Experiment
Body mass index (BMI) has been implicated as a primary factor influencing cancer development. To gain insight into the genetic factors linking BMI and cancer, chemical carcinogenesis was performed on a genetically heterogeneous cohort of interspecific backcross mice created by crossing tumor susceptible Mus musculus (FVB/N) mice with tumor resistant Mus spretus (SPRET/Ei) mice to developed a genetically heterogeneous cohort of (FVB/N x SPRET/Ei)F1 x FVB/N backcross mice that varies significantly in both tumor susceptibility and BMI.
Chemical carcinogenesis was initiated using the two-stage DMBA/TPA protocol. Mice were weighed at seven weeks. Lengths were measured by first anesthetizing mice and then by measuring the nasal/anal distance. BMI was calculated by taking the weight (in grams) at seven weeks and dividing by nasal to anal length squared (in cm). The resulting phenotype values were mean centered by sex prior to use in analysis.
Papillomas were counted every two weeks starting at the tenth week of TPA administration until the 20th week. At carcinoma formation, the date and location were recorded. The time from DMBA/TPA treatment start to carcinoma emergence was then used to calculate carcinoma-free survival. At sacrifice, dorsal skin, all tumors, and other tissues were harvested.
Genotypes were generated using pre-treatment tail DNA with a custom panel of TaqMan assays
consisting of 276 markers tiling the autosomes and the X chromosome at approximately 10 cM spacing. RNA was generated from tail and dorsal skin samples taken at seven weeks. QTL analysis was performed using the scanone function from the R package qtl (version 1.27-10). Significance was evaluated by comparing observed values to those obtained in 1000 random permutations for autosomes and 27,000 for the X chromosome. Initial investigations compared the results of assessing each phenotype with no covariates, with sex as an additive covariate, and again as an interactive covariate. LOD peak thresholds were defined by taking the 95 % Bayesian credible interval for each peak. To assess the significance of sex/genotype interactions at individual QTL, linear modeling was performed incorporating marker genotype, sex, and the sex/genotype interaction term as predictors of phenotype values. Significance was evaluated by ANOVA at 0.05.
To evaluate the relationship between pre-treatment physiological parameters and carcinogenesis, pre-treatment BMI, weight, and length were compared to papilloma burden in two independent skin carcinogenesis cohorts of (FVB/N x SPRET/Ei)F1 x FVB/N backcrossed mice. Cohort 1 consisted of 141 mice, 139 of which had whole-genome genotyping data available. Cohort 2 had 237 mice, 235 of which had genotyping data available. There were approximately as many males as females in each group.
In these cohorts, BMI and weight, but not length, were significantly positively correlated with papilloma burden (Fig. 1b). In a combined analysis, BMI was positively correlated to papilloma burden (p < 1e-5, rho = 0.25). The relationship appeared to be driven primarily by male mice, which showed a correlation p value < 1 e-7and a rho of 0.38 (Fig. 1b). BMI was not significantly correlated with papilloma burden in female mice. Similar results were observed for weight, which showed a significant positive correlation with papilloma burden in male mice (rho = 0.33; p < 1e-5), and no correlation in female mice.
The relationship between BMI and carcinoma formation was also investigated. Both cohorts were merged and treated as a single dataset in order to maximize statistical power. Using the set of 378 mice, of which 262 developed a carcinoma, it was observed that BMI was significantly associated with increased risk of carcinoma development in male mice (p < 1e-3), but not in female mice.
FVB/N and SPRET/Ei mice vary significantly in average weight and adiposity, with FVB mice typically weighing more and having a lower percentage of body mass attributable to fat.
To identify the genetic factors influencing BMI, weight, and length in the backcross population QTL mapping was performed using mice from both cohorts. Multiple significant loci for each phenotype were identified; locations GRCm38, Table 2:
QTL Bmiq5 (body mass index QTL 5) mapped to Chromosome 3 with a LOD score of 3.81 at peak location 96.0 Mb between 41-116 Mb explaining 0.031% of trait variance.
QTL Bmiq6 (body mass index QTL 6) mapped to Chromosome 4 with a LOD score of 5.05 at peak location 126.0 Mb between 111-126 Mb explaining 0.055% of trait variance.
QTL Bmiq7 (body mass index QTL 7) mapped to Chromosome 6 with a LOD score of 3.81 at peak location 80.0 Mb between 25-80 Mb explaining 0.024% of trait variance.
QTL Bmiq8 (body mass index QTL 8) mapped to Chromosome 9 with a LOD score of 3.88 at peak location 42.0 Mb between 25-69 Mb explaining 0.027% of trait variance.
QTL Bmiq9 (body mass index QTL 9) mapped to Chromosome 10 with a LOD score of 6.0 at peak location 15.0 Mb between 13-25 Mb explaining 0.059% of trait variance.
QTL Bmiq10 (body mass index QTL 10) mapped to Chromosome 11 with a LOD score of 3.89 at peak location 51.0 Mb between 36-85 Mb explaining 0.027% of trait variance.
QTL Bmiq11 (body mass index QTL 11) mapped to Chromosome 12 with a LOD score of 3.58 at peak location 8.0 Mb between 8-40 Mb explaining 0.039% of trait variance.
QTL Bmiq12 (body mass index QTL 12) mapped to Chromosome X with a LOD score of 6.76 at peak location 36.0 Mb between 36-78 Mb explaining 0.098% of trait variance.
QTL Bmiq5 and Bmiq8 had significant interactions between sex and genotype. Bmiq5 displayed an increase in BMI for heterozygous female mice relative to homozygous females, but no difference in male mice. Heterozygous male mice at Bmiq8 had approximately 8% higher BMI than homozygous males, while there was no difference between females.
Of the eight Bmiq identified QTL, six (Bmiq5, Bmiq6, Bmiq7, Bmiq8, Bmiq10 and Bmiq11) were significantly associated with papilloma burden in male mice. Of the six QTL, four (Bmiq6, Bmiq7, Bmiq8 and Bmiq10) were not significantly associated with papilloma burden in female mice. Table S3.
QTL Wght7 (weight 7) mapped to Chromosome 4 with a LOD score of 3.51 at peak location 112.0 Mb between 91-148 Mb explaining 0.044% of trait variance.
QTL Wght8 (weight 8) mapped to Chromosome 10 with a LOD score of 4.74 at peak location 15.0 Mb between 5-44 Mb explaining 0.028% of trait variance.
QTL Wght9 (weight 9) mapped to Chromosome 10 with a LOD score of 3.58 at peak location 107.0 Mb between 60-121 Mb explaining 0.052% of trait variance.
QTL Wght10 (weight 10) mapped to Chromosome 11 with a LOD score of 5.31 at peak location at 51.0 Mb between 4-89 Mb explaining 0.052% of trait variance.
QTL Wght11 (weight 11) mapped to Chromosome X with a LOD score of 9.13 at peak location 36.0 Mb between 12-60 Mb explaining 0.194% of trait variance.
QTL Wght12 (weight 12) mapped to Chromosome X with a LOD score of 4.35 at peak location 120.0 Mb between 95-151 Mb explaining 0.137% of trait variance.
QTL Lgth10 (body length 10) mapped to Chromosome 7 with a LOD score of 3.18 at peak location 66.0 Mb between 52-126 Mb explaining 0.033% of trait variance.
QTL Lgth11 (body length 11) mapped to Chromosome 10 with a LOD score of 4.18 at peak location 107.0 Mb between 91-121 Mb explaining 0.07% of trait variance.
QTL Lgth12 (body length 12) mapped to Chromosome 11 with a LOD score of 4.62 at peak location 11.0 Mb between 4-95 Mb explaining 0.06% of trait variance.
QTL Lgth13 (body length 13) mapped to Chromosome X with a LOD score of 5.61 at peak location 48.0 Mb between 12-60 Mb explaining 0.175% of trait variance.
QTL Lgth14 (body length 14) mapped to Chromosome X with a LOD score of 5.62 at peak location 120.0 Mb between 95-151 Mb explaining 0.151% of trait variance.
For the majority of QTL (14 out of 19), the SPRET/Ei allele conferred increased phenotype values, with the FVB/N allele associated with reduced values for QTL Bmiq6, Bmiq11, Lgth11, Wgth7 and Wgth9 (Fig. 3a). The high-BMI allele was also the high-papilloma burden allele, except for the Chromosome 3 QTL Bmiq5.
Using a combined approach involving linkage analysis, gene expression analysis, and gene co-expression network analysis, the Panx3 gene was implicated as simultaneously influencing both BMI and tumorigenesis phenotypes. In skin pre-treatment Panx3 expression levels were correlated with both BMI and tumor burden. Panx3 is also present in a well-conserved lipid metabolism network and Pannexin function attenuates inflammatory signaling. These results strongly suggested that Panx3 links tumorigenesis and BMI at the genetic level.
Analysis Tools