Chromosome Mapping of Multiple Loci Affecting the Genetic Predisposition to Rat Liver Carcinogenesis¹

Increasing evidence (1) indicates that individual risk of liver cancer reflects the amount of exposure to environmental agents, such as hepatitis B and hepatitis C viruses, Aflatoxin B1, ethanol consumption, and so forth, combined with the genetic predisposition of an individual. It may be hypothesized that the existence in the human population of several susceptibility and resistance alleles contributes to the risk of liver cancer. The identification of these genes could be of primary importance to additionally clarify the pathogenesis of HCCs,³ as well as for new preventive and therapeutic approaches. However, the difficulties in directly dissecting the genetic risk factors in humans makes the use of rodent models attractive for the discovery of some possible mechanisms of polygenic control of hepatocarcinogenesis in humans.

Previous studies on a murine model have led to the identification of seven Hcs loci (Hcs1 to Hcs7), and two resistance loci (Hcr1 and Hcr2) in crosses between susceptible and resistant mice strains (reviewed in Refs. 1, 2). We have generated, in our laboratory (3), the resistant BFF1 rat strain by crossing the BN (B), resistant, to the F344 (F), susceptible, rats. Linkage analysis of BFF1 × F344 backcross rats revealed the presence of two Hcs loci (rat Hcs1 and Hcs2) mapping to chromosomes 7 and 1, and three Hcr loci (rat Hcr1 to Hcr3), mapping to chromosomes 10, 4, and 8 (4). Two QTLs mapping to chromosomes 4 and 1, in genetic linkage with the development of early preneoplastic foci and the progression of late lesions, respectively, have been identified in (F344xDRH)F2 rats (5, 6). The first QTL corresponds to Hcr2 of BFF1 × F344 rats (4). LOH studies (7) have suggested the presence of a cluster of putative oncosuppressor genes at the Hcr1 locus. A putative suppressor gene (rcc+), mapping to chromosome 12p, critical for determining the sensitivity of rats to DENA-induced liver carcinogenesis, has been identified in MHC-recombinant rat ACP strains, congenic for the MHC-linked growth reproduction complex (grc) region (8).

The genetic trait of resistant BFF1 rats results in relatively low DNA synthesis, increase in remodeling, and relatively small volume of preneoplastic and neoplastic lesions (3, 4), whereas the resistance of Copenhagen rats mainly depends on high remodeling of preneoplastic liver lesions (9). The initiation stage is not affected in these strains. A great decrease in lesion volume, associated with a decrease in number, has been seen in the resistant DRH rats (6), whereas in ACP rats resistance to hepatocarcinogenesis essentially depends on sharp decrease in lesion number (8).

The above observations suggest the existence of a biological complexity of hepatocarcinogenesis in rats, indicating that genetic control of tumor development might also be complex, and the loci discerned thus far in BFF1 × F344 backcross progeny can probably not completely account for this complexity. To additionally analyze the genetic susceptibility of rats to liver cancer, we subjected the RH model of hepatocarcinogenesis to a QTL analysis to map cancer susceptibility genes in BFF2 intercrosses, in which linkage analysis is more sensitive than in backcross progeny.

F344 and BN rats (140–160 g, at the beginning of the experiment; Charles-River Italia, Calco, LC, Italy) were crossed to generate BFF1 and BFF2 rats. Animals were fed a standard diet (type 48; Piccioni, Gessate, Milan, Italy) and tap water ad libitum, and were housed individually in suspended wire-bottomed cages in a room with constant temperature (22°C) and humidity (55%), and with a 12-h light (6 a.m. to 6 p.m.)/dark cycle. All of the animals received humane care, and the study protocols were in compliance with the guidelines of our institution for use of laboratory animals.

Neoplastic nodules were induced in 20 F344, 20 BN, and 20 BFF1 rats, and 116 intercross rats by the RH protocol (10) that included initiation with a necrogenic dose of DENA (150 mg/kg) followed, after repair, by a 15-day feeding a hyperproteic diet (type 52; Piccioni) containing 0.02% 2-acetylaminofluorene, with a partial hepatectomy at the midpoint of this feeding. All of the rats were killed 32 weeks after initiation. The livers were resected, and small portions of gross neoplastic nodules and small pieces of liver were processed for H&E staining, and GST-P immunohistochemistry as published (3). Number/cm³ and mean volume of lesions, and volume fraction were determined by computer-assisted morphometric analysis (3).

Genomic DNA from spleens of intercross rats was extracted from isolated nuclei and purified as published (4). Genotyping was made by PCR, using as primers 179 polymorphic microsatellite markers (Roche Diagnostic S.p.A., Monza, Italy), distributed throughout all of the autosomes, leaving gaps <26 cM, excepting one gap of 30.2 cM on chromosome 1, 28 cM on chromosome 3, and 33 cM on chromosome 12. PCR products were run on 3.5% agarose gels or in an Alfexpress Automated Sequencer (Amersham, Pharmacia Biotech).

Linkage maps were constructed using the MAPMAKER/EXP 3.0 program. Association of tumor susceptibility with alleles of microsatellite markers was evaluated by LOD score (4). Threshold LOD score values at 2.8 and 4.3 were considered for “suggestive” and “significant” linkage, respectively (11). The proportion of total intercross variability, explained by the association between the marker and the trait (R²), was taken as an index of the importance of each locus. QTL analysis was made using parametric and nonparametric methods with MAPMAKER/QTL 1.1. ANOVA procedure (SAS Institute Inc., Cary, NC) was used to analyze potential interactions between genetic loci, and Ps were corrected for multiple comparisons according to Lander and Schork (12). Differences between parental strains, and between homozygous and heterozygous intercross rats for phenotypic parameters were analyzed by ANOVA, and multiple comparisons were made by TK test using GraphPad InStat 3.⁴

Body and liver weights did not significantly differ in BN, BFF1, and intercross rats throughout the experiment with respect to F344 rats (data not shown). The number of GST-P-positive lesions per cm³ of liver 32 weeks after initiation was 36–41% higher in BN and BFF1 rats, with respect to F344 rats (Table 1). V and VF were 33–36 and ∼6-fold higher, respectively, in F344 than in BN and BFF1 strains, without any difference between BN and BFF1 rats. Nodules developing in F344 rats showed atypical features exhibiting distortion of plate arrangement, thickened plates, and mild nuclear atypia, suggesting initial malignant transformation (Fig. 1,A). Some nodules showed microscopic features of well-differentiated carcinomas. In BN and BFF1 rats, most nodules did not exhibit atypical features and were constituted by clear-acidophilic cells (Fig. 1 B). As expected, the distribution of nodular N, V, and VF values in BFF2 rats indicated large variations of susceptibility, roughly in the range of parental strain values (not shown).

Linkage analysis of N, V, and VF values (Fig. 2) identified a QTL on chromosome 1, in significant linkage with VF, and with a LOD score peak of 4.9 and a 1-LOD unit support interval of 18 cM, between D1Mit3 and D1Mgh23. The profile of LOD score for V reproduced at a lower level that of VF. Another QTL affecting nodule volume with significant linkage on chromosome 16 had a LOD score peak of 4.43 and 1-LOD unit support interval of 14 cM, between D16Mit2 and D16Mgh5. The profile of LOD score for VF reproduced at a lower level than that of V. Two additional QTLs affecting V were identified. The first one, on chromosome 4, had a LOD score of 3.7, suggestive of genetic association, and a 1-LOD unit support interval of ∼14 cM, between D4Mgh6 and D4Mgh11. The second QTL, on chromosome 3, was characterized by a genetic linkage with V less than suggestive (LOD score peak, 2.6) but showed additive interactions with a QTL on chromosome 4 (see below). A QTL on chromosome 6, affecting nodule number, showed two peaks. The highest LOD score of 4.4 indicated a significant genetic linkage. A 18 cM 1-LOD unit support interval with respect to this LOD score peak, located between D6Rat22 and D6Rat11, included the second peak. Two QTLs found during the analysis of chromosome 8 were both suggestive of genetic linkage with V (LOD score peaks, 2.8) and N (LOD score peaks, 2.9). These QTLs showed a 1-LOD unit support intervals of 28 cM and 22 cM, respectively, located between D8Mit2 and D8Rat21, and D8Rat21 and D8Mgh3. The two peaks were 40.7 cM distant from each other. The profiles of LOD scores for V and VF showed several peaks less than suggestive of genetic linkage. Finally, a putative QTL showing association less than suggestive with VF and LOD score peak of 1.6 was identified on chromosome 10 at D10Rat51 (not shown). This QTL corresponds to Hcr1 identified previously in significant association with VF (4).

We next investigated whether the phenotypic behavior of neoplastic nodules in BFF2 progeny was accounted for by the presence of susceptible or resistant genes located in the QTLs identified and contributed by the parental strains. This was evaluated by ANOVA of average phenotypic values, and allelic distribution patterns in homozygous and heterozygous progeny (Table 2). Rats carrying one or two B alleles at the QTLs on chromosomes 1 and 16 exhibited V and VF values significantly higher than rat homozygous for F allele. Thus, these QTLs were defined as rat Hcs3 and Hcs4, respectively. For the QTL mapping to chromosome 4, a significant dosage-negative effect of the Ballele was observed for V in rats heterozygous or homozygous for this allele. This locus was named rat Hcr4. A different situation was seen for the QTLs on chromosome 8, of which the markers closest to the LOD score peaks were positioned 84.4 and 43.7 cM downstream with respect to the first marker used on the chromosome. These loci were characterized by a dosage-negative effect of the B allele for V/VF only in homozygous rats. The rats bearing homozygous or heterozygous F allele showed significantly higher V and VF values and lower N values. The QTL closer to telomere (at 84.4 cM) overlapped with a QTL identified previously on the same chromosome in BFF1 × F344 backcross rats (Hcr3), which exhibited a dosage-negative effect of the B allele. The other QTL on chromosome 8 was tentatively defined as rat Hcr5 (compare “Discussion”). Finally, the QTL on chromosome 6, named rat Hcr6, exhibited N values significantly lower in rats bearing at least one B allele, and the QTL mapping to chromosome 3 showed a dominant dosage-negative effect of F allele on nodule volume and was named rat Hcr7.

The analysis of R² values (Table 2) shows that phenotypic effects range between 11% and 19.2% for all of the loci, excepting Hcs3, of which the phenotypic effect is ∼50% for VF. The phenotypic effect of Hcs/Hcr loci is largely influenced by additive interactions (4). To address this point, we evaluated V values, in relation to the allelic distribution patterns in BFF2 rats for the markers closest to LOD score peaks of Hcs3 and Hcs4, at chromosomes 1 and 16, and of Hcr4 and Hcr7, at chromosomes 4 and 3. Additive interactions involving more than two QTLs could not be evaluated because such type of interaction implicates allele distribution into at least 24 subgroups consisting of a number of rats insufficient for statistical analysis. High V values occurred in BFF2 rats inheriting only B alleles at markers D1Rat70 and D16Rat6, on chromosomes 1 and 16, respectively. Rats inheriting two or three B alleles at the same markers showed intermediate values not significantly different from those of double BB homozygous rats. Significantly lower V values (P < 0.001) were seen in all of the rats bearing one to two B alleles at only one of the two markers, or in double FF homozygous rats, namely in rats in which additive effects of the B alleles were not possible. A more complex situation occurs for Hcr4 and Hcr7 loci of which the phenotypic effects depend on B alleles for Hcr4 and F alleles for Hcr7. Relatively low V values were found in rats homozygous for F alleles at D3Rat48 and B alleles at D4Mgh7, as well as in rats bearing at least one F allele at chromosome 3 and one B allele at chromosome 4. All of the other rats, in which no additive effects between the F allele at D3Rat48 and Ballele at D4Mgh7 were possible, showed V values significantly higher (P < 0.001) than those of rats homozygous for F and B alleles at chromosomes 3 and 4, respectively.

No additive interactions occurred between QTLs on chromosome 8 and Hcr7 for V, and Hcr6 for N. Analysis of additive interactions of B alleles at D4Mgh7 (Hcr4) and D8Rat18 (Hcr3) did not reveal any significant difference between V values of BFF2 rats carrying BB/BB or BB/FBalleles at D8Rat18 and D4Mgh7, respectively. All of the other allelic patterns at these marker loci were associated with significantly higher V values, with respect to the BB/BBdouble homozygous rats, indicating the absence of additive interactions even in rats carrying FB alleles at D8Rat18. This confirms that the B allele at this marker locus is recessive as suggested by ANOVA (Table 2).

Additional QTLs affecting hepatocarcinogenesis that lack phenotypic effects may be detected by the study of epistatic interactions inducing phenotypic effects not predictable on the basis of the sum of their separate effects. Thus, we performed two-by-two ANOVA with all of the markers against all of the other markers. Among interactions calculated on the basis of V, only those having P < 0.05 (values corrected for genome-wide comparisons; 12) were considered. D3Rat4 and D20Mit4 had significant reciprocal interaction (corrected P = 0.033). Significant interactions were also detected between D5Mcw1 and D20Rat1 (corrected P = 0.04), and between D1Mgh23, which is in the region of Hcs3 (LOD score 3.3), and D10Rat25 (corrected P = 0.05) and D8Mit5 (corrected P = 0.05). Neither of the microsatellite marker loci involved showed significant individual effect. Highest V values were associated with homozygosity of B allele at D3Rat4/D20Mit4, homozygosity of B allele and Fallele at D5Mcw1 and D20Rat1, respectively, and homozygosity of F allele at D1Mgh23, and of B allele at D10Rat25 and D8Mit5. However, the distribution of genotype combinations among possible subgroups was characterized by great variation in the number of BFF2 rats that did not allow reliable multiple comparisons among all of the genotype subgroups. Thus, the identification of allelic combinations with main epistatic effects was hypothetical.

Genetic predisposition to rodent liver carcinogenesis influences growth rate, volume, and progression capacity of neoplastic nodules (3, 4). Previous research (4) identified, in BFF1 × F344 backcross rats, two Hcs loci affecting nodule V/VF (Hcs1) and n (Hcs2), respectively, and three Hcr loci affecting V/VF (Hcr1-3). We have now identified in BFF2 rats two new loci named Hcs3 and Hcs4 on chromosomes 1 and 16, respectively, in significant genetic linkage with VF and V, showing a dosage-positive effect of the B alleles. Total variability explained by the association between the marker at LOD score peak and the character (R²) was ∼50% for Hcs3, indicating that this locus gives a major contribution to nodule volume, sufficient to elicit per se a sensitive phenotype (4). Additive interaction between Hcs3 and Hcs4 brings the total variability of the trait up to 70%. A QTL mapping to chromosome 4 showed a dosage-negative effect of the B allele, and a LOD score peak in suggestive linkage with V, positioned 18 cM downstream with respect to the LOD score peak of Hcr2 identified previously (4). The newly discovered locus, named Hcr4, shows additive interaction with a putative QTL in less than suggestive linkage with V mapping to chromosome 3. This locus, tentatively named Hcr7, exhibited a dosage-negative effect of F allele on V. Additive interaction between Hcr4 and Hcr7 accounted for ∼30% of the variability of the character, suggesting that Hcr7 represents an important region deserving additional investigation. Two other loci on chromosome 8 apparently were in suggestive linkage with N and V. The LOD score peaks in correspondence of D8Rrat18 and D8Mit2 were positioned 84.4 and 43.7 cM from the first marker used in the chromosome, respectively. The telomeric QTL corresponds to Hcr3 discovered previously in backcross rats (4) and characterized by a dominant, dosage-negative effect of the B allele on V. However, the B allele at this QTL showed a recessive dosage-negative effect in BFF2 rats. This apparent discrepancy cannot be explained by our results. It should be noted that changes in genetic substrate implicate differences in gene-gene interaction (13) and, consequently, in the combined effects of all of the resistance/susceptibility loci responsible for the expression of the trait and of tumor development in backcross and BFF2 rats. Taking into account these and previous observations (4), we still consider Hcr3 and, by analogy, the centromeric QTL, named Hcr5, as resistance loci. A dosage-negative effect of the B alleles for nodule number was seen at a locus on chromosome 6 in significant linkage with N, named Hcr6. This observation and identification of QTLs affecting positively lesion number on chromosomes 8 and 1 indicate the presence in BN and BFF1 rats of various genes controlling nodule number with a prevalence of an overall positive effect (3, 4). Generation of recombinant congenic rats is presently underway in our laboratory to render the polygenic trait oligogenic. This may result in more precise positioning of QTLs, better characterization of phenotypic effects of B and F alleles in each QTL, and, eventually, restriction of chromosomal segments to attempt positional or candidate gene cloning.

Previous (4) and present data show a prevalence of susceptibility B alleles in backcross and intercross progeny. Determination of the genealogic tree of various rat strains showed the occurrence of at least five genetic events during the evolution of F344 rats from an ancestor common to BN rats (14). This is consistent with the generation of susceptible rat strains from a common resistant feral ancestor as a result of selective mutation of resistance alleles that consequently are not activated by carcinogen treatment. Maintenance of unaltered resistance alleles in BN rats may result in the inactivation (modifier effect) of susceptibility alleles. B alleles associated with a susceptible phenotype in backcross and intercross subpopulations are identified as susceptibility alleles. The validation of this hypothesis awaits cloning of susceptibility and resistance genes. It is interesting to note that c-myc amplification and disruption of the pRb-E2F pathway, leading to fast G₁ phase progression and G₁-S transition, occurs in preneoplastic and neoplastic liver lesions of F344 rats, whereas the lesions of resistant rats show low c-myc activity, p16^INK4A up-regulation, and restraint in cell cycle activity (15). c-myc maps to Hcs1 on rat chromosome 7, which undergoes allelic imbalance in the HCC of susceptible LFF1 rats (7) but not in BFF1 rats (16). This suggests that c-myc and/or cell cycle key genes are susceptibility genes or targets of these genes. They are not up-regulated in carcinogen-treated resistant rats, probably as a consequence of the modifier effect of Hcr genes.

Interstrain comparison shows that Hcs3 maps to the same segment of chromosome 1 where it is located the resistance Drh1 locus, affecting the development of preneoplastic liver foci induced in (F344xDRH)F2 rats (5, 6). A second locus, Drh2, controlling the progression of preneoplastic foci to carcinoma in the same rats, maps to chromosome 4 in correspondence of the Hcr2 identified previously (4). Thus, it appears that at least two Hcr loci have been conserved during the evolution of DRH and BN rats from a common ancestor. The relatively high number of Hcs and Hcr loci, and additive and epistatic interactions discovered in BFF1 rats confirms the previous hypothesis (3, 4, 5, 6) of a high complexity of rat liver carcinogenesis in terms of a number of genetic factors involved and of interplay between these factors. Taking advantage of these findings, interspecies comparisons for genomic alterations at Hcs/Hcr loci may be made. Rat Hcs2 and Hcs3 are positioned in chromosomal segments syntenic to mouse Hcs1 and Hcs6, respectively, affecting lesion size (2), as well as to human chromosomal segments where frequent duplications (Hcs3) or LOH (Hcs4) occur (17). Rat Hcr loci are located in chromosomal segments syntenic to mouse and/or human chromosomes where frequent LOH occur (17). Various genes potentially involved in hepatocarcinogenesis map⁵ in correspondence of Hcs3 (H19, H-ras, Cdnk1c, Igf2. Gstp1, and Cyp17), Hcs4 (Gstp15), Hcr4 (v-raf1, Tgfα), Hcr3 and Hcr5 (Hnf, Rbp2, Ccnd3, Cyp19, 1a1, and 1a2), Hcr6 (Fos, Hnf3a, and Esr2), and Hcr7 (Gstp11 and Sp3). Additional work to discover interstrain polymorphisms of these genes, consistent with differences in phenotypic behavior, could help in discerning their candidacy as susceptibility/resistance genes. Most of these genes are deregulated in early and/or late stages of rodent and human hepatocarcinogenesis (17). Among them, Tgfα, located at Hcr4/Drh2 locus (4, 5) and at Xhs1 locus controlling X-ray hypersensitivity of LEC rats (18), is involved in growth of initiated hepatocytes, and its up-regulation together with c-myc amplification contributes to hepatocarcinogenesis in rodents (19) and, probably, in humans (17). These observations taken together suggest the commonalty of some genetic mechanisms of hepatocarcinogenesis in rodents and humans, and polygenic predisposition to HCC in rodents represents a useful model of the genetics of human hepatocarcinogenesis. An ∼2-fold rise in cancer risk occurs in the relatives of patients with HCCs (20). This is consistent with a situation similar to that of rat, in which multiple loci affecting HCC development are segregated, and the final effect results from interactions among loci. Because humans are characterized by assortative mating, low penetrance of the trait may be expected in the progeny of genetically predisposed individuals, which explains the rarity of familial clusters of HCC, even in high-risk areas.

We thank Drs. Nicolò P. Maciotta and Giuliana Solinas for expert assistance for statistical analysis of epistatic interactions.