Decomposing genomic variance using information from GWA, GWE and eQTL analysis.

A commonly used procedure in genome-wide association (GWA), genome-wide expression (GWE) and expression quantitative trait locus (eQTL) analyses is based on a bottom-up experimental approach that attempts to individually associate molecular variants with complex traits. Top-down modeling of the entire set of genomic data and partitioning of the overall variance into subcomponents may provide further insight into the genetic basis of complex traits. To test this approach, we performed a whole-genome variance components analysis and partitioned the genomic variance using information from GWA, GWE and eQTL analyses of growth-related traits in a mouse F2 population. We characterized the mouse trait genetic architecture by ordering single nucleotide polymorphisms (SNPs) based on their P-values and studying the areas under the curve (AUCs). The observed traits were found to have a genomic variance profile that differed significantly from that expected of a trait under an infinitesimal model. This situation was particularly true for both body weight and body fat, for which the AUCs were much higher compared with that of glucose. In addition, SNPs with a high degree of trait-specific regulatory potential (SNPs associated with subset of transcripts that significantly associated with a specific trait) explained a larger proportion of the genomic variance than did SNPs with high overall regulatory potential (SNPs associated with transcripts using traditional eQTL analysis). We introduced AUC measures of genomic variance profiles that can be used to quantify relative importance of SNPs as well as degree of deviation of a trait's inheritance from an infinitesimal model. The shape of the curve aids global understanding of traits: The steeper the left-hand side of the curve, the fewer the number of SNPs controlling most of the phenotypic variance.

Genetic influences on growth and body composition in mice: multilocus interactions.

BACKGROUND: The genetic architecture of body weight and body composition is complex because these traits are normally influenced by multiple genes and their interactions, even after controlling for the environment. Bayesian methodology provides an efficient way of estimating these interactions. SUBJECTS AND MEASUREMENTS: We used Bayesian model selection techniques to simultaneously estimate the main effects, epistasis and gene-sex interactions on age-related body weight (at 3, 6 and 10 weeks, denoted as WT3wk, WT6wk and WT10wk) and body composition (organ weights and fat-related traits) in an F(2) sample obtained from a cross between high-growth (M16i) mice and low-growth (L6) mice. RESULTS: We observed epistatic and main-effect quantitative trait loci (QTL) that controlled both body weight and body composition. Epistatic effects were generally more significant for WT6wk than WT10wk. Chromosomes 5 and 13 interacted strongly to control body weight at 3 weeks. A pleiotropic QTL on chromosome 2 was associated with body weight and some body composition phenotypes. Testis weight was regulated by a QTL on chromosome 13 with a significantly large main effect (2log(e)BF approximately 15). CONCLUSION: By analyzing epistatic interactions, we detected QTL not found in a previous analysis of this mouse population. Hence, the detection of gene-gene interactions may provide new information about the genetic architecture of complex obesity-related traits and may lead to the detection of additional obesity genes.

Mapping of 443 porcine EST improves the comparative maps for SSC1 and SSC7 with the human genome.

Numerous mapping studies of complex traits in the pig have resulted in quantitative trait loci (QTL) intervals of 10-20 cM. To improve the chances to identify the genes located in such intervals, increased expressed sequence tags (EST)-based marker density, coupled with comparative mapping with species whose genomes have been sequenced such as human and mouse, is the most efficient tool. In this study, we mapped 443 porcine EST with a radiation hybrid (RH) panel (384 had LOD > 6.0) and a somatic cell hybrid panel. Requiring no discrepancy between two-point and multipoint RH data allowed robust assignment of 309 EST, of which most were located on porcine chromosomes (SSC) 1, 4, 7, 8 and X. Moreover, we built framework maps for two chromosomes, SSC1 and SSC7, with mapped QTL in regions with known rearrangement between pig and human genomes. Using the Blast tool, we found orthologies between 407 of the 443 pig cDNA sequences and human genes, or to existing pig genes. Our porcine/human comparative mapping results reveal possible new homologies for SSC1, SSC3, SSC5, SSC6, SSC12 and SSC14 and add markers in synteny breakpoints for chromosome 7.

Physical mapping of genes in the porcine ovarian transcriptome.

Reproductive efficiency and associated traits are of major economic importance to the swine industry and have been more difficult to improve genetically than other production traits. Integration of phenotypical data with gene mapping and expression studies provides a powerful approach for dissection of the genetic basis regulating complex traits. We developed a total of 101 polymerase chain reaction-based markers, representing 91 unique genes, for expressed sequence tags previously reported to be putatively differentially expressed in the porcine ovarian transcriptome of a swine line selected on an index of high ovulation rate and embryonic survival. These were subsequently used in physical mapping experiments with a porcine radiation hybrid and somatic cell hybrid panels. Our results increased the information content of the porcine physical map useful for comparative mapping by c. 10%. Moreover, the mapped genes are likely to be biologically relevant to the molecular mechanisms that control ovulation rate in the pig. A total of 12 differentially expressed genes were mapped to regions previously reported to contain quantitative trait loci affecting swine ovulation rate.

A genome scan for quantitative trait loci and imprinted regions affecting reproduction in pigs.

Quantitative trait loci for reproductive traits in a three-generation resource population of a cross between low-indexing pigs from a control line and high-indexing pigs from a line selected 10 generations for increased index of ovulation rate and embryonic survival are reported. Phenotypic data were collected in F2 females for birth weight (BWT, n = 428), weaning weight (WWT, n = 405), age at puberty (AP, n = 295), ovulation rate (OR, n = 423), number of fully formed pigs (FF, n = 370), number of pigs born alive (NBA, n = 370), number of mummified pigs (MUM, n = 370), and number of stillborn pigs (NSB, n = 370). Grandparent, F1, and F2 animals were genotyped for 151 microsatellite markers. Sixteen putative QTL (P < 0.10) for reproductive traits were identified in previous analyses of these data with single QTL line-cross models. Data were reanalyzed with multiple QTL models, including imprinting effects. Data also were analyzed with half-sib models. Permutation was used to establish genome-wide significance levels ( = 0.01, 0.05, and 0.10). Thirty-one putative QTL for reproductive traits and two QTL for birth weight were identified (P < 0.10). One Mendelian QTL for FF (P < 0.05), one for NBA (P < 0.05), three for NSB (P < 0.05), three for NN (P < 0.05), seven for AP (P < 0.10), five for MUM (P < 0.10), and one for BWT (P < 0.10) were found. Partial imprinting of QTL affecting OR (P < 0.01), BWT (P < 0.05), and MUM (P < 0.05) was detected. There were four paternally expressed QTL for NN (P < 0.10) and one each for AP (P < 0.05) and MUM (P < 0.10). Maternally expressed QTL affecting NSB (P < 0.10), NN (P < 0.10), and MUM (P < 0.10) were detected. No QTL were detected with half-sib analyses. Multiple QTL models with imprinting effects are more appropriate for analyzing F2 data than single Mendelian QTL line-cross models.

Quantitative genomics: exploring the genetic architecture of complex trait predisposition.

Most phenotypes with agricultural or biomedical relevance are multifactorial traits controlled by complex contributions of genetics and environment. Genetic predisposition results from combinations of relatively small effects due to variations within a large number of genes, known as QTL. Well over 200 QTL have been reported for growth and body composition traits in the mouse, which likely represent at least 50 to 100 distinct genes. Molecular biology has yielded significant advances in understanding these traits at the metabolic and physiological levels; however, little has been learned regarding the identity and nature of the underlying polygenes. In addition to the significantly poor precision inherent to QTL localization, it is very difficult to differentiate between co-localization and coincidence when comparing QTL with other QTL and with potential candidate genes. The wide gap between our knowledge of physiological mechanisms underlying complex traits and the nature of genetic predisposition significantly impairs discovery of genes underlying QTL. Identification and genetic mapping of key transcriptional, proteomic, metabolomic, and endocrine events will uncover large lists of significant positional candidate genes for growth and body composition. However, integration of experimental approaches to jointly evaluate predisposition and physiology will increase success of QTL identification by merging the power of recombination with functional analysis. Measuring physiologically relevant subphenotypes within a structured QTL mapping population will not only facilitate pathway-specific prioritization among candidate genes, but may also directly identify genes underlying QTL. This would advance our understanding of the genetic architecture of complex traits by testing the central hypothesis that genes controlling predisposition to a quantitative trait are primarily involved in trans-regulation of the primary physiological pathways that regulate the trait.

Evaluation of gene expression in pigs selected for enhanced reproduction using differential display PCR and human microarrays: I. Ovarian follicles.

Differential display PCR (ddPCR) and complementary DNA microarray analyses were used to evaluate gene expression differences in porcine ovarian follicles between a line of pigs selected for an index of ovulation rate and embryo survival (Line I) and its randomly selected control line (Line C). Follicles (4.0 to 7.0 mm) were dissected from ovaries of multiparous sows (n = 27) at either 2 or 4 d following PGF2alpha analog injection on d 12 to 14 of the estrous cycle. Using ddPCR, differentially expressed bands (n = 282) were excised from gels and 107 were sequenced, yielding 84 unique porcine follicle expressed sequence tags. Northern hybridization confirmed differential expression (between lines, days, or follicle sizes) for messenger RNA representing the calpain I light subunit, cytochrome C oxidase subunit III, cytochrome P450 aromatase, and cytochrome P450 side chain cleavage genes. For microarray analysis, two mRNA pools representing follicles (d 2; 4.50 to 4.75 mm) from Line I and Line C sows were hybridized to the Incyte UniGEM V1.0 human chip (approximately 7,000 gene probes). A second analysis was performed using mRNA from follicles (d 2; 4.50 to 5.00 mm) hybridized to the Incyte UniGEM V2.0 human chip (approximately 9,100 gene probes). A total of 33 and 21 genes were identified with significant expression differences using UniGEM V1.0 and V2.0, respectively (twofold or greater relative expression following adjustment for expression of control probes). However, there was little overlap between results of the two hybridizations. Expression differences between lines for two genes, follistatin and nuclear receptor subfamily 4, group A, member 1, were confirmed using Northern hybridization. These results demonstrate changes in follicular gene expression as the result of long-term selection for enhanced reproduction. These correlated responses may directly represent allelic variation utilized by selection (e.g., quantitative trait loci), or more likely, transcriptional changes in other genes that interact with reproductive QTL. This work represents one of the first applications of gene expression analysis to evaluate long-term selection response in livestock populations.

Evaluation of gene expression in pigs selected for enhanced reproduction using differential display PCR: II. Anterior pituitary.

The objective of this study was to identify differentially expressed genes in the anterior pituitary (AP) of sows selected for enhanced reproductive phenotypes. Selection in the Index (I) line was based on an index of ovulation rate and embryo survival, whereas random selection was used in the Control (C) line. Average numbers of fully formed piglets at birth were 12.5 +/- 1.5 and 9.9 +/- 2.0 for Line I and C sows used in this study, respectively. In order to induce luteolysis and synchronize follicle development, sows were injected (i.m.) with 2 mL of prostaglandin F2alpha analog between d 12 and 14 of the estrous cycle. Tissue was harvested 2 d (d2) or 4 d (d4) after injection, resulting in four experimental groups: Cd2 (n = 6), Cd4 (n = 4), Id2 (n = 6), and Id4 (n = 7). Differential display PCR (ddPCR) was used to search for transcriptional changes between selection lines in the AP, using samples within line but pooled across days. Northern hybridization was used to confirm ddPCR results. For ddPCR, two pools were used from each line (C and I). Three genes were confirmed to be differentially expressed between Lines I and C: G-beta like protein, ferritin heavy-chain, and follicle stimulating hormone beta subunit, whereas many other expressed sequence tags were observed to be differentially expressed but still require confirmation. Our findings indicate that long-term selection to increase ovulation rate and decrease embryo mortality has altered transcriptional patterns in the anterior pituitary, most likely as correlated responses.

Mapping of porcine ESTs obtained from the anterior pituitary.

We report the physical mapping of porcine expressed sequence tags (ESTs) from anterior pituitary clones isolated by differential display PCR in a study using lines selected for reproduction. These ESTs were mapped using a somatic cell hybrid panel (SCHP) and a radiation hybrid panel (IMpRH) as follows (SCHP position, nearest marker on the RH map): SPARCL1 (8q23-q27, SSP1); ATF4 (5p11-p15, AC02); MEF2C [2(1/2q21)-(1/2q22), SW2134]; FTH1 (2p14-p17, SWR783); FRAP1 (6q22-q23, SW1355); PBP (14, SW2508); LOC92004 [13q23-(1/2q41), CP]; and PGRMC1 [Xq22, SW1943]. All RH assignments were at LOD score >6.0 except for PGRMC1 at LOD score 5.4. ESTs TCP1 [12p11-(2/3p13)], SF3B1 (15q23-q26) and Clock (8q11-q12) were assigned using only the SCHP. The map position of SPARCL1 coincides with a quantitative trait loci (QTL) for age at puberty found in the University of Nebraska selection lines. Physical mapping of ESTs reported in the present study contributes to characterization of the transcriptome of anterior pituitary of pigs, adds new information to the public database of the porcine genome expression map, and further develops the porcine-human comparative map.

Predictors of marker-informativeness for an outbred F2 design.

Generalization of the polymorphism information content (PIC) index to represent marker informativeness (MI) for a three-generation F2 design requires that two additional sources of non-informativeness be added to the PIC formula: the probability of matings between like-heterozygous F1 individuals, of which one is non-informative; and that of matings between like-heterozygous F1 individuals, which are both fully informative but where line of origin of the same alleles is reciprocal. Given the dense marker-maps currently available for some species, this F2 informativeness parameter constitutes the natural criterion for marker selection in F2 designs, and two computer programs to predict MI from grandparental marker-genotypes were developed for an F2 population originating from two divergent selection lines of outbred mice (F approximately 0.2). A total of 403 markers had been genotyped for the F0 grandparents (n=31), and 14 markers had also been genotyped in the complete pedigree including 559 F2 individuals. One program was based on assumptions of random-mating (RM), while the other (PED) accounted for the pedigreed mating structure. For the 403 markers, the correlation between MI from RM and from PED was 0.95, and the average deviation between the two predictions was 0.005 MI units (MI ranged from 0 to 1). Correlations between predicted and realized MI for the 14 fully genotyped markers were 0.97 for PED and 0.94 for RM, while the corresponding average of deviations between predicted and actual values were 0.01 and 0.04, respectively. Absolute deviations from realized MI never exceeded 0.09 and 0.16 for PED and RM, respectively. Simulated optimization of the mating system to maximize average MI of 28 markers on one chromosome led to improvements in the range of 15-20% average MI (0.07-0.09 MI units). The degree of relative advantage conferred by the F2 generalization of the PIC index over the traditional index was found to be of minor significance.

Localization of a recessive juvenile cataract mutation to proximal chromosome 7 in mice.

OBJECTIVE: To localize the chromosomal position of a novel cataract mutation (juvenile recessive cataract; jrc) in mice. METHODS: A mapping population was developed by crossing cataract males (albino MH) to wild-type females (black C57BL/6J). F1 females were backcrossed to albino MH males with cataracts. RESULTS: The results were consistent with a model of a single autosomal recessive gene [153 cataract, 169 wild-type; chi2 = 0.8, 1 degree of freedom (d.f.), p > 0.35]. Linkage with the albino (tyrosinase; Tyr) locus was evident (chi2 = 61.5, 1 d.f., p < 0.0001), implicating chromosome 7 as the location of jrc. Recombination percentages (+/- SE) between jrc and D7Mit340 (1.2 cM location), D7Mit227 (16.0 cM) and D7Mit270 (18.0 cM) were 17.1 +/- 2.1, 3.7 +/- 1.1 and 6.2 +/- 1.3%, respectively. Multi-point mapping determined that the most likely order of these loci is D7Mit340 - jrc - D7Mit227 - D7Mit270 - Tyr. Although animals with the mutant phenotype appeared to have little or no sense of sight, their growth was not different (p >0.20) from that of normal mice. CONCLUSION: The jrc mutation model may be useful in the study of the genetics of cataracts in other animal species, including humans.

Identification of quantitative trait loci affecting reproduction in pigs.

The objective of this research was to identify chromosomal regions harboring QTL affecting reproduction in pigs. A three-generation resource population was developed by crossing low-indexing pigs from a randomly selected control line (C) with high-indexing pigs of a line selected for increased index of ovulation rate and embryonic survival (I). Differences between Lines I and C at Generation 10 were 6.7 ova and 3.3 fetuses at 50 d of gestation and 3.1 fully formed and 1.6 live pigs at birth. Phenotypic data were collected on F2 females, born in three replicates, for ovulation rate (n = 423), age at puberty (n = 295), litter size (n = 370), and number of nipples (n = 428). Litter-size data included number of fully formed, live, stillborn, and mummified pigs. Grandparent, F1, and F2 animals were genotyped for 151 microsatellite markers distributed across all 18 autosomes and the X chromosome. Genotypic data were available on 423 F2 females. Average spacing between markers was 19.3 Kosambi centimorgans. Calculations of logarithms of odds (LOD) scores were by least squares, and fixed effects for sire-dam combination and replicate were included in the models. Genome-wide significance level thresholds of 5% and 10% were calculated using a permutation approach. There was evidence (P < 0.05) for QTL affecting ovulation rate on SSC9, age at puberty on SSC7 and SSC8, number of nipples on SSC8 and SSC11, number of stillborn pigs on SSC5 and SSC13, and number of fully formed pigs on SSC11. There was evidence (P < 0.10) for additional QTL affecting age at puberty on SSC7, SSC8, and SSC12, number born live on SSC11, and number of nipples on SSC1, SSC6, and SSC7. Litter size is lowly heritable and sex-limited. Therefore, accuracy of selection for litter size may be enhanced by marker-assisted selection. Ovulation rate and age at puberty are laborious to measure, and thus marker-assisted selection may provide a practical and efficient method of selection.

Candidate gene analysis for loci affecting litter size and ovulation rate in swine.

A candidate gene approach was used to determine whether specific loci explain responses in ovulation rate (OR) and number of fully formed (FF), live (NBA), stillborn, and mummified pigs at birth observed in two lines selected for ovulation rate and litter size compared with a randomly selected control line. Line IOL was selected for an index of OR and embryonic survival for eight generations, followed by eight generations of two-stage selection for OR and litter size. Line C was selected at random for 16 generations. Line COL, derived from line C at Generation 8, underwent eight generations of two-stage selection. Lines IOL and C differed in mean EBV by 6.1 ova and 4.7 FF, whereas lines COL and C differed by 2.2 ova and 2.9 FF. Pigs of Generation 7 of two-stage selection lines were genotyped for the retinol binding protein 4 (RBP4, n = 190) and epidermal growth factor (EGF, n = 189) loci, whereas pigs of Generations 7 and 8 were genotyped for the estrogen receptor (ESR, n = 523), prolactin receptor (PRLR, n = 524), follicle-stimulating hormone beta (FSHbeta, n = 520), and prostaglandin-endoperoxide synthase 2 (PTGS2, n = 523) loci. Based on chi-square analysis for homogeneity of genotypic frequencies, distributions for PRLR, FSHbeta, and PTGS2 were different among lines (P < 0.005). Differences in gene frequencies between IOL vs C and COL vs C were 0.33 +/- 0.25 and 0.16 +/- 0.26 for PRLR, 0.35 +/- 0.20 and 0.15 +/- 0.24 for FSHbeta, and 0.16 +/- 0.16 and 0.08 +/- 0.18 for PTGS2. Although these differences are consistent with a model of selection acting on these loci, estimates of additive and dominance effects at these loci did not differ from zero (P > 0.05), and several of them had signs inconsistent with the changes in allele frequencies. We were not able to find significant associations between the polymorphic markers and phenotypes studied; however, we cannot rule out that other genetic variation within these candidate genes has an effect on the traits studied.

Comparative mapping of RPL3, a gene overexpressed in multiple obesity models.

The ribosomal protein 3 gene is differentially expressed in hypothalamus and brown adipose tissue between mouse lines divergently selected for heat loss, and in skeletal muscle of the ob/ob mouse model. Unfortunately, multiple Rpl3-processed pseudogenes have hampered mapping of the functional gene copy in mammalian species. Using PCR amplification with intronic primer binding, we have mapped Rpl3 to MMU15, and have also localized RPL3 to BTA5 in cattle. Comparative mapping implicates a previously mapped copy of RPL3 on HSA22 as the functional copy of human RPL3, while predictive mapping places the porcine homologue on SSC5.

Applying functional genomics research to the study of pig reproduction.

Functional genomics is an experimental approach that incorporates genome-wide or system-wide experimentation, expanding the scope of biological investigation from studying single genes to studying potentially all genes at once in a systematic manner. This technology is highly appealing because of its high throughput and relatively low cost. Furthermore, analysis of gene expression using microarrays is likely to be more biologically relevant than the conventional paradigm of reductionism, because it has the potential to uncover new biological connections between genes and biochemical pathways. However, functional genomics is still in its infancy, especially with regard to the study of pig reproduction. Currently, efforts are centred on developing the necessary resources to enable high throughput evaluation and comparison of gene expression. However, it is clear that in the near future functional genomics will be applied on a large scale to study the biology and physiology of reproduction in pigs, and to understand better the complex nature of genetic control over polygenic characteristics, such as ovulation rate and litter size. We can look forward to generating a significant amount of new data on differences in gene expression between genotypes, treatments, or at various temporal and spatial coordinates within a variety of reproductively relevant systems. Along with this capability will be the challenge of collating, analysing and interpreting datasets that are orders of magnitude more extensive and complex than those currently used. Furthermore, integration of functional genomics with traditional genetic approaches and with detailed analysis of the proteome and relevant whole animal phenotypes will be required to make full use of this powerful new experimental paradigm as a beneficial research tool.

Gene expression in hypothalamus and brown adipose tissue of mice divergently selected for heat loss.

Gene expression was evaluated in mice divergently selected for 16 generations for heat loss, measured by direct calorimetry. The high (MH) heat loss line has approximately 50% greater heat loss, approximately 35% less body fat, approximately 20% greater feed intake, and twofold greater activity levels than the low (ML) heat loss line. At 11 wk, inbred males (developed from MH and ML) were euthanized 3 h after dark for dissection of tissues and extraction of RNA. Differential display PCR (DD-PCR) was used to evaluate transcriptional differences between lines in hypothalamus and brown adipose tissue (BAT). Evaluation was replicated within and across lines, using family pools of mRNA. Two genes were confirmed by competitive RT-PCR and/or Northern analysis to have greater levels of mRNA present in ML relative to MH mice. In both hypothalamus and BAT, the ribosomal protein L3 (RPL3) gene was expressed at higher levels in ML, whereas an unknown expressed sequence tag (EST) was also found at higher levels in the hypothalamus of ML mice. These results implicate RPL3 in regulation of energy balance and extend the genetic dissection of response to selection to the transcriptional level.

Quantitative trait loci for directional but not fluctuating asymmetry of mandible characters in mice.

Non-directional variation in right minus left differences in bilateral characters, referred to as fluctuating asymmetry (FA), often has been assumed to be largely or entirely environmental in origin. FA increasingly has been used as a measure of developmental stability, and its presumed environmental origin has facilitated the comparisons of populations believed to differ in their levels of stability. Directional asymmetry (DA), in which one side is consistently larger than the other, has been assumed to be at least partially heritable. Both these assumptions were tested with interval mapping techniques designed to detect any quantitative trait loci (QTLs) affecting FA or DA in 15 bilateral mandible characters in house mice resulting from a cross of the F1 between CAST/Ei (wild strain) and M16i (selected for rapid growth rate) back to M16i. For purposes of the analysis, all mandibles were triply measured and 92 microsatellite markers were scored in a total of 350 mice. No significant QTLs were found for FA, but three QTLs significantly affected DA in several characters, confirming both assumptions. The QTLs for DA were similar in location to those affecting the size of several of the mandible characters, although they accounted for an average of only 1% of the total phenotypic variation in DA.