* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download Complexity of Lung Cancer Modifiers: Mapping of
Survey
Document related concepts
Nutriepigenomics wikipedia , lookup
Microevolution wikipedia , lookup
Polycomb Group Proteins and Cancer wikipedia , lookup
Genomic imprinting wikipedia , lookup
Heritability of IQ wikipedia , lookup
Dominance (genetics) wikipedia , lookup
Minimal genome wikipedia , lookup
Site-specific recombinase technology wikipedia , lookup
Genome evolution wikipedia , lookup
History of genetic engineering wikipedia , lookup
Population genetics wikipedia , lookup
Pathogenomics wikipedia , lookup
Genome (book) wikipedia , lookup
Oncogenomics wikipedia , lookup
Transcript
Complexity of Lung Cancer Modifiers: Mapping of Thirty Genes and Twenty-five Interactions in Half of the Mouse Genome Nikos Tripodis, Augustinus A. M. Hart, Remond J. A. Fijneman, Peter Demant Background: Numerous low-penetrance genes control susceptibility to cancer in experimental animals, but the overall genetic information on this group of genes (i.e., number of loci and their mutual interactions) is missing. We performed a systematic search, scanning roughly half of the mouse genome for lung cancer susceptibility (Sluc) genes affecting tumor size or number by using mouse recombinant congenic (RC) strains. In each RC strain (OcB), approximately 12.5% of the genome is derived from the lung cancer-resistant strain B10.O20, whereas the rest is derived from the lung cancer-susceptible strain O20. Methods: A total of 730 F2 hybrids from five (OcB × O20) crosses were tested. Pregnant mice were treated on day 18 of gestation with a single dose of N-ethyl-N-nitrosourea. When offspring were 16 weeks old, whole lungs were removed and sectioned semiserially, and the size of all lung tumors (n = 2658) was determined. Analysis of variance was used for detection of linkage, and models (including main effect and two-way interactions) were tested with a statistical program. Results: We detected a total of 30 Sluc loci (16 new plus 14 previously reported) and 25 two-way interactions. Some of these interactions are counteracting (e.g., Sluc17 and Sluc20), resulting in the partial or total masking of the individual independent effect (main effect) of each involved locus. Seven loci (Sluc1, Sluc5, Sluc12, Sluc16, Sluc18, Sluc20, and Sluc26) and two interactions (Sluc5 × Sluc12 and Sluc5 × Sluc26) were detected in more than one RC strain. Conclusions: The extrapolation of our results to the whole genome suggests approximately 60 Sluc loci (90% confidence intervals = 42 to 78). Despite the genetic complexity of lung cancer, use of appropriate map1484 REPORTS ping strategies can identify a large number of responsible loci and can reveal their interactions. This study provides an insight into the genetic control of lung tumorigenesis and may serve as a paradigm for investigating the genetics of other cancer types. [J Natl Cancer Inst 2001;93:1484–91] Understanding genetic predisposition to nonfamilial (sporadic, common) cancer requires identification of the majority of involved genes. A recent extensive study on twins (1) indicated that genetic predisposition is responsible for an unexpectedly large proportion of apparently nonhereditary cancers, supporting the indications from numerous observations of familial clustering of specific cancers. This study, however, also indicated that the relevant genes do not have a strong effect and hence are difficult to identify. Lung cancer, the leading cause of cancer deaths worldwide (2), also exhibits familial clustering both in smokers and in nonsmokers (3), indicating the involvement of presently unknown genes. The comparison of a number of inbred mouse strains revealed large differences in susceptibility to both spontaneous and chemically induced lung tumors and resulted in the mapping of several susceptibility loci (4– 6) and their mutual interactions. Most of these susceptibility or modifier loci seem to influence the “intrinsic” cellular properties, such as growth, invasion, differentiation or vascularization potential or stimulation to appropriate immune responses, control of genomic stability, and DNA repair damage rather than carcinogen activation or inactivation (4). Little is known, however, about the total number of genes predisposing to lung cancer in humans or in animals. To estimate this number, we used the mapping efficiency of the mouse recombinant congenic (RC) strains to screen systematically for genes affecting number and size of lung tumors in the mouse genome (7–10). After N-ethyl-N-nitrosourea (ENU) treatment, the lung tumor-susceptible strain O20/A (denoted O20) develops statistically significantly larger tumors than the resistant strain B10.O20/Dem (denoted B10.O20) (5,8–10). RC strains (O20-congenic-B10.O20/Dem, abbreviated as OcB) have each only a random approximately 12.5% subset of genes derived from the lung tumor-resistant strain B10.O20, yet strains OcB-3, OcB-6, OcB9, and OcB-16 do not develop statistically significantly larger lung tumors than strain B10.O20 (10). On the other hand, strain OcB-4 is more susceptible and, like strain O20, is statistically significantly different from strain B10.O20. With these five OcB strains, it is possible to screen approximately half of the mouse genome for susceptibility to lung cancer genes. F2 crosses between each of these OcB strains and strain O20 were generated, and linkage analysis of susceptibility to lung cancer was performed. Originally, we mapped 14 lung cancer susceptibility (Sluc) loci (8–10) in strains OcB-4, OcB-6, and OcB-9 by using the MQM (i.e., multipleQTL [quantitative trait loci] model)mapping program (11–13). At the present time, higher computing power and advanced statistical packages allow for a more thorough analysis of these crosses. As a result, we now can simultaneously assay the possible main effects and pairwise interactions of markers from almost all segregating segments. MATERIALS AND METHODS Mice and Carcinogen Treatment Strain O20, the H-2 congenic strain B10.O20 that carries the H-2pz haplotype of strain O20 on the C57BL/10 background, and the OcB series of RC strains that consists of 19 homozygous strains are maintained in The Netherlands Cancer Institute, Amsterdam. The mice used in this study (7) were subjected to a strict light–dark regimen and received acidified drinking water and a standard laboratory diet (Hope Farms, Woerden, The Netherlands) ad libitum. F2 crosses were generated between the common background strain O20 and each of the following OcB strains: OcB-3, OcB-4, OcB-6, OcB-9, and OcB-16. At day 18 of gestation, pregnant F1 mice were treated intraperitoneally with 40 mg/kg body weight of ENU, as described previously (8,9). The ENU-treated offspring were killed at 16 weeks of age, and their whole lungs were removed. The animal treatment and the setup of the experiments have been approved by the Animal Experimentation Ethics Committee of The Netherlands Cancer Institute. Affiliations of authors: N. Tripodis, P. Demant (Division of Molecular Genetics, H5), A. A. M. Hart (Division of Radiotherapy), The Netherlands Cancer Institute, Amsterdam; R. J. A. Fijneman, Department of Cell Biology and Immunology, Faculty of Medicine, Vrije Universiteit, Amsterdam. Correspondence to: Peter Demant, M.D., Ph.D., Division of Molecular Genetics (H5), The Netherlands Cancer Institute, Plesmanlaan 121, 1066CX Amsterdam, The Netherlands (e-mail: demant@ nki.nl). See “Notes” following “References.” © Oxford University Press Journal of the National Cancer Institute, Vol. 93, No. 19, October 3, 2001 Phenotyping and Genotyping The whole lungs were fixed in AEF solution (i.e., 40% [vol/vol] ethanol, 5% [vol/vol] acetic acid, 4% [vol/vol] formaldehyde, and 0.41% NaCl) and embedded in histowax. The lungs were sectioned semiserially (5-m sections at 100-m intervals) and stained with hematoxylin–eosin. The size of the lung tumor was determined as described previously (8,9). Briefly, the tumor size is the sum of all measured surfaces (calculated with the aid of a grating in the ocular) in the semiserial sections where the tumor is present, and it corresponds to tumor volume. Tumors that did not exceed a diameter of 300 m in any of the sections were not included in the data (total number of tumors used ⳱ 2658). Variation in tumor size within each mouse was quite high in each cross. For example, in the OcB-4 cross, in the same mouse, tumors ranged in size from 8 × 106 m3 to 1380 × 106 m3, although, more frequently, the within-mouse tumor size ranged between 6 and 60 × 106 m3. Histologically, tumors were alveolar, papillary, or, most frequently, mixed alveolar and papillary, with relatively small mitotic figures and varying degrees of heterogeneity. The DNA of the F2 hybrid mice was isolated from their tails and genotyped with simple sequence-length polymorphic markers (Mouse MapPairs; Research Genetics, Huntsville, AL) (9,10). Marker positions are based on Version 3.2 of the mouse genome database (www. informatics.jax.org). The (O20 × OcB-6)F2 cross was used only for confirmation of loci mapped in the other crosses because of its small size (81 F2 mice were used in the analysis). The genomic segments that segregate in each F2 cross were identified previously (14,15). Statistical Analysis The chromosomal position of genes affecting tumor size in different chromosomal regions (with the use of individual microsatellite markers) was determined by analysis of variance. The size of the tumors was loge (natural logarithm) transformed (for the purpose of approximation to a normal distribution) and was nested per individual mouse. The effect of each marker, sex, and interactions between pairs (marker–marker and marker–sex) on tumor size was tested by the PROC GLM (general linear models) statement of the SAS® 6.12 for Windows (SAS Institute, Inc., Cary, NC). All statistical tests were two-sided. We use the term “susceptibility” to denote larger tumors and “resistance” to denote smaller tumors. The number of markers as well as sex screened for each F2 cross and the size of each cross are as follows: The OcB-3 cross (130 F2 animals) had seven markers (D1Mit170, D7Mit57, D8Mit3, D10Mit51, D13Mit139, D14Mit120, and D19Mit3); the OcB-4 cross (157 F2 animals) had 15 markers (D1Mit36, D2Mit5, D2Mit56, D4Mit4, D4Mit27, D6Mit158, D8Mit35, D9Mit2, D9Mit12, D10Mit28, D13Mit78, D15Mit13, D15Mit96, D15Nds3, and D18Mit17); the OcB-6 cross (81 F2 animals) had 10 markers (D1Mit16, D2Mit5, D2Mit200, D4Mit23, D4Mit66, Kras2, D12Nds2, D14Mit120, D16Mit9, and D19Mit3); the OcB-9 cross (193 F2 animals) had 13 markers (D2Mit56, D4Mit158, D6Mit218, D7Mit32, D7Nds2, D8Mit3, D10Mit122, D11Mit15, D16Mit9, D18Mit17, D19Mit61, D19Mit9, and D19Mit33); and the OcB-16 cross (169 F2 animals) had 14 markers (D1Mit221, D2Mit200, D4Mit5, D4Mit70, D7Mit55, D7Mit105, D8Mit3, D8Mit15, D10Mit122, D12Nds2, D15Mit13, D15Mit96, D16Mit19, and D18Mit7). Each known segregating segment in each cross is represented by at least one marker (or more if the segment is longer than 20 cM) The OcB-6 cross was used only for confirmation purposes in this study because of its small size. In each statistical model, we tried to include, in addition to all main effects (typically, approximately 14 markers and sex) from each segregating segment, as many two-way interactions as possible. Since it is not possible to include all two-way interactions in a single model (91 in the case of 14 markers) because of the limiting size of each cross, we first tested the effect of each marker and all of its two-way interactions with all other markers (13 interactions per marker in the case of a total of 14 markers). A backward-elimination procedure was followed to exclude statistically nonsignificant interactions (P>.05). In constructing the test model, typically, the five worst interactions from each marker are excluded, and all remaining interactions together with the main effects of all tested markers are combined. A backwardelimination procedure follows until all remaining pairs have statistical significance below 5%. In certain cases (the OcB-4 and the OcB-16 crosses), because of the large number of tested markers, one (in the OcB-4 cross) or two (in the OcB-16 cross) markers had to be left out when the test model was constructed. Therefore, two overlapping statistical models (in which a randomly selected marker and all of its interactions are substituted by the one left out in the other model) were used for the OcB-4 and three for the OcB-16 cross to have a more complete coverage of all segregating segments and a better representation of all possible interactions. The percentage of explained variance is approximately similar in all crosses (e.g., 39.3% for the OcB-3 cross), as is the goodness of fit (e.g., R2 for the OcB-3 cross is .315). The calculated P values were then corrected according to a formula recommended by Lander and Kruglyak (16). One of the variables in this formula is the length of the segregating genome. Theoretically, in each cross, approximately 200 cM is segregating (12.5% of the total genome length), but that differs in each OcB strain. To use the appropriate value, we calculated the total length of the segregating segments per OcB strain. For each segregating segment, the extreme pair of markers (one proximal and the other distal) that have a donor allele determines the length of the minimal donor segment. Similarly, the length of the maximal segment is that of the minimal segment plus the distance between the next closest pair of markers (one proximal and the other distal) that has the allele of the background strain. The maximal and minimal segregating donor segment lengths (first and second values, respectively, in centimorgan) per OcB tested are as follows: OcB-3 (183–109), OcB-4 (323–146), OcB-9 (291–168), and OcB-16 (302–158). The more stringent maximal values were used in the correction formula (since the corrected P value is proportional to the tested genome length). The number of statistical models performed per cross was also taken into consideration in a final correction by multiplying the corrected P value by the number of models used per cross. Journal of the National Cancer Institute, Vol. 93, No. 19, October 3, 2001 For confirmation, the statistical models included the tested locus and all of the other common loci between the original and the tested cross. One analysis per cross per tested locus was carried out. The pointwise P values from confirmatory analyses are reported as recommended previously (16). To compare the effect of a to-be-confirmed locus (A) between the cross in which it was originally mapped (the original cross) and the confirmatory cross, we used the following procedure: If locus A is involved in an interaction with locus B, but the interacting locus is not segregating in the confirmatory cross, then all of the mice in this cross are B°/°, i.e., homozygous for the O20 allele. In that case, the weighted mean of the original cross is recalculated only for all mice that have the B°/° genotype, with the use of mendelian frequencies (1 : 2 : 1), and the effect of locus A is then given as a percent difference from this weighted mean (therefore, these values are different from the ones in Table 1, which are derived from the cross mean). This weighted mean is then compared with the effect of locus A in the confirmation cross. RESULTS Using one or more statistical models per OcB cross to analyze the genetics of tumor size, we confirmed the previously detected lung tumor size Sluc loci (8–10) (data not shown), and we report 16 new lung cancer susceptibility loci, Sluc15 through Sluc30. All of the loci were detected in pairwise interactions, in which the effect of each locus depends on the genotype of the interacting locus (Table 1; Fig. 1, A and B). For example, the Sluc12b/b (homozygote for the B10.O20 allele) genotype is associated with susceptibility if the genotype at the interacting locus is Sluc22b/b and with resistance in Sluc22°/° (homozygote for the O20 allele) mice. Thus, the prediction of the phenotypic effect of a particular locus requires knowledge of any possible interactions that this locus may be involved in and determination of the allelic status of the interacting loci. A statistically significant main effect was observed only for Sluc20 (Table 1). The magnitude of the effects observed here is within the range of effects detected in other studies of mouse lung tumor susceptibility (17–19). Some genotypic combinations are associated with extremely susceptible phenotypes, such as Sluc5b/b Sluc25° / °, Sluc9° / °Sluc25° / °, Sluc9 b/ ° Sluc25°/°, Sluc10b/bSluc25°/°, Sluc10°/° Sluc25°/°, and Sluc18b/bSluc26b/b, which all result in a greater than fivefold increase in tumor size, compared with the mean size of the cross, whereas genotypes Sluc5 b/b Sluc16 b/ °, Sluc5 b/b Sluc25 b/b , Sluc5 b/b Sluc25 b/ °, Sluc5 b/ °Sluc25 b/b , Sluc5 b/b Sluc26° / °, Sluc9° / °Sluc25 b/b , REPORTS 1485 Table 1. Linkage data and estimated effects of the lung cancer susceptibility (Sluc) loci on lung tumor size** **The data are presented as percent deviations from the mean tumor size of each cross, adjusted for the effect of the remaining markers in the model (least-squares mean from the SAS output). The loge (i.e., natural logarithm)-transformed means of tumor size for the crosses (± standard deviation) are as follows: OcB-3, 3.15 (±0.69); OcB-4, 2.91 (±0.74); OcB-9, 2.86 (±0.52); and OcB-16, 3.24 (±0.70). Only statistically significant new linkage data are presented. Confirmatory results are in the text. The genotypes of loci A and B are b/b for homozygotes for the B10.O20 allele, b/o for heterozygotes, and o/o for homozygotes for the O20 allele. The Sluc2 locus was detected as a main effect dependent on the sex of mice. A statistically significant main effect for Sluc20 was also detected (shown in the last line). The (OcB × O20)F2 cross, in which loci A and B were detected, is shown in the second to the last column. The final corrected P value given in the last column was determined according to reference (16) after taking into consideration the number of models used per cross. *Loci A and B are interacting. †Effects on female mice. §Effects on male mice. Sluc10b/bSluc25b/b, Sluc10b/bSluc25b/°, and Sluc10°/°Sluc25b/b result in a greater than fivefold decrease in tumor size (Table 1). It is interesting that Sluc25 is 1486 REPORTS frequently involved in the most resistant (when at least one B10.O20-derived allele is present) and most susceptible (when homozygote for the O20 allele) pheno- types. Importantly, statistically significant linkage was also found for loci with weak effects (e.g., Sluc20 main effect), where the difference between the most extreme Journal of the National Cancer Institute, Vol. 93, No. 19, October 3, 2001 Fig. 1. Lung cancer susceptibility (Sluc) loci in the mouse. A) Schematic idiogram of the mouse genome and the location of the 30 Sluc loci detected by using the recombinant congenic strain system. The Kras2 and the H-2 loci are also included. The average minimal candidate Sluc locus region (± standard deviation) is 10.5 (±8) cM, and the average maximal region (± standard deviation) is 19.9 (±9.9) cM, and eight of these Sluc loci are mapped in regions of 12 cM or less. B) Schematic representation of all detected interactions between Sluc loci (numbers correspond to Sluc locus numbers from Table 1 and from references (8–10). Sluc6, Sluc8, and Sluc13 (9) (main effect only) influence the number of lung tumors, whereas all of the other loci influence the size of the lung tumors. The linking lines denote interactions. phenotypes is as low as 20% of the mean tumor size of the cross. The reliability of the RC strain mapping is shown by our ability to map the same Sluc loci in two or three different RC strains. The Sluc20 locus, found independently in the OcB-9 and OcB-16 crosses, has similar phenotypic effects (in Journal of the National Cancer Institute, Vol. 93, No. 19, October 3, 2001 the sense that the same genotypes in the two crosses have effects in the same direction, i.e., that susceptibility and resistance are associated with the same genoREPORTS 1487 types, although the magnitude of the effect is different) in the two crosses (Table 1; Fig. 2, A). The Sluc1 locus, originally mapped in the OcB-9 strain (8), was also detected in the OcB-3 (P ⳱ .0016) cross, with a consistent association between the genotype and the phenotype (Fig. 2, B), and in the OcB-6 (P ⳱ .0262) cross. Similarly, we confirmed Sluc16 (P ⳱ .0025) and Sluc18 (P ⳱ .036), both originally detected in the OcB-4 cross, as main effects in the OcB-6 cross but with different genotype–phenotype relationships in the two crosses (data not shown). Since these loci are involved in interac- tions, differences in their effects between crosses may be due to the presence of additional interactions in the confirmatory crosses. The interaction between Sluc5 and Sluc12 that was originally detected in an OcB-6 cross, where the animals were killed after 35 weeks (9), was independently found in the OcB-16 cross (Table 1). The same genotypes have effects in the same direction in both crosses; e.g., Sluc5b/bSluc12b/b is more resistant than Sluc5b/bSluc12°/° or Sluc5°/°Sluc12b/b, in both crosses (Table 1) (9). These confirmatory results not only increase the confidence of mapping but also reduce the original candidate chromosomal regions by limiting them to the part of the donor strain segments shared by both RC strains. If one assumes a single Sluc locus in the tested regions of chromosome 8, Sluc20 maps in the 25-cM B10.O20derived segregating segment shared by strains OcB-9 and OcB-16 (Fig. 3, A). Similarly, the shared segment of chromosome 19 in the OcB-3, OcB-6, and OcB-9 crosses is within the peak of Sluc1 of the QTL-likelihood plot (8), reducing the original candidate region (45.8–57.9 cM) to the length of the common shared segment (8.8–14.3 cM; Fig. 3, B). The lung Fig. 2. Independent detection and confirmation of lung cancer susceptibility (Sluc) loci in different OcB F2 crosses. A) Sluc20 was independently detected in the OcB-9 (interacting with Sluc1) and OcB-16 (interacting with Sluc17) crosses. Each panel shows the phenotypic effect (in percent difference from the mean of each cross for the mice that are homozygous for the O20 allele at the interacting loci) of the Sluc20 genotypes indicated at the top right corner. The left panel shows the effect of Sluc20 in the OcB-9 cross, in mice homozygous for the O20 allele at the interacting locus Sluc1. In this way, it is possible to compare these effects with the effects seen in the OcB-16 cross (in the right panel) in mice homozygous for the O20 allele at the interacting locus Sluc17. Sluc1, which interacts with Sluc2 in the OcB-9 cross (8), was confirmed in the OcB-3 and the OcB-6 crosses (B), where Sluc2 does not segregate because it is in all of the mice homozygous for the O20 allele. Each panel shows the phenotypic effect (expressed in percent difference from the mean of each cross) of the Sluc1 genotypes indicated at the top right corner. The left panel shows the effect of Sluc1 when the interacting locus Sluc2 is homozygous for the O20 allele (the percent difference from the mean of the cross was recalculated only for mice in which the interacting locus Sluc2 is homozygous for the O20 allele). The central panel shows the main effects of Sluc1 in the OcB-3 cross, and the right panel shows the main effects of Sluc1 in the OcB-6 cross. 1488 REPORTS Journal of the National Cancer Institute, Vol. 93, No. 19, October 3, 2001 Fig. 3. Refinement of mapping of lung cancer susceptibility (Sluc) loci by independent detection or confirmation in different OcB F2 crosses. In each chromosome (chr.) idiogram, the maximal B10.O20derived donor (gray and black), the minimal donor (black only), and the O20derived background (white) segments are shown. Representative polymorphic markers that define these regions are shown over the chromosomes, and the distances between them are indicated at the bottom in centimorgans (each group of chromosomes and segments is drawn to scale). The abbreviations “cen” and “tel” indicate the orientation of the segment in relation to the centromere and telomere, respectively. For the definition of maximal and minimal segments, see the “Materials and Methods” section. Chromosome 8 is shown in the OcB-9 and OcB-16 strains (A). The maximal common shared segment that includes the Sluc20 (D8Mit3) region between the two crosses is 25 cM. Chromosome 19 is shown in the OcB-3, OcB-6, and OcB-9 strains (B). The Sluc1 (D19Mit9) region is segregating in all three crosses. The maximal common segregating region between the three crosses is 14.3 cM, and the minimal region is 8.8 cM. The distal part of chromosome 1 is shown in the OcB-4, OcB-6, and OcB-16 strains (C). The Sluc5 (D1Mit16) region is segregating in the OcB-4 and OcB-6 crosses. Linkage results with marker D1Mit221 in the OcB-16 cross also suggest a Sluc locus in this part of chromosome 1. It is possible that a small as-yet-unidentified segment is shared by all three crosses, carrying the Sluc5 gene, most likely in the region common to the OcB-4 and OcB-6 strains. Alternatively, Sluc5 and an additional Sluc locus map in distal chromosome 1. tumor susceptibility locus Pas3 (19) maps outside the OcB-3 and OcB-6 segregating segments of chromosome 19, suggesting the presence of at least two lung cancer susceptibility loci in this chromosome (Pas3 proximal and Sluc1 distal). The distal part of chromosome 1, where Sluc5 has been mapped previously (9), interacts with Sluc26 in the OcB-4 and OcB-16 crosses (Table 1) and with Sluc12 in the OcB-16 and OcB-6 crosses (9). Possibly, an as-yet-unidentified short common segment of donor strain origin is shared between all three strains (Fig. 3, C), suggesting the involvement of a single locus, Sluc5, in these interactions; alternatively, the OcB-16 results indicate an additional Sluc locus. A susceptibility to the colon cancer locus (Scc3) and a radia- tion-induced leukemia sensitivity locus (Ril3) also maps in this region (20,21). DISCUSSION We have shown here that susceptibility to lung cancer is controlled by many genes, acting either alone or through interactions with other genes. Similar results have been observed in other studies on tumor susceptibility (17,18,22,23). In our study and in two previous studies (8,9), scanning of approximately one half of the genome with a total of 887 (OcB × O20)F2 hybrid mice—730 for lung tumor size here and 887 for lung tumor number and size (9)—provided statistically significant evidence for 30 Sluc loci on 17 chromosomes (Fig. 1, A) and the Kras2 Journal of the National Cancer Institute, Vol. 93, No. 19, October 3, 2001 locus on chromosome 6 (9). Perhaps even more important is the fact that 25 unique pairwise interlocus interactions were detected (Fig. 1, B). This result is equivalent to approximately one locus per approximately 29 F2 mice and one interaction per 35 F2 mice. These results are all the more impressive, considering that all other mouse lung cancer susceptibility loci mapped to date (17–19,23–28) required an excess of 1500 test mice (or about one statistically significant locus per 133 mice) and, in those studies, only two interactions were identified (17,18). These results demonstrate the high resolving power of the RC system. A possible explanation for the success of this system may be that, given a complex multigenic trait with numerous interacting loci, it allows the sysREPORTS 1489 tematic testing not only of main effects but also of a high percentage of all possible two-way interactions with a reasonable number of test animals (e.g., in this study, there were, on average, 146 animals per F2 cross). In contrast, for crosses in which the whole genome is segregating, it is practically impossible to systematically scan for interactions (an average of 14 markers and 91 two-way interactions for an RC cross compared with 80 markers and 3160 two-way interactions for a cross where the whole genome is segregating), unless specific selection criteria are applied to study only small subsets of two-way interactions; e.g., Festing et al. (24) tested for interactions only between markers for which they detected a statistically significant main effect. Several Sluc loci (Sluc1, Sluc5, Sluc12, Sluc16, Sluc18, Sluc20, and Sluc26) and interactions (Sluc5 × Sluc12 and Sluc5 × Sluc26) were confirmed in more than one RC strain. Because of the random segmentation of the donor genome in the RC strains, the average estimated length of the Sluc loci candidate regions (± standard deviation) is 15.2 cM (±8.2), and eight of the Sluc loci are mapped in regions of 12 cM or less. These Sluc loci are (maximal region indicated by first number and minimal region indicated by second number): Sluc1 (12 cM–9 cM), Sluc3 (12 cM–2.8 cM), Sluc4 (12 cM–8 cM), Sluc12 (3 cM–3 cM), Sluc17 (11.5 cM–3 cM), Sluc23 (5 cM–3 cM), Sluc26 (11.7 cM–1 cM), and Sluc28 (11 cM–8 cM). We are currently refining the mapping of the boundaries of the segregating segments of all candidate regions of Sluc loci. In addition, new recombinants are currently being generated and tested, with the aim of further reducing and finemapping the candidate regions, thus facilitating loss of heterozygosity studies and direct cloning of these genes. Furthermore, the homologous regions of human chromosomes are also better defined, providing an excellent tool for focused genetic studies in the human population and the cloning of their human counterparts. Extrapolating the number of Sluc loci to the total mouse genome, we predict approximately 60 (90% confidence interval ⳱ 42 to 78) loci and multiple interactions. This number is probably an underestimation of the true number of Sluc loci, since it is derived from a limited number of F2 hybrid mice and a single pair of inbred strains. The high frequency of these interactions suggests the existence of synergic pathways and molecular inter1490 REPORTS actions (e.g., protein–protein interactions), as well as the existence of higher order interactions. These interactions may play a role in buffering of genetic variation (29) and may be subsets of more complicated functional modules (30). We show that, despite the apparent complexity of a multigenic trait, with the right choice of genetic and statistical tools, it is possible to detect the relevant loci and their interactions. To our knowledge, this is the first time that a direct calculation of loci involved in a polygenic trait has been performed by systematically screening half of the murine genome. Although the number of these loci is large, new technologies (e.g., single nucleotide polymorphism analysis, complementary DNA arrays) and the sequence of the human and mouse genome and syntenic comparisons between the two will greatly facilitate the identification of these genes. Elucidation of the action of these genes will provide an insight into the molecular mechanisms of tumorigenesis and may serve as a paradigm for other multigenic traits. REFERENCES (1) Lichtenstein P, Holm NV, Verkasalo PK, Iliadou A, Kaprio J, Koskenvuo M, et al. Environmental and heritable factors in the causation of cancer—analyses of cohorts of twins from Sweden, Denmark, and Finland. N Engl J Med 2000;343:78–85. (2) Pisani P, Parkin DM, Ferlay J. Estimates of the worldwide mortality from eighteen major cancers in 1985. Implications for prevention and projections of future burden. Int J Cancer 1993; 55:891–903. (3) Bepler G. Lung cancer epidemiology and genetics. J Thorac Imaging 1999;14:228–34. (4) Balmain A, Nagase H. Cancer resistance genes in mice: models for the study of tumour modifiers. Trends Genet 1998;14:139–44. (5) Oomen LC, van der Valk MA, Demant P. MHC and non-MHC genes in lung tumor susceptibility in the mouse: implications for the study of the different lung tumor types and their cell of origin. Exp Lung Res 1991;17: 283–304. (6) Heston WE. Genetic analysis of susceptibility to induced pulmonary tumors in mice. J Natl Cancer Inst 1942;3:69–78. (7) Demant P, Hart AA. Recombinant congenic strains—a new tool for analyzing genetic traits determined by more than one gene. Immunogenetics 1986;24:416–22. (8) Fijneman RJ, de Vries SS, Jansen RC, Demant P. Complex interactions of new quantitative trait loci, Sluc1, Sluc2, Sluc3 and Sluc4, that influence the susceptibility to lung cancer in the mouse. Nat Genet 1996;14:465–7. (9) Fijneman RJ, Jansen RC, van der Valk MA, Demant P. High frequency of interactions between lung cancer susceptibility genes in the mouse: mapping of Sluc5 to Sluc14. Cancer Res 1998;58:4794–8. (10) Fijneman RJ, van der Valk MA, Demant P. Genetics of quantitative and qualitative aspects of lung tumorigenesis in the mouse: multiple interacting Susceptibility to lung cancer (Sluc) genes with large effects. Exp Lung Res 1998; 24:419–36. (11) Jansen RC, Stam P. High resolution of quantitative traits into multiple loci via interval mapping. Genetics 1994;136:1447–55. (12) Jansen RC. Controlling the type I and type II errors in mapping quantitative trait loci. Genetics 1994;138:871–81. (13) Jansen RC. Complex plant traits: time for polygenic analysis. Trends Plant Sci 1996;1: 89–94. (14) Groot PC, Moen CJ, Hart AA, Snoek M, Demant P. Recombinant congenic strains: genetic composition. In: Lyon MF, Rastan S, Brown SD, editors. Genetic variants and strains of the laboratory mouse. Oxford (U.K.): Oxford University Press; 1996. p. 1660–70. (15) Stassen AP, Groot PC, Eppig JT, Demant P. Genetic composition of the recombinant congenic strains. Mamm Genome 1996;7:55–8. (16) Lander E, Kruglyak L. Genetic dissection of complex traits: guidelines for interpreting and reporting linkage results. Nat Genet 1995;11: 241–7. (17) Manenti G, Gariboldi M, Fiorino A, Zanesi N, Pierotti MA, Dragani TA. Genetic mapping of lung cancer modifier loci specifically affecting tumor initiation and progression. Cancer Res 1997;57:4164–6. (18) Manenti G, Gariboldi M, Elango R, Fiorino A, De Gregorio L, Falvella FS, et al. Genetic mapping of a pulmonary adenoma resistance locus (Par1) in mouse. Nat Genet 1996;12:455–7. (19) Devereux TR, Wiseman RW, Kaplan N, Garren S, Foley JF, White CM, et al. Assignment of a locus for mouse lung tumor susceptibility to proximal chromosome 19. Mamm Genome 1994;5:749–55. (20) van Wezel T, Stassen AP, Moen CJ, Hart AA, van der Valk MA, Demant P. Gene interaction and single gene effects in colon tumour susceptibility in mice. Nat Genet 1996;14: 468–70. (21) Meruelo D, Offer M, Flieger N. Genetics of susceptibility for radiation-induced leukemia. Mapping of genes involved to chromosomes 1, 2, and 4, and implications for a viral etiology in the disease. J Exp Med 1981;154:1201–11. (22) Nagase H, Mao JH, Balmain A. A subset of skin tumor modifier loci determines survival time of tumor-bearing mice. Proc Natl Acad Sci U S A 1999;96:15032–7. (23) Festing MF, Yang A, Malkinson AM. At least four genes and sex are associated with susceptibility to urethane-induced pulmonary adenomas in mice. Genet Res 1994;64:99–106. (24) Festing MF, Lin L, Devereux TR, Gao F, Yang A, Anna CH, et al. At least four loci and gender are associated with susceptibility to the chemical induction of lung adenomas in A/J×BALB/c mice. Genomics 1998;53:129–36. Journal of the National Cancer Institute, Vol. 93, No. 19, October 3, 2001 (25) Ryan J, Barker PE, Nesbitt MN, Ruddle FH. KRAS2 as a genetic marker for lung tumor susceptibility in inbred mice. J Natl Cancer Inst 1987;79:1351–7. (26) Gariboldi M, Manenti G, Canzian F, Falvella FS, Radice MT, Pierotti MA, et al. A major susceptibility locus to murine lung carcinogenesis maps on chromosome 6. Nat Genet 1993; 3:132–6. (27) Obata M, Nishimori H, Ogawa K, Lee GH. Identification of the Par2 (Pulmonary adenoma resistance) locus on mouse chromosome 18, a major genetic determinant for lung carcinogen resistance in BALB/cByJ mice. Oncogene 1996;13:1599–604. (28) Pataer A, Nishimura M, Kamoto T, Ichioka K, Sato M, Hiai H. Genetic resistance to urethaninduced pulmonary adenomas in SMXA recombinant inbred mouse strains. Cancer Res 1997;57:2904–8. (29) Hartman JL 4th , Garvik B, Hartwell L. Principles of the buffering of genetic variation. Science 2001;291:1001–4. (30) Hartwell LH, Hopfield JJ, Leibler S, Murray AW. From molecular to modular cell biology. Nature 1999;402(6761 Suppl):C47–52. Journal of the National Cancer Institute, Vol. 93, No. 19, October 3, 2001 NOTES Supported by grant 925-01-01 from the Dutch Foundation for Scientific Research (Nederlandse Organisatie voor Wetenschappelijk Onderzoek) and by contracts CHRXCT930181 and CTBIO-98-0445 from the Commission of European Communities. We thank M. Treur-Mulder and E. DelzenneGoette (Division of Molecular Genetics, The Netherlands Cancer Institute, Amsterdam) for their excellent technical assistance. Manuscript received January 29, 2001; revised July 17, 2001; accepted August 10, 2001. REPORTS 1491