Genetic linkage analysis identifies Pas1 as the common locus…

Top of page

Introduction

Inflammation is an important risk factor for various types of cancer, which frequently originate from sites of infection and chronic irritation.¹ The inflammatory response involves several processes, including the recruitment of immune and inflammatory cells (for example, macrophages, dendritic cells and leukocytes), the release of chemical mediators (for example, cytokines, chemokines and reactive oxygen species) and tissue repair.^{2, 3} In the tumor microenvironment, infiltrating leukocytes and soluble mediators actively contribute to tumor initiation, promotion and progression by sustaining cancer cell proliferation, survival and/or migration. In addition, cancer cells often constitutively produce inflammatory molecules such as cytokines and chemokines, indicating a strong relationship between inflammation and cancer.¹ In the lung, several studies have indicated an association between the risk of cancer and pre-existing conditions of non-malignant pulmonary inflammation, such as chronic bronchitis, emphysema, bronchial asthma, pneumonia, lung tuberculosis and chronic obstructive pulmonary disease (COPD).^{4, 5, 6} COPD poses the greatest risk for lung cancer among smokers, and hence the pathogenetic links between the diseases require further study.⁴

An experimental model to study the role of inflammation in the development of cancer is represented by unique mouse lines characterized for their acute inflammatory response (AIR). Specifically, AIR maximum (AIRmax) and AIR minimum (AIRmin) mice have been phenotypically selected on the basis of their high and low responses, respectively, to acute inflammation induced by subcutaneous injection of Bio-Gel P-100 polyacrylamide beads.⁷ The progressive phenotypic divergence between the two mouse lines is explained by the fixation to homozygosity of alleles regulating the respective high and low AIRs, while maintaining a heterogeneous background.⁸ AIRmax and AIRmin mouse lines show a ~30-fold difference in the number of infiltrated leukocytes in the inflammatory exudate after treatment.⁹ These lines also differ widely in their genetic predisposition to several chemically induced tumors, with a high inflammatory response being associated with a low susceptibility to skin and lung tumors.^{10, 11} These findings suggest that a subset of loci fixed during the selection for inflammatory response also regulate cancer susceptibility.

Using a model of urethane-induced lung tumorigenesis, we previously reported that lung tumor multiplicity is up to 40-fold higher in AIRmin than in AIRmax mice.¹¹ That study also showed that the AIRmax and AIRmin lines have significantly different genotypes at the polymorphic pulmonary adenoma susceptibility 1 (Pas1) locus, the major lung tumor modifier locus in mice.¹² These lines exhibited differential lung inflammation dynamics after urethane injection and, accordingly, their lung transcriptome was found to be distinct for genes of various pathways related to the inflammatory response.¹³ Recently, we performed linkage analysis of 290 (AIRmax × AIRmin)F2 intercross mice using genome-wide single-nucleotide polymorphism (SNP) arrays, and identified a quantitative trait locus (QTL) on chromosome 7 (Irm1, Inflammatory response modulator 1) that is associated with the number of infiltrated cells in Bio-Gel P-100-induced exudates.¹⁴ However, in that work we did not determine the lung tumor susceptibility of these intercross mice.

As inflammation is a complex phenotype, it is likely controlled by several distinct loci, some of which may have undergone selection to fixation or to differential allelic frequency during the generation of AIR mice. To further investigate the genetic control of inflammatory processes and its relationship with lung cancer susceptibility, we set out to identify additional QTLs in AIR mice. Here, we report the results of a combined lung tumorigenesis and inflammatory response experiment in a large pedigree of 802 mice, including 43 founder mice, 66 first-generation (AIRmax × AIRmin)F1 mice and 693 second-generation (AIRmax × AIRmin)F2 intercross mice.

Top of page

Results and discussion

AIR mouse lines show divergent inflammatory and lung tumorigenesis phenotypes

To assess the inflammatory response, AIRmax and AIRmin parental (founder) mice and the (AIRmax × AIRmin)F2 intercross mice were injected subcutaneously with polyacrylamide beads and evaluated 24 h later. The median number of leukocytes ( × 10⁵ per ml, Ncell) of AIRmax mice in their inflammatory exudates was ~33-fold higher than that of AIRmin mice (square-root-transformed data are shown in Figure 1a); this difference was significant (P<0.001, Kruskal–Wallis test). The (AIRmax × AIRmin)F2 intercross mice had Ncell values intermediate between those of the AIRmax and AIRmin parental lines. When analyzed separately by sex, F2 males (n=350) had 2.7-fold higher median values than F2 females (n=340; Figure 1b; P<0.001, Kruskal–Wallis test).

The susceptibility to lung tumorigenesis was evaluated at 35 weeks of age in mice that had been treated with urethane (300 mg per kg body weight) when they were 1-week old; this analysis was done in all (AIRmax × AIRmin)F2 intercross mice and in 54 non-parental AIRmax and AIRmin mice. Overall, the median number of lung tumors (Nlung) in the 22 AIRmax mice (Nlung=1.0) was 3.5-fold lower than in the 32 AIRmin mice (Nlung=3.5; Figure 2a; P<0.01, Kruskal–Wallis test). The (AIRmax × AIRmin)F2 mice had Nlung values that were not significantly different from those of AIRmin mice. When analyzed separately by sex (Figure 2b), male mice displayed a clear phenotypic pattern, with the F2 animals (n=335) having intermediate Nlung median values between the low values of male AIRmax and the higher values of AIRmin mice. In females, this pattern was absent, with F2 animals (n=329) having Nlung median values similar to AIRmin females. Given the broad distributions in Nlung values, the difference between male and female F2 mice was not significant.

The difference in Nlung between AIRmax and AIRmin mice observed here is much lower than what we previously reported (40-fold).¹¹ In that earlier study, lung tumorigenesis was initiated in 2-month-old mice by injecting urethane at 1000 mg per kg body weight. The different treatment schedules and observation periods (10–25 weeks after urethane treatment in the previous study versus 35 weeks in the present one) may explain, at least in part, the different results. Another important variable that may have influenced the results is the possibility of genetic drift in these non-inbred AIRmax and AIRmin lines. In fact, the AIRmin mice of the present study had lung tumor multiplicity values similar to those of intermediate susceptibility strains (for example, BALB/c and 129/J strains) rather than the high-susceptibility strains (for example, A/J strain, up to 30 lung tumors per mouse upon urethane treatment).^{12, 15}

Genetic linkage mapping of multiple inflammatory modifier loci

To identify loci controlling the inflammatory response, we analyzed genetic linkages in 801 mice of the AIR population (DNA was not available for one F1 female mouse). First, we genotyped the mice using a DNA array representing 1449 SNP markers. At quality control, 169 SNPs (11.7%) were excluded for having a GenTrain score <0.75. We removed an additional 390 SNPs with a minor allele frequency <0.1 and 11 SNPs whose genotype was identical to that of a flanking SNP (r²=1). Thus, there were 879 informative markers (824 autosomal) for study. No Mendelian inconsistencies were detected. Then, we determined genome-wide statistical thresholds for Ncell by permutation analysis. This calculation obtained logarithm of odds (LOD) scores of 3.34 and 3.64, respectively, for α=0.1 and α=0.05 genome-wide statistical probabilities (5000 permutations).

Figure 3 shows the results of the genetic linkage analysis for the Ncell phenotype. Two major peaks above the α=0.05 statistical threshold are seen, plus three additional peaks above the α=0.1 cutoff. The large peak on chromosome 7 had a LOD score of 15.74 and a 1-LOD confidence interval ranging from markers rs3673653–rs3682038. This QTL corresponds to the Irm1 locus, as the chromosomal region is almost identical, thus confirming its linkage with Ncell. In our previous study of 290 (AIRmax × AIRmin)F2 intercross mice,¹⁴ the Irm1 locus had LOD=3.6; this lower LOD score most likely depends on the smaller size of the previous study population. Another QTL mapped on chromosome 11, between rs13481119 and rs3090212, with LOD=7.74; in this same chromosomal region, a previous study mapped the collagen-induced arthritis (Cia40) QTL, also modulating inflammation.¹⁶ Markers on chromosome 4 were also significantly linked with the Ncell phenotype; the QTL peak (LOD=3.56) mapped at ~102.6 Mb, between rs13477883 and rs6324470. A significant locus on chromosome 6 (LOD=3.52) is located adjacent to that of Pas1. Finally, a QTL reaching LOD=3.34, just above the α=0.1 threshold, was identified on chromosome 13, near rs13481783, at 42.936 Mb where no known QTLs related to inflammation or tumorigenesis map.

The phenotypic variances explained by the linked loci on chromosomes 4, 6, 7, 11 and 13 were 1.8%, 1.8%, 9.5%, 4.4% and 1.7%, respectively. Thus, each locus provided a small contribution to the overall phenotypic variance, as expected for a trait-like inflammatory response, which is under complex genetic control. These results therefore confirm the major role played by the Irm1 locus on chromosome 7 in the control of inflammatory response¹⁴ and suggest that other minor undetected QTL and non-genetic environmental effects may also contribute to the total phenotypic variance.

The analysis for epistatic QTL found no evidence of epistasis at the α=0.1 genome-wide level. In many ways, a large F2 intercross population such as ours (with alternative alleles fixed in the original lines) is considered ideal for detecting epistasis as all genotype classes are present. In this study, notwithstanding the large sample (that is, ~700 mice), the effects of epistasis were not high enough to be detected. This negative result suggests that epistasis does not have a role in the modulation of inflammatory response in our mouse population or, alternatively, that the possible epistatic effects are too weak to be detected in our F2 intercross.

Pas1 is the major locus modulating lung tumor susceptibility in AIR mice

Repeating the genome-wide analysis to identify loci associated with tumor susceptibility (Figure 4), we calculated the statistical threshold for Nlung as 3.63 (P<0.05; 5000 permutations). The analysis identified two peaks above the cutoff: a major peak on chromosome 6 (LOD=12.18) at rs6265387 and an additional QTL (LOD=4.69) on chromosome 18 between rs6161154 and rs13459193.

The QTL on chromosome 6 corresponded with the Pas1 locus, indicating that this locus has the strongest linkage with lung tumor multiplicity. At rs6265387, mice carry either the G allele of the A/J-like susceptible strain or the A allele of the C57BL/6J-like resistant strain. Phenotype by genotype analysis in the (AIRmax × AIRmin)F2 population showed that the G allele conferred higher Nlung values, whereas the A allele was associated with lower values: G/G, 4.0±0.2, n=349; A/G, 3.1±0.2, n=259; A/A, 1.0±0.3, n=53 (mean±s.e.). As can be noted by the paucity of animals with the A/A genotype, most of the mice in the (AIRmax × AIRmin)F2 population carried the A/J-derived susceptibility allele at the Pas1 locus (allelic frequency of G=0.7). Despite the predominance of the Pas1 susceptibility allele, lung tumor multiplicity values were relatively low (that is, mean Nlung=3.1 and 4.0 in heterozygous and homozygous mice for the Pas1 susceptibility allele), probably due to the presence of dominant lung tumor-resistant alleles^{17, 18} at homozygosity or near homozygosity in both parental lines AIRmax and AIRmin mice, or to other unknown reasons.

The QTL on chromosome 18 between rs6161154 and rs13459193 maps to the region that is known to harbor the pulmonary adenoma resistance-2 (Par2) locus.^{15, 19} In this locus, 4930503L19Rik was recently identified as a novel lung tumor susceptibility gene.²⁰ Also called lung adenoma susceptibility 2 (Las2), this gene was strongly associated with resistance to tumor formation and had epistatic interactions with the Pas1 locus. However, in the present study, we did not detect such epistatic interaction (data not shown).

The phenotypic variances explained by the linked loci on chromosomes 6 and 18 were 8.0% and 2.8%, respectively. The 8.0% obtained for the Pas1 locus confirms its role in the control of lung tumorigenesis in AIR mice,²¹ even though this value is lower than the ~30–40% of phenotypic variance explained by Pas1 in crosses of inbred mice.^{21, 22} In genetic analyses of complex phenotypes, such as lung tumorigenesis, conducted in genetically heterogeneous populations, like that studied here, the fraction of phenotypic variance explained by genotypes may be low. This phenomenon gives rise to the so-called missing heredity, whose underlying reasons are still a matter of debate.^{23, 24}

Pas1 locus controls inflammatory response and lung tumorigenesis in AIR mice

The two genetic linkage analyses presented here highlight the complex genetic control of both the inflammatory response and lung tumorigenesis in AIR mice. In the F2 intercross population, there was no correlation between the Ncell and Nlung phenotypes (Spearman’s rho=−0.06; P=0.15). This result may be due to the likelihood that Ncell and Nlung phenotypes are controlled by multiple loci but only one of these loci is in common. Indeed, only on chromosome 6 did we identify similarly positioned loci linked to both phenotypes (Figure 5). The resolution power of QTL analysis, notwithstanding the large size of the present population, did not allow us to establish whether both phenotypes are controlled by the same locus or by two closely positioned loci. The linkage of the chromosome 6 locus (Pas1) with Ncell was weak (LOD=3.52), whereas the linkage of the same locus with Nlung was strong (LOD=12.18).

Previously, we fine-mapped Pas1 to a narrow region of ~0.5 Mb and found six genes as candidates for the locus: branched chain aminotransferase 1, cytosolic (Bcat1); lymphoid-restricted membrane protein (Lrmp); cancer susceptibility candidate 1 (Casc1); LYR motif-containing 5 (Lyrm5); v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog (Kras); and intermediate filament tail domain-containing 1 (Ifltd1).²⁵ Interestingly, Kras mutational activation in lung bronchiolar cells has been associated with an inflammatory response characterized by infiltration of alveolar macrophages and neutrophils.²⁶ Moreover, in a rat model of renal fibrosis, caused by unilateral ureteric obstruction, Kras expression increased, whereas treatment with antisense oligonucleotides silencing Kras reduced renal Kras to levels below basal and inhibited interstitial fibrosis.²⁷ Thus, Kras may represent the common mediator of inflammatory response and lung tumor susceptibility in AIR mice. Alternatively, the Sox5 gene, mapping just proximal to the Pas1 locus, may be a candidate for inflammatory response susceptibility. Sox5 knock-out mice showed abnormal lung development, and the human homologous SOX5 gene is associated with the lung inflammatory pathology COPD.²⁸

Our results are in agreement with several findings in humans showing the co-segregation of loci modulating both lung inflammation-related pathologies and lung cancer risk, although there are important interspecies differences. For example, the region of human chromosome 15q25 is associated with both COPD and lung cancer risk; this region contains six genes and it is not yet clear whether the same candidate genes are involved in both pathologies.^{29, 30} The human 15q25 region is homologous to a region of mouse chromosome 9; however, in this study, we did not identify inflammation or lung tumor modifier loci on chromosome 9. Also in humans, advanced glycosylation end product-specific receptor (AGER) and palmitoyl-protein thioesterase 2 (PPT2) genes (both associated with the risk of COPD³¹) and the major histocompatibility complex class II DQ beta 1 (HLA-DQB1) gene (associated with asthma³²) map close to the BCL2-associated athanogene 6 (BAG6) gene, a candidate for lung cancer risk.³³ The human chromosomal region containing the BAG6 is homologous to a mouse chromosome 17 region where we did not map any inflammation- or lung tumor-related loci.

This study suggests that the differential responses to lung tumorigenesis of AIRmin and AIRmax mouse lines, selected for their differential responses to acute inflammation, derive from the Pas1 locus that modulates both phenotypes. Overall, our findings may be relevant in understanding the relationships between inflammatory response and tumorigenesis in animal models and also in other species, including humans. Further studies, including specific recombinant crosses in the Pas1 region, analyzed for their inflammatory response phenotype, and candidate gene analyses, are required to clarify the relationships between Pas1 and the inflammatory response.

Top of page

Materials and methods

Mouse crosses

AIRmax and AIRmin lines (formally designated Ibut:AIRH and Ibut:AIRL at the Institute for Laboratory Animal Research, National Research Council) and crosses were developed and maintained at the animal facilities of the Laboratory of Immunogenetics of the Butantan Institute, São Paulo, Brazil.⁷ All procedures were approved by the Institutional Animal Care and Use Committee of Butantan Institute, and all animals received humane care according to the criteria outlined in the Guide for the Care and Use of Laboratory Animals of the US National Academy of Sciences and the US National Institutes of Health.

Parental AIRmax (n=22; 9 males) and AIRmin (n=21; 11 males) mice of selection generation F48 were intercrossed, generating about 200 (AIRmax × AIRmin)F1 offspring. From these offspring, 66 animals (one or two animals per litter; total, 33 males) were arbitrarily selected and mated, avoiding endogamy, to generate 693 (AIRmax × AIRmin)F2 mice (352 males and 341 females). The three generations, for a total of 802 mice from 33 families, were used in the genetic linkage study. An additional group of AIRmax (n=22) and AIRmin (n=32) mice of the same selection generation F48, from the same litters as the pedigree’s founder mice, was phenotyped for lung tumorigenesis and AIR, to avoid possible deleterious effects of the treatments on the offspring.

Treatments and phenotype evaluation

To initiate lung tumorigenesis, mice were treated subcutaneously at 1 week of age with urethane (300 mg per kg body weight). This treatment was done in all 693 (AIRmax × AIRmin)F2 mice and in the 54 non-parental AIRmax and AIRmin mice from the same litters as the pedigree’s founder mice. At 16 weeks of age, the urethane-treated (AIRmax × AIRmin)F2 mice and the 43 parental AIRmax and AIRmin mice were injected subcutaneously with 750 μl Bio-Gel P-100 polyacrylamide beads (Bio-Rad, Hercules, CA, USA) to induce an AIR.

Treated mice were phenotyped for acute inflammation as described.¹⁴ Briefly, local exudates at the site of injection of Bio-Gel P-100 were harvested after 24 h and leukocytes were counted in Malassez hemocytometers. The inflammatory phenotype was expressed as the number of infiltrating cells × 10⁵ per ml (Ncell).

At 35 weeks of age, that is, 34 weeks after urethane injection, F2 mice and non-parental AIRmax and AIRmin mice were killed with CO₂ for lung tumor evaluation. Lungs were inflated with 1 ml RNAlater solution (Ambion, Life Technologies, Carlsbad, CA, USA), excised and submerged in additional RNAlater. Each lung was visually inspected for the presence of tumors, which were measured with a digital micrometer (Digimess, São Paulo, Brazil). The total number of tumors (Nlung) was recorded for each mouse.

Genome-wide SNP genotyping

Genomic DNA was extracted from tail tips with E.Z.N.A. Tissue DNA kit (Omega Bio-Tek Instruments, Watford, UK) and quantified by fluorimetry with the Quant-iT PicoGreen dsDNA quantification kit (Invitrogen, Carlsbad, CA, USA). SNPs were genotyped with the BeadArray Platform (Illumina, San Diego, CA, USA) using a mouse linkage panel of 1449 SNP loci, as described.¹⁴ Briefly, allelic discrimination with the GoldenGate Genotyping Assay (Illumina) is obtained through the generation of a synthetic allele-specific template from genomic DNA followed by PCR amplification using three universal primers designed for each SNP.³⁴ Arrays were hybridized and scanned using the BeadArray Reader (Illumina) at a resolution of 0.8 μm. Illumina GenCall Data Analysis Software was used to analyze genotype data; SNPs with a GenTrain score >0.75 were kept in the analysis. Then, PLINK software, version 1.07 (http://pngu.mgh.harvard.edu/~purcell/plink/) was used for filtering of genotype data.³⁵ SNPs were eliminated from analysis if they had a minor allele frequency <0.1 or if they were in perfect linkage disequilibrium with a flanking SNP (r²=1).

Statistical analysis

Differences between groups in the phenotypes Ncell and Nlung were analyzed using the Kruskal–Wallis test. The correlation between phenotypes was expressed as Spearman’s rank correlation coefficient rho.

Because the distributions of Ncell and of Nlung phenotypes were skewed to the right, we transformed the original values to their square roots to improve normality. The transformed values were used for the interval-mapping QTL analyses that were adjusted by sex because of the statistically significant sex difference in the Ncell phenotype. Searches for QTLs affecting the phenotypes under study were carried out with genome-wide linkage analysis between genotypes and phenotypes by interval mapping using GridQTL version 1.5.2,³⁶ which uses a linear model to fit phenotype data according to genotypes. Additive and dominant effects at the QTL were included along with other the explanatory variables of sex and family. The significance thresholds of phenotype–genotype associations were estimated by genome-wide permutation analysis. The phenotypic variances explained by the linked loci were measured as a reduction in the residual variance when the QTL was included in the model in the presence of all the other QTL; these values were expressed as a percentage of the residual mean square after adjusting for sex and family.

Epistatic QTLs were mapped simultaneously by fitting a model with additive and dominant QTL effects for each QTL and the four possible pairwise interactions.³⁷ Two complementary searches were used. In the first, for each pre-identified QTL, the interactions with all other genome positions were tested and, in the second, an exhaustive two-dimensional genome scan was conducted. Two loci were declared epistatic if both the full two-QTL model including marginal and interacting effects and just the interacting effects were significant. Thresholds were determined by permutation.