancer risk, there is reason to believe that the presence of these rare, highly penetrant mutations will not appreciably distort current estimates of radiation cancer risk in the population. Stated simply, only a very small fraction of excess cancers in an irradiated human population are expected to arise in individuals carrying familial cancer genes. The ICRP (1998) and NRPB (1999) stressed, however, that this conclusion took no account of the presence of potentially more common cancer genes of low penetrance that do not express familial cancer. The ICRP and NRPB reports also commented on the problems inherent in identifying and making judgments about radiation cancer risk in genetic subgroups carrying such weakly expressing genes and considered the issue of genetically imposed risk to individuals. These matters are discussed in subsequent sections. Genes of Low Penetrance As noted earlier in this chapter, knowledge of heritable factors in tumorigenesis stems largely from studies on strongly predisposing autosomal dominant familial traits and autosomal recessive disorders having unambiguous phenotypes. The problem of estimating the heritable impact on cancer risk from weakly expressing genes of low penetrance and other genetic modifiers of the cancer process has been with us for some time. However, not unexpectedly, an understanding of this issue is proving difficult to obtain. To a large measure this is due to the likelihood that, individually, polymorphic variant genes probably contribute small additional cancer risks to each carrier in a largely tissue-specific manner. These will tend to escape detection by conventional medical genetic and epidemiologic studies. A combination of such genes and their interaction with environmental risk factors may, however, provide a substantial genetic component to both spontaneous and radiation-associated risk. The magnitude of this risk in a given human population would then be determined by gene frequencies together with the pattern or strength of gene-gene and gene-environment interactions. These issues of population cancer risk have been discussed widely in the context of epidemiologic and molecular genetic findings (Hoover 2000; Houlston and Tomlinson 2000; Lichtenstein and others 2000; Peto and Mack 2000; Shields and Harris 2000; Dong and Hemminki 2001; Nathanson and Weber 2001; Ponder 2001). Here it is sufficient to illustrate some of the progress being made in respect of the weakly expressing genetic component of human and animal tumorigenesis. Where possible, emphasis is placed on data having some connection with cancer risk after ionizing radiation. Human Breast Cancer BRCA1 and BRCA2 genes have been identified as the principal genetic determinants of the 2–5% of breast cancer that expresses in multiple-case families; other, more weakly expressing genes involved in familial breast cancer remain
OCR for page 86
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2 to be uncovered (Nathanson and Weber 2001; Ponder 2001). However, epidemiologic evidence is highly suggestive of a more extensive genetic component to breast cancer risk (Peto and Mack 2000), and much effort is being expended to identify the functional gene polymorphisms that might be involved. Although some of the evidence remains controversial, Dunning and colleagues (1999) and Nathanson and Weber (2001) note the potential involvement of polymorphic genes that encode steroid hormone receptors and paracrine growth factors (e.g., AR, CYP19) together with genes involved in the metabolism of chemical species (e.g., GSTP1) and in DNA damage response (e.g., ATM, RAD51, TP53). The most persuasive evidence on breast cancer genes other than BRCA1 and BRCA2 concerns the cell cycle checkpoint kinase gene CHEK2. A truncating germline deletion of this gene is present in around 1% of healthy individuals and is estimated to result in about a twofold increase of breast cancer risk in women and about a tenfold increase in men (Meijers-Heijboer and others 2002). Two data sets have some association with cancer risk after radiation. First is the question of breast cancer risk in individuals who are heterozygous carriers of the ATM mutation of the highly radiosensitive disorder AT. ATM carriers (ATM+/−) might represent 0.25–1% of the general population, and there is evidence of modestly increased cellular radiosensitivity in ATM+/− genotypes. It is therefore reasonable to consider an increased risk of radiogenic breast cancer in these carriers. Considerable effort has been expended on molecular epidemiologic analysis of spontaneous breast cancer risk in ATM+/− women (Bishop and Hopper 1997; ICRP 1998; Broeks and others 2000; Laake and others 2000; Geoffroy-Perez and others 2001; Olsen and others 2001; Teraoka and others 2001). Although the position remains somewhat uncertain, it seems reasonable to conclude that while increased breast cancer risk may be associated with ATM+/− in some cohorts, the relative risk is likely to be modest (<3), and the overall impact on spontaneous breast cancer risk in the population is rather small. Some data suggest, however, that it is only certain dominant negative missense mutations of ATM that predispose to cancer (Khanna 2000; Chenevix-Trench and others 2002), and for these, the relative risk may be substantially higher. The critical question is whether the ATM+/− genotype may more specifically and significantly increase breast cancer risk after radiation. For good scientific reasons, some early claims on substantial risks at low doses are not regarded as being well founded (see ICRP 1998). While a modestly increased contribution of the ATM+/− genotype to radiogenic cancer risk should not be discounted, three recent studies on patients developing second cancers after RT argue against a major impact from the ATM gene (Nichols and others 1999; Broeks and others 2000; Shafman and others 2000). In total, these studies considered 141 patients with second cancers; the studies of Shafman and colleagues (2000) and Broeks and colleagues (2000) specifically considered a total of 89 second breast cancer cases. None of the cases studied carried ATM mutations. The second line of evidence concerns the inheritance of chromosomal radiosensitivity and its association with breast cancer risk (Roberts and others 1999). In brief, in studies on cultured blood lymphocytes, up to around 40% of unselected breast cancer cases were shown to exhibit an abnormal excess of chromatid aberrations following X-irradiation in the G2 phase of the cell cycle. By contrast, this chromosomal trait was seen in only around 5% of age-matched controls. Follow-up family studies provided evidence on the heritability of the trait, which, although not of a simple Mendelian form, could be genetically modeled. As yet there is no evidence on the specific genes involved. In summary, advances in breast cancer genetics do allow the construction of a general scheme to describe the interactive genetic component of familial risk, including some allowance for common genes of low penetrance (Ponder 2001). Polygenic computational models describing the overall genetic component of spontaneous breast cancer risk in the population are also under development (Antoniou and others 2002). Although gene candidates and cellular phenotypes may prove to be instructive, there is at present little to guide specific conclusions on the question of the common genetic component of radiation-associated cancer risk. The evidence available would tend to argue against a major overall impact on radiation breast cancer risk from the ATM gene in its heterozygous form, although specific ATM genotypes may, in principle, carry substantially increased risk. Human Colonic and Other Neoplasms There is evidence that the genetic component of colonic cancer also includes a significant contribution from genes of low penetrance. In a recent review of 50 studies on the potential impact of common polymorphisms, Houlston and Tomlinson (2001) identified significant associations with risk for APC-I1307K, HRAS1-VNTR, and MTHFR-Val/Val. For TP53, NAT1, NAT2, GSTM1, GSTT1, and GSTP1 polymorphisms, the evidence was weaker. Specific data relating to gene polymorphisms and radiation risk are lacking although, as for breast cancer, there is some evidence of an association between colon cancer risk and lymphocyte chromosomal radiosensitivity (Baria and others 2001). Finally, in illustration of ongoing work, it is relevant to mention polymorphic associations between GSTP1 and chemotherapy-related leukemia (Allan and others 2001), MCUL1 and uterine fibroma (Alam and others 2001), GFRalpha1 and medullary thyroid carcinoma (Gimm and others 2001), PPARG and endometrial carcinoma (Smith and others 2001), and TP53 and adrenal cortical carcinoma (Ribeiro and others 2001). In their review of gene-environment interactions, Shields and Harris (2000) focus on lung cancer risk, and in this area, Bennett and colleagues (1999)
OCR for page 87
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2 have provided evidence on the potential impact of GSTM1 allelic status on tobacco-related lung cancer risk. The broad but incomplete picture that emerges from these studies is of some associations between gene polymorphisms and risk for a range of human tumor types, as well as the clear need for larger and more definitive studies. Human DNA Repair Gene Polymorphisms It has already been noted that DNA repair genes play a crucial role in cellular responses to radiation and that major germline deficiencies in these genes can lead to heritable predisposition to cancer. Accordingly, considerable effort is being expended in the search for common functional polymorphisms that might act as low-penetrance cancer susceptibility genes. A series of studies have identified common and less common polymorphisms in around ten DNA repair genes, some of which appear to have cellular consequences (Price and others 1997; Shen and others 1998; Mohrenweiser and Jones 1998; Duell and others 2000). The associations between these polymorphisms and radiosensitivity and/or tumor risk remain unclear, although there are some positive indications (Duell and others 2001; Hu and others 2001). Much of this work has centered on genes involved in base- or nucleotide-excision repair (Miller and others 2001). Studies on genes controlling DNA DSB repair are less well developed. However, there are indications that a relatively common (in ~6% of the population) functional polymorphism in the XRCC2 gene of the homologous recombinational repair pathway for DNA DSBs associates with a modestly increased risk of breast cancer (Kuschel and others 2002; Rafii and others 2002). A significant association between breast cancer risk and certain polymorphisms of NHEJ DNA repair has also been reported (Fu and others 2003). A recent review of DNA repair gene polymorphisms and cancer risk recommends large, well-designed studies that include consideration of relevant exposures (Goode and others 2002). Genetic Studies with Animals The recognized difficulties of resolving the modifying effects of low-penetrance genes on human cancer risk have prompted experimental genetic studies with rodent models in which genetic-environmental interactions can be more closely controlled. This approach has been applied principally in mice for the study of naturally arising polymorphic variation that influences spontaneous cancer risk and the risk after exposure to chemical carcinogens and, in a few instances, ionizing radiation (Balmain and Nagase 1998). These studies have the capacity to provide proof-of-principle evidence of the impact of such common loci, together with their possible interactions and tissue specificity, as well as the classes of genes and mechanisms involved. Thus, although specific functional gene polymorphisms identified in mice may not predict those of humans precisely, the overall pattern of cancer risk modification should provide broad guidance on the potential for such effects in humans. Much of the research on the role of germline polymorphic loci in mouse tumorigenesis has centered on spontaneous and chemically induced neoplasms. These studies include tumors of the skin (e.g., Nagase and others 2001; Peissel and others 2001), lung (e.g., Lee and others 2001; Tripodis and others 2001), and intestinal tract (e.g., van Wezel and others 1996; Angel and others 2000). The most important messages to emerge from these studies are that multiple common loci can exert complex patterns of control over tumor susceptibility and resistance (synergistic and antagonistic interaction), that the loci tend to be relatively tissue specific in their activity, and that genetic determinants of spontaneous and induced tumorigenesis are often shared. A particularly revealing conclusion from the study of Tripodis and colleagues (2001) is that as many as 60 loci may interact to determine the risk of a single tumor type; specific pairwise interaction of a proportion of these loci was also demonstrated. A second approach used in mouse genetic studies is to seek evidence of natural polymorphic loci that modify the tumorigenic expression of a major cancer-predisposing germline mutation. In this way, evidence has been obtained for substantial genetic modification of tumorigenesis in Trp53- (Backlund and others 2001) and Apc-deficient mice (van der Houven van Oordt and others 1999; Moser and others 2001). In the case of Apc, one of these modifier genes (Pla2g2a) has been identified provisionally (Cormier and others 2000). In general, these effects of genetic modifiers are again consistent with the potential interaction of multiple tissue-specific loci, and some of the data relate to tumors induced by ionizing radiation. Some studies in this area have the specific objective of mapping and characterizing the polymorphic loci that influence tumorigenic radiosensitivity and tumor characteristics. Multiple loci have been shown to influence susceptibility to radiation-induced lymphoma and leukemia (Balmain and Nagase 1998; Szymanska and others 1999; Saito and others 2001; Santos and others 2001). One study of Boulton and colleagues (2001) provided evidence that the AML loci determining leukemia or lymphoma susceptibility were distinct from those that influenced genomic instability in bone marrow cells. However, no candidate genes were identified. Genetic loci influencing the susceptibility of mice to α-particle (227Th)-induced osteosarcoma have also been mapped (Rosemann and others 2002), but again, no candidate genes were specifically identified. By contrast, another set of investigations has associated a strain-specific functional polymorphism of the gene Prkdc encoding DNA PKcs with induced genomic instability, DNA DSB repair deficiency, and susceptibility to radiation-induced breast cancer (Okayasu and others 2000; Yu and oth-
OCR for page 88
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2 ers 2001). This same Prkdc polymorphism has also been implicated in radiation-induced lymphomagenesis, as a modifier of induced intestinal neoplasia in Apcmin mice (Degg and others 2003), and as a candidate gene for the Rapop1 apoptosis-controlling locus (Mori and others 2001). Other tissue-specific loci that control apoptosis have also been genomically mapped (e.g., Weil and others 2001). With respect to breast cancer susceptibility in mice, it is already clear that loci other than Prkdc can be involved (Moser and others 2001). From recent studies, it seems likely that one such gene is ATM, which in the heterozygous form can enhance the frequency of both genomic instability and ductal dysplasia of the breast of irradiated mice (Weil and others 2001). Conclusions Although much remains to be learned about genetic susceptibility to the tumorigenic effects of radiation, it is possible to frame some interim conclusions of the role it may play in determining radiation cancer risk at the individual and population levels. The principal point to emphasize is that cancer is a multifactorial set of diseases, and as such, there is expected to be a complex interplay between multiple germline genes and a plethora of other host- and environment-related factors. The data available, although far from complete, tend to support this basic expectation. The key issues and arguments are given here in brief summary. For rare major gene deficiencies in humans and mice, there can be strong effects on radiation cancer risk, and for individual carriers, it seems likely that the greatest implications may be for the risk of second cancers after RT (see ICRP 1998). Although the data are sparse, such high-dose radiation exposure in childhood may carry the greatest risk. However, due to differences in genetic background, a uniformity of tumorigenic response in RT patients with major gene deficiencies should not be expected. The fact that strongly expressing cancer-prone disorders are so rare argues against a significant impact and distorting effect on estimates of cancer risk in irradiated populations; population genetic modeling fully supports this view (see ICRP 1998). By contrast, at the level of whole populations it is feasible that certain inherited combinations of common low-penetrance genes can result in the presence of subpopulations having significantly different susceptibilities to spontaneous and radiation-associated cancer. In due course, the accumulation of sufficient molecular epidemiologic data may allow for some meaningful theoretical modeling of the distribution of radiation cancer risk and the possible implications for radiological protection. Irrespective of such modeling, risk estimates based on epidemiologic evaluation of whole populations will encompass this projected genetic heterogeneity of response. Therefore, the key issue is not whether the estimate of overall cancer risk is genetically confounded, but rather the extent to which genetic distortion of the distribution of this risk might lead to underprotection of an appreciable fraction of the population. In this respect, some initial guidance for thought is already available from the data discussed in this chapter. These data suggest large numbers of loci of low penetrance with relatively small individual effects and a significant degree of locus-specific interaction and tissue specificity that may apply to their activity. Projecting this scenario to a range of radiogenic tumors in a genetically heterogeneous human population would tend to lead to a situation in which the balance between a certain set of tumor susceptibility (S) and resistance (R) loci in a given subgroup might serve to emphasize risk in a given set of organs. Equally, however, the balance of additional S and R locus combinations might provide a degree of resistance to the induction and development of cancer in other organs. Thus, with this first genetic scenario, major distortions of the distribution of overall cancer risk after radiation might not apply simply because different genetic susceptibilities would tend to “average out” across organs. By contrast, a second hypothetical scenario involves a small subset of common polymorphic loci that exert organ-wide effects on tumor susceptibility or resistance, which might be particularly strong in the specific instance of radiation exposure (e.g., functional polymorphisms for genes involved in initial tissue-wide cellular response to radiation damage). In this instance, genetically determined distortion of the distribution of overall cancer risk might be expected. At present, the data available are insufficient to distinguish the likely contributions from these two genetic scenarios. Finally, the large study of cancer concordance in 90,000 Nordic twin pairs should be noted. Lichtenstein and colleagues (2000) and Hoover (2000) make some important points about the difficulties that exist in separating the genetic and environmental components of cancer. In essence, Hoover notes that this Nordic study, like others, is consistent with the presence of low-penetrance cancer-predisposing genes in the general population. However, the confidence intervals for the heritable component of cancers at common sites were wide—all ranged from around 5 to 50%. It was also pointed out that for cancer at common sites, the rate of concordance in monozygotic twins was generally less than 15%. Thus, the absolute risk of concordance of site-specific cancer in identical genotypes sharing some common environmental factors is rather low. In addition to this, a study based on the Swedish Family Cancer Database (Czene and others 2002) has provided further information on the genetic component of organ-specific cancer. With the exception of the thyroid, the environment appears to have the principal causal role for cancer at all sites. One important message that emerges from current data on cancer genes of low penetrance and the overall genetic component of cancer is that predictive genotyping of individuals for the purposes of radiological protection may not
OCR for page 89
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2 be feasible in the medium term. The likely involvement of multiple and relatively organ-specific sets of polymorphisms and gene-gene or gene-environment interactions makes the provision of meaningful judgments on risk most uncertain. For these reasons it may be more realistic at this stage of knowledge to focus attention on general patterns of gene-radiation interactions and their implications for population risk, rather than risk for specific individuals. SUMMARY In this chapter, the committee has reviewed cellular-molecular and animal studies relevant to the complex multistage process of radiation tumorigenesis. Attention has also been given to evidence from various studies on the inherited factors that influence radiation cancer risks. The principal objective of this work was to provide judgments on radiation cancer risk of prime importance to radiological protection, particularly where these judgments serve to couple information about the action of radiation on cells (Chapters 1 and 2) with the epidemiologic measures of risk considered in subsequent chapters. Mechanisms of Radiation Tumorigenesis A critical conclusion on mechanisms of radiation tumorigenesis is that the data reviewed greatly strengthen the view that there are intimate links between the dose-dependent induction of DNA damage in cells, the appearance of gene or chromosomal mutations through DNA damage misrepair, and the development of cancer. Although less well established, the data available point toward a single-cell (monoclonal) origin for induced tumors and indicate that low-dose radiation acts predominantly as a tumor-initiating agent. These data also provide some evidence on candidate, radiation-associated mutations in tumors. These mutations are predominantly loss-of-function DNA deletions, some of which are represented as segmental loss of chromosomal material (i.e., multigene deletions). This form of tumorigenic mechanism is broadly consistent with the more firmly established in vitro processes of DNA damage response and mutagenesis considered in Chapters 1 and 2. Thus, if as judged in Chapters 1 and 2, error-prone repair of chemically complex DNA double-strand damage is the predominant mechanism for radiation-induced gene or chromosomal injury involved in the carcinogenic process, there can be no expectation of a low-dose threshold for the mutagenic component of radiation cancer risk. One mechanistic caveat explored was that novel forms of cellular damage response, collectively termed induced genomic instability, might contribute significantly to radiation cancer risk. The cellular data reviewed in Chapter 2 identified uncertainties and some inconsistencies in the expression of this multifaceted phenomenon. However, telomere-associated mechanisms did provide a coherent explanation for some in vitro manifestations of induced genomic instability. The data considered in this chapter did not reveal consistent evidence for the involvement of induced genomic instability in radiation tumorigenesis, although telomere-associated processes may account for some tumorigenic phenotypes. A further conclusion was that there is little evidence of specific tumorigenic signatures of radiation causation, but rather that radiation-induced tumors develop in a tumor-specific multistage manner that parallels that of tumors arising spontaneously. However, further cytogenetic and molecular genetic studies are needed to reduce current uncertainties about the specific role of radiation in multistage radiation tumorigenesis; such investigations would include studies with radiation-associated tumors of humans and experimental animals. Quantitative Studies of Experimental Tumorigenesis Quantitative animal data on dose-response relationships provide a complex picture for low-LET radiation, with some tumor types showing linear or linear-quadratic relationships while other studies are suggestive of a low-dose threshold, particularly for thymic lymphoma and ovarian cancer. However, since the induction or development of these two cancer types is believed to proceed via atypical mechanisms involving cell killing, it was judged that the threshold-like responses observed should not be generalized. Radiation-induced life shortening in mice is largely a reflection of cancer mortality, and the data reviewed generally support the concept of a linear dose-response at low doses and low dose rates. Other dose-response data for animal tumorigenesis, together with cellular data, contributed to the judgments developed in Chapters 10 and 12 on the choice of a DDREF for use in the interpretation of epidemiologic information on cancer risk. Adaptive responses for radiation tumorigenesis have been investigated in quantitative animal studies, and recent information is suggestive of adaptive processes that increase tumor latency but not lifetime risk. However, these data are difficult to interpret, and the implications for radiological protection remain most uncertain. Genetic Susceptibility to Radiation-Induced Cancer The review of cellular, animal, and epidemiologic or clinical studies on the role of genetic factors in radiation tumorigenesis shows that there have been major advances in understanding, albeit with some important knowledge gaps. An important conclusion is that many of the known, strongly expressing, cancer-prone human genetic disorders are likely to show an elevated risk of radiation-induced cancer, probably with a high degree of organ specificity. Cellular and animal studies suggest that the molecular mechanisms underlying these genetically determined radiation effects largely mirror those that apply to spontaneous tumorigenesis
OCR for page 90
Health Risks from Exposure to Low Levels of Ionizing Radiation: Beir VII Phase 2 and are consistent with knowledge of the somatic mechanisms of tumorigenesis reviewed earlier in this chapter. In particular, evidence has been obtained that major deficiencies in DNA damage response and tumor-suppressor-type genes can serve to elevate radiation cancer risk. Limited epidemiologic data from follow-up of second cancers in gene carriers receiving radiotherapy were supportive of the above conclusions, but quantitative judgments about the degree of increased cancer risk remain uncertain. However, since major germline deficiencies in the genes of interest are known to be rare, it is possible to conclude from published analyses that they are most unlikely to create a significant distortion of population-based estimates of cancer risk. The major practical issue associated with these strongly expressing cancer genes is judged to be the risk of radiotherapy-related cancer. A major theme developing in the whole field of cancer genetics is the interaction and potential impact of more weakly expressing variant cancer genes that may be relatively common in human populations. Knowledge of such gene-gene and gene-environment interactions, although at an early stage, is developing rapidly. The animal genetic data reviewed in this chapter provide proof-of-principle evidence of how such variant genes with functional polymorphisms can influence cancer risk, including limited data on radiation tumorigenesis. Attention has also been given to recent molecular epidemiology data on associations between functional polymorphisms and cancer risk, particularly with respect to DNA damage response genes. Some issues of study design have been discussed, and although much work has been reported on cancer risk in heterozygous carriers of the ATM gene, clear judgments about radiation risks remain elusive. Given that functional gene polymorphisms associated with cancer risk may be relatively common, the potential for significant distortion of population-based risk was explored, with emphasis on the organ specificity of the genes of interest. A preliminary conclusion is that common polymorphisms of DNA damage response genes associated with organ-wide radiation cancer risk would be the most likely source of major interindividual differences in radiation response. Although good progress is being made, there are important gaps in understanding the extent of genetic influences on radiation cancer risk. Accordingly, further work is needed in humans and mice on gene mutations and functional polymorphisms that influence radiation response and cancer risk. Human molecular genetic studies should, where possible, be coupled with epidemiologic investigations.
Representative terms from entire chapter: