Low-penetrance genetic susceptibility and resistance loci implicated in the relative risk for radiation-induced acute myeloid leukemia in mice

Cancer is the result of the accumulation of genetic lesions (mutations) in a single cell over time. The lifetime relative risk for cancer therefore increases with age and is determined by the mutation rate, the number of genetic lesions required for malignant transformation, the number of target cells, and intercurrent mortality. Exposure to a carcinogen or the inactivation of gene(s) involved in maintaining genetic stability increases the mutation rate and the probability of malignant transformation. There is also a genetic component that determines relative susceptibility to de novo or induced cancer, and though this is exemplified in the germline transmission of rare, highly penetrant alleles in familial cancer predisposition syndromes, there is increasing evidence that a high proportion of cancers arise in a susceptible subpopulation that carries low-penetrance gene(s).1-3 

Studies of Japanese atomic bomb survivors and of radiotherapy and chemotherapy patients have shown that the main long-term health consequence of exposure is acute myeloid leukemia (AML) and myelodysplasia (MDS). Radiation- and therapy-related acute myeloid leukemias (r-AML and t-AML) are clonal malignancies that arise in most cases from a multipotent hemopoietic stem cell (HSC) and commonly exhibit allelic loss (5q− or monosomy 7), and 10% to 15% of newly diagnosed cases of 5q− or monosomy 7 AML/MDS arise in patients with a history of chemotherapy or radiotherapy.4-8 The extensive use of combination chemotherapy/radiotherapy and/or autologous stem cell transplantation has increased long-term cancer survival, but this has resulted in a 1.5% to 24% cumulative risk for secondary t-AML up to 10 years after treatment, depending on the primary cancer and therapy.5,7-10 Efforts to elucidate the genetic factors that modulate susceptibility to therapy-related AML have concentrated on enzymes involved in carcinogen detoxification, and polymorphic variations in the genes encoding NAD(P)H:quinone oxidoreductase, cytochrome P450 3A4, epoxide hydrolase, and glutathione S-transferase, found to be associated with susceptibility to induced AML.11-15 

r-AML in inbred CBA/H mice is widely considered the most appropriate mouse model of human r-AML. Mouse r-AMLs arise after a long latency (mean latency, 480 days after irradiation), indicating that malignant transformation of the target HSC is a multistage process and there is a low (less than 0.1%) incidence of de novo AML in control CBA/H mice.16,17 A curvilinear radiation dose–response curve in mice and humans is composed of an initial increase in AML risk according to (dose),2 followed by a decrease at higher doses,6,16 and is presumably the net result of 2 opposing dose-dependent processes—mutation induction and cell death. In contrast, inbred C57BL/6 mice are susceptible to radiation-induced thymic lymphoma, a thymus-dependent T-cell malignancy that has no human counterpart,18 and C57BL/6 mice can be considered to be resistant to r-AML.

The differences in susceptibility to r-AML in the CBA/H and C57BL/6 inbred mouse strains can be exploited in genetic linkage analyses to map r-AML susceptibility or resistance loci as the first step in the identification of the gene(s) involved. The lifetime incidence of r-AML in 3 Gy X-irradiated CBA/H mice is approximately 20%, but the incidence of r-AML in 3 Gy X-irradiated (CBA/H × C57BL/6)F1 and F1 backcross and F1 intercross mice is approximately 7%.16 17 r-AML susceptibility is a complex polygenic and partially dominant mouse genetic trait, and the risk following exposure is dependent on genetic background. We have therefore carried out a genome-wide screen for linkage between phenotype (r-AML) and genotype in affected and unaffected 3 Gy X-irradiated (CBA/H × C57BL/6)F1 × CBA/H backcross mice and identified at least 2 suggestive loci on chromosomes 1 and 6. Compared with the (CBA/H × C57BL/6)F1 × CBA/H backcross cohort as a whole, the loci individually increase the relative risk for r-AML by approximately 2-fold, suggesting that r-AML susceptibility in mice is determined by at least 2 low-penetrance genetic loci in the inbred mouse genetic backgrounds used in this study.

Because the long-term repopulating HSC (LT-HSC) is the target cell in radiation-leukemogenesis, the susceptibility/resistance gene(s) products must either have a role in the regulation of LT-HSCs or their response to ionizing radiation, or they must confer a proliferative advantage during the radiation leukemogenic process. An excess of chromosome 6 aberrations in r-AMLs and irradiated bone marrow cells implicates the locus in preleukemic cell proliferation. The stem cell frequency regulator 1 (Scfr1) locus maps to the r-AML–susceptibility locus on distal chromosome 1, so genetically determined target size (LT-HSC numbers) may be one factor in the susceptibility to induced AML.

CBA/H and C57BL/6 mice were from the Harwell colony, and DBA/2 and BALB/c mice were from Harlan UK Limited. Eight- to 12-week-old (CBA/H × C57BL/6)F1 × CBA/H mice were exposed to a single acute dose of 3.0 Gy x-rays at 0.5 Gy/min (constant potential, 250 kV; 1.2 mm Cu). Animal studies were carried out under guidance issued by the Medical Research Council in its publication, Responsibility in the Use of Animals for Medical Research (July 1993), and Home Office Project license no. PPL 30/689 and 30/1272.

Moribund mice were killed by cervical dislocation. Blood and bone marrow smears were taken for microscopic analyses, and tails, leukemic spleens, and kidneys were snap frozen. DNA was prepared from control (tail, kidney, or both) and leukemic spleen essentially as described.17 19 

Myeloid leukemias were initially diagnosed by leukemic blood cell morphology, and leukemic spleen was then further analyzed for immunoglobulin heavy chain (IgH) gene rearrangements and for the expression of lineage-specific/restricted markers to identify r-AML and early pre-B lymphomyeloid (L-ML) leukemias.17 19 

Microsatellite primer sequences were from the Whitehead Institute,20 and genetic positions were from the 2000 Chromosome Committee Reports (Jackson Laboratory21). Five hundred eighty-seven microsatellite markers across the 19 mouse autosomes were screened for informative polymorphisms in the inbred CBA/H and C57BL/6 mice. One hundred thirty-eight were identified to give a genome-wide screen at approximately 20–centimorgan (cM) intervals. Statistical significance of excess heterozygosity or homozygosity at individual microsatellite markers in the affected mice was evaluated using the χ²test for homogeneity and was compared with unaffected mice using the χ² test of independence.22 Relative risk (RR) was estimated essentially as described.23 

The lifetime incidence of r-AML in 3 Gy X-irradiated CBA/H mice is approximately 20%, whereas it is undetectable in X-irradiated C57BL/6 mice.16-18 We have previously shown that there is no statistically significant difference (P = .3) in the proportions of mice with r-AML in irradiated (CBA/H × C57BL/6)F1, F1 backcross, and F1 intercross mice (approximately 7% lifetime incidence), but they all differ significantly (P = .0000013) from the r-AML incidence in the parental CBA/H strain.17 To further elucidate the complex genetics underlying r-AML susceptibility in CBA/H mice, 1087 (CBA/H × C57BL/6)F1 × CBA/H mice (F1 × CBA/H) were exposed to 3 Gy x-rays, and causes of death were determined over their lifetimes. Sixty-seven r-AMLs were diagnosed.

Fifty-eight to 67 affected mice were genotyped in a genome-wide screen of the 19 autosomes at intervals smaller than 22 cM using polymorphic chromosome-specific microsatellite markers (Table1).

Sex chromosomes were not analyzed because susceptibility to r-AML is not sex linked.17 In the first instance, excess homozygosity or heterozygosity was estimated using the χ²test of homogeneity, and genetic markers that gave a χ²value of less than 1.0 were excluded. Genetic markers that gave a χ² value of more than 1.0 were examined in further detail by genotyping all affected mice at the marker and at additional markers on either side. As illustrated in Table 1, linkage between genotype and phenotype was observed with more than 99.9% pointwise probability on chromosomes 1 (excess homozygosity) and 6 (excess heterozygosity), and weaker linkage (95%-99% probability) was observed on chromosomes 4 (excess heterozygosity) and chromosomes 2 and 13 (excess homozygosity). However, only the chromosome 1 and 6 loci meet the criterion for suggestive linkage in a backcross analysis.24 

Two hundred eighty-eight to 357 mice of the 1020 irradiated F1 × CBA/H cohort that died of other causes (DOC) were also genotyped at the chromosomal intervals of interest to exclude segregation distortion and to test for the enrichment of specific genotypes in the affected mice using the χ² test of independence. As illustrated in Table 2, a comparison of the affected and unaffected (DOC) mice reveals more than 99% confidence of linkage on chromosomes 1 (D1Mit150) and 6 (D6Mit384).

The excess homozygosity detected on chromosome 2 (D2Mit237) in the affected mice was not significantly different than the D2Mit237 genotype of 272 unaffected mice (P = .28), so linkage to proximal chromosome 2 is excluded (data not shown). Comparison of the affected and unaffected mice using the χ² test of independence indicates that linkage at chromosomes 4 (D4Mit292;P = .021) and 13 (D13Mit248; P = .043) cannot be excluded. Given the weak penetrance of r-AML (approximately 7% lifetime incidence) in the F1 and F1 × CBA/H backcross mice and the relatively low numbers of affected mice analyzed, the statistical significance achieved is remarkable, with minimum P values occurring at D1Mit150 and D6Mit384 (Table 2).

Neither locus is essential for r-AML induction in the backcross mice. For example, compared with the 67 affected F1 × CBA/H mice, the RR of r-AML in the F1 × CBA/H mice that are homozygous at D1Mit150 on chromosome 1 is 2.17 (P = .0072), and the RR for excess heterozygosity on chromosome 6 is 2.75 (P = .00046). To determine whether each locus contributes additively to r-AML risk, the 4 genotype combinations for the most informative loci on chromosomes 1 (D1Mit150; 1^hom or 1^het) and 6 (D6Mit384; 6^hom or 6^het) were assessed in the affected mice and in 322 mice that died of other causes. As illustrated in Table3, the χ² test of homogeneity reveals no statistically significant deviation from the 1:1:1:1 ratio expected for the 4 genotype combinations in unaffected DOC mice (P = .57). In contrast, the affected mice do show a statistically significant deviation from the expected (P = .00013). The χ² test of independence reveals a statistically significant difference of the relative proportions of the 4 genotypes in the affected mice compared with the unaffected mice (χ² = 18.64; df = 3;P = .000325), and RR estimates indicate that this statistical significance is driven by enrichment of the most favored 1^hom6^het genotype (RR = 2.99) and by depletion of the least favored 1^het6^homgenotype (RR = 0.27) in the affected mice (Table 3).

During the course of these studies, 54 early pre-B mixed-lineage lymphomyeloid leukemias (L-MLs) were diagnosed19 in the irradiated backcross mice. The r-AML and L-ML loss of heterozygosity profiles on chromosomes 2 and 4 differ,19 suggesting that they represent 2 distinct hemopoietic malignancies, and mice affected with L-ML are included in the unaffected mice that died of other causes (Tables 2-3). However, further evidence to support the proposition that 2 distinct leukemogenic processes are involved can be inferred when the 4 genotype combinations for the most informative loci on chromosomes 1 (D1Mit150; 1^hom or 1^het) and 6 (D6Mit384; 6^homor 6^het) were compared in r-AML– and L-ML–affected mice. As shown in Table 4, the χ²test of homogeneity reveals no statistically significant deviation from the 1:1:1:1 ratio expected for the 4 genotype combinations in L-ML–affected mice (P = .28), and the χ²test of independence reveals a statistically significant difference of the relative proportions of the 4 genotypes in the r-AML– and L-ML–affected mice (χ² = 14.6; df = 3;P = .0022). RR estimates indicate that this statistical significance is again driven by the enrichment of the most favored r-AML 1^hom6^het genotype (RR = 3.01) and depletion of the least favored r-AML 1^het6^homgenotype (RR = 0.176) in the r-AML–affected mice, further evidence that 2 distinct leukemogenic processes are involved.

Because the lifetime incidence of r-AML in the F1 × CBA/H mice as a whole is 6.7% and the F1 × CBA/H subpopulation with the most favored 1^hom6^het genotype has a 2.99 RR, the incidence of r-AML in the hybrid 1^hom6^hetgenotype F1 × CBA/H mouse subpopulation is similar to the 20% lifetime r-AML incidence in the parental inbred CBA/H strain.16 17 

The r-AML–susceptible CBA/H and resistant C57BL/6 inbred mouse strains have been exploited in a genome-wide genetic linkage analysis of (CBA/H × C57BL/6)F1 × CBA/H backcross mice to identify at least 2 suggestive loci (chromosomes 1 and 6), each of which makes a modest approximately 2-fold contribution to the RR of r-AML in the (CBA/H × C57BL/6)FI × CBA/H backcross mice following exposure to 3 Gy x-rays. Together, a RR of 3 for both genotypes accounts for the difference in the lifetime r-AML incidence in the irradiated F1 × CBA/H backcross cohort as a whole (6.7% incidence) and in the irradiated parental inbred CBA/H mouse strain (20% incidence).

Although the results of the approximately 1500 cM genetic linkage analysis has focused attention on 2 suggestive, approximately 7-cM chromosomal intervals, it is self-evident that neither is necessary nor sufficient for r-AML and that other low-penetrance genes are clearly involved. The statistical power of the current study of (CBA/H × C57BL/6)F1 × CBA/H backcross mice is constrained by the low (approximately 7%) lifetime incidence of r-AML in the 1037 X-irradiated mice analyzed. For example, there is weak evidence that loci on chromosomes 4 and 13 may also be involved (Table 1), but the statistical power is insufficient to satisfy the accepted criterion for suggestive linkage.24 Nevertheless, the intervals on mouse chromosomes 1, 6, and 13, defined by the minimum P values in the χ² test of homogeneity (Tables 1, 2, and data not shown) are syntenic with human chromosomes 1q41-42, 7p21, and 5q23-35, respectively,21 an intriguing observation given that induced AML in humans commonly involves monosomy 7, 5q−, or both.4,5 Furthermore, because there is no statistically significant difference in the incidence of r-AML in 3 Gy X-irradiated (CBA/H × C57BL/6)F1 × CBA/H and (CBA/H × C57BL/6)F1 × C57BL/6 backcross mice,17 additional loci should be revealed in a genome-wide genetic linkage analysis of affected and unaffected (CBA/H × C57BL/6)F1 × C57BL/6 backcross mice.

Although each interval contains more than 100 genes, there must be a difference in the biologic activity of the susceptibility/resistance gene product(s) in the r-AML–susceptible and the r-AML–resistant mouse strains. Our current understanding of the radiation myeloid leukemogenic process suggests that the gene product(s) probably has a role in the regulation of the target LT-HSC, the cellular response to ionizing radiation and/or confer a proliferative advantage to the preleukemic LT-HSC. Furthermore, r-AML susceptibility/resistance loci do not appear to be involved in the leukemogenic process leading to mixed-lineage early to pre-B lymphomyeloid leukemia in the same irradiated backcross mice.

The stem cell frequency 1 locus (Scfr1), which determines the frequency of bone marrow long-term culture-initiating cells (LTC-IC) in inbred mice, maps to the interval between D1Mit113 (90 cM) and D1Mit17 (113 cM),25 a significant overlap with the 92.3 to 100 cM chromosome 1 interval defined by D1Mit111 and D1Mit150, which have the minimum P values in the genetic linkage analysis (Table 2). Because LTC-ICs are phenotypically and functionally indistinguishable from LT-HSCs25 and are therefore the target cell in r-AML, Scfr1 is an obvious candidate. Independent studies have shown that adult C57BL/6 mice have lower bone marrow LT-HSC numbers than other inbred mouse strains. For example, AKR/J mice have 5 times more Lin⁻Thy-1^low Sca-1⁺c-kit⁺bone marrow cells than C57BL/6 mice26; DBA/2 mice have 11 times more LTC-IC bone marrow cells than C57BL/6 mice25; and 8-week-old DBA mice have 2.75 times more day 35 cobblestone area-forming cells (CAFCs) than C57BL/6 mice.27,28 Given that the AML-susceptible inbred CBA/H mice used in this study are closely related to DBA/2 mice,29 which have higher numbers of LTC-ICs and day 35 CAFCs than the AML-resistant C57BL/6 strain,25,27,28 modest differences in stem cell numbers may contribute to the relative risk for induced AML. It is highly likely that the size of the stem cell compartment is regulated by multiple genes in addition to Scfr1 and that differences in cell survival and radiosensitivity following exposure to 3 Gy x-rays are a further potential factor in determining effective target size in the subsequent multistage leukemogenic process. Interindividual variation in bone marrow CD34⁺ stem and granulocyte macrophage–colony-forming unit (CFU-GM) progenitor cell numbers has been reported in humans,30-33 and though some of the variation observed may be caused by technical difficulties in the assays and the assays may not necessarily accurately reflect target LT-HSC numbers, the magnitude of the interindividual variation observed suggests that target size does vary within the human population.

The excess heterozygosity detected on chromosome 6 (Table 2) suggests the presence of a CBA/H r-AML resistance locus. However, loss of heterozygosity (LOH) studies did not detect allelic loss on chromosome 6 in r-AMLs34 (and data not shown), indicating that the locus does not appear to be subject to the inactivation processes normally associated with tumor-suppressor genes. More than 100 r-AMLs that arose in irradiated inbred CBA, C3H, RFM, and B6C3F1 genetic backgrounds have been analyzed for clonal chromosomal abnormalities by conventional G-banding or fluorescence in situ hybridization (FISH).35-39 As summarized in Table 1, allelic loss (terminal or interstitial deletions) on chromosome 2 is observed in more than 95% of r-AMLs, and aberrations on chromosome 6 represent the next most frequent clonal aberration (33.5%), followed by chromosome 15 (22%). Significantly, trisomy 6 (15%) is observed at a greater than 2-fold higher frequency than trisomy 15 (7.1%) or 1 (6%) and at a greater than 3.5-fold higher incidence than trisomy of any of the other 15 autosomes. Chromosomal abnormalities involving chromosome 15, and trisomy in particular, are frequently observed in mouse radiation-induced thymic lymphomas and have been implicated in tumor progression.40 In addition to the relative excess of chromosome 6 aberrations in the r-AMLs, an excess of CBA/H mouse bone marrow cells carrying aberrations of chromosomes 2 or 6 have also been reported in mice up to 6 months after exposure to 3 Gy x-rays,41 suggesting that these aberrations confer a proliferative advantage in vivo. This is consistent with the genetic analyses that indicate that though excess heterozygosity on chromosome 6 is neither necessary nor sufficient for radiation leukemogenesis, it does increase the RR of r-AML.

Ionizing radiation induces genetic instability in the clonal descendants of a single irradiated cell that persists for many cell generations in vitro and in vivo, so exposure potentially results in an elevated mutation rate.42,43 Radiation-induced genetic instability is induced at a higher efficiency in the clonal descendants of irradiated short-term repopulating HSCs (ST-HSCs) from the r-AML–susceptible CBA/H mouse than in ST-HSC from the r-AML–resistant C57BL/6 mouse, and it appears to be a recessive CBA/H genetic trait because it is induced at the lower efficiency in the (CBA/H × C57BL/6)F1 hybrid.44 A modest increase in mutation rate caused by a genetically determined susceptibility to radiation-induced genetic instability could result in a modest increase in the risk for r-AML without its being absolutely necessary for the radiation–leukemogenic process. Significantly, a genetic component to radiation-induced genetic instability has also been observed in human ST-HSC45 and ongoing chromosomal instability detected in mouse and human r-AML.36 46 

The genetic data presented here are further supportive17evidence that r-AML susceptibility in the CBA/H mouse model is a complex polygenic trait involving low-penetrance resistance and susceptibility loci. Because this is consistent with the proposal that most cancers and leukemias arise in a susceptible human subpopulation,3 the CBA/H mouse model may be more relevant to human-induced AML than might have been expected. Furthermore, the colocalization of the suggestive chromosome 1 r-AML susceptibility locus and Scfr1 raises the possibility that in addition to the number of genetic lesions required for malignant transformation and the mutation rate, target cell numbers may be part of the equation defining target size in oncogenesis. This is a provocative but testable hypothesis, and it has potential implications in our understanding of low-penetrance genetic risk factors.