CTC
2nd Annual Meeting
1-
List
of Abstracts
·
Abstracts
not marked as [Poster Presentation] will be presented as 20 minute talks.
·
The
person presenting each talk is indicated in bold in the author list.
·
Abstracts
are sorted by the surname of the presenting author, except for the Catalyst and
the Keynote Address.
Catalysts
for an Open Discussion by CTC Members
Moderated by Robert W. Williams, Ken Paigen, David Threadgill, and Kent Hunter
Goal of the CTC: To increase the visibility of complex traits, and the quantitative trait loci (QTL) that contribute to these traits, in contemporary biomedical research through 1. improved awareness and appreciation for the natural variation that exists within species that is responsible for significant phenotypic variation; 2. recognition by the broader community of scientists that the vast majority of human phenotypes have complex multigenic components whose contributions need to be understood to obtain comprehensive understanding of relevant human phenotypes; 3. acceptance that the environment, through complex interactions with genetic variation, is a critical determinant of phenotypic diversity; and 4. development of community-wide resources that enable efficient experimental dissection of the causes of phenotypic diversity.
These goals will be met by planning, generating, and freely distributing more powerful and accessible resources, methods, and reagents for the analysis of complex traits. We need these resources to understand the basis of pervasive human ailments that result from still poorly defined interactions of genetic and environmental variables. Current approaches in functional genomics typically exploit single gene variants on single genetic backgrounds. The CTC will advance complementary approaches for analysis of polygenic traits on more complex genetic backgrounds, even in multiple environments. Although statistically more demanding, this complementary strategy will provide researchers with realistic and rewarding resources to model human populations and to identify the causes and potential treatments for human ailments that are not realistically alleviated by single-gene manipulation. Researchers are rapidly acquiring many of the key computational and genomics tools to successfully analyze complex multifactorial variation and disease. The ultimate goal is to accurately predict genetic susceptibility and phenotypes in humans as a function of allelic variation and environmental exposure.
Outreach and translation: A major emphasis of any community-project must be inclusion. The resources and tools assembled and developed by the CTC will be done so with the expectation that they will encourage adoption by a larger community of biologists, statisticians and clinicians who appreciate the translational potential of mouse models but do not necessarily have the genetic, experimental or computational support to exploit mouse models of human diseases efficiently.
Specific goals: that the CTC intends to reach include: 1. significant improvements in the precision of QTL mapping; 2. resources to characterize 2- and 3-way1 epistatic interactions with good power; 3. means to efficiently study gene pleiotropy and genetic heterogeneity; 4. resources that encompasses great genetic diversity; and 5. renewable resources that can be studied under different conditions, treatments, environments and, equally important, can be distributed between laboratories. Implicit in these specific goals is the continued development and improvement of statistical and computational methodologies for investigating and modeling complex traits. [The reason I keep going back to 3-way interactions that that we must keep the future in mind. An interactions can be genetic or non-genetic so a 2-way will only allow gene x environment while a 3-way will support gene x gene x environment or gene x environment x environment which will be critical in understanding the role of the environment in determining phenotypes.]
Multiple projects: The CTC should support several resources to achieve long-term and near-term goals. One key resource is the adoption and generation of one or more large reference panels of isogenic mice (inbred, hybrid, or even cloned) that can be used for collaborative and integrative genetic analyses. Two current, complementing but not exclusive implementations are the 1K Collaborative Cross (1st CTC Workshop Report) discussed at last year's meeting, and the Phenome Project (Paigen and Eppig, 2000; Bouge et al. 2002 CTC).
In addition to these mouse resources, the CTC encourages generation and dissemination of better genetic, genomic, and bioinformatic reagents and resources. Specific examples include support for sequencing, genotyping, and phenotyping lines of mice. New SNPs and microsatellites are already being validated and typed on large panels of inbred strains by members of the CTC. The CTC supports full sequencing of 10 or more strains of mice in the next decade. Finally, we actively support the continued development of very high throughput phenotyping methods of the type being developed by mutagenesis programs.
3. Generating Community and Financial Support: CTC-sponsored projects need support of 1. CTC members; 2. the broader community of mouse geneticists and molecular biologists; 3. the clinical and human genetics community; and 4. funding agencies. We have made great headway on #1, and are beginning to have some impact on #2, but now need to increase our profile and communicate our ideas more forcefully and effectively to #3 and #4. How do we do this? What do we offer that is unique, that is essential? How high do we shoot? How many projects can our community support? How much money do we need to ensure our efforts are successful?
Some technical hurdles: What are the current technical, bioinformatic, and computational hurdles preventing truly efficient and powerful complex trait analysis? Can we demonstrate the power and precision that we can achieve with different types of resources? We need simulations. A proposal by RW for a Mapping Challenge Data Set based on the 1K Collaborative Cross or a heterogeneous population such as represented by the Phenome Strain Set with a suitable prize.
Collaborative Cross Design Criteria: and the Gritty/Crucial Details
1. Criteria for inclusion of parents in generating a Collaborative Cross.
2. How much genetic diversity can we encompass and handle?
3. Numbers of parental strains as a function of numbers of final RI lines.
4. Issues regarding breeding efficiency and inbreeding loss.
5. Are we planning for an RI set that by itself can reach single gene identification of QTL or are we planning for a set that clones QTLs in combination with other sources of analysis (expression profiling, informatics, sequence data, etc.)
6. What other resources are likely to be most useful for gene identification?
7. What genetic resolution is desirable in terms of a cost/benefit analysis.
8. How far do we want to go in identifying epistatic interactions - pairs, triples, quadruples, more?
9. To what extent do RIX mice add to our ability to define the location of QTL?
10. How do we propose to use RIX mice?
11. What is the optimal number of generations and parallel crosses to keep going in creating RI lines. Is it better to use more pens and reduce the probability of losing lines, or keep the number of pens per line down and accept that more lines are lost from a larger pool of lines.
12. What does the cost analysis of generating lines look like when we include both fixed costs and variable costs?
Keynote Address:
Polygenes and the prevention of Cancer
Bruce Ponder, University of Cambridge, CRUK Department of Oncology,
Strangeways Research Laboratories, Worts Causeway, Cambridge CB1 8RN.
Most common cancers in man show a tendency to familial clustering. The risk to a close relative of a case is increased 2- to 3-fold. Most of this effect is genetic in origin. Our analysis of familial clustering of breast cancer suggests that known strongly predisposing genes such as BRCA1/2 account for only 20% of the total genetic predisposition. A substantial part of the remaining 80% may be attributed to the combination of many common genetic variants, individually of small effect.
Modelling based on population based epidemiological studies suggests a log-normal distribution of risk, such that women in the top 20% have a 30-40 fold higher risk of breast cancer than those in the lowest 20%. If this is even approximately true, there are obvious and important implications for screening and prevention.
I will discuss these data, and the progress and obstacles in finding the genes and other factors that would make up an individual risk profile.
Genome Resources and Identification of Complex Trait Genes
Timothy J Aitman
Physiological Genomics and Medicine Group, MRC Clinical Sciences Centre, Imperial College, Hammersmith Campus, London W12 0NN
A wealth of genome resources, including whole genome sequences, is now available for a wide range of species. These have already contributed to identification of the genes underlying hundreds of essentially Mendelian traits in humans, mice and other species. The present challenge of identifying complex trait genes is facilitated by genome resources, but most strategies still depend on accurate localisation of susceptibility genes by genetic linkage, and restriction of the critical interval using, for example, congenic lines, linkage disequilibrium, outbred heterogeneous stocks and progeny testing. Genome resources help principally in identification of positional candidate genes in regions of linkage. The imprecise relationship between genotype and complex trait phenotype usually demands some form of complementation test for proof of gene identity. We have used microarrays to help identify positional candidate genes in regions of linkage, and transgenic complementation to assist in proof of gene identity. Our finding of a chromosomal deletion at the Cd36 locus as a cause of insulin resistance and dyslipidaemia in spontaneously hypertensive rats raises the question of what type of sequence variants are frequent causes of genetically complex traits. The small number of complex trait genes identified to date makes this issue hard to resolve at the present time, but efficient strategies for identifying complex trait genes rest on better knowledge of the range of sequence variants that commonly underlie complex traits.
Genetics of Promotion Susceptibility in
the Two-Stage Skin Tumor Model
JM Angel, M Caballero, and J DiGiovanni.
The University of
Epidemiologic
data suggest that cancer susceptibility in the general population is a function
of multiple, poorly penetrant modifier genes each of
which contributes to, but is not solely responsible for predisposition to
developing a particular type of cancer after exposure to certain environmental
carcinogenic agents. Genetic differences in susceptibility to two-stage skin
carcinogenesis have been known for many years and the major contribution to
susceptibility appears to be at the level of tumor promotion. Studies of
crosses of sensitive DBA/2 and C3H with resistant C57BL/6 mice suggest that
susceptibility to 12-O-tetradecanoylphorbol-13-acetate (TPA) skin tumor
promotion is a multigenic trait. Furthermore, the data suggest that one locus
increases promotion susceptibility when inherited from the resistant C57BL/6
parent. One promotion susceptibility
locus, Psl1, was mapped to an ~40 cM region of distal chromosome 9 and other loci were
tentatively mapped to several unlinked chromosomal regions. Results from tumor studies using interval
specific congenic mouse strains suggest that at least
three genes that modify the response to TPA skin tumor promotion map within
this 40 cM region of distal chromosome 9.
Inheritance of the DBA/2 allele of two of these genes results in
increased sensitivity to TPA while inheritance of the DBA/2 allele of the third
gene results in decreased sensitivity. In addition, analysis of TPA sensitivity
in (C57BL/6
BD27)F2, in which Psl1
does not segregate, supports the mapping of TPA susceptibility loci to mouse
chromosomes 1 (Psl3), 2 (Psl2), and 19 (Psl4). Inheritance of the DBA/2 allele of Psl2 results in increased sensitivity to TPA. In contrast, inheritance of the C57BL/6
allele of Psl3 results in increased
sensitivity to TPA. Inheritance of the
C57BL/6 allele of Psl4 results in
decreased sensitivity to TPA. The identification and characterization of the
specific genes that modify the response to TPA skin tumor promotion will aid in
tumor diagnosis, prognosis, and intervention.
Supported by NIEHS grant ES08355,
Genetic dissection of sweet taste
AA Bachmanov1, DR Reed1, X Li1, M. Inoue2, GK Beauchamp1, MG Tordoff1 [ bachmanov@monell.org ]
1Monell Chemical Senses Center, Philadelphia, PA, USA 2Laboratory of Cellular Neurobiology, School of Life Science, Tokyo University of Pharmacy and Life Science, Hachioji, Tokyo, Japan
Sweet taste perception is initiated when a sweetener molecule interacts with a G protein-coupled receptor in the taste buds of the oral cavity. The taste information is conveyed through afferent gustatory nerves to the brain and evokes physiological and behavioral responses. We have used genetic analyses to gain insight into mechanisms of sweet taste. Examples of specific biological questions that we address are: What are the sweet taste receptors? What are the ligands for these receptors? What is the relative contribution of peripheral taste sensitivity and central processes in determining behavioral consummatory responses to sweeteners? How does variation in sweet taste responsiveness affect other traits, for example alcohol consumption?
To address these questions we have used several genetic approaches. A genome scan of the offspring of mice from the C57BL/6ByJ and 129P3/J strains phenotyped with several sweeteners and ethanol showed individual genetic architecture for consumption of different taste compounds. The Sac (saccharin preference) locus identical to the Tas1r3 taste receptor gene affects, to different degrees, the behavioral responses to some but not all sweeteners, and also to ethanol. Consistent with its role in the initial step of the taste transduction, allelic variation of the Tas1r3 gene affects afferent activity of the chorda tympani gustatory nerve in response to lingual application of certain sweeteners. To assess the functional role of the Tas1r3 sequence variants, we have analyzed genotype-phenotype associations in multiple inbred mouse strains, and across species, from rodents to primates. The ligand specificity of the taste receptor encoded by the Tas1r3 gene was examined using mice from the 129.B6-Sac congenic strain tested in behavioral and electrophysiological experiments with a wide array of chemically diverse sweeteners. The role of genes other than Sac/Tas1r3 in response to sweeteners and ethanol is being studied using a genome scan of the F2 intercross between 129.B6-Sac congenic and C57BL/6ByJ inbred strains, and subsequent short-term phenotype-based selection.
Acknowledgments: Supported by NIH
grants R01DC00882 (GKB), R03DC03509, R01DC04188 and R01DK55853 (DRR),
R01AA11028 (MGT) and R03DC03853 (AAB).
Identification of Aurora2/Stk6/STK15 as a candidate low penetrance tumor susceptibility gene in mouse and man.
A.Balmain,
Mouse models have been used to identify
Quantitative Trait Loci (QTLs) that control
susceptibility to cancer and other complex diseases. Locations of QTL are
frequently determined by linkage analysis of crosses between inbred mouse
strains, an approach that usually identifies regions of 10-20cM containing the
critical gene(s). We have combined linkage analysis with haplotype mapping in interspecific (M.musculus X M.spretus) mouse crosses, which can
refine the regions of interest from 20cM to about 1-2cM. Using this approach,
we localised a skin tumor susceptibility locus on
distal mouse chromosome 2 to an interval of 1.5cM that contained the gene
encoding Aurora2 (also known as Stk6 in mouse and STK15 in human) as a candidate skin tumor
susceptibility gene. The Aurora2/Stk6
allele inherited from the susceptible musculus parent was over-expressed in normal cells, and
preferentially amplified in tumor cells from F1 hybrid mice. A common genetic
variant was identified in the human orthologue of Aurora2/STK15 (Ile31) that is
preferentially amplified and associated with the degree of aneuploidy
in human colon tumors. The Ile31 variant transforms rat 1 cells more potently
than the more common Phe31 variant allele. These results are consistent with an
important role for the Ile31 variant of Aurora2/STK15 in human cancer
susceptibility.
A
statisticians point of view on the significance of QTL
Yoav Benjamini,
Professor of Statistics,
Over the last year a lively discussion went on among the participants of the CTC regarding QTLs. One of the major issues discussed was related to the assessment of the significance of a QTL. I shall make a few comments on some of the arguments raised during the discussion, and in particular on the use of the False Discovery Rate (FDR) approach to the problem. I shall highlight fundamental difficulties in the current approach, and point at possible solutions in other directions.
Mouse Phenome
Database: Research tool and data resource
Molly Bogue, Jill Marcus, Stephen Grubb
The Jackson Laboratory,
mollyb@jax.org http://www.jax.org/phenome
The discovery of defined regions of high and low genetic variation inherited from various progenitor subspecies among inbred strains holds great promise to make mapping and positional cloning of disease related genes more feasible. High-resolution haplotype maps of a defined set of strains combined with sophisticated delineation of their phenotypic variation and gene expression patterns will enable complex trait analysis on an unprecedented scale.
Recombinant inbred lines, F1 hybrids, consomics, and other advanced crosses are critical for complex trait analysis and high-resolution mapping. Because substantial investments are required for producing these designer strains, it is cost effective to make well-informed decisions that will maximize community resources. To knowledgeably select progenitor strains for generating new lines, inbred strains must be comprehensively characterized by measuring parameters relevant to human health.
In combination with empirical data, sufficient genotypic profiles are required to fully tap the potential of inbred strains for choosing optimal strains for selective breeding strategies and many other research applications.
The Mouse Phenome Project is an international collaborative effort to promote the quantitative phenotypic characterization of mouse strains under standardized conditions and to make the data publicly available. Members of the research community are contributing valuable datasets for inclusion in the Mouse Phenome Database (MPD). Data are already available for a number of metabolic, developmental, and behavioral parameters on as many as 40 different inbred strains. In addition, we are collaborating in a number of phenotyping initiatives and coordinating with large-scale genotyping consortia. So far, over 600 SSLP for 47 strains and 150,000 SNPs have been contributed to the MPD for worldwide access.
Among those strains with dense haplotype maps, the MPD has identified compelling differences in a number of vital traits, including white blood cell differentials, gallstone susceptibility, heart rate variability, diet-related HDL cholesterol levels, lung function, and anxiety-related behaviors.
The Mouse Phenome Project is providing and maintaining a public database for data storage, access, and analysis. We will coordinate with large-scale consortia to develop integrated datasets that enable complex trait analysis and functional annotation of the mouse genome. We are eager to assist in community efforts to build novel genetic resources and are seeking community input to define directives for these activities and to expand the MPD repertoire of tools that will permit more sophisticated queries and analyses.
Genetical Genomics to Identify Gene Pathways
Regulating Hematopoietic Stem Cells.
Leonid Bystrykh1, Ellen Weersing1, Sue Sutton5, Bert Dontje1, Edo Vellenga2, Jintao Wang3 Kenneth F. Manly3, Robert W. Williams4, Michael Cooke5, Gerald de Haan1
1Department of Stem Cell Biology,
2Department of Hematology,
3Molecular & Cellular Biology, Roswell Park Cancer Institute,
4Department of Anatomy and
Neurobiology, University
of
5Genome Institute of the Novartis
Research Foundation,
We have used a genetical genomics approach to identify gene pathways regulating hematopoietic stem cell (HSC) proliferation and frequency. Large-scale mRNA expression studies were performed using purified HSC isolated from a fully genotyped pedigree of recombinant inbred BXD mouse strains, and quantitative trait loci associated with variation in expression of each transcript were mapped. This analysis immediately identified cis-acting polymorphic stem cell genes, but in addition, multiple strongly trans-acting controlling loci were discovered that were associated with most of the differentially expressed genes. Detailed analysis of these co-regulated transcripts suggests that they very likely constitute pathways specifying stem cell gene networks. We identified two pathways that were biologically linked with stem cell proliferation and frequency. Candidate genes affecting these traits map to chromosome 11 and 18 respectively, but in human are highly linked at 5q31.1. Deletions in this region are associated with acute myeloid leukemia and myelodysplastic syndrome.
Our results, together with recent reports using similar approaches, document the power of genetical genomics to dissect complex traits. Entire gene pathways critically associated with phenotypic differences immediately become identifiable as collections of co-regulated genes controlled by a single locus, and key candidate genes within such a locus will be revealed by their physical position.
Mouse lines, long-term selected on growth, fatness and metabolic
rate - a unique resource for the genetic dissection of QT’s
1 Scottish Agricultural College, Sustainable Livestock Systems, Growth genetics Section, Sir Stephen Watson Building, Bush Estate, Penicuik, Midlothian, EH26 0PH
2 ICAPB, The University of
In
order to identify the genetic and physiological basis of growth, metabolic rate
and fatness, the establishment of inbred lines derived from divergently
selected lines is an important tool. It is necessary first to develop highly
divergent selection lines for the traits of interest, and then derive inbred
lines from them. Such selection lines, followed by inbreeding have been
developed in
Using
candidate gene approaches the expression of some genes (leptin
production, leptin reception, GHRH receptor) was
manipulated by the introgression of natural mutations
(Lepob, Leprdb,
lit) into the fat lines and the growth lines, respectively. In each case,
these loci were found to contribute little of the selection response.
To
test for epistatic interaction, a growth QTL
identified on Chr. X of PH vs. PL was introgressed into DUH, the heaviest known mouse line. The
effects of the QTL appear to be smaller on the high DUH background. Similarly
the murine mutation in the myostatin
gene compact has been introgressed into this line to measure its effects on
another genetic background and to evaluate possible negative side effects.
Males homozygote for compact were
compared with wt animals leaner (11.6 vs. 17.4% fat), had shorter tails (10.4
vs. 11.3 cm), had a higher “dressing out percentage (48.7% vs. 38.7%), had
higher muscle weights and relative and absolute smaller organ weights.
These
lines and some results on QTL mapping, on candidate gene approaches using these
resources as well as the development of some congenic
lines will be reviewed. Further details can be found on our website (http://www.ed.ac.uk/~eang17/mice.html)
and in recent publications (see below). Details on more growth lines can be
found in (Bünger et al. 2001a, Bünger et al. 2001b).
§
Bünger L, Hill WG.
(1999) Inbred lines derived from long-term divergent selection on fat content
and body weight. Mammalian Genome 10, 645-648.
§
Bünger L, Laidlaw AH, Bulfield G, Eisen EJ, Medrano JF, Bradford GE, Pirchner
F, Renne U, Schlote W, Hill
WG. (2001a) Inbred lines of mice derived from long-term on growth selected
lines: unique resources for mapping growth genes. Mammalian Genome 12, 678-686.
§
Bünger,L., Renne,U., Buis,R.C.
(2001b) Body weight limits in mice - Long-term selection and single genes. In E.C.R.Reeve (Ed.), Encyclopedia of Genetics. (pp. 337-360).
The Role of Epistasis in the Control
of Complex Traits
Örjan Carlborg and Chris Haley
Roslin Institute,
Historically, epistasis has only received limited attention in quantitative genetic research. The role of gene interactions in the control of complex traits has been relatively neglected because of their apparent low contribution to the genetic variance as determined in studies based on classical quantitative genetic approaches. Epistasis has received more attention in the study of qualitative traits, such as coat colour in animals, and a number of epistatic mechanisms affecting these traits have been investigated in the literature. Although most of the qualitative genetic work was done a number of years ago, it clearly shows the biological importance of epistasis on a “locus by locus” level. With the current knowledge of the similarity of the molecular mechanisms underlying quantitative and qualitative traits, it is reasonable to expect that epistasis is an important mechanism underlying variation in complex traits.
Most QTL mapping studies have been analysed using methods based on detection of marginal genetic effects (i.e. additive and dominance effects) of individual QTL. It is only recently that epistasis has been considered. The most powerful strategy to map epistatic QTL involves fitting the locations for QTL to be evaluated without assuming that either of the QTL has marginal effects. This procedure is called a simultaneous search. We have used simultaneous mapping of epistatic QTL in several experimental populations of mice and chicken and have shown that epistasis is an important mechanism for quantitative traits. For example, we used simultaneous mapping of epistatic QTL pairs to analyse five growth traits in a broiler x layer chicken F2 intercross. On average 2.6 additional interacting QTL were detected for each analysed trait. These novel epistatic QTL and their interactions increased the amount of explained residual variation by on average 50%. A total of 15 QTL were identified in this study and 13 of these were involved in significant epistatic interactions with at least one other QTL. When the genotype class means for the 22 significant epistatic QTL pairs were visually inspected, four clusters of similar types of genotype-phenotype patterns were identified. Three of the clusters represent dominance-by-dominance, heterosis and multiplicative epistasis, and in total 17 of the detected pairs could be assigned to one of the four clusters. The results clearly illustrate the increase in power obtained by using this novel method for simultaneous detection of epistatic QTL, and also how visualization of genotype-phenotype relationships for epistatic QTL pairs provides new insights to biological mechanisms underlying complex traits.
Genetic correlations and associative
networks for CNS transcript abundance and neurobehavioral phenotypes in a
recombinant inbred mapping panel.
1Chesler, E.J.; 2Wang, J.; 1Lu, L.; 1Qu, Y.; 2Manly, K.; 1Williams, R.W.
1Dept. of Anatomy & Neurobiology and Center for
Genomics & Bioinformatics. Univ. of
Genetic correlation analysis uses associations of phenotypes in similar or identical genetic background to evaluate shared genetic mediation of traits. The BXD recombinant inbred lines are an isogenic strain set made by sib mating of the cross progeny of C57BL/6J and DBA/2J strains. In addition to the original BXD/Ty strains, ourselves and others are in the process of creating several new lines from these progenitors so the resulting strain set will include approximately 100 strains. The BXD RI set has been widely characterized on numerous traits, largely related to behavioral and immunological phenotypes. Transcript abundance in the forebrain and cerebellum was assayed in these mice using the Affymetrix U74Av2 and 430AB microarrays respectively. Previously reported BXD strain phenotypes were obtained from the literature and from personal communication with individual investigators. The current database includes over 430 traits. The public can browse this data in depth at www.nervenet.org. Family-means based associations of these phenotypes with expression levels of the thousands of transcripts on the microarrays can be generated using WebQTL. Spearman's rank correlations or Pearson product moment correlations can be calculated. Tools are also available to evaluate univariate normality and linearity of the observed trait relations. The results are available in a public database and analysis engine at www.webqtl.org. This Internet resource provides mapping of gene transcription and other complex traits using a high-density marker panel. The correlation of phenotypic values of gene expression and other traits can be used to narrow down the many hundreds of genes in a quantitative trait locus (QTL) to a handful of promising candidates for further evaluation. However, correlated transcripts need not reside in QTL regions. Transcription regulatory QTLs for highly correlated genes may also be the site of genetic variation responsible for behavioral variation. Gene expression associative networks can be constructed from transcript to transcript or trait correlations, and shared genetic mediators of multiple traits can be identified. The community wide assembly of data for the study of shared trait mediation will cumulatively enhance the utility of recombinant inbred strains as a resource.
EMAGE – The Edinburgh Mouse Atlas of Gene Expression
J.
Christiansen1,
MRC Human Genetics
Unit,
One challenge of the post-genomics
era is to develop a comprehensive understanding of the expression patterns of
specific gene products in both space and time. This effort will help in
identifying and studying networks of interactions that exist between these
molecules. One resource to aid in these analyses is the Edinburgh Mouse
Atlas of Gene Expression (EMAGE), a database of spatially mapped gene
expression data in the developing mouse embryo that has been developed as part
of the Edinburgh Mouse Atlas Project (EMAP). EMAGE is part of the Mouse
Gene Expression Information Resource (MGEIR) which is being built in a
collaborative effort with the Jackson Laboratory,
All EMAGE data is mapped into a standard framework: the EMAP Digital Atlas of Mouse Development. This Atlas consists of a standardised nomenclature for the anatomical structures that are present at every Theiler stage of embryo development and at least one representative three-dimensional digital embryo model for most post-implantation Theiler stages. These 3D embryo models have been constructed by either stacking and aligning entire sets of serial histological sections or by imaging representative embryos using Optical Projection Tomography (OPT: Sharpe et al, (2002) Science 296: 541-545). As the embryo models are 3D objects, it is possible to take virtual sections through these in any plane to reveal internal anatomical detail.
EMAGE expression data is mapped into the EMAP Atlas framework both using text (to the anatomical nomenclature) and spatially (to corresponding regions within the embryo models). Whole mount data imaged by photographing specimens is mapped in 2D as a domain that is projected onto the 'surface' of an EMAP embryo model. Section and OPT data is mapped into the 3D space of the digital EMAP embryo models. Searching EMAGE data that has been mapped into the standard framework models is performed by text based methods or by spatial based approaches whereby 2D or 3D query domains are defined by the user. The software required to search EMAGE can be downloaded from the EMAGE website. The same software can also be used to prepare private databases for in-lab data management or to prepare remote electronic submissions to the EMAGE central database. Submissions to the central database can alternatively be made by sending specimens directly to the EMAGE Editorial Office for data entry. Editorial staff can help with the submission process and all other aspects of the database. EMAGE is publicly accessible from http://genex.hgu.mrc.ac.uk/
1,3Peter
Demant, 1,3Claudia A. L. Ruivenkamp, 1,3Carlo
Zanon, 1Tamas Csikos,
1,3Tom van Wezel, 1,4, Nikos Tripodis, 5Hugo Horlings,
6Jana Mullerova
1Division of Molecular Genetics, The Netherlands
Cancer Institute,
Cancer susceptibility in humans, with the
exception of a small group of familial cancers representing under 5 – 10
percent of cases, is under strong genetic influence, which recent studies
indicate is most likely exerted by multiple low penetrance
genes. As their identification is virtually impossible in human families or
populations, the most accessible strategy is to search for them in laboratory
animals and then define their human homologues. In the course of analysis of
colon and lung tumor susceptibility in mice we mapped 15 colon tumor
susceptibility genes and 30 lung tumor susceptibility genes, all involved in
the determination of tumor number or tumor size, and that inter-locus
interactions are an essential and frequent component of cancer susceptibility
genes.
After the general features of cancer
susceptibility genetics have been established, identification of these genes
and elucidation of the control of biologically relevant features of tumors –
their capacity of progression and metastasis - are the critical tasks for the
coming period. As proof of principle, we cloned positionally
the Scc1 gene, which controls susceptibility to colon cancer in mice. We have
identified the Protein tyrosine phosphatase receptor
type J (Ptprj) as the mouse Scc1 and shown that its
human homologue, PTPRJ, is involved in several human cancers: colorectal, lung,
and breast and that its somatic mutations in colorectal cancers predispose to
one of the several pathways of malignant progression, indicating the
contribution that cancer susceptibility genes can make for a better assessment
of prognosis of cancer patients. We also studied genes influencing the
malignant progression of lung tumors in mice. We developed tools for the quantitation of lung tumor histopathology and mapped 7
genes affecting tumor progression.
The demonstration that cancer susceptibility QTLs can be positionally cloned,
and that QTL analysis applies to qualitative characteristics of the malignant
state of tumors, indicate that studying these genes helps us understand cancer
progression pathways, which are essential the optimal individual clinical
management of cancer.
Type 2 Diabetes- A complex trait
Roger D. Cox, Michelle Goldsworthy, Ayo Toye, Liz Bentley, Lee Moir, Alison Hugill, Vesna Miijat, Alison Haynes and Julie Quarterman.
Medical Research
Council, Mammalian Genetics Unit, Harwell,
We are using the mouse to develop models of type 2 diabetes with a view to identifying genes and pathways involved in development of this common disease. In our presentation we will describe our approach which uses ENU in simple dominant screens as well as in sensitised screens where insulin resistant mice (due to knockout mutations in IR and IRS-1) are mutagenised. We will show data on new models identified using glucose tolerance tests and our progress in mapping and cloning these. We will also present some of our data on mapping glucose tolerance QTLs in a C57BL/6J x C3H cross.
A Haplotype Map of the Laboratory Mouse: Acceleration of Positional
Cloning
Claire M. Wade, Andrew Kirby, E.J. Kulbokas, Eric S. Lander, Kerstin Lindblad-Toh, and Mark J. Daly
The recent completion and annotation of the genome sequence of Mus musculus (C57BL/6J) offer tremendous potential for geneticists studying complex phenotypes using the mouse as a model system. However at this point, beyond offering a quick assessment of gene content and potential candidates in large regions of interest, the data does not instantly offer a rapid acceleration for positional identification of genes underlying complex phenotypes. Decades of exquisite phenotyping and detailed analysis of crosses of inbred mice have resulted in initial localization of hundreds of loci involved in complex disease and quantitative phenotypes, but very few genes have been conclusively identified from these studies. A clear understanding of the origin and structure of genetic variation in these strains provides a key missing piece of this puzzle.
Recent analysis of polymorphism across the genome (Wade et al., 2002) has established that the genomes of commonly used inbred strains are by and large, mosaics of long segments (~ 2Mb) which in most cases are clearly derived from either western European M. m. domesticus or Asian M. m. musculus ancestry. A critical implication of this observation is that nearly all variation among these strains is comprised of ancestral differences between these highly diverged subspecies and can be easily tracked with a sparse map of SNPs given the observed size of the ancestral segments. To fully realize the power of the genome sequence, we have embarked upon a complete genetic dissection of the ancestral segments making up the most commonly used inbred lines and data from several genomic loci examined already will be presented.
I will also discuss the immediate uses of this map to focus positional cloning of QTLs through
1) the reduction of regions obtained through linkage analysis via identification of segments shared by the strains used for the cross
2) the selection of ideal strain combinations for further reduction of critical regions through simple intercross/backcross experiments
3) the use of correlation between phenotype and ancestral sequence origin across many inbred strains to identify very short genomic regions most likely to harbor responsible genes
Such a haplotype map will enable a dramatic acceleration in positional cloning, allowing the entire community of mouse geneticists to quickly and fully take advantage of decades of phenotyping and linkage mapping studies which have conclusively mapped but not yet identified genes responsible for countless medically important phenotypes. Moreover, this effort provides a comprehensive genetic component to programs such as the Mouse Phenome Project, already underway to collect complete and diverse phenotype data for the same commonly used mouse strains.
QTL on Mouse Chrs 1, 4 and 6 that
Influence Light-induced Retinal Degeneration
[Poster Presentation]
M. Danciger1, J.E. Lyon1, D.M. Worrill1, J.Lem2, C.Grimm3, A.Wenzel3, C.E. Remé3.
1Biology
Department,
2Ophthalmology &
Cardiology, Tufts-New
3Laboratory of Retinal
Cell Biology,
Purpose:
BALB/cByJ (BALB/c) retinas are significantly more
susceptible to light insult than those of 129S1/SvImJ (129). The purpose of
this work was to determine the quantitative trait loci that influence intense
light-induced retinal degeneration (LRD) as a first step toward identifying the
genes/alleles they represent.
Methods: 289 F2 progeny of an intercross between the 129 and
BALB/c strains aged to approximately 6 weeks were exposed to 15,000 LUX of
light for 1 hour after their pupils were dilated, and then placed in the dark
for 16 hours. After 10-12 days of dim cyclic light, the amount of rhodopsin remaining in the retinas was measured spectrophotometrically. This was used as the trait quantitating the degree of retinal degeneration. Among the
F2 progeny, neither gender nor pigmentation had a significant influence on the
amount of rhodopsin loss after light exposure. For
genetic study, DNAs from F2 progeny are being genotyped with 80 dinucleotide repeat markers spanning the genome. For
screening purposes, the markers are first tested in 27-36 F2 progeny with the
most rhodopsin remaining and the 27-36 F2 progeny
with the least rhodopsin. Any marker with a 95%
probability of being associated with phenotype is tested in all F2 progeny.
Data were analyzed with the Map Manager QTX program.
Results: With ~ 65% of the genome covered so far, three QTL on
mouse Chrs 1, 4 and 6 have been identified. Each
accounts for a substantial portion of the total genetic effect influencing LRD,
and each represents genes with BALB/c susceptible (or 129 protective) alleles.
The Chr 6 QTL is in the same region as a highly
significant QTL influencing age-related retinal degeneration.
Conclusions: Identification of the gene modifiers represented by these QTL may be important for human retinal diseases that are accelerated by light exposure.
Genetic
dissection of common diseases in the post-genomic era
Ariel Darvasi
The identification of genes affecting common diseases such as diabetes, asthma, cancer, schizophrenia, etc. as well as other complex traits, is considered to be one of the major challenges of contemporary genetics. In the past decade several attempts were made to identify such genes with relatively little success. This is now changing due to the advance in the knowledge of the human genome (a product of the Human Genome Project), new high throughput molecular technologies and advanced computational (statistical) approaches. We studied these elements to suggest an efficient strategy for the genetic dissection of complex traits in humans. Linkage disequilibrium patterns across the human genome and across populations, has been examined. It is suggested that the use of homogeneous populations may provide significant advantages for gene discovery. For the molecular analysis of the DNA, we have considered several technologies for estimating allele frequencies from DNA pools, thus somewhat overcoming the genotyping bottleneck. These approaches have been put to work and have successfully established gene-disease associations in schizophrenia and other common diseases.
Epistatic QTL for gene expression in mice;
potential for BXD expression data
Dirk-Jan de Koning*,[1], Örjan Carlborg*,1, Robert Williams†, Lu Lu†, and Chris Haley*
* Roslin
Institute,
†Center for Genomics and Bioinformatics,Department of
Anatomy and Neurobiology, University of
Microarray studies on segregating populations offer the opportunity to map QTL affecting gene expression (Jansen & Nap, 2001; Jansen, 2003). Post-hoc analyses of significant QTL can be used to infer hypothetical interactions between loci and genes. A more direct approach would be to model interactions between QTL during the QTL mapping process. A testing framework for epistatic QTL, as well as efficient algorithms for simultaneous mapping of epistatic QTL, have recently been described and used to map epistatic QTL in experimental data from line-cross experiments (Carlborg et al 2000; Carlborg and Andersson 2002; Carlborg et al 2003).
Using the BXD genotypes that were made available by Williams et al. (2001) and expression data from 10 genes/ESTs from http://webqtl.roswellpark.org/search.html we tested the concept of mapping epistatic QTL in these RI lines. The marker data were used in two different ways: 1) an analysis at the marker positions only using the single marker genotypes. 2) 1-cM interval mapping using marker derived probabilities for line origin. For the detection of epistatic QTL, we simultaneously fit the marginal (additive) effects and interactions for all possible QTL-pairs in the genome. Significance testing for QTL-pairs and their interactions are based on three randomisation tests further described in Carlborg and Andersson (2002). A summary of the results from mapping of epistatic QTL pairs for IGF2 is given in Figure 1.
The single-QTL genome scan revealed a strongly suggestive QTL on chromosome 4 and a genome-wide significant QTL on chromosome 19. The testing procedure for epistatic QTL revealed four QTL pairs that reached 5% genome-wide significance and one additional pair that reached 20% genome-wide significance. All five pairs showed strong evidence for epistatic interactions. Five novel QTL regions were identified. Four of these interacted with the QTL on chromosome 4, which reached 10% genome-wide significance based on analysis of its marginal effects. One novel QTL-pair (Chr 4 and 16), where neither of the involved QTL had significant marginal effects, was also detected.
The four most significant QTL pairs have very similar genotype-phenotype relationships. Two possible explanations for the pattern are 1) that the first QTL only has an effect in one of the genotypes of the second QTL or 2) that the second QTL has an effect that changes direction depending on the genotype of the first QTL. The genotype-phenotype pattern displayed by these QTL pairs is a clear example where simultaneous fitting of epistatic QTL dramatically increases statistical fit. For the single QTL analysis, the marginal effect that can be detected for the first QTL is decreased by 50% due to the genetic effect of the second QTL and the marginal effects of the second QTL is non-existent given the genetic effects of the first QTL. Several markers that are part of a significant QTL pair had very low evidence for a QTL in the single QTL genome scan. The results from the other nine genes that we analysed, uncovered epistatic interactions for three of them, three showed only a single suggestive QTL while for the remaining three we did not detect any QTL.
Using our QTL mapping procedures, we are able to 1) repeat the results obtained by the WebQTL resource and 2) identify novel interacting QTL with significant effects on gene expression. We strongly believe that these results clearly point out the usefulness of our QTL mapping procedures on this type of data: 1) We detect more markers that affect the expression of a gene compared to using a single QTL genome scan. 2) Epistatic interactions between markers can help resolve metabolic pathways by indicating which genomic regions that are jointly affecting expression of the analysed gene. It would be extremely valuable to analyse a larger set of genes with these methods in order to quantify the contribution of epistasis to the regulation of gene expression.
We are now in the process of analysing the complete experiment containing expression data for > 12,000 genes, using epistatic models. In the initial stages, we will explore the optimum strategies for weighting the observations and width of the scanning interval on a subset of data
References
Carlborg and Andersson, 2002, Genet Res. 79:175-84.
Carlborg et al.2000, Genetics. 155:2003-10
Carlborg et al. 2003, Genome Res. 13:413-21.
Jansen, 2003, Nat Rev Genet. 4:145-51
Jansen & Nap, 2001,Trends Genet.17:388-91.
Williams et al.,2001, Genome Biol. 2:RESEARCH0046
[1] These authors contributed equally to this research
Evidence for modifier genes in the
fragile X syndrome mouse model
[Poster Presentation]
Vanessa Errijgers1,
Rudi D’Hooge2, Peter P. De Deyn2, R. Frank Kooy1
1Department
of Medical Genetics,
2Department of Neurochemistry and Behavior, Born-Bunge Foundation, University of
Fragile X syndrome is the most common form of X-linked mental retardation. It's caused by a CGG triplet expansion in the first exon of FMR-1, resulting in transcriptional silicing of the gene due to abnormal hypermethylation. The absence of the FMR-1 protein leads to mental retardation, aberrant behavior, and macroorchidism.
We have compared the acoustic startle respons (ASR) of male fragile X knockout mice bred in different genetic backgrounds. The ASR is characterized by a coordinated contraction of the muscles of the eyelid, neck and extremities, elicited by a sudden, loud acoustic stimulus. It’s mediated by an oligosynaptic neuronal pathway.
We found a 12% difference in ASR to a sound level of 100 dB between 13 knockout and 20 control C57BL/6J inbreds. As it has been hypothesized that genetic background might influence the test results, we have tested the ASR in C57BL/6J x 129P2/OlaHsd F1 mice. We measured a larger, 28%, difference between 9 KO’s and 10 controls in respons to the same stimulus. Surprisingly, F2 intercross mice showed no difference in ASR between 51 knockouts and 58 controls. As both F1 and F2 intercross mice consists for 50% of genetic material from the 129P2/OlaHsd strain and for 50% of genetic material from the C57BL/6J strain, the different distribution of the genetic background seems to influence the relative difference in ASR between knockouts and controls.
A backcross of B129F1 mice with C57Bl/6 resulted in a 13% difference in ASR between 26 knockouts and 34 controls. We selected mice with the lowest and largest ASR respectively for further backcrossing and ASR testing.
Our results suggest a strong influence of the genetic background on the ASR of the fragile X knockout mouse and suggest that modifier genes play a role in the fragile X syndrome.
Identifying the genetic basis of high growth (hg) modifier quantitative trait loci (QTL): Integration of genetic and bioinformatic approaches.
[Poster Presentation]
Charles R. Farber and Juan F. Medrano
Department of
Animal Science,
The hg deletion is a partially recessive mutation that results in a 30-50% increase in mature body size in mice and is due to absence of the suppressor of cytokine signaling-2 (Socs-2) gene. The loss of Socs-2, a negative regulator of cytokine signaling, prolongs activation of the growth hormone (GH) signaling pathway, leading to an overgrowth phenotype. Previous research within our laboratory identified quantitative trait loci (QTL) which modify the effect of the high growth (hg) deletion in a CAST x C57Bl6/J-hg/hg F2 population. QTL on Chrs. 2, 9, 11 and 17 were found to interact with hg and influence one or more of the following traits: growth rate from 2-9 weeks, carcass protein, carcass ash, femur length and carcass fat. To fine map and ultimately identify the genetic basis of each QTL we have developed two integrative approaches, combing genetic and bioinformatic strategies. For the first approach congenic lines are being developed for each QTL by introgressing CAST alleles onto a C57Bl6/J-hg/hg background using the “speed congenic” approach. Congenic lines for each QTL are currently in the final stages of production. Characterization of each line will provide confirmation of each QTL and a resource for candidate gene identification. The second approach is a multifaceted bioinformatics strategy aimed at exploiting the knowledge of epistasis between the modifier QTL and hg and the comparative genome maps of human and mouse. In this context we have constructed GH signaling pathway maps, enabling rapid identification of candidate genes which map to QTL regions and possibly interact with Socs-2. The pathways have been created using the GenMAPP program. Our current pathways contain 146 genes and serve as the foundation for candidate gene identification. These GenMAPP GH pathways integrate information from human and mouse literature, mapping experiments and expression databases, providing support to our selection of candidate genes. Selected candidates genes are sequenced to look for functional mutations. Real-time quantitative PCR studies will identify genes that are differential expressed. It is our expectation that these integrative approaches will not only lead to the identification of the genes underlying the QTL modifying the effect of hg, but also serve as a model to accelerate the discovery rate of modifier loci in mice.
Inter-species consomic
strains and accompanying tools for mapping and identification of quantitative
trait loci
Jiri Forejt1, Sona Gregorova, Radka Aixnerova, Petr Jansa, Petr Divina
Most of the genetic variation among current inbred strains of mice is believed to reflect differences between Mus musculus musculus and Mus musculus domesticus (sub)species of the house mouse. Here we report on our progress towards generating a collection of 21 consomic (chromosome substitution) strains and a conplastic strain, using a Mus m. musculus derived PWD/Ph inbred strain as a chromosome donor and C56BL6/J (B6) strain (mostly of domesticus origin) as a recipient. A panel of 245 pre-selected SSLP (Mit) markers, polymorphic between strains PWD/Ph and B6, was used in marker-assisted introgression of individual PWD chromosomes into the B6 background. The frequency of non-recombinant PWD chromosomes in backcross progenies varied between 0.14 and 0.28. The strain was considered consomic at backcross-10 generation (N11). At this stage, over 99.9% of the genetic background should be of B6 origin. As of April 2003, nine chromosomes reached this goal, Chrs 2, 3, 4, 6, 11, 14, 16 and 19. From these, Chrs 2, 12 and 14 are already available as viable and fertile homosomics (homozygotes for a donor chromosome). The remaining 12 chromosomes are lagging 1 to 3 backcross generations behind. The B6-XPWD males are sterile, the genetics of sterility is in progress. All data for mice, markers, genotypes and crosses is maintained in the CONSOMIC database (Sun Solaris mySQL server) with Perl Web interface.
The positional cloning of a QTL often requires transgenesis and/or gene targeting of the candidate gene. For this purpose genomic libraries and embryonic stem cell lines of PWD/Ph origin would be desirable. We prepared a PWD/Ph cosmid library, using the LAWRIST7 vector, with 700 000 primary clones of 44 kb average insert size, representing 12-fold coverage of the mouse genome. For construction of the first BAC genomic library of the Mus m. musculus species (PWD/Ph) the BAC vector pBACe3.6 was used. As of April 2003, 39000 primary clones with average insert size 105 kb were obtained, representing 1.6-fold mouse haploid genome.
After completion of consomics and the conplastic
strain, the mice will be extended to The Jackson Laboratory to make them
available to researchers, to characterize them for a wide range of
physiological phenotypes, and to deposit the data in the Mouse Phenome Database (with Dr. B Paigen).
1International Fellow of Howard Hughes Medical Institute. Supported partly by EC grant QLRT-2000-00233
Congenic-based strategies for dissecting the
architecture of QTL and their biological functions
Dominique Gauguier,
Stephan C Collins, Karin Wallace, Robert H Wallis, Steven P Wilder
The Wellcome Trust Centre for Human Genetics, University of Oxford, Roosevelt Drive, Headington, OXFORD OX3 7BN, United Kingdom
Mapping genetic loci that underlie quantitative variations of phenotypes relevant of human multifactorial diseases, whose etiology associates both genetic and environmental factors, is relatively straightforward in animal models. Robust statistical methods that have been optimised for the definition of statistical thresholds for evidence of significant genetic linkage and the general characterisation of QTL architecture, have very little impact on the actual implications of statistically defined QTL characteristics in the biology and pathophysiology of complex phenotypes in model organisms. Even in classical experimental crosses (intercross and backcross) derived from rat strains, which are the models of choice for extensive and accurate screenings of complex phenotypes, including for example glucose homeostasis and blood pressure, phenotypic information that can be associated with a QTL is often very limited.
Congenic lines represent essential tools for validating the existence of a QTL, fine QTL mapping, and characterising the phenotypes controlled by sequence variants at the locus. Congenics have generally been produced for random chromosomal segments covering the original QTL and have often provided evidence for the presence of several independent genetic alterations underlying the QTL effect. We present an integrated approach based on investigations in congenic lines that generates information on the robustness and significance of a QTL originally mapped in a classical experimental cross, allows a rationale strategy for fine QTL mapping, and provides insights into the pathophysiological consequences of sequence variants at the locus that are required for subsequent gene expression profiling and disease gene identification.
Cardiopulmonary
phenotypes induced by mainstream cigarette smoke in inbred strains of
mice.
[Poster
Presentation]
Terry Gordon, PhD,
Christine Nadziejko, PhD, and Maarten Bosland, DVSc., PhD.
Voluntary and involuntary exposure to cigarette smoke is an important
health, social, and economic concern. A number of epidemiologic studies
have demonstrated that 10 to 15% of cigarette smokers develop lung
cancer. In a similar fashion, the cardiopulmonary effects of mainstream
cigarette smoke vary amongst individuals. The reason for this variable
phenotypic response to cigarette smoke inhalation is unclear, but it is
becoming increasingly clear that genetic background is a host factor for
susceptibility to a number of airborne particles and gases. There is a
need for predictive genetic animal models of inter-individual variation (i.e.,
genetic predisposition) for lung cancer and other cardiopulmonary
diseases. Although studies have been conducted for spontaneous tumor
production and for chemical-induced lung cancer (e.g., NNK, ENU, and urethane),
a systematic examination of murine interstrain differences in response to cigarette smoke
inhalation has not been conducted. We addressed this research gap by
examining the strain distribution pattern of cardiopulmonary phenotypes in 8
inbred strains of mice exposed to cigarette smoke. We observed increases
in the number of lung tumors in cigarette smoke-exposed A/J, A/HeJ, and CAST/E mice. Significant interstrain
increases in lung volume displacement, a crude index of emphysema, were
observed in 4 out of 8 strains exposed to cigarette smoke. Similarly, in
a strain-dependent manner, the olfactory epithelium of the nasal cavity was
replaced by respiratory epithelium including mucous-containing cells.
Finally, we observed a strain-dependent increase in right heart size, an index
of pulmonary hypertension, as illustrated by an increase in the right/left
ventricle mass ratio. We thus hypothesize that genetic factors contribute
to several of the adverse cardiopulmonary effects of mainstream cigarette
smoke, including pulmonary hypertension, lung cancer, and emphysema. We
are using both recombinant inbred mice and a computational approach, using
algorithms and SNP databases developed by other investigators, to identify
candidate genes in these disease models.
Unravel the behavior/genome interface
using SEE software and database
[Poster Presentation]
Guy Horev*, Dan Yekutieli+, Neri Kafkafi$, Anat Sakov+, Dina Lipkind*, Greg Elmer$, Ilan Golani*
* Department
of Zoology, George Wise Faculty of Life Sciences,
+ Department
of Statistics and Operations research, The Raymond and Beverly Sackler Faculty of
$ Department
of Psychiatry,
e-mail: horevguy@post.tau.ac.il
For the student of behavior RI strains are a most appropriate group for dealing with the behavior/genome interface. The variance between the RI strains associated with genetic factors is high, yet they are genotyped extensively and that information is freely available. The limitation is that the behaviour be measurable, and that the parent lines have to be enough separated in their behaviour.
SEE is a Strategy for the Exploration of Exploratory behavior, by analyzing the path traced by a rodent (a rat or a mouse), using the x,y coordinates file exported from any tracking system. SEE was originated from ethologically-oriented studies, which found that locomotor behavior is highly structured and consists of typical behavior patterns. Once these patterns were isolated, they were also found useful in psychopharmacological and psychobiological studies. Moreover, SEE emphasizes measurable properties of exploratory behaviour, and includes sophisticated data analysis tools. (SEE software and path databases are already freely available in the SEE website http://www.tau.ac.il/~ilan99/see/help/).
We show that, as a part of the broader pool of 10 inbred strains of mice, C57 and DBA, which are the parental lines of the BXD set, have sufficiently distant behavior in many behavioral endpoints pertaining to motor, motivational, and cognitive aspects of exploratory behavior. These differences in behavior are replicable across laboratories – an issue receiving much attention in behaviour genetics in recent years. In view of these results, we suggest that appropriate use of the SEE software and multi-lab database will enable to derive replicable results also from single-lab QTL analysis. We demonstrate a hierarchical design of such an experiment, and discuss its inherent difficulties.
Integrative multi-dimensional transomic
analysis strategies reveal novel insights into potentially important molecular
pathways for mammary tumor dissemination
Laboratory of Population Genetics,
NCI, NIH, 41/D702,
Our laboratory has demonstrated that the genetic background of tumor significantly influences the ability of the tumor to disseminate. Using conventional QTL mapping crosses and RI strains, we demonstrated the presence of at least 5 genomic regions that were significantly or suggestively associated with metastatic efficiency. Conventional congenic mapping experiments are currently in progress to achieve high resolution mapping of candidate loci. In parallel, to supplement and accelerate the candidate gene identification process, we have been performing an integrative, trans-disciplinary strategy utilizing a variety of “-omic” technologies to develop a more comprehensive pathway-based understanding of the metastatic process. Combining genetic, somatic cell genomics, bioinformatics, haplotype mapping, microarray analysis, proteomics and genome-wide transcriptional QTL mapping, we have identified a molecular process that plays an important role in metastatic progression. The identification of this pathway as a potentially important element in tumor dissemination has recently been supported by the confirmation that the murine ortholog of the breast cancer metastasis suppressor gene Brms1 is an integral member of the complex. Additionally, this complex is potential at the nexus of a number of metastasis associated genes. Finally, all of our genetically defined QTL peaks contain members of this molecular process, suggesting that they may be interesting candidate genes. Our data to date suggest that integrative “transomic” strategies may provide a vehicle to significantly accelerate QTL gene identification and analysis. In vitro molecular and in vivo complementation assays are currently in progress to evaluate the significance of our transomic findings.
On the Integration of QTL, gene
expression and sequence analysis
Robert Hitzemann, Barry Malmanger , Cheryl Reed , Maureen Lawler, Barbara Hitzemann, Shannon Coulombe, Kari Buck, Brooks Rademacher, Nicole Walter, Yekatrina Polyakov , James Sikela, Brenda Gensler, Sonya Burgers, Robert W. Williams,, Ken Manly, Jonathan Flint and Christopher Talbot.
Research
Service,
Although hundreds if not thousands of quantitative trait loci (QTL) have been described for a wide variety of complex traits, only a very small number of these QTLs have been reduced to quantitative trait genes (QTGs) and quantitative trait nucleotides (QTNs). A strategy is described for detecting QTGs and QTNs that is based on leveraging the information contained within the haplotype structure of the mouse genome. The strategy utilizes the 6 F2 intercrosses that can be formed from the C57BL/6J (B6), DBA/2J (D2), BALB/cJ (C) and LP/J (LP) inbred mouse strains. Focusing on the phenotype of basal locomotor activity, it was found that in all three B6 intercrosses, a QTL was detected on distal Chromosome 1; no QTL was detected in the other three intercrosses and thus, it was assumed that at the QTL, the C, D2 and LP strains had identical alleles. These intercross data were used to form a simple algorithm for interrogating microsatellite, single nucleotide polymorphism (SNP), brain gene expression and sequence databases. The results obtained point to Kcnj9 (which has a markedly lower expression in the B6 strain) as being the likely QTG. Further, it is suggested that the lower expression in the B6 strain results from a polymorphism in the promoter region that disrupts the binding of at least 3 transcription factors. Overall, the method described should be widely applicable to the analysis of QTLs. Supported in part by MH 51372 and AA 11043.
Multiple approaches for dissecting
quantitative trait loci and cloning the underlying genes
Fuad A. Iraqi1, Harry Noyes2, David Mburu1, Claire Wade3, Steve Kemp2 and John Gibson1 [ Fuad A. Iraqi, f.iraqi@cgiar.org ]
1Biotechnology Program. International Livestock Research Institute (ILRI), P. O. Box 30709, Nairobi, Kenya; 2School of Biological Sciences, University of Liverpool, Liverpool, UK; 3 Whitehead/MIT Center for Genome Research, 9 Cambridge Center
Cloning genes underlying quantitative trait loci (QTL) still appears not to be a trivial exercise. The recent success of sequencing the mouse genome obviously will assist the process, but the major challenge facing geneticists in cloning these genes is dissecting the QTL to a sufficiently small genomic interval that allows molecular cloning. In mice, most QTL detection and gene discovery has been accomplished by using F2 or backcross experimental designs. This has only allowed mapping at a rather wide confidence intervals (CI) because of the limited number of recombination events in small chromosomal regions. A number of experimental designs have been proposed to dissect a given QTL, including the use of advanced intercross lines (AIL), interval specific congenic strains (ISCS), recombinant inbred strains (RIS), recombinant sire progeny test (RSPT), SNP haplotype mapping and functional genomic approaches. In addition, a comparative mapping approach of the QTL of the same trait between different species may provide additional fine mapping. In the course of a search for QTL for resistance to Trypanosoma infection in mice, most of the above approaches were applied.
Initially, a genome-wide search was performed for QTL for resistance to Trypanosoma infection in two F2 resource populations by crossing resistant, C57BL/6, with two susceptible strains, A/J and BALB/cJ, respectively. Strong evidence was found for QTL on three chromosomes, which accounted for much of the genetic difference between resistant and susceptible mice. The QTL were mapped within genomic regions of 20-40 cM. Following the F2 mapping, two F6 AIL populations were developed from the founders of the F2 generations. Results confirmed the three QTL in the C57BL/6 x A/J AIL, but the QTL on chromosome 5 was apparently lost from C57BL/6 x BALB/c AIL. The reduction of confidence interval for the three QTL ranged from 2.5 to more than 10 fold in F6 AIL by comparison with F2 results. The chromosome 1 QTL resolved into three sub-QTL. Subsequently, ISCS, haplotype mapping, RSPT and functional genomic approaches were applied to further fine map the QTL. By comparing the mapped QTL regions between mouse and cattle for the same trait, we were able to narrow down to a possible candidate region shared between the two species. Results for most of the additional approaches are being collated at time of writing. Current indications are that the ISCS have added less information to refine map intervals than we had hoped for. The RSPT is time consuming and relatively expensive, and results should be available by time of the CTC meeting. Haplotype mapping is looking very promising, but results indicate that the haplotypes that trace the sub-species origins in inbred lines are finely divided and require a very high density of SNP to resolve. A small pilot trial of microarray based gene expression assay of infected C57BL/6 and A/J mice yielded two possible candidate genes mapping to our QTL intervals. In future we plan much more detailed gene expression assays, including experiments that examine gene expression in QTL congenic lines in order to tie gene expression patterns to individual QTL.
Morphological
integration in the mouse mandible: multivariate analysis of QTL effects
Christian Peter Klingenberg
Morphological
structures are integrated, that is, their parts are co-ordinated
to form a functioning unit and reflect interactions of the developmental
processes involved in assembling the structure. This topic has traditionally
been investigated with phenotypic data or genetic covariance components from
breeding experiments, but it is just as relevant for the effects of individual
loci. I report results of a study in collaboration with J. M. Cheverud (
FVBS/Ant, A sigted
variant of the FVB/N
strain suitable for behavioural
analysis
[Poster Presentation]
R. Frank Kooy1, Vanessa
Errijgers1, Ilse Gantois1, Guy Nagels2, Aaron
W. Grossman3, Peter P. De Deyn2, Rudi D’Hooge2
1Department of Medical Genetics, and 2Department of Neurochemistry and Behavior, Born-Bunge Foundation, University of Antwerp, 2610 Antwerp, Belgium; 3Neuroscience Graduate Program, Beckman Institute, University of Illinois, Illinois 61801 IL, Urbana-Champaign, USA
Mice of the FVB/N strain are severely visual impaired as a result of genetic defects in the tyrosine kinase gene and the cGMP phosphodiesterase gene, resulting in albinism (c/c) and retinal degeneration (rd/rd), respectively. Nevertheless, FVB/N mice are commonly used for the generation of transgenic animals because of their large, strong pronuclei and high breeding performance. However due to visual impairment of the FVB/N animals, the resulting transgenic animals cannot be used in tests that depend on vision, including tests of cognitive behavior. Therefore we have bred a sighted version of FVB/N by intercrossing the FVB/N strain with 129P2/OlaHsd followed by repeated backcrossing with FVB/N mice while selecting against albinism and homozygosity of the retinal degeneration mutation. After 11 generations of backcrossing, sighted animals were intercrossed to generate the congenic FVBS/Ant strain, that is pigmented (c-ch/c-ch) and devoid of the genetic predisposition to retinal degeneration. The visual abilities of the FVBS/Ant mice were demonstrated by eye histology, a clear visual evoked potential in response to light stimuli and by increased performance in the Morris water maze test.
Assessing Anxiety in Heterogeneous Stock (HS) Mice
L. Liu1, C.
Fernandes1, J. L. Paya-Cano1, M. J. Galsworthy1, F.
Rijsdijk1,
1 Social, Genetic and Developmental
Psychiatry Research Centre, Institute of Psychiatry, King's College London, De Crespigny Park, London SE5 8AF, UK.
2 Area de Psicobiologia,
Facultad de Psicologia, Universitat de Valencia, Avda,
Anxiety and related exploratory
activity are complex traits with a genetic contribution likely to be equally complex.
A large variety of mouse tests of anxiety have been developed and attempts have
been made to define the trait within, and across, different models. However, it
is unclear whether a common underlying phenotype of anxiety can be extracted
from these multiple tests or whether each test measures distinct situation
specific anxiety and/or other exploratory behaviours. Moreover, properly
assessing mouse behaviours in order to extract accurate information for genetic
studies requires recognition of factors such as activity that can affect mouse
performance.
We have developed a mouse anxiety
test battery (including open field, elevated plus maze, light/dark box and
SHIRPA primary screen) to assess the structure of anxiety behaviours and to
extract a common anxiety factor for genetic studies. Heterogeneous stock (HS)
mice derived from outbred crosses from eight inbred strains were run through
the battery, as these mice provide individual differences in behaviour as well
as greater genetic variability for association studies. In our study of 444
mice (222 full sib pairs), we found that anxiety and activity measures highly
correlate between sib pairs, suggesting that there is a strong familial
background for such behaviours. Factor analysis and model fitting was performed
on data collected from 444 mice (222 sib pairs), to illustrate the relation
between factors underlying the common anxiety phenotype and environmental
factors. In addition, comparison between this behavioural construct and those
suggested by recent QTL evidence regarding the structure of these phenotypes
will be presented.
An algorithm for simultaneous search for
multiple QTL
Kajsa Ljungberg, Sverker Holmgren and Orjan Carlborg
Department of
Scientific Computing, Information Technology,
We have studied algorithms for simultaneous search for multiple QTL. Forward selection fails to detect epistatic QTL pairs where both loci lack significant marginal effects. Such QTL have been found in real data, see e.g. Carlborg et al., 2003. In order to investigate whether QTL pairs, triplets, etc of this type are generally important for quantitative traits it is necessary to simultaneously look for multiple QTL. Since the commonly used exhaustive search is too computationally demanding in higher dimensions, more efficient search algorithms are needed.
Carlborg et al used a general purpose genetic optimization algorithm for the searches in two dimensions. However, to obtain good results with a genetic algorithm it is necessary to fine-tune a significant number of parameters, and the optimal settings can be very different for different problems. By using an algorithm which is specially adapted to the QTL search problem more reliable results can be obtained in less time requiring fewer user-determined parameters. Here we present such an algorithm, a slight modification of DIRECT which is a global optimization algorithm presented by Jones et al. in 1991.
We compared DIRECT with the genetic optimization algorithm and with exhaustive search, investigating reliability in finding the correct optimum and speed. The tests have been made on a real data set from an F2 intercross between red jungle fowl and white leghorn with 820 individuals and a genome size of 2300 cM, using a resolution of 1cM. Both two- and three-QTL models for nine different traits with and without interaction parameters have been used. We have also performed searches using randomized data, as when determining empirical significance thresholds. In the analyses we used a linear regression mapping method, but we stress that DIRECT and the genetic algorithm can be used together with any method of QTL mapping based on a genomic search.
DIRECT was very reliable, finding the correct optimum for all non-randomized data. The search in two dimensions was 2 orders of magnitude faster than exhaustive search, taking about 30 seconds, and in three dimensions DIRECT was four orders of magnitude faster, one search taking under five minutes. Compared with DIRECT, the genetic algorithm performance was comparable for two-dimensional searches, but significantly worse in three dimensions. With randomized data DIRECT was not quite as reliable in finding the true optimum, but the results were always sufficient for obtaining empirical significance thresholds. We conclude that DIRECT a reliable and fast algorithm when searching for multiple QTL.
Quantitative trait locus analysis of prion disease incubation time in mice
S. E. Lloyd1,
1MRC Prion Unit and
Department of Neurodegenerative Diseases, Institute of Neurology,
2The Wellcome Trust
Centre for Human Genetics,
Prion disease incubation time in mice is a quantitative trait varying from 100-500 days. The main genetic determinant of incubation time is the prion gene, Prnp, where allele a (Leu 108, Thr 189) and allele b (Phe 108, Val 189) are associated with short and long incubation times respectively. Significant differences in incubation times occur between inbred lines with the same Prnp alleles suggesting that other genes are contributing to the observed variation. To identify these loci we analysed an F2 intercross between two strains of mice, CAST/Ei and NZW/OlaHsd, with significantly different incubation periods when challenged intracerebrally with either RML/ scrapie prions or mouse passaged BSE. Interval mapping identified three highly significantly linked regions on chromosomes 2, 11 and 12 for RML/scrapie and overlapping regions on chromosomes 2 and 11 for mouse passaged BSE, suggesting that these loci may act independently of prion strain. Because the regions of linkage identified in these crosses were broad (10-20cM), we are now employing a number of different approaches for fine mapping and candidate gene identification. These strategies include the use of a heterogeneous stock of mice and the generation of congenic mice to reduce the size of the regions and the use of microarrays and sequencing to identify the molecular basis of the variation.
WebQTL: Complex trait analysis beyond QTL
mapping
Kenneth F. Manly1, Jintao Wang1, Robert W. Williams2
1Roswell Park Cancer Institute, Department of
Molecular and Cellular Biology,
2University of
WebQTL is a Web-based software service for complex trait analysis and transcriptional network analysis in recombinant inbred lines of mice. WebQTL combines fast software for mapping quantitative trait loci (QTLs) with software for searching for correlations among simple and quantitative traits. These functions are supported by genotype and trait data that are not easily obtainable elsewhere. WebQTL includes curated genotype data for five sets of mouse recombinant inbred lines and one advanced intercross. Thus, to identify QTLs, geneticists need provide only quantitative trait data from one of the supported populations. WebQTL has an integrated database of trait data from BXD recombinant inbred lines, including a variety of physiological and behavioral data and gene expression data from brain and hemopoietic stem cells. The expression data is derived from hybridizations with mouse oligonucleotide microarrays (Affymetrix U74Av2). Thus, WebQTL can define RNA expression that correlates with biological phenotypes and QTLs that control either RNA expression or biological phenotypes. That is, it can find associations between phenotypes, RNA steady-state levels in one or more tissues, and genotypes. To make WebQTL more versatile and to help integrate it with other resources, recent enhancement of WebQTL includes development of a programmable interface. This interface returns information as tables of text in response to a query containing a keyword list. This interface allows automated searches and retrieval of information under control of simple scripts, and it returns results in a form that can easily be processed by scripts.
WebQTL is supported by the
Bayesian MCMC QTL mapping in outbred
mice
Andrew Morris, Binnaz Yalcin, Jan Fullerton and Jonathan Flint.
Wellcome Trust Centre for
Human Genetics,
We present a novel method for mapping QTL across candidate
regions using large sibships of outbred mice. We
begin by reconstructing marker haplotypes within sibships
to determine recombination breakpoints and inheritance vectors, allowing for
any uncertainty in the estimation procedure. The mapping method is developed in
a Bayesian framework, approximating the posterior distribution of the location
of QTL(s) across the candidate region. We employ Markov chain
We illustrate the method by application to an anxiety related trait using MF1 mice obtained from Harlan. Our results suggest evidence of multiple QTLs in a 4Mb region of mouse chromosome 1, with potential overlap with the RGS gene cluster.
Strain Differences in Stem/Progenitor Cell Proliferation and Survival During Adult Neurogenesis in the
Dentate Gyrus
N.L.
Hayes and R.S. Nowakowski.
Department of
Neuroscience and Cell Biology.
In
the adult central nervous system cell proliferation and neuron production
persists into adulthood in a two distinct regions of the brain. As the first
step toward identifying the genetic controls on this process, we have examined
cell proliferation in one of these regions, the dentate gyrus,
in several inbred strains of mice. All of the mice used for this study were
purchased from the Jackson Laboratories (
A novel
approach to genetically dissect complex quantitative traits using a diallel cross of recombinant inbred
strains
Alexander V. Osadchuk1, David C. Airey2, Lu Lu3, David W. Threadgill4, Robert W. Williams3
1Institute of Cytology and Genetics,
3Center for Neuroscience and Department
of Anatomy and Neurobiology,
4Dept of Genetics,
Introduction We describe a novel approach for the genetic dissection of complex traits. This approach exploits two advantages of the diallel cross of recombinant inbred lines: (i) the accurate estimation of hybrid line means and cross variance components; and (ii) immediate knowledge of hybrid genotypes from parental genotypes. The first advantage provides an estimate of average within-line environmental variance with which to test the adequacy of multilocus epistatic segregation models of the line means. The second advantage serves to define segregation model QTLs. Below we describe methods and results using measures of the weight of the cerebellum.
Method
The method used to construct the optimal least squares linear segregation
model is based on multiple regression analysis of the diallel
matrix and subsequent enumeration of all possible genotype variants in the
parental RI strains. A 4-locus model of the cerebellum weight adjusted for
animal age illustrates the method. It is assumed that the genotype of each
parental RI strain is represented by a vector of four variables designating
four loci with values of 0 or 1, i.e. each locus has only two alleles, 
,
the values of 0 or 1 indicating whether the allele is inherited from BALB/cByJ or C57BL/6ByJ progenitor strain, respectively. Let all
four loci be linked to autosomal chromosomes and each
locus has three digenic epistatic
interactions with other loci. Then, according to Van der
Veen’s pure-line metric [Mather
& Jinks, 1982] the genotypic contribution to the phenotype for the progeny
of qr
x qs
crosses will be the linear function of 32 indicator variables designated by
capital letters and origin – m:
where 
are indicator variables for the additive (dn)
and dominance (hn)
effects of locus number n. 
are indicator variables for the homozygote x
homozygote (ank),
homozygote x heterozygote (bnk), heterozygote x homozygote(cnk),
and heterozygote x heterozygote (gnk) interaction effects between n and k loci. Each indicator variable in the linear equation is
multiplied by the corresponding coefficients (genetic effects). Furthermore,
each indicator variable is the integer function of the genotype value, and each
may assume values of either 1, 0, or –1. Thus, the set
of 94 phenotypes derived from crosses of 13 parental RI and the 2 progenitor
strains is represented by a set of 94 linear equations with 32 variables and
one origin:
Having set the arbitrary genotype values for the
parental RI strains, the values of the indicator variables are unambiguously
defined in these equations. Using a multiple regression procedure, the task is
now reduced to a search for values of the genetic effects that minimize the
deviations from the observed phenotypic values. To test the adequacy of the
solution, the residual variance (the one discounted in multiple regression) is compared with the environmental variance
(average within-cross variation). If the Fisher’s test values for the two
variances do not differ significantly, then the obtained solution adequately
describes the among-cross variation. In such a case, the coefficient of
multiple determination, R2, indicates the proportion of the
among-cross variation explained by the fitted multiple regression. The choice
of genotype values for the parental strains is iteratively evaluated to obtain
acceptable candidate solution(s). If no variant adequately describes the
experimental data, a more complex model is needed. The total number of all
possible genotype variants for the set of 13 RI strains equals 24*13
= 252. In such a case it is currently difficult to perform
an exhaustive search for all possible genotype variants in the parental RI
strains.
To
overcome this computational difficulty, a two-step procedure was
developed. As a first step, an initial (“bootstrap”) solution was obtained by
sequential complication of a segregation model. It consists of scaling up the
number of RI strains and the genetic effects included in the model until a
first adequate solution is produced. To obtain a “bootstrap” solution, a
special PC program was developed. It was able to perform
an exhaustive search of up to 230 different genotype variants of RI
strains. The second step exploited the “bootstrap” solution to search for all
adequate solutions. For this purpose a second PC program was developed for an
exhaustive search around the initial or selected point (about 20000 different
genotype variants was examined under each iteration).
The
strongest solutions provide a means to predict the genotypic values for all
possible genotypes (there are 34 = 81 genotypic values). This makes
it possible to predict the cerebellum weights of a set of conventional F2 intercross progeny made using the same progenitor strains
and assuming an equal survival of all 81 genotypes. A log-likelihood ratio
goodness-of-fit procedure was employed to test expected and observed data of
the F2 phenotypes. This is an independent test of model validity that also
serves as a criterion for further selection of solutions. This combined
approach allowed us to derive optimal solutions that model both the 94 isogenic RI and RIX crosses as well as an F2 generation.
Finally, each adequate solution is characterized by specific strain distribution patterns (SDPs) for each of four “model” loci. These SDPs can be compared to known SDPs among well-typed marker loci in the CXB RI set. This mapping procedure for model loci is similar to that used to map any Mendelian locus. Although standard software can be used for this procedure, we wrote code to test for concordance and to find the most satisfactory matches rapidly and with a graphic output.
Results The weight of the cerebellum was measured in 13 RI strains of the CXB set, both progenitor strains, F1 and F2 testcross progeny, and in a full set of 78 RIX F1 diallel lines between the 13 RI strains (13 x 12/2). The data consisted of 1338 mice, 685 males and 653 females, ~12 mice per isogenic cross, plus 249 F2 mice. A two-way ANCOVA demonstrated highly significant among-cross variation in cerebellar weight adjusted for age. Although there was no significant sex effect or cross-by-sex interaction effect, the between-sex correlation was significant but not very high. Segregation models were therefore developed for each sex separately and combined.
Using 94 isogenic line means we defined 84 adequate 4-locus solutions. An additional selection procedure exploiting the F2 data, reduced this to 14 solutions. Subsequent analysis of the “champion” variant among 14 solutions was carried out in detail. Genetic effects that were small and insignificant were omitted from the regression analysis of the champion variant. The champion model accounted for ~ 80% of the between-cross variance, and its residual variation did not differ significantly from environmental noise. The correlation between observed and predicted hybrid line means was very high, > 0.90.
Every adequate solution is
characterized by four putative SDPs, each of which
should ideally match that of a known marker. Using a special PC program, genome
regions closely linked with model loci are determined. This was done separately
for the sets of 14 and 84 adequate solutions. Similar mapping results were
obtained for both sets of solutions. The first model locus was linked with 5
putative genome regions, with a small preference of D16Mit34. The second SDP locus was strongly linked with D17Mit160. The third SDP was linked with
D4Mit236. The forth SDP was less
strongly linked with D11Mit1.
The novel approach presented here provides a compelling alternative method to disclosing the internal genetic architecture of complex traits, as it allows for the presence of both independently acting QTLs and higher order interacting sets of QTLs. More powerful diallel data sets and computing platform changes will further improve the approach. (The work was partly supported by the grant RFBR N 01-04-49523 to AO and MH 62009 to RW)
Some considerations regarding program content and strategy
Kenneth Paigen, the Jackson Laboratory
I suggest we define our program as developing a set of resources to facilitate the high resolution mapping, cloning and functional analysis of QTL important for understanding mammalian physiological processes and human health. In addition to RI lines this program should consider whatever resources will be cost effective. In this, one of our primary goals should be providing these resources in as user friendly a manner as possible to the biomedical research community at large. We want the largest possible community of researchers taking advantage of the powerful, new genetic strategies emerging. The larger our user base, the more our project can contribute and the more likely it will be funded.
We also want to be clear that facilitating the cloning and functional analysis of QTL is as important as their original identification. This being so, the “utility” of a QTL becomes important. This depends on the magnitude and variability of its effect, considered in isolation. Given large crosses and sophisticated statistics, it is possible to identify QTL that are very difficult to work with subsequently. It will be better to define success by the number of useful QTL we find, rather than the fraction of the variance that can be accounted for.
In thinking about useful QTL, we could emphasize the value of QTL as identifying entry points and/or framework genes of important pathways, but this suffers by comparison with the various ENU mutagenesis programs which are more efficient at identifying and then cloning single gene mutations. Alternatively. it would be advantageous to focus on the concordant chromosome locations of human and mouse QTL for the same phenotype. The existence of concordance requires that there is a modest number of physiologically critical genes for these phenotypes, that it is possible to identify nearly all of these in mice, and, by implication, that their protein products are prime targets for preventive or therapeutic intervention when they involve human health. i.e. these are particularly valuable genes to understand.
Some of the major factors we need to evaluate are (a) the cost effectiveness of our proposals compared to alternative uses of money – such as individual projects doing QTL mapping in standard crosses and/or mutagenesis programs. (b) The maintenance costs of the program relative to its projected level of use. (c) The access pattern – mice to investigator (only works if mice are “clean”) or investigator to mice (requires lab facilities and phenotyping equipment at mouse house). Given the difficulties of the latter, it will have to be predominantly mice to investigator. And (d) The ease of use – this is especially important if we are to bring these tools to the entire community of mammalian biologists
A list of possible resources for creating a comprehensive resource would include:
1. A set of RI lines from multiple progenitor strains (my own preliminary cost analysis suggests that the creation of 300+ strains may well be the financial maximum for funding). For a variety of biological reasons we should consider the possibility of 48 lines from each of the six possible pairwise crosses of four progenitor strains (288 lines) as an alternative to one four-way cross.
2. Consomic lines between pairs of strains. Sixty-six strains would be required for one way consomics.
3. Including existing RI sets with particular utility, especially the very large sets between strain pairs and the set made from the HS 8-way cross.
4. An expanded phenome project that exhaustively phenotypes 50 strains to identify starting strains for individual crosses
5. A bioinformatics resource to combine QTL mapping data with other data sources to optimize candidate gene selection.
6. A user service group to help non-geneticists design and analyze QTL experiments so extend the use of the resources beyond the mouse genetics community
7. Regular meetings to discuss the experiences and problems of the user community.
And finally, we should consider attempting to fund our project in phases, where intitial successes using existing resources and success at meeting milestones in the creation of new resources will engender additional support
Generation of a SNP-based mouse
haplotype map
Pletcher, M.T.
1, Batalov, S. 2, Barnes, S.W.
2,
1The Scripps Research
Institute,
Haplotype analysis, or determination of ancestral inheritance in chromosome evolution, has the potential to increase the speed and efficiency in the way phenotypic traits are mapped in both humans and model organisms. A high-resolution haplotype map will allow for genome-wide linkage disequilibrium studies in humans and in silico QTL analysis in mice. To this end, we have previously assembled a haplotype map of 6 common inbred strains of mice based on sample sequencing of loci relatively evenly spaced across the genome. This type of map produced haplotype blocks with a resolution of roughly + 2 Megabases. To increase the scope and precision of this map, we are currently typing the 48 mouse strains of Jackson Labs’ Phenome Project with 10,000 SNP-based markers. At this date, we have approximately 1000 markers typed for these DNAs. Because of the biallelic nature of the SNP markers, a completely accurate haplotype map will require a higher density of markers but the final set should provide sufficient data to identify associations with small physical map regions for many of the Phenome project phenotypes as well as suggest the best mapping pairs to use for further refining these loci.
A screen
for modifier genes of astrocytoma and malignant
peripheral nerve sheath tumors in inbred mice mutant for Nf1 and p53.
Karlyne M. Reilly,1
Robert B. Tuskan,1 Michelle Perella,1 C. Dahlem Smith,1 Roderick T. Bronson,2 Dagan Loisel,3 Emily Christy,3 Jeremy
Ledger,3 and
1 Mouse Cancer Genetics Program, NCI@Frederick,
2 Department of Pathology, Tufts
University Schools of Medicine and Veterinary Medicine,
3 Center for
Cancer Research/HHMI, MIT,
The clinical heterogeneity of neurofibromatosis type 1 (NF1), as well as a study comparing twins and siblings with NF1 (Easton et al, 1993), suggests the importance of modifier genes in the severity of this disease. It is not clear, however, whether modifier genes similarly affect the cancers associated with NF1, or what the genes modifying NF1 might be. Mice mutant for both Nf1 and Trp53 (p53) develop tumors associated with NF1 (Cichowski et al, 1999; Vogel et al, 1999; Reilly et al, 2000) and comparison of these mice on different inbred strain backgrounds shows evidence for modifier genes acting on several tumor phenotypes. We are mapping modifier genes for resistance to several different NF1-associated tumors, and preliminary evidence suggests the presence of several modifier genes acting on this system. Cloning and identification of these genes will lead to a better understanding of NF1-associated tumorigenesis and may give rise to new therapies to prevent cancer or reduce the severity of NF1 generally.
Hippocampal gene expression profiling across eight mouse
strains: towards understanding the molecular basis for behaviour
Cathy Fernandes, Jose Paya-Cano, Frans Sluyter, Ursula DSouza, Robert Plomin, Leonard C Schalkwyk
Social, Genetic and Developmental Psychiatry Research Centre, PO 82, Institute of Psychiatry, Kings College London, De Crespigny Park, London SE5 8AF, U.K.
Abstract: Mouse inbred strains differ in many aspects of their phenotypes, and it is known that gene expression does so too. We have made a carefully controlled eight strain expression profile comparison using one well defined brain tissue (the hippocampus) in quadruplicate on Affymetrix MGU74av2 chips, and obtained highly significant differences. Many such differences would correspond to functional differences, and across a sufficient panel of strains this could be used to identify genes involved in particular phenotypes, similar to genetic correlation. This expression correlation is potentially a very powerful method, especially given the large scale generation of phenotypic data on a panel of strains in the Phenome project. As an example, correlation of the over 200 probesets with Bonferroni significant strain differences with a ranking of strains by aggression phenotype yielded COMT and FGF1 as the most significant rank correlations.
Bayesian Analysis of Multilocus
Association in Quantitative and Binary Traits
Mikko J. Sillanpää (mjs@rolf.helsinki.fi)
Rolf Nevanlinna
Institute,
A Bayesian model-based method for multilocus association analysis of quantitative and binary
traits is presented (Kilpikari & Sillanpää, 2003). The method selects a trait-associated
subset of markers among candidates and is equally applicable for analyzing wide
chromosomal segments and small candidate regions (even areas within the gene).
The method can be applied in situations involving missing values in genotype
data. The number of associated markers, their positions and strengths of
association are all estimated simultaneously. The posterior distributions of
parameters are obtained through Markov chain
REFERENCES:
KILPIKARI R, SILLANPÄÄ MJ. 2003. Bayesian analysis of multilocus association in quantitative and qualitative traits. Genetic Epidemiology (in press).
SILLANPÄÄ MJ, KILPIKARI R, RIPATTI S, ONKAMO P, UIMARI P. 2001. Bayesian association mapping for quantitative traits in a mixture of two populations. Genetic Epidemiology 21(suppl 1): S692-S699.
Towards QTL Mapping & Microarrays to
Identify Modifiers of Caspase 3 Induced Exencephaly
Kathleen G. Banks1, Caroline
Houde2, Colleen C. Nelson3
Sophie Roy2, Donald W. Nicholson2, and
Elizabeth M. Simpson1
1Centre for Molecular Medicine and Therapeutics, British Columbia Research Institute for Children's & Women's Health, and Department of Medical Genetics, University of British Columbia, Vancouver, British Columbia, Canada; 2Merck-Frosst Canada & Co., Pointe-Claire-Dorval, Québec, Canada; 3The Prostate Centre at Vancouver General Hospital, Department of Surgery, University of British Columbia, Jack Bell Research Centre.
Caspase 3 is one of the key effectors in apoptosis, which plays an important role in brain development. Mice deleted for the caspase 3 gene (Casp3) displayed defects in brain development, prominent exencephaly, and premature death between 1-5 weeks of age (Kuida et al., 1998; Woo et al., 1998). Notably, this phenotype was variable when described on a mixed B6129 genetic background. Subsequent backcrossing to C57BL/6J (B6) and 129X1/SvJ strains resulted in background-dependent phenotypes, i.e. resistant on B6 and sensitive on 129X1 (Leonard et al., 2002). Our aim has been to further explore this phenotypic dependence on genetic background and to identify genes that modify caspase 3-mediated apoptosis using a modifier screen. Towards this end we have generated congenic strains of Casp3-null mice: B6 (N15), 129S1/SvImJ (N11), FVB/N (N15), DBA/2J (N8), and CAST/Ei (N8). Our results show CAST-Casp3-/- mice (n = 6) are highly resistant to the presence of the mutation as none have shown a gross phenotype to date. Similarly, B6-Casp3-/- mice (n = 165) are relatively resistant, with survival up to 2 years and showing, on average, less severe brain malformations without exencephaly.
In total, twelve percent of B6-Casp3-/- adults developed enlarged brains and a single mouse had exencephaly at e18.5. Conversely, the 129-Casp3-/- mice (n = 32) were highly sensitive to the mutation. These mice were present at wean at a lower frequency than expected (p<0.001); most died perinatally, 3 barely survived past wean, and all showed phenotypic abnormalities ranging from exencephaly to cleft face. Of fifty-nine 129-Casp3-/- e18.5 embryos 100% were similarly affected. Finally, phenotypic expression was highly variable on the FVB/N background indicating major non-genetic influences. Prior to initiating a mapping cross we are exploring a variety of ways to phenotype including quantification of brain tissue, DNA fragmentation in telencephalic vesicle tissue cultures, and microarray analysis. Having established this resource, we are set to begin the screen to identify modifiers of this important gene in brain development.
A multi-phenotype procedure for fine scale
mapping of QTL in outbred heterogeneous stock mice
1University of
2University of
Multiple genetic components as well as the environment influence human diseases such as generalized anxiety disorder, asthma and diabetes. Such complexity has made it difficult to identify genes in humans involved in these disorders. Using animal models of disease, on the other hand, bypasses several problems inherent in studying human populations such as genetic heterogeneity and environmental variation. Conventional quantitative trait loci (QTL) methods (studying the segregating F2 generation of a cross between two inbred strains) have successfully identified QTL for many phenotypes. However the mapping resolution for QTL is poor, making gene identification laborious and often unfruitful. As an alternative to conventional QTL mapping methods using inbred strains, we are phenotyping 2000 mice from a heterogeneous stock (HS). These HS mice are outbred from eight inbred founder strains, enabling us to use ancestral haplotypes in the identification and fine mapping of QTLs. Using a dynamic programming algorithm on simulated data, we have shown that the genotyping of 2000 HS mice with 3000 molecular markers will enable us accurately to identify even small effect QTL (explaining 2.5 – 5% of the phenotypic variance) and map them at a sub-centimorgan resolution. To maximize the information gained from such a large undertaking, we are measuring multiple phenotypes in the HS mice, including measures for anxiety, asthma and metabolic disorders, among others. Once QTL are identified for these phenotypes, the high resolution analysis will allow us to determine candidate genes more easily than conventional QTL methods. In this presentation, the phenotyping protocol and data collection will be discussed.
Allelic
series and extreme background dependent genetic heterogeneity associated with
alterations in epidermal growth factor receptor signaling.
Karen E. Strunk, Reade B. Roberts, Daekee Lee, and David W. Threadgill
Department of Genetics,
Four mutations, a targeted null mutation (Egfrtm1Mag), a spontaneous recessive hypomorph (Egfrwa2), and two ENU-induced (EgfrDsk5, dominant hypermorph and EgfrWa5, dominant hypomorph), exist for the epidermal growth factor receptor. Phenotypic analysis of the mutations reveals that the phenotypes are modified by many other genes. For example, the timing of lethality due to perturbed placental development in mice homozygous for Egfrtm1Mag is strongly dependent on the genetic background harboring the mutation. The first indication for a complex set of modifier genes was evident when initial attempts to genetically map background modifiers using outbred Swiss-derived CD-1 mice were unsuccessful. Therefore, to investigate the genetic architecture contributing to survival of Egrtm1Mag homozygous embryos, we partitioned the genetic variability segregating within the outbred population by surveying nine Swiss-derived inbred strains for viability of the Egfrtm1Mag null mutants; the strains included ALR/LtJ, ALS/LtJ, APN, APS, ICR/HaRos, NOD/LtJ, NON/LtJ, SJL/J, and SWR/J. We observed that these independently derived strains support varying levels of survival of Egfrtm1Mag homozygous embryos, suggesting that outbred CD-1 mice are genetically heterogenous for Egfrtm1Mag modifier alleles. To further characterize modifiers of Egfrtm1Mag and to test the role of heterosis, the mutation was also analyzed on eight congenic strains, derived from 129S6/SvEvTAC, AKR/J, BALB/cJ, BTBR, C3H/HeJ, C57BL/6J, DBA/2J, and FVB/NJ inbred genetic backgrounds. By intercrossing the congenic lines to create hybrid F1 embryos, we observed that different genetic backgrounds have complementary modifiers, further supporting the notion that the phenotypic variation observed with Egfrtm1Mag homozygous embryos is due to extensive genetic heterogeneity. Furthermore, a detailed analysis of the crosses suggests that there exists three distinct groups of modifiers, functioning at different stages of development, one to support survival of Egfrtm1Mag homozygous embryos to mid-gestation, one to support development through the mid-gestation transition from yolk-sac to placental-derived nutrient sources, and the third during later stages of gestation.
Similar observations have been observed for congenic lines carrying the Egfrwa2 mutation. In fact, only one genetic background surveyed supports a robust survival of Egfrwa2 homozygous mice. Three other congenic lines, derived from A/J, BTBR, and 129S6/SvEvTAC, do not support survival. Interestingly, hybrids between BTBR and 129S6 do not complement. Numerous physiological abnormalities associated with Egfr mutations are also restricted to specific genetic backgrounds. We have confirmed the existence of extensive genetic heterogeneity across Egfr lines at the molecular level via gene profiling suggesting that microarrays will be a generally useful technique to partition genetic heterogeneity contributing to other complex traits in the mouse.
These results have significant clinical ramifications. Small molecule inhibitors of Egfr are currently being used to treat numerous types of cancer. However, we would predict that significant individual differences would exist in the ability of these inhibitors to reduced cancer growth depending on patient background genetics. As such, the existing mouse Egfr mutations should provide a tool to partition patients that will positively respond to Egfr-inhibitor therapy.
QTL fine-mapping with Recombinant-Inbred Heterogeneous Stocks and In-Vitro Heterogeneous Stocks
William Valdar,
Jonathan Flint, Richard Mott
Wellcome
Trust Centre for Human Genetics
We
compare strategies to fine-map Quantitative Trait Loci (QTL) in mice using
Heterogeneous Stocks (HS). We show that a panel of about 100 Recombinant Inbred
Lines (RIL) derived from an HS, and which we call an RIHS, is ideally suited to
fine-map QTL to very high resolution, without the cost of additional
genotyping. We also investigate a strategy based on in-vitro fertilisation of
large numbers of F1 offspring of HS males crossed with an inbred line (IVHS).
This method requires some additional genotyping but avoids the breeding delays
and costs associated with the construction of a RI panel. We show that QTL
detection is higher using RIHS than with IVHS, and that it is independent of
the number of RI lines, provided the total number of animals phenotyped is
constant. However, fine-mapping accuracy is slightly better using IVHS. We also
investigate the effects of varying the number of HS generations and using
microsatellites instead of SNPs. We find that quite modest generation times of
10-20 generations are optimal. Microsatellites are only superior to SNPs when
the generation time is 30 or more and when the markers are widely spaced.
Detailed genetic
characterization of an anxiety susceptibility QTL in mouse model : how far are we from discovering a QT
gene?
Binnaz Yalcin, Jan Fullerton, Sue Miller, Richard Mott and Jonathan Flint.
Wellcome Trust Centre for Human Genetics,
Using HS mice, we have fine mapped an open field behaviour QTL on mouse chromosome 1 to an interval of 0.8 cM. We have established a 4.8-Mb high-resolution integrated BAC-based map underlying this region, encompassing 10 genes: 8 of which are known (B3Galt2, Glrx2, Ssa2, Uchl5) including 4 members of the RGS gene family and 2 unknown genes CDC73 homologue and retinoic acid inducible neural specific protein homologue. We sequenced all the genes in each of the 8 HS founders strains. We did not find any mutations that alter the genes. We have used a statistical approach to assign probabilities to all non coding polymorphisms identified as possible QTNs. Using a combination of different approaches such as haplotype structure and higher resolution mapping with the use of MF1 mice, it appears that two loci might be involved with different modes of action. The first region correspond to the RGS genes cluster, and the second region points out a 48 Kb inversion that we have recently found.
In this presentation I will show that a genetic study of a behavioural trait in an animal model takes us close to the quantitative trait nucleotide but does not identify a gene for certain. We need functional studies to help us identify the variants that affect this phenotype.
A multi-strain, high-resolution mouse
haplotype map reveals the heterogeneity of mouse haplotype block structure
Laboratory of Population Genetics, National Cancer Institute, National Institutes of Health, 8424 Helgerman Court, Room 101, MSC 8302, Bethesda, Maryland 20892-8302, U. S. A.
We have constructed a multi-strain, high-resolution haplotype map for the 100-Mb mouse chromosome 16 using ~70,000 SNPs derived from Celera whole genome shotgun sequencing and from MIT whole genome shotgun sequencing. Haplotype block sizes, haplotype block structure and the distribution of genomic regions with extremely low or high SNP density revealed by this map are strikingly different from what has been described in previous publications1,2. Less than 5% of SNPs were included in haplotype blocks >=1Mb, and the average and the median haplotype block sizes are 39kb and 500kb respectively. Only 18% of the genomic regions are SNP-poor, and some of these are included in large haplotype blocks in which regions with extremely high and low SNP density have the same segregation pattern. The proportion of non-synonymous SNPs that change protein sequences is 50% lower in the mouse genome than in the human, suggesting that selection pressure on the inbred strains may have eliminated deleterious alleles. We are currently attempting to verify and potentially extend these results by resequencing approximately 32-kb of the genome in more than 20 inbred strains.
References
1. Wiltshire, T. et al. Genome-wide single-nucleotide polymorphism analysis defines haplotype patterns in mouse. Proc Natl Acad Sci U S A 100, 3380-5 (2003).
2. Wade, C.M. et al. The mosaic structure of variation in the laboratory mouse genome. Nature 420, 574-8 (2002).