Abstract As a result of nonequivalent genetic con- tribution of maternal and paternal genomes to off- springs, genomic imprinting or called parent-of-origin effect, has been broadly identified in plants, animals and humans. Its role in shaping organism’s develop- ment has been unanimously recognized. However, statistical methods for identifying imprinted quantita- tive trait loci (iQTL) and estimating the imprinted ef- fect have not been well developed. In this article, we propose an efficient statistical procedure for genome- wide estimating and testing the effects of significant iQTL underlying the quantitative variation of inter- ested traits. The developed model can be applied to two different genetic cross designs, backcross and F 2 families derived from inbred lines. The proposed pro- cedure is built within the maximum likelihood frame- work and implemented with the EM algorithm. Extensive simulation studies show that the proposed model is well performed in a variety of situations. To demonstrate the usefulness of the proposed approach, we apply the model to a published data in an F 2 family derived from LG/S and SM/S mouse stains. Two par- tially maternal imprinting iQTL are identified which regulate the growth of body weight. Our approach provides a testable framework for identifying and estimating iQTL involved in the genetic control of complex traits. Keywords EM algorithm Genomic Imprinting Inbred Lines Maximum likelihood Quantitative trait loci Introduction Gene imprinting refers to the phenomenon that the expression of certain gene is determined by the non- equivalent genetic contribution of maternal and paternal genomes to offsprings (Pfeifer 2000). As a result of genomic imprinting, monoallelic expression of a gene is expected depending on the parental origin of the alleles. Genomic imprinting is thought of as a particular sub-type of dominant modification (Sapienza 1990). Therefore, a dosage difference would be expected when one or more of these modifiers are sex- linked. For example, the reciprocal heterozygotes may have different average phenotypes if the underlying gene is imprinted. Representing a totally different genetic inheritance pattern compared to the traditional Mendelian’s inheritance, imprinting-like phenomena have been ubiquitously observed in a wide range of phyla span- ning from plants, animals to humans. Since the first imprinted gene (Igf2) was identified in mouse (DeChiara et al. 1991), there are over 80 imprinted genes being identified in mammals (Morison et al. 2005). Imprinted genes were also identified in other organisms such as plants (Alleman and Doctor 2000) Y. Cui (&) Department of Statistics and Probability, Michigan State University, A-411 Wells Hall, East Lansing, MI 48824, USA e-mail: [email protected]J. M. Cheverud Department of Anatomy and Neurobiology, Washington University Medical School, St. Louis, MO 63110, USA R. Wu Department of Statistics, University of Florida, Gainesville, FL 32611, USA Genetica (2007) 130:227–239 DOI 10.1007/s10709-006-9101-x 123 ORIGINAL PAPER A statistical model for dissecting genomic imprinting through genetic mapping Yuehua Cui James M. Cheverud Rongling Wu Received: 3 March 2006 / Accepted: 24 July 2006 / Published online: 6 September 2006 ȑ Springer Science+Business Media B.V. 2006
13
Embed
A statistical model for dissecting genomic imprinting through genetic mapping
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Abstract As a result of nonequivalent genetic con-
tribution of maternal and paternal genomes to off-
springs, genomic imprinting or called parent-of-origin
effect, has been broadly identified in plants, animals
and humans. Its role in shaping organism’s develop-
ment has been unanimously recognized. However,
statistical methods for identifying imprinted quantita-
tive trait loci (iQTL) and estimating the imprinted ef-
fect have not been well developed. In this article, we
propose an efficient statistical procedure for genome-
wide estimating and testing the effects of significant
iQTL underlying the quantitative variation of inter-
ested traits. The developed model can be applied to
two different genetic cross designs, backcross and F2
families derived from inbred lines. The proposed pro-
cedure is built within the maximum likelihood frame-
work and implemented with the EM algorithm.
Extensive simulation studies show that the proposed
model is well performed in a variety of situations. To
demonstrate the usefulness of the proposed approach,
we apply the model to a published data in an F2 family
derived from LG/S and SM/S mouse stains. Two par-
tially maternal imprinting iQTL are identified which
regulate the growth of body weight. Our approach
provides a testable framework for identifying and
estimating iQTL involved in the genetic control of
complex traits.
Keywords EM algorithm Æ Genomic Imprinting ÆInbred Lines Æ Maximum likelihood Æ Quantitative trait
loci
Introduction
Gene imprinting refers to the phenomenon that the
expression of certain gene is determined by the non-
equivalent genetic contribution of maternal and
paternal genomes to offsprings (Pfeifer 2000). As a
result of genomic imprinting, monoallelic expression of
a gene is expected depending on the parental origin of
the alleles. Genomic imprinting is thought of as a
particular sub-type of dominant modification (Sapienza
1990). Therefore, a dosage difference would be
expected when one or more of these modifiers are sex-
linked. For example, the reciprocal heterozygotes may
have different average phenotypes if the underlying
gene is imprinted.
Representing a totally different genetic inheritance
pattern compared to the traditional Mendelian’s
inheritance, imprinting-like phenomena have been
ubiquitously observed in a wide range of phyla span-
ning from plants, animals to humans. Since the first
imprinted gene (Igf2) was identified in mouse
(DeChiara et al. 1991), there are over 80 imprinted
genes being identified in mammals (Morison et al.
2005). Imprinted genes were also identified in other
organisms such as plants (Alleman and Doctor 2000)
Y. Cui (&)Department of Statistics and Probability, Michigan StateUniversity, A-411 Wells Hall, East Lansing, MI 48824, USAe-mail: [email protected]
J. M. CheverudDepartment of Anatomy and Neurobiology, WashingtonUniversity Medical School, St. Louis, MO 63110, USA
R. WuDepartment of Statistics, University of Florida, Gainesville,FL 32611, USA
Genetica (2007) 130:227–239
DOI 10.1007/s10709-006-9101-x
123
ORIGINAL PAPER
A statistical model for dissecting genomic imprintingthrough genetic mapping
Yuehua Cui Æ James M. Cheverud Æ Rongling Wu
Received: 3 March 2006 / Accepted: 24 July 2006 / Published online: 6 September 2006� Springer Science+Business Media B.V. 2006
The square roots of the mean squared errors of the MLEs are given in parentheses
The locations of the the QTL is described by the map distances (in cM) from the first marker of the linkage group (100 cM long). Thehypothesized r2 value is 1.22 for H2 = 0.1, 0.42 for H2 = 0.25 and 0.20 for H2 = 0.4
232 Genetica (2007) 130:227–239
123
In general, our iQTL interval mapping model can
provide reasonable estimates of QTL locations, QTL
effects and residual variances for various imprinting
mechanism, no imprinting (Table 2), completely
maternal imprinting (Table 3), and partially maternal
imprinting (Table 4). Several interesting findings are
summaries as follows. First of all, the QTL location can
be reasonably estimated with the precision depending
on the sample size and heritability level and also the
imprinting mechanism. As we expected, better esti-
mation can be obtained with large sample size and
higher heritability level. The square root of mean
square root (SMSE) decreases as sample size increases
from 200 to 400, and similar pattern can be observed
for increased heritability level. Comparing the two
designs, design B1 provides more precise estimation
than design F2 when the QTL is completely maternal
imprinted. However, poor estimation is observed for
design B1 when the QTL is partially imprinted.
Second, most genetic parameters can be quite reli-
ably estimated using our model. As expected, the
additive genetic effect (a) can be more reliably esti-
mated than the imprinting effects (c1 and c2). Increased
H2 and sample sizes can always improve the precision
of the parameter estimation (Tables 2–4). But relative
to the effect of increasing sample sizes, the increase of
H2 from 0.1 to 0.4 can lead to more significant
improvement for the estimation precision than the
increase of sample size n from 200 to 400. For example,
the SMSE of the MLE of the maternal imprinting
parameter c1 for design B1 in Table 3 would decrease
by more than 2-fold when increasing H2 from 0.1 to 0.4
than increasing n from 200 to 400. This suggests that in
practice well-managed experiments, through which
Table 3 The MLEs of the QTL position and effect parameters with completely maternal imprinting based on 200 simulation replicates
Design H2 n Position (48 cM) l = 5 a = 0.5 c1 = – 1 c2 = 1 r2
The square roots of the mean squared errors of the MLEs are given in parentheses
The locations of the the QTL is described by the map distances (in cM) from the first marker of the linkage group (100 cM long). Thehypothesized r2 value is 2.25 for H2 = 0.1, 0.75 for H2 = 0.25 and 0.38 for H2 = 0.4
Table 4 The MLEs of the QTL position and effect parameters with partially maternal imprinting based on 200 simulation replicates
Design H2 n Position (48 cM) l = 5 a = 0.5 c1 = 0.5 c2 = – 0.4 r2
The square roots of the mean squared errors of the MLEs are given in parentheses
The locations of the the QTL is described by the map distances (in cM) from the first marker of the linkage group (100 cM long). Thehypothesized r2 value is 1.35 for H2 = 0.1, 0.45 for H2 = 0.25 and 0.23 for H2 = 0.4
Genetica (2007) 130:227–239 233
123
residual errors are reduced and therefore H2 is
increased, are more important than simply increasing
sample sizes. The similar findings are obtained in
in terms of the estimation precision for most genetic
parameters (indicated by SMSE).
A case study
Currently, we do not have real data for the backcross
design to test our model. We applied our model to a
published data set (Vaughn et al. 1999) based on de-
sign F2 aimed to map iQTL that contribute to varia-
tion in body growth in an animal model system –
mouse. The F2 progeny was measured for their body
mass at 10 weekly intervals starting at age 7 days. A
regular genetic linkage map was constructed. The
sex-specific genetic distance is reconstructed based on
the marker data and the fact of 1.25:1 genetic distance
ratio between female and male chromosome (Dietrich
et al. 1996).
The similar LR profile plot is obtained by using the
model proposed in this study as the one shown in Cui
et al. (2006). To save space, we only give the likelihood
profile plot for data observed at Week 4 (Fig. 1). In
Fig. 1, the solid curve corresponds to the LR profile
and the dashed line corresponds to the 5% significant
threshold out of 1000 permutations. Clearly, there are
one QTL detected on chromosome 6 (Q6) and one
detected at chromosome 15 (Q15). Table 5 tabulates
the maximum likelihood estimates (MLEs) of the QTL
position and effect parameters, their asymptotic sam-
pling error and the P-values for iTest 1–3.
Even though the current model and our previous
model (Cui et al. 2006) agree on the number and
location of QTL detected, the imprinting property for
the identified QTL is different. It can be seen from
the testing result that Q6 shows evidence of imprinting
0
10
20
30
40D1 D2 D3 D4 D5
10
20
30
40D6
LR
D7 D8 D9 D10 D11
0
10
20
30
40D12 D13 D14 D15 D16 D17 D18 D19
10 cM
Test Position
Fig. 1 The profiles of the log-likelihood ratios (LR) between thefull (there is a QTL) and reduced model (there is no QTL) forbody mass growth observed at Week 4 across the 19 mousechromosomes using the linkage map constructed from microsat-ellite markers (Vaughn et al. 1999). The genomic positionscorresponding to the peak of the curve are the MLEs of the QTL
localizations (indicated by the arrows). The genome-widethreshold values for claiming the existence of QTL is given asthe horizonal dashed lines. Tick marks on the x-axis representthe positions of markers on the linkage group, whose names aregiven in Vaughn et al. (1999)
(P-value = 0.025 in iTest 1), and this QTL is a partially
maternal iQTL (P-value = 0.03 in iTest 2 and 0.039 in
iTest 3, c1\c2). However, different from previous
findings, Q6 shows a time-related (age-dependent)
effect which shows partial imprinting only during week
4 and 5. After week 6, the imprinting effect switches off
and this QTL functions as a regular Mendelian QTL
until the end of the observation period, week 10. An-
other inconsistency is that QTL in chromosome 15
(Q15) does not show evidence of imprinting (P-va-
lue = 0.72 in iTest 1).
Another iQTL detected is located on chromosome
10, and it shows partial maternal imprinting effect
starting from week 7. One Mendelian QTL detected is
on chromosome 7. These two QTL are consistent with
our previous findings (Cui et al. 2006). Totally, we
identified four major QTL and two of them are Men-
delian QTL and two of them are iQTL.
Discussion
The discovery of genomic imprinting leads to a totally
different area of epigenetic study, epigenetics, which is
different from the traditional Mendelian landscape. As
a normal mechanism of genetic regulation, genomic
imprinting has provided remarkable insights into pre-
viously puzzling human diseases such as Prader-Willi,
Angelman, and Beckwith-Wiedemann syndromes
(Falls et al. 1999). The identification of imprinted
genes in plants, animals and humans also greatly
changed the way scientists think of traditional Men-
delian genetics. Advanced biotechnology such as
positional cloning and candidate gene testing has sig-
nificantly facilitated the identification of imprinted
genes with monoallelic expression (Pfeifer 2000).
While considering that these techniques are still
expensive to implement in lab, genetic mapping pro-
vide an alternative way for identifying potentially im-
printed genes responsible for quantitative variation
(de Koning et al. 2000, 2002; Hanson et al. 2001;
Knapp et al. 2004; Shete et al. 2003; Strauch et al.
2000; Tuiskula-Haavisto et al. 2004). These methods
are focused on family based study or outbred species.
However, family based data is not available for plants
and using outbred lines might introduce confounding
genetic effects for imprinting due to high heterozy-
gosity at a given QTL (Lin et al. 2003).
To overcome these problems for mapping iQTL,
inbred lines serve as a good alternative since one can
easily select the lines with interests of specific pheno-
type. Here we propose a statistical mapping approach
to detect genome-wide significant iQTL in experi-
mental crosses of inbred species. A quantitative genetic
model is presented through which two different genetic
designs can be applied to map iQTL. This model
considers two imprinting parameters which measure
the degree of imprinting, and hence provide a complete
dissection of the imprinting effect. Several hypothesis
tests can be conducted focusing on these two parame-
ters to provide a complete dissection of imprinting
property of any detected significant QTL. As such, the
developed model is both statistically and biologically
interpretable. The partial imprinting effect shown by
the iQTL detected in this study is consistent with cur-
rent findings (Sandovici et al. 2003, 2005; Naumova
and Croteau 2004; Spencer 2002).
Simulation study demonstrates the usefulness of the
developed model based on backcross and F2 design in
which robust parameter estimation can be obtained
with modest heritability level and sample size. Low
heritability (H2 = 0.1) and small sample size (n = 200)
result in poor parameter estimation. This information
is useful in practice as we need to be cautious about the
interpretation of imprinting effect detected with real
data when the proportion of variance explained by the
QTL is low. Although backcross design outperforms F2
design in most cases in which backcross design pro-
vides more precise genetic parameter estimation than
the F2 design, the trade-off is poor variance and QTL
Table 5 The MLEs of the QTL position and effect parameters in the F2 progeny derived from the LG/J and SM/J strains estimated bythe imprinting model at week 4
Chrom QTL position l a c1 c2 r2 iTest 1 iTest 2 iTest 3 H2
location estimation in backcross design. However, the
bias could be greatly reduced if the two design popu-
lations ( B1 and B2) are combined together. Statisti-
cally, design B2 can be viewed as a replicate of design
B1. With increased total sample size, we would expect
increased mapping power and reduced bias for
parameter estimation. A number of studies has shown
the advantage of combing different related crosses in
QTL mapping study (Xu 1998; Zou et al. 2001).
Therefore, backcross design is more preferable to the
F2 design in real experiment.
As shown in the paper, the genetic difference be-
tween male and female chromosomes provides the
theoretical basis for distinguishing the effects of two
heterozygotes Amaf and amAf. However, the difference
is just a genomewide averaged difference and real
distance may vary by intervals. Moreover, many plants
and animals do not show evidence of differences for
two homologous chromosomes. These greatly restrict
the usefulness of design F2. For these concerns, we
would recommend researchers to use the backcross
instead of the F2 design when they design their
experiments. For existing F2 data, it is still worth to try
our F2 model to explore some potential imprinting
information.
In our previous paper (Cui et al. 2006), the genetic
model is derived by partition the additive effect into
two parts and hence it can be defined as an additive
model. The genetic model proposed in this study only
put restrictions on the two reciprocal heterozygotes
and can be defined as a multiplicative model. Even
though the results obtained by both models are broadly
consistent in which both models agree on the same
number of QTL detected spanning across the entire
genome, disagreement arises for the imprinting prop-
erty of detected QTL Q6 and Q15 as indicated by the
case study. These discrepancies may be due to the
difference in the threshold criteria used to claim the
existence of an iQTL between the two approaches or
could be due to the nature of the genes. For example,
Q6 shows some evidences of time-related (or age-
dependent) effect. One of the possible explanations for
this phenomenon is to use the loss of imprinting (LOI)
theory where the imprinting effect could be removed
by epigenetic modification (Feinberg 2001). Evidence
of LOI has been reported on chromosome 6 in mouse
genome for a maternally imprinted gene DLX5 (Hor-
ike et al. 2005). This gene is located on a well identified
imprinted gene cluster (Ono et al. 2003). Since im-
printed genes tend to cluster together on the genome
(Pferfer 2000), it is possible that Q6 is also located on
this region and experiences LOI. Therefore, before
seeking for further lab verification of the imprinting
property of any detected iQTL, it is safe to try both
models.
Our presentation of this paper is focused on the
genetic model and its implementations on two genetic
designs. The model can be applied to plant or animal
systems. Despite its usefulness of identifying and
estimating the effect of genomic imprinting, the
model developed in this study can be considered as
simple. Several studies have shown that maternal ef-
fect plays a very important role in many species
(Wade 1998; Wolf 2000, 2002; Agrawal et al. 2001).
Its effect in shaping the development of the offspring
is beyond the direct inheritance of alleles and can be
thought of as an indirect contribution of maternal to
its offspring. Both genomic imprinting and maternal
effect may have similar contribution to the phenotype
of offsprings from the same mother. Having a clear
dissection of these two effect could result in a better
understanding of the genetic architecture of any bio-
logical traits caused by epigenetic modification. An-
other limitation of the current model is that gene
epistasis is not considered where complicated inter-
actions could occur among the same or different
generations. Evidence of interaction between im-
printed genes in mouse has been reported (Cattanach
et al. 2004). Our follow-up papers are concerned with
developing models to dissect the contribution of
maternal effect and genomic imprinting involving
complicated genetic interactions.
Acknowledgements We thank the anonymous referees and theeditor for their valuable comments on the manuscript. Thisresearch was supported by a start-up fund from Michigan StateUniversity.
Appendix A: EM algorithm for estimating the genetic
parameters
The following algorithm is applied to Design I. Similar
algorithm can be obtained for Design II with little
modification and hence is omitted here.
The log-likelihood function of Eq. (5) can be written
Clapcott SJ, Teale AJ, Kemp SJ (2000) Evidence for genomicimprinting of the major QTL controlling susceptibility totrypanosomiasis in mice. Parasite Immunol (Oxf) 22:259–264
Cui YH, Casella G, Wu RL (2004) Mapping quantitative traitlocus interactions from the maternal and offspring genomes.Genetics 167:1017–1026
Cui YH, Lu Q, Cheverud JM, Littell RC, Wu RL (2006) Modelfor mapping imprinted quantitative trait loci in an inbred F2design. Genomics 87:543–551
de Koning D-J, Rattink AP, Harlizius B, van Arendonk JAM,Brascamp EW et al (2000) Genome-wide scan for bodycomposition in pigs reveals important role of imprinting.Proc Natl Acad Sci USA 97:7947–7950
de Koning D-J, Bovenhuis H, van Arendonk JAM (2002) On thedetection of imprinted quantitative trait loci in experimentalcrosses of outbred species. Genetics 161:931–938
DeChiara TM, Robertson EJ, Efstratiadis A (1991) Parentalimprinting of the mouse insulin-like growth factor II gene.Cell 64:849–859
Dempster AP, Laird NM, Rubin DB (1977) Maximum likeli-hood from incomplet data via the EM algorithm. J R StatistSoc B 39(1):1–38
Dietrich WF, Miller J, Steen R, Merchant MA, Damron-Boles Det al (1996) A comprehensive genetic map of the mousegenome. Nature 380:149–152
Falls JG, Pulford DJ, Wylie AA, Jirtle RL (1999) Genomicimprinting: implications for human disease. Am J Pathol154:635–647
Feinberg AP (2001) Genomic imprinting and cancer. In: ScriverCR, Beaudet al, Sly WS, Valle D (eds) The metabolic andmolecular bases of inherited disease. McGraw-Hill, NewYork, pp 525–537
Haldane JBS (1922) The part played by recurrent mutation inevolution. Am Nat 67:5–9
Hanson RL, Kobes S, Lindsay RS, Kmowler WC (2001)Assessment of parent-of-origin effects in titative traits. Am JHum Genet 68:951–962
Horike S, Cai S, Miyano M, Cheng J-F, Kohwi-Shigematsu T(2005) Loss of silent-chromatin looping and impairedimprinting of DLX5 in Rett syndrome. Nat Genet 37:31–40
Jeon J-T, Carlborg O, Tornsten A, Giuffra E, Amarger V et al(1999) A paternally expressed nd cardiac muscle mass inpigs maps to the IGF2 locus. Nat Genet 21:157–158
Knapp M, Strauch K (2004) Affected-sib-pair test for linkagebased on constraints for identical-by-descent distributionscorresponding to disease models with imprinting. GenetEpidemiol 26:273–285
Lander ES, Botstein D (1989) Mapping Mendelian factorsunderlying quantitative traits using RFLP linkage maps.Genetics 121:185–199
Lin M, Lou X-Y, Chang M, Wu RL (2003) A general statisticalframework for mapping quantitative trait loci in nonmodelsystems: issue for characterizing linkage phases. Genetics165:901–913
Louis TA (1982) Finding the observed information matrixwhen using the EM algorithm. J Roy Stat Soc Ser B44:226–233
Luedi PJ, Hartemink AJ, Jirtle RL (2005) Genome-wide pre-diction of imprinted murine genes. Genome Res 15:875–884
McInnis MG, Lan TH, Willour VL, Mcmahon FJ, Simpson SGet al (2003) Genome-wide scan of bipolar disorder in 65pedigrees: supportive evidence for linkage at 8q24, 18q22,4q32, 2p12, and 13q12. Mol Psychiatry 8:288–298
Morison IM, Ramsay JP, Spencer HG (2005) A census ofmammalian imprinting. Trends Genet 21:457–465
Naumova AK, Croteau S (2004) Mechanisms of epigeneticvariation: polymorphic imprinting. Curr Genomics 5:417–429
Nezer C, Moreau L, Brouwers B, Coppieters W, Detilleux J et al(1999) An imprinted QTL with major effect on muscle massand fat deposition maps to the IGF2 locus in pigs. NatGenet 21:155–156
Ono R, Shiura H, Aburatani H, Kohda T, Kaneko-Ishino T,Ishino F (2003) Identification of a large novel imprintedgene cluster on mouse proximal chromosome 6. GenomeRes 13:1696–1705
Paterson AD, Naimark DMJ, Petronis A (1999) The analysis ofparental origin of alleles may detect susceptibility loci forcomplex disorders. Hum Hered 49:197–204
Pfeifer K (2000) Mechanisms of genomic imprinting. Am J HumGenet 67:777–787
Sandovici I, Leppert M, Hawk PR, Suarez A, Linares Y,Sapienza C (2003) Familial aggregation of abnormalmethylation of parental alleles at the IGF2/H19 and IGF2Rdifferentially methylated regions. Hum Mol Genet12:1569–1578
Sandovici I, Kassovska-Bratinova S, Loredo-Osti JC, Leppert M,Suazez A et al (2005) Interindividual variability and parentof origin DNA methylation differences at specific humanAlu elements. Hum Mol Genet 14:2135–2143
Sapienza C (1990) Sex-linked dosage-sensitive modifiers asimprinting genes. Dev Suppl 107–113
Shete S, Zhou X, Amos CI (2003) Genomic imprinting andlinkage test for quantitative trait loci in extended pedigrees.Am J Hum Genet 73:933–938
Spencer HG (2002) The correlation between relatives on thesupposition of genomic imprinting. Genetics 161:411–417
Strauch K, Fimmers R, Kurz T, Deichmann KA, Wienker TF,Baur MP (2000) Parametric and nonparametric multipointlinkage analysis with imprinting and two-locus-trait models:application to mite sensitization. Am J Hum Genet 66:1945–1957
Tuiskula-Haavisto M, de Koning DJ, Honkatukia M, SchulmanNF, Maki-Tanila A, Vilkki J (2004) Quantitative traitloci with parent-of-origin effects in chicken. Genet Res84:57–66
Van Laere AS, Nguyen M, Braunschweig M, Nezer C, ColletteC et al (2003) A regulatory mutation in IGF2 causes amajor QTL effect on muscle growth in the pig. Nature425:832–836
Vaughn TT, Pletscher LS, Peripato A, King-Ellison K, Adams E,Erikson C, Cheverud JM (1999) Mapping quantitative traitloci for murine growth. A closer look at genetic architecture.Genet Res 74:313–322
Wade MJ (1998) The evolutionary genetics of maternal effects.In: Mousseau TA, Fox CW (eds) Maternal effects as adap-tations. Oxford University Press, New York, pp 5–21
Wolf JB (2000) Gene interactions from maternal effects. Evo-lution 54:1882–1898
Wolf JB, Vaughn TT, Pletscher LS, Cheverud JM (2002) Con-tribution of maternal effect QTL to genetic architecture ofearly growth in mice. Heredity 89:300–310
Xu S (1998) Mapping quantitative trait loci using multiple fam-ilies of line crosses. Genetics 148:517–524
Zou F, Yandell BS, Fine JP (2001) Statistical issues in theanalysis of quantitative traits in combined crosses. Genetics158:1339–1346