Mapping novel genetic loci associated with female liver weight variations using Collaborative Cross mice

Abstract Background Liver weight is a complex trait, controlled by polygenic factors and differs within populations. Dissecting the genetic architecture underlying these variations will facilitate the search for key role candidate genes involved directly in the hepatomegaly process and indirectly involved in related diseases etiology. Methods Liver weight of 506 mice generated from 39 different Collaborative Cross (CC) lines with both sexes at age 20 weeks old was determined using an electronic balance. Genomic DNA of the CC lines was genotyped with high‐density single nucleotide polymorphic markers. Results Statistical analysis revealed a significant (P < 0.05) variation of liver weight between the CC lines, with broad sense heritability (H 2) of 0.32 and genetic coefficient of variation (CVG) of 0.28. Subsequently, quantitative trait locus (QTL) mapping was performed, and results showed a significant QTL only for females on chromosome 8 at genomic interval 88.61‐93.38 Mb (4.77 Mb). Three suggestive QTL were mapped at chromosomes 4, 12 and 13. The four QTL were designated as LWL1‐LWL4 referring to liver weight loci 1‐4 on chromosomes 8, 4, 12 and 13, respectively. Conclusion To our knowledge, this report presents, for the first time, the utilization of the CC for mapping QTL associated with baseline liver weight in mice. Our findings demonstrate that liver weight is a complex trait controlled by multiple genetic factors that differ significantly between sexes.

known that common diseases associated with liver enlargement may be either due to hepatocyte fatty infiltration and hepatocyte enlargement (alcoholic hepatitis and other causes of fatty liver), or to infiltration of cancer cell deposits, which are growing rapidly (metastatic cancer, lymphoma, hepatoma). Other cases of congestive hepatomegaly could be due to hepatic venous outflow obstruction (congestive heart failure). [6][7][8][9] Liver weight is known to be correlated with body growth, internal organ weight, metabolic trait, age and sex, which are assumed to be controlled by polygenic effects. [10][11][12] Previous studies of body growth and its composition, for purposes of either medical knowledge on growth or meat-producing industries, affirmed and significantly contributed to the understanding of the complex genetic components underlying those age-related traits (ie, selection experiments, quantitative trait loci [QTL] mapping, genomewide association studies [GWAS]). [13][14][15][16][17][18][19] However, the complex genetic background controlling growth in the context of liver weight is still obscure and requires the use of advanced animal models that may enable narrowing the mapped QTL intervals.
Rodent and human physiology are very similar, and various mouse models have been used widely in the study of liver anatomy and function in healthy and diseased forms, noting that animal models known so far in the study of human liver diseases manage to mimic specific features of the human disease but not all. 20,21 Given the wide genetic variation existing between human populations alongside the multiple limitations in human study (eg, weak control for standardized investigations), studying a complex human trait or disease requires a highly genetically diverse mouse population rather than a single mouse model. For this purpose, the Collaborative Cross (CC) mouse population model was designed to provide a new model population dedicated to genetic analysis of complex traits as needed for understanding complex human diseases. 22,23 This unique reference genetic resource comprises a set of approximately 350 recombinant inbred lines (RILs) created from full reciprocal matings of eight divergent strains of mice: A/J, C57BL/6J, 129S1/SvImJ, NOD/ LtJ, NZO/HiLtJ, CAST/Ei, PWK/PhJ and WSB/EiJ. Aiming to create a unique and inexhaustible resource of RILs presenting a large phenotypic and genetic diversity, a controlled randomization was carried out during the breeding process to disband large linkage disequilibrium blocks and to recombine the natural genetic variation of the inbred strains. 24

| Ethical statement
All experimental mice and protocols were approved by the Institutional Animal Care and Use Committee (no. M-10-073 and M-14-007) of Tel-Aviv University (TAU), which adhered to the Israeli guidelines that follow the National Institutes of Health of USA animal care and use protocols.

| Phenotype recording
At 20 weeks old, following 12 weeks of standard rodent diet, mice were killed by cervical dislocation after i.p. injection of anesthetic solution (ketamine/xylazine). Thereafter, mice were dissected and livers collected, and the liver weight of each mouse was determined using an electronic balance.

| Availability of data and materials
Phenotype data presented in this study will be publically available in the Mouse Phenome Database (http://phenome.jax.org) and all SNP genotype data is available at http://mtweb.cs.ucl.ac.uk/mus/www.

| Statistical analysis
Phenotypic variations between the CC lines were calculated by oneway ANOVA using the SPSS version 23 software (SPSS, Chicago, IL, USA). Significant variations were considered at P ≤ 0.05. Estimated heritability (H 2 ) and genetic coefficient of variation (CV G ) were calculated for the phenotypic traits using the ANOVA output ( as shown in our previous report. 52

| CC line marker genotyping
Collaborative Cross lines were genotyped using three different arrays at four inbreeding generation intervals, first with the MDA, consisting of 620 000 SNPs, 27 and later with MUGA, consisting of 7500 markers, and finally with Mega-MUGA genotype array, consisting of 77 800 markers to confirm their genotype status. 51

| Genotype-phenotype linkage analysis
Quantitative trait locus mapping was performed using the baseline liver weight phenotypic data and the genotypic data of the CC lines using HAPPY software. 28 The QTL mapping was performed in three directions, once for the overall mouse population, then separately by sex. The mapping of QTL at SNP interval (Llocus) of CC line (k) was tested using the linear regression framework below, in which the HMM probability of descent from founder strain (s) is denoted by P LK(s) : Significance level presented as the negative log 10 of the P-value of null hypothesis test (R ANOVA). Estimation of genome wide significance was performed by permutation test, in which CC line labels were permuted between the phenotypes. Further details of the QTL approach used in this study are available in our previous studies. [31][32][33][34][35]37 3 | RESULTS

| Liver weight
Our findings demonstrate a significant profile of phenotypic variations between the CC lines for the liver weight, suggesting empirical evidence for the strong genetic component controlling the liver weight. Two-way ANOVA for sex * line interactions was significant (P = 0.01) indicating that females and males differ significantly across CC lines in their liver weight; therefore, we analyzed data separately by sex. Although the tested cohort was kept under controlled, common environmental conditions, one-way ANOVA for variations of mean liver weight in grams (g) between the CC lines showed highly significant variations for the overall population and also separately for both sexes (overall population P = 8.71e −18 , female mice P = 5.65e −09 , male mice P = 2.08e −15

| QTL mapping and founder effect
Initially, 1000 randomization tests were performed to calculate the 5%, 10% and 50% genome-wide significant thresholds. The three thresholds are presented on a Manhattan plot in Figure 2, and found to be logP = 6.43, 6. 16  Finally, the effect of each founder haplotype on liver weight at the QTL on chromosome 8 was calculated as deviation relative to WSB/EiJ, which is arbitrarily assigned the trait effect of 0. Results of this analysis are presented in Figure 3. The locus showed a complex pattern of haplotype effects of the founders, with the wild-derived strains, mainly PWK, playing a major role but other strains also contributing to the overall QTL effect. QTL analysis was not significant neither for males nor for the overall population.  The FTO gene (MGI: 1347093), encoding the fat mass and obesity-associated protein, was the first GWAS-identified obesity and obesity-related trait-associated gene (ie, type 2 diabetes, hip circumference, bodyweight index, bodyweight). 61,62 Owing to the various studies of FTO association with obesity and obesity-related phenotypes, it is now evident that FTO is highly expressed in adipose tissues, playing a crucial role in adipogenesis (cross-talk with Irx3) and extremely involved in early development, and yet its function is still obscure beyond the adipose tissue. 63 In relevance to liver weight phenotype, studies of FTO homozygous, null mice reported postnatal growth retardation accompanied by decreased bodyweight (lean and fat weight), suggesting the complex and major role of FTO in body development and composition, whether independently or by co-regulatory mechanisms with the IrxB cluster and Rpgrip1l gene. 64   Beside protein coding genes and QTL, abundant 5′-C-phosphate-G-3′ islands (CpG islands) (39 features) were located within the significant genomic interval of LWL1, which is known to be associated with promoters of housekeeping genes and genes with a tissuerestricted pattern of expression. 74 CpG island distribution varies F I G U R E 3 Estimated haplotype effect size at chromosome 8 quantitative trait loci (QTL) for liver weight (g) trait. Effects are shown as deviations relative to WSB/EiJ, which is arbitrarily assigned the trait effect of 0. The X-axis represents eight founder strains of the CC lines; Y-axis represents the estimated haplotype effect size of the CC founders within the whole genome to be in preference for gene-rich loci;

| Candidate genes underlying the mapped QTL
advanced studies of epigenetics and DNA methylation propose a possible important role of CGI methylation in mammalian development and cellular differentiation. 75 Table S1). 14  | 217

CONF LICTS OF INTEREST
None.

AUTHOR CONTRIBU TI ONS
HJA participated in the design of the study, carried out the mice assessment, participated in data analysis and drafting the manuscript.