Quantitative Trait Loci for Maternal Performance for Offspring Survival in Mice

Corresponding author: Andréa C. Peripato, Instituto de Biociências, Universidade de São Paulo, Rua do Matão 277 sala 300, São Paulo, SP 05508-900 Brazil., peripato{at}ib.usp.br (E-mail)

Communicating editor: G. A. CHURCHILL

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Maternal performance refers to the effect that the environment provided by mothers has on their offspring's phenotypes, such as offspring survival and growth. Variations in maternal behavior and physiology are responsible for variations in maternal performance, which in turn affects offspring survival. In our study we found females that failed to nurture their offspring and showed abnormal maternal behaviors. The genetic architecture of maternal performance for offspring survival was investigated in 241 females of an F₂ intercross of the SM/J and LG/J inbred mouse strains. Using interval-mapping methods we found two quantitative trait loci (QTL) affecting maternal performance at D2Mit17 + 6 cM and D7Mit21 + 2 cM on chromosomes 2 and 7, respectively. In a two-way genome-wide epistasis scan we found 15 epistatic interactions involving 23 QTL distributed across all chromosomes except 12, 16, and 17. These loci form several small sets of interacting QTL, suggesting a complex set of mechanisms operating to determine maternal performance for offspring survival. Taken all together and correcting for the large number of significant factors, QTL and their interactions explain almost 35% of the phenotypic variation for maternal performance for offspring survival in this cross. This study allowed the identification of many possible candidate genes, as well as the relative size of gene effects and patterns of gene action affecting maternal performance in mice. Detailed behavior observation of mothers from later generations suggests that offspring survival in the first week is related to maternal success in building nests, grooming their pups, providing milk, and/or manifesting aggressive behavior against intruders.

MATERNAL performance is one of the major components of fitness (FALCONER and MACKAY 1996 ) because reproductive success is a consequence not only of the fecundity of the parents, but also of the survival of the offspring. Indeed, environmental effects determined by the mother are more important than any other single factor in determining variation in early offspring size, growth, and survival (LEE et al. 1991 ). Variations in the environment provided by individual mothers for their offspring are due, in part, to genetic differences among mothers. The effect that a mother has on her offspring's phenotypes, independent of the genes she has transmitted, is referred to as her maternal performance and is measured in terms of the effect she has on her offspring's characteristics (CHEVERUD and MOORE 1994 ). Variations in maternal performance are likely to be due to variations in maternal physiology and behaviors, such as nest building, grooming, and milk production.

Despite recent interest in maternal effects (e.g., see MOUSSEAU and FOX 1998 ; WOLF et al. 1998 ), maternal performance often has been overlooked in the evolutionary genetic literature. This has occurred because of the difficulty of collecting critical data and because many researchers considered this trait as environmental with regard to the offspring phenotypes. While this is true, variation among mothers in their performance can be due to both genetic and environmental effects operating on the mother. Even so, as a fitness component, maternal performance is expected to have a relatively low additive genetic variance and low heritability (FALCONER and MACKAY 1996 ). Despite this expectation, some maternal performance constituents, such as nest-building behavior (h² = 0.27) in mice (BULT and LYNCH 2000 ) and egg volume (h² = 0.55) in passerine birds (POTTI 1999 ), have moderate heritability, indicating a substantial genetic basis. A survey of the agricultural literature indicates that maternal performance for offspring growth in mammals has a moderate heritability (CHEVERUD 1984 ). Considering the importance of maternal performance, if only one or a few genes were responsible for these traits, most of the genetic variation would have been eliminated by selection and drift (BULT and LYNCH 2000 ), which suggests that maternal performance is a complex quantitative trait.

Observation of certain maternal behaviors allows us to investigate the causes of variation in maternal performance for offspring survival. Many different factors affecting maternal behavior have already been identified. During pregnancy, females have extensive hormonal alterations that enhance neural activity and contribute to changes in maternal behavior (KINSLEY et al. 1999 ). In rodents, hormonal changes during pregnancy and parturition are an additional factor that contributes to the rapid appearance of maternal behavior (ROSENBLATT 1967 ). Oxytocin has different roles in rats and mice, but it is essential for maternal performance in both, as is estradiol (PEDERSEN et al. 1982 ; NISHIMORI et al. 1996 ; YAMAMURO and SENSUI 1998 ). Prolactin is required for normal reproduction and mammary gland development in mice (HORSEMAN et al. 1997 ; ALSTON-MILLS et al. 1999 ), and the prolactin receptor (PRLR) is a regulator of maternal behavior (LUCAS et al. 1998 ). The neurochemical mechanisms involved in maternal behavior have been described as involving interactions among critical areas in the brain such as the hippocampus, the media preoptical area, and several areas of the hypothalamus (FLANNELLY et al. 1986 ; BERNARDIS and BELLINGER 1996 ; LONSTEIN and STERN 1997 ; OLAZABAL and FERREIRA 1997 ; NUMAN et al. 1998 ).

Individual gene effects on behaviors affecting maternal performance have been studied through the knockout of various genes. Several genes involved in maternal care have been identified in this way, all of which are associated with the central nervous system, and particularly with the hypothalamus (BROWN et al. 1996 ; THOMAS and PALMITER 1997 ; LEFEBVRE et al. 1998 ; LUCAS et al. 1998 ; LI et al. 1999 ). Although all knockout females (FosB, Dbh, Mest, PRLR, and Peg3 deficient) presented normal olfactory capacity, a prerequisite for maternal behavior in mice (GANDELMAN et al. 1971 ), they showed various abnormal behaviors associated with these genes. Lack of pup retrieval and crouching over the nest were the only behaviors in common among these knockouts. Mest- and Peg3-deficient mice performed poorly in nest building (LEFEBVRE et al. 1998 ; LI et al. 1999 ), a problem that Dbh-deficient females did not exhibit (THOMAS and PALMITER 1997 ). FosB- and Dbh-deficient females had lactation problems (BROWN et al. 1996 ; THOMAS and PALMITER 1997 ), and the lack of placentophagia was reported in Mest- and Dbh-deficient females (THOMAS and PALMITER 1997 ; LEFEBVRE et al. 1998 ).

Inbred mouse strains are often difficult to maintain because of reduced litter size and maternal failure to nurture offspring (SILVER 1995 ; FALCONER and MACKAY 1996 ). This failure can be due to environmental disturbance (especially diseases) or, in some cases, to a deleterious mutation becoming fixed in the strains during the inbreeding process (FESTING 1979 ). During the early stages of inbreeding, many of the animals may be infertile or in poor condition due to the occurrence of deleterious recessive alleles in homozygous states (SILVER 1995 ). This reduction in fitness is usually overcome in F₁ hybrids between different inbred lines. The hybrids often show heterosis and better maternal performance than that of their inbred parents. The recovery of fertility and effective maternal care in hybrids indicates a genetic basis for variations in maternal performance among strains.

While we were establishing recombinant inbred and random bred mouse strains from an intercross of the Large (LG/J) and Small (SM/J) inbred mouse strains (CHEVERUD et al. 1999 ) for genetic mapping studies, we found females that failed to nurture their offspring and showed abnormal maternal performance. This aberrant behavior often led to loss of the entire litter. Although these behaviors occur in both parental strains, it is much less frequent in F₁ hybrids, indicating a genetic basis for the behavioral complex.

We investigate the genetic basis of maternal performance by studying the association between individual quantitative trait loci (QTL) and abnormal maternal performance for offspring survival in the cross of two inbred strains of mice. Maternal performance was determined on the basis of success or failure of the mother in maintaining at least one offspring alive in the first week after birth. F₂ females from an intercross between the LG/J and SM/J mouse strains were randomly mated to F₂ males and the success or failure of their litters was scored. This study allowed a QTL search on all 19 murine autosomes and the identification of the number and relative size of gene effects and patterns of gene action affecting maternal performance in this cross. Additionally, we examined maternal behavioral components in later generations of the same cross to characterize which maternal behaviors modulated maternal performance for offspring survival in the intercross population.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Mouse strains and breeding:
The mouse strains used in this study are the LG/J and the SM/J inbred lines. The history and details of animal husbandry are available in CHEVERUD et al. 1996 , KRAMER et al. 1998 , and VAUGHN et al. 1999 . We utilize these strains because their genomes have been characterized extensively and because they show heterosis in the cross for the traits we are investigating. Ten SM/J males were mated with 10 LG/J females, producing 41 hybrids autosomally identical to one another and heterozygous at each locus that differs between the parental strains. The F₁ hybrids were randomly mated, producing 510 F₂ progeny. Animals born were individually identified within the first week. Three weeks after birth they were weaned and placed in single-sex cages with at most five animals per cage. Animals were fed ad libitum with Purina PicoLab rodent chow 20 (5353; St. Louis). At 10 weeks 241 F₂ females were randomly mated with F₂ males to form F₃ progeny. Males were removed from the breeding cage when their mate was determined to be pregnant. Each female's litter size and litter survival through the first week was recorded. These intercrosses have continued over generations to produce recombinant inbred (RI) and advanced intercross (AI) lines. To produce these lines, two different systems of mating were used: brother-sister and random mating to produce, respectively, RIs and AIs. Furthermore, we continue to breed the parental strains LG/J and SM/J in our mouse facility.

Scoring maternal performance for offspring survival:
The first week of life is an important period for offspring survival. The major factor determining the life or death of the litter in this phase is maternal performance (LEE et al. 1991 ). In our cross we examined data from these females and scored whether or not they lost the whole litter in the first week after birth. Usually, in our cross, primiparous females with loss of the entire litter maintained this pattern in consecutive parities, so this abnormal maternal performance was not just random failure due to stress or inexperience. Maternal performance for offspring survival was scored as the ability of the female to maintain the litter alive through the first week of life. If only part of the litter was lost, the mother still was considered successful, whereas mothers that lost the entire litter were considered unsuccessful. We scored litter size and maternal performance on the basis of offspring survival in mothers from LG/J and SM/J inbred lines, 21 F₁ and 241 F₂ generation females. Offspring survival for these 241 F₂ females was scored as an all-or-none trait and used in a QTL study of maternal performance for offspring survival in this cross.

Molecular genotyping:
Total cellular DNA was extracted from liver using DNA QIAamp tissue kit (QIAGEN, Chatsworth, CA). PCR amplification of microsatellite loci was performed according to the protocol described by DIETRICH et al. 1992 and modified by ROUTMAN and CHEVERUD 1994 . PCR product was visualized using 5–6% agarose gels and ethidium bromide staining. Ninety-six polymorphic loci were scored in our study to cover all 19 autosomes as completely as possible (VAUGHN et al. 1999 ; CHEVERUD et al. 2001 ). The X chromosome was excluded from this study because of a relative lack of microsatellite marker variability (ROUTMAN and CHEVERUD 1995 ; CHEVERUD et al. 1996 ). The relative positions of these markers are given in the MOUSE GENOME DATABASE (2001). However, map distances are known to vary between crosses so for this study map distances were calculated using MAPMAKER 3.0b (LANDER et al. 1987 ; LINCOLN et al. 1992 ) as described in VAUGHN et al. 1999 and CHEVERUD et al. 2001 .

Statistical procedures: Interval mapping:
The presence of potential QTL and their relative positions were determined by interval mapping (LANDER and BOLSTEIN 1989 ) using multiple regression analysis (HALEY and KNOTT 1992 ). This analysis consists of regressing the phenotype (maternal performance, or MP) on genotype scores every 2 cM along each chromosome. Genotype scores are calculated using the probability that an individual is homozygous for either parent allele or heterozygous at the specified location. These probabilities are multiplied by -1, 0, and 1 for homozygous SM/J or heterozygous or homozygous LG/J, respectively, and summed to obtain an additive genotype score (X_a). Likewise, the dominance genotypic score (X_d) is calculated as the probability of heterozygosity at the arbitrary intermediate location. Genotypic scores are imputed using the genotypes measured at the flanking markers and the rates of recombination between the location of interest and the flanking markers (HALEY and KNOTT 1992 ). We used the model,

where µ is a constant, a is the additive genotypic value, X_a is the additive genotype score, d is the dominance genotypic value, X_d is the dominance genotype score, and e is the residual. These regression coefficients are estimates of the additive (a) and dominance (d) genotypic values if a QTL occurs at the tested position (FALCONER and MACKAY 1996 ). The probability of a gene affecting the character at that specific location was obtained using the SETCOR procedure in SYSTAT 8.0 (COHEN and COHEN 1983 ; COHEN and WILKINSON 1997 ).

We use parametric models here even though the dependent and independent variables—maternal performance for offspring survival and genotype score—are not in interval scale. HILTON 1976 argues that the significance tests obtained from this model are accurate as established by KORT 1973 . Other studies have confirmed the feasibility of using interval mapping for QTL studies of binary traits in intercrosses and backcrosses (VISSCHER et al. 1996 ; REBAI 1997 ; MCINTYRE et al. 2001 ) or even in multifamily half-sib designs (KARDAMIDEEN et al. 2000).

Significance levels:

Statistical significance of one-QTL models was evaluated using LOD scores. Because of multiple comparison problems (LANDER and KRUGLYAK 1995 ), we adjusted pointwise probabilities by calculating the number of statistically independent regressions performed on each chromosome and dividing the appropriate probability level by the effective number of independent tests performed (CHEVERUD 2001 ). The effective number of independent tests was calculated using the variance of the eigenvalues of the intermarker correlation matrix calculated separately for each chromosome. A Bonferroni-corrected significance threshold was then obtained by dividing the pointwise threshold (0.05) by the number of independent tests. We calculated two threshold levels to distinguish between chromosome-wide and genome-wide significance levels. The chromosome-wide significance threshold was obtained by dividing 0.05 by the number of independent tests calculated for each chromosome, while the genome-wide threshold was obtained by dividing 0.05 by the number of independent tests summed over all the chromosomes. This approach has been validated using simulation results (see CHEVERUD 2001 for details). A chromosome-wide significance value is very informative in that it strikes a balance between eliminating false-positive results and avoiding false negatives (RAO 1998 ; WELLER et al. 1998 ). The genome-wide threshold is an important value because LOD scores above the genome-wide threshold indicate highly significant evidence of linkage, while those exceeding only the appropriate chromosome-wide threshold suggest linkage, requiring confirmation from additional data (LANDER and KRUGLYAK 1995 ).

We also evaluated statistical significance of the single-locus QTL genome scan by running a permutation test (CHURCHILL and DOERGE 1994 ; DOERGE and CHURCHILL 1996 ) in QTL Cartographer (BASTEN et al. 1994 , BASTEN et al. 2002 ). Experiment-wise statistical significance is obtained by rearranging phenotypes to genotypes 1000 times. The single-locus QTL detected in these analyses were the same as we obtained using the HALEY and KNOTT 1992 model, and their likelihood estimates were also very similar. Thus the use of the HALEY and KNOTT 1992 model with a binary dependent variable does not compromise the analysis.

Epistasis:

The interaction among single-locus QTL detected by interval mapping was tested using the "physiological" epistasis model (CHEVERUD and ROUTMAN 1995 ; ROUTMAN and CHEVERUD 1997 ; CHEVERUD 2000 ) in SAS (SAS INSTITUTE 1998). In this model, epistatic genotypic values are defined and their contributions to additive, dominance, and interaction variances are all included in significance testing for epistasis. For this study, each pair of maternal performance QTL was investigated for epistatic interactions, where multiple regressions were performed at these loci using the additive and dominance genotype scores and their products as the independent variables. The forms of interaction, additive by additive (X_a1 x X_a2), additive by dominance (X_a1 x X_d2), dominance by additive (X_d1 x X_a2), and dominance by dominance (X_d1 x X_d2), were considered as independent variables and maternal performance as the dependent variable. Statistical significance was tested as described in ROUTMAN and CHEVERUD 1997 and CHEVERUD 2000 .

We used an interchromosomal two-way genome-wide scan performed at every 2 cM along the mouse chromosomes to test for epistasis across the whole genome (CHEVERUD 2000 ). At each pair of locations, a test for genic epistasis was performed. We used a generalization of multiple regression procedures that allows us to regress the dependent variable(s) on the four interaction components (X_a1 x X_a2, X_a1 x X_d2, X_d1 x X_a2, X_d1 x X_d2) holding the main effects ("a" and "d" for each of the two loci) constant (COHEN and COHEN 1983 ). This was done to obtain a single, joint significance test for the interaction terms in the model independent of their direct effects. The model is

where MP is the dependent variable and µ is the constant. The independent variables are the interaction terms X_a1X_a2, X_a1X_d2, X_d1X_a2, and X_d1X_d2, while the independent partial variables are the genotype scores X_a1, X_d1, X_a2, and X_d2 at the specified location. The aa, ad, da, and dd regression coefficients measure additive-by-additive, additive-by-dominance, dominance-by-additive, and dominance-by-dominance genotypic values for epistasis, respectively. The "|" indicates that the independent variables listed to the right are partialed out of the independent variables to the left. This provides a joint test for the interaction terms independent of tests for the single-locus scores. The probability obtained and parameters estimated are the same as in a standard multiple regression (COHEN and COHEN 1983 ). We tested the two-locus interactions separately from the single-locus effects because we had already done a single-locus scan, in this case finding only two single-locus QTL.

The number of independent tests in the two-way genome-wide scan was estimated by summing the products of the numbers of independent tests for each pair of chromosomes over all chromosome pairs (CHEVERUD 2000 ). This is likely to be an overestimate of the true number of independent tests because there will be complex correlations between tests involving different chromosome pairs that share a chromosome. This makes our Bonferroni correction for epistasis conservative. We calculated 2736 independent tests in the two-way genome-wide scan. Given this number of tests, we expect 136 false-positive significant results at the 0.05 pointwise level even in the absence of epistasis. To control for "false-positive" results, we used a significance threshold correction based on the Bonferroni test (CHEVERUD 2001 ). A significant epistatic interaction in this study was considered when epistasis at the pairs of positions reached the Bonferroni threshold level of 0.1, which in our case corresponds to the probability value of . We also considered epistasis significant if one of the four modes of epistasis (additive by additive, additive by dominance, dominance by additive, and dominance by dominance) was significant at the 9.3 x 10^-6 level (3.7 x 10^-5/4), even if the overall epistasis model was not significant.

A randomization procedure was performed to corroborate the significance values obtained for epistatic interactions. This randomization consisted of creating 1000 data sets by sampling without replacement from the original maternal performance data and keeping the genotypes unchanged. Because of the extremely large number of interactions in a two-way genome-wide scan, the randomization test among all chromosomes was prohibitive; therefore we analyzed the interactions on a subsample of the two-way tests, those involving chromosomes 6 and 10. These chromosomes were chosen because they are of average size and had one significant epistatic interaction detected using the procedures described above. Every possible interaction between locations on these two chromosomes was investigated for each of the 1000 randomly created data sets using the model previously described. When a significant interaction was detected, a full regression model was prepared and the probability of significant epistasis estimated. We expected 50 of the 1000 randomizations to reach significance by chance at a probability level corresponding to the 0.05 Bonferroni-corrected level. However, only 31 tests displayed interaction that surpassed this level (0.00276 for these two chromosomes) and none achieved the probability level obtained in the analysis of the original data. The observation of 31 false-positive results rather than the expected 50 indicates that our methods are conservative when evaluating a two-way genome-wide scan with a binary dependent variable. The probability of obtaining 31 false positives rather than the expected 50 is <0.006.

Behavior observation:
An ancillary behavioral observation study was undertaken to provide a better understanding of the specific behaviors related to maternal performance for offspring survival in the intercross population. We could not perform these observations in the F₂ generation since these females were no longer alive; hence we studied later generations of the same intercross that were being maintained in our mouse facility. The information obtained from these later generations suggests which maternal features would have been associated with offspring survival in the F₂ generation. Thus, we observed 199 pregnant females from matings performed to produce RI lines and an advanced intercross line (random matings) in the F₉–F₁₄ generations. We calculated inbreeding levels using pedigree data and the PEDSYS (Pedigree Data Management System) program (DYKE 1996 ) using the Quaas-Henderson algorithm (BOYCE 1983 ). The maternal features we investigated included nest building before and after delivery, placentophagia and pup grooming, presence of milk in the stomach of the pup (indicating presence of milk in the mother), aggressive behavior against intruders, and pup retrieval. Maternal behavioral monitoring began when pregnancy was detected and was completed 7 days after delivery. Observation was performed daily for 10–15 min per female.

The analyses of behavioral data were undertaken as follows: association among nominal variables was tested by cross-tabulation using the Pearson chi-square test and the phi coefficient, while association among scalar variables was examined using the general linear model procedures in SYSTAT 8.0. The phi coefficient is a standard measure of association for two-way tables corresponding to the Pearson product moment correlation between binary variables. Because reproductive success has been negatively associated with inbreeding, we first tested the association of inbreeding with behavioral variables using the following model:

(1)

A second model was used to test the association among variables and female maternal performance for offspring survival (Success):

(2)

where µ is the constant, N1 is the prepartum nest, N2 is the postpartum nest, G is placentophagia and grooming pups after birth, M is pup stomachs with milk, AB is aggressive behavior to intruders, F is the inbreeding coefficient, and e is the residual.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Maternal performance for offspring survival:
Data from litter size at birth and litter survival in the first week after birth are represented in Table 1. Data were collected from mothers across generations F₁ and F₂ and also from females from LG/J and SM/J inbred lines. To increase sample size of the inbred parental strains, these values were estimated from females currently maintained in our laboratory. The LG/J strain shows significantly lower rates of maternal success than those of the SM/J strain (40 vs. 74%, respectively). However, each parental strain has a lower success rate than that of the F₁ (100%) or F₂ (85%) hybrids.

Interval mapping:
The regression of maternal performance for offspring survival on their genotype scores in the 241 F₂ females at 96 microsatellite loci allowed a genome scan for QTL for this trait. We found highly significant linkage at the genome-wide level on chromosome 7 at D7Mit21 + 2 cM (Fig 1) using both Bonferroni-correction and randomization-based significance thresholds. This locus is underdominant and explains 6.2% of the phenotypic variance in maternal performance (Table 2). A suggestive linkage at the chromosome-wide level was found on chromosome 2 at D2Mit17 + 6 cM (Fig 2) using the Bonferroni-corrected significance threshold. This result exceeded the genome-wide 5% significance threshold as determined by randomization. As can be seen in Table 2, this locus is overdominant and accounts for 4.4% of the variance in maternal performance.

View larger version (20K): In this window In a new window Download PPT slide   Figure 1. LOD plot of chromosome 7. Significant LOD score (3.21) at genome-wide level (3.18) at 2 cM downstream of the marker D7Mit21 indicates highly significant evidence of QTL in this position. The chromosome-wide significance threshold for this chromosome is 2.12.

View larger version (20K): In this window In a new window Download PPT slide   Figure 2. LOD plot of chromosome 2. The chromosome-wide significance threshold is 2.02. LOD score at significant peak is 2.43, showing suggestive linkage at a chromosome-wide level at D2Mit17 + 6 cM.

Epistasis between markers on chromosome 2 and 7 was not significant for the overall model (P = 0.123), although there is borderline significance (P = 0.025) at the Bonferroni-corrected level for dominance-by-dominance epistasis (Fig 3).

View larger version (33K): In this window In a new window Download PPT slide   Figure 3. Maternal performance genotypic values for D2Mit17 + 6 cM and D7Mit21 + 2 cM.

Epistasis:
In the two-way genome-wide epistasis scan, a total of 346 epistasis tests are significant at the 0.05 pointwise level, 2.5 times the number expected by chance (136). Furthermore, 15 of these 346 exceed the Bonferroni-corrected significance criterion, 5 at 0.05 (1.8 x 10^-5) and an additional 10 at the 0.1 (3.7 x 10^-5) levels. Therefore, maternal performance differs among genotypes at one locus, depending on which genotypes are present at another locus. These pairs of loci are summarized in Table 3. Twenty-three chromosomal regions were involved in these 15 interactions (markers with overlapping confidence regions were conservatively considered to be a single locus). Of the two QTL identified in the single-locus analysis, D2Mit17 + 6 cM interacts significantly with other regions across the genome, but D7Mit21 + 2 cM does not, although a separate region on chromosome 7 (D7Nds1 + 0 cM) is involved in epistatic interactions. Chromosomes 12, 16, and 17 were the only ones that did not present significant interactions for epistatic QTL for maternal performance. Most loci participate in only one interaction although 7 of 23 loci participated in two or more interactions (Fig 4). These loci are not involved in a single unified network, but form several small separate sets of interacting loci.

View larger version (10K): In this window In a new window Download PPT slide   Figure 4. Epistatic interaction patterns for maternal performance. QTL are represented by respective chromosome number or by chromosome number and a letter when more than one QTL occurred in the same chromosome. QTL 1a refers to D1Mit3, 1b to D1Mit14, 2a to D2Mit1, 2b to both D2Mit38 and D2Mit17, 3a to D3Mit54, 3b to D3Mit194, 5a to D5Mit61, 5b to D5Mit26, 6a to D6Mit1, 6b to both D6Mit58 and D6Mit5, 11a to D11Mit15, 11b to D11Mit333, 15a to D15Mit5, and 15b to D15Mit2. For other chromosomes, check Table 2.

All four forms of epistasis are represented in the results. Since we found 25 significant epistasis coefficients among the 15 significant interactions and since we have four forms of epistasis, we would expect to find each form approximately 6 times. Our results did not deviate from this expectation. Additive-by-additive epistasis (e.g., Fig 5) occurred 6 times, additive-by-dominance and dominance-by-additive epistasis (e.g., Fig 6) appeared 11 times, and dominance-by-dominance epistasis (e.g., Fig 7) occurred 8 times (7 of them as negative dominance-by-dominance epistasis). A multiple regression model involving only the direct-effect loci on chromosomes 2 and 7 as independent variables has an adjusted multiple r² of 10%. A second multiple regression model containing all significant direct effects, the direct effects of loci involved in epistatic interactions, and significant epistatic effects accounts for 34.5% of the phenotypic variation after adjusting for the number of independent variables. Thus a fairly significant proportion of phenotypic variation in maternal performance is due to epistasis.

View larger version (44K): In this window In a new window Download PPT slide   Figure 5. Genotypic values for maternal performance indicating additive-by-additive interaction between D1Mit3 + 8 cM and D5Mit61 + 38 cM.

View larger version (29K): In this window In a new window Download PPT slide   Figure 6. Maternal performance genotypic values for the two-locus genotype at D8Mit25 + 22 cM and D15Mit2 + 20 cM illustrates additive-by-dominance interaction.

View larger version (29K): In this window In a new window Download PPT slide   Figure 7. Dominance-by-dominance interaction between D3Mit54 + 26 cM and D19Mit16 + 16 cM.

Behavior observation:
The frequency of maternal performance features observed in this study is presented in Fig 8. Observations from later generations indicate that pregnant female mice usually start to build a nest before delivery and maintain it postpartum. This behavior was found in most of the 144 successful females in our behavioral study. Although unsuccessful females (55) sometimes showed pre- and postpartum nest building, the nests built generally were of poor quality (data not shown). Successful and unsuccessful females differed significantly for these traits ( = 0.152 and P = 0.033 for prepartum and = 0.343 and P = 1.5 x 10^-7 for postpartum nest building). Placentophagia and cleaning pups was undertaken more frequently by successful rather than unsuccessful females, but the difference is of only borderline significance ( = 0.127, P = 0.0739). Presence of milk in the stomach of the pup is a major distinction between the two kinds of females ( = 0.886, P = 7.57 x 10^-36). Unsuccessful females fail to provide milk for their offspring. This failure could also be due to absence of suckling behavior in pups; however, unsuccessful females in most cases do not handle the pups. Consequently, this absence of milk is most likely due to the mothers' failure rather than to failed suckling behavior in their offspring. Aggressive behavior against intruders and pup retrieval was measured together because all females that performed aggressive behavior also retrieved their pups. Unsuccessful females, generally, did not respond to external stimulus and did not rescue pups removed from the nest ( = 0.439, P = 1.0 x 10^-9), in contrast to successful females that would usually attack our hands and, after carrying all pups back to the nest, immediately crouch over them.

View larger version (35K): In this window In a new window Download PPT slide   Figure 8. Frequency of maternal features significantly associated with maternal performance for offspring survival. The frequency given is for lack of the feature. "Nest 1" refers to lack of nest building before birth and "Nest 2" to nest building after birth; "Clean" indicates absence of placentophagia and pup grooming; "Milk" means failure in milk provision; and "AGB" refers to lack of aggressive behavior with retrieval of pups after disturbance. Successful and unsuccessful females differed significantly for these traits.

Inbreeding values for females used in the behavioral study are on average 0.561 ± 0.026. Association between inbreeding and success raising a litter was significant (F = 9.1728, P = 0.0028). When we examined the relationship between inbreeding and all the observed maternal behavior variables, we found a significant association among them (P = 0.048). Nevertheless, this significant result was due primarily to association between inbreeding and pup cleaning and milk provision (P = 0.034 and P = 0.040, respectively), but not to pre- and postpartum nest building and aggressive behavior against intruders.

When we tested for the association among different behavior variables and female maternal performance for offspring survival, a highly significant value was found (P = 5.85 x 10^-65). This analysis indicates that variables primarily responsible for this association are milk provision and aggressive behavior/pup retrieval (P = 1.0 x 10^-17 and P = 0.0004, respectively). Interestingly, the level of inbreeding is not significantly associated with success independent of the behavioral factors (P = 0.716).

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Interval mapping revealed two QTL affecting maternal performance. The genetic architecture of these QTL indicates a contrasting pattern. Underdominance at the chromosome 7 locus indicates that heterozygous females are, on average, less successful than either homozygote at this locus, while overdominance at the chromosome 2 QTL indicates that the heterozygotes display more successful maternal performance for offspring survival than do the parental genotypes. Although of only borderline significance after Bonferroni correction, maternal performance differences seem to be affected by interactions between these two loci (Fig 3). If the QTL at D7Mit21 + 2 cM is homozygous, the QTL at D2Mit17 + 6 cM is slightly additive; only when the former is heterozygous does the chromosome 2 locus present overdominance. Likewise, underdominance at the chromosome 7 locus is minimal in chromosome 2 heterozygotes but strong in the two homozygous genotypes.

The presence of significant QTL suggests a gene or genes associated with maternal performance for offspring survival at those chromosomal positions. It may be instructive to consider whether genes known to map to those regions include those likely, a priori, to have an effect on maternal performance. We searched the MOUSE GENOME DATABASE (2001) for candidate genes that affect maternal performance located close to these QTL. Although this identification is preliminary, it may suggest potential genes affecting the trait. The FBJ osteosarcoma oncogene B (FosB) and the paternally expressed gene 3 (Peg3) are appropriate candidate genes because they are in the proximal region of chromosome 7, 3.0–4.5 cM away from the position of the QTL at D7M21 + 2 cM. These candidate genes are associated with the lack of nurturing maternal behavior in knockout mice (BROWN et al. 1996 ; LI et al. 1999 ). Both FosB- and Peg3-deficient females fail to retrieve and crouch over their pups; the former also show poor nest-building performance and the latter have lactation problems. These phenotypes are similar to the abnormal maternal behaviors observed in our study. Naturally, it is possible that this chromosome 7 QTL could represent the combined effects of multiple, closely linked loci. Oxytocin (Oxt) is a candidate gene for the QTL at D2Mit17 + 6 cM since it is located in the confidence interval of the QTL at D2Mit17 (the intermediary region of chromosome 2) and has an essential role in stimulating maternal behavior. The release of oxytocin during delivery triggers maternal behavior in rats (PEDERSEN et al. 1982 ); oxytocin-deficient female mice failed to provide milk to the offspring (NISHIMORI et al. 1996 ), although they had normal maternal behaviors.

Many regions across the genome were found to participate in epistatic interactions for maternal performance for offspring survival. Only one of the QTL identified in the single gene mapping (D2Mit17 + 6 cM) participated in significant epistatic interactions. Although some loci participated in more than one interaction, the 23 chromosomal regions did not show a unified network of interactions. Rather, many separate interaction sets were identified (Fig 4). It is noteworthy to contrast these results with a study investigating QTL for adiposity in the same population, which found a single network connecting all epistatic QTL (CHEVERUD et al. 2001 ). Maternal performance is a complex trait and distinct mechanisms may be operating separately, affecting maternal behaviors, physiology, or both. These mechanisms combine in a chain of events that lead to the distinct maternal behaviors conducive to offspring survival. Some candidate genes for QTL showing epistatic interactions are shown in Table 4. All five gene knockouts linked to maternal behavior (BRIDGES 1998 ; LI et al. 1999 ) are potential candidate genes in our study. Two of them are candidates for single QTL and three of them are found near regions involved in epistatic interactions (Table 4). Other potential candidate genes for QTL are associated with abnormal maternal behavior and complex behavioral traits. Candidate genes suggested here do not cover all QTL found in our study. Furthermore, it is possible that other genes linked to the candidate genes we identified are actually responsible for the observed variation. At any rate, the use of the QTL approach enables the identification of regions in the genome affecting the trait of interest that may be investigated further.

All forms of epistasis are represented in the interactions found here (see Table 3). Additive-by-additive epistasis (e.g., Fig 5) is a pattern that can interfere with our ability to find single-locus QTL in a genetic mapping study. This "epistatic nullification" occurs because epistasis cancels out the effects of each single locus at the intermediate allele frequencies found in an F₂ population. It causes QTL analyses to underestimate the number of loci involved in complex traits (ROUTMAN and CHEVERUD 1997 ). In additive-by-additive epistasis, additivity is detected only within classes of genotypes at the interacting locus. In Fig 5, for example, if D5Mit61 + 38 cM is SS, D1Mit3 + 8 cM is additive with SS having the highest maternal performance. On the other hand, if D5Mit61 + 38 cM is LL, the pattern is reversed. D1Mit3 + 8 cM is still additive, but LL has the best maternal performance. When we average genotypic effects at one locus across all genotypes at the second locus, the effects nullify each other and may have led to our failure to detect single-locus effects on one or both of these chromosomes.

Additive-by-dominance epistasis shows the same pattern as dominance-by-additive epistasis, but with the roles of the loci reversed. Fig 6 represents additive-by-dominance epistasis between chromosome 8 and chromosome 15. Note that D8Mit25 + 2 cM has additive effects with the LL homozygote displaying superior maternal performance for both D15Mit2 + 20 cM homozygotes but that this effect is reversed in D15Mit2 + 20 cM heterozygotes. Likewise, D15Mit2 + 20 cM shows overdominance among D8Mit25 + 2 cM SS animals and underdominance among D8Mit25 + 2 cM LL animals.

Most observed cases of dominance-by-dominance epistasis are negative (Table 3). An example of this kind of interaction (Fig 7) shows that double heterozygotes have a lower genotypic value than that of single heterozygotes for each locus. Despite this, double heterozygotes still have better maternal performance than that of parental or recombinant homozygotes. This heterosis may explain why no unsuccessful females were observed in the F₁ generation. While investigating body weight in this same population we found a similar pattern in which the majority of the dominance-by-dominance epistasis was negative (CHEVERUD 2000 ). However, our findings do not predict hybrid dysgenesis because double heterozygotes were better than the parentals, although worse than single heterozygotes at either locus. A separate study on maternal effects on early growth in this same population found five single-locus QTL and 10 interactions for maternal performance for offspring growth (WOLF et al. 2002 ). Seven of these QTL were found in similar positions to those described here. Maternal performance for offspring survival and offspring growth may be directly related. The same maternal features that affect offspring survival may also affect offspring growth. Therefore we expect that some genes may modulate both traits.

The results of this genome-wide scan for maternal performance contrast strikingly with those obtained for a wide variety of morphological traits, including mandibular morphology (CHEVERUD et al. 1997 ), cranial morphology (LEAMY et al. 1999 ), growth and age-specific body weights (CHEVERUD et al. 1996 ; VAUGHN et al. 1999 ), adiposity (CHEVERUD et al. 2001 ), and binary skeletal nonmetric traits (LEAMY et al. 1998 ). Each of these genome scans found large numbers of main-effect QTL spread throughout the genome. Most QTL discovered had significant additive effects. For maternal performance we found only two direct-effect QTL and neither had a significant additive effect. The only comparable result has been for fluctuating asymmetry for mandibular morphology, which had no direct genetic effects (LEAMY et al. 1997 ) but many significant epistatic interactions (LEAMY et al. 2002 ).

Maternal performance was inferred from the analysis of litter survival. To corroborate this inference, we must show that females with low maternal performance present specific maternal behaviors that affect their success in rearing litters. Behaviors significantly associated with maternal performance include suckling, nest building, placentophagia and pup grooming, and retrieval of pups after disturbance. In our study, females whose pups survived the first week built a good nest before and kept it after delivery. Such females usually performed placentophagia, groomed pups, provided milk, and protected their offspring against intruders. Significant differences found between successful and unsuccessful females for these variables point out the lack of these maternal behaviors in the latter. Mice with knockout genes involved in maternal care displayed similar abnormal behaviors (BROWN et al. 1996 ; THOMAS and PALMITER 1997 ; LEFEBVRE et al. 1998 ; LUCAS et al. 1998 ; LI et al. 1999 ).

Inbreeding is also involved with this behavioral alteration, since maternal failure to nurture offspring is one of the causes of inbred mouse strain failure (SILVER 1995 ; FALCONER and MACKAY 1996 ). In our study, inbreeding was negatively associated with female success (P = 0.0028) through its effects on some of the maternal behaviors we studied, such as pup cleaning and milk provision. It had no significant effect independent of these behaviors.

Even though the behaviors were observed in later generations, the strong association between these variables and maternal performance suggests that unsuccessful females in the F₂ generation shared the same behaviors found in later generations; i.e., they did not display maternal care after parturition and had problems with lactation.

Maternal performance, as a fitness component, is under stringent selection pressure. Consequently, it is expected to have low additive genetic variation (FALCONER and MACKAY 1996 ). However, high levels of total genetic variation have been associated with fitness traits (HOULE 1992 ; FOWLER et al. 1997 ; POTTI 1999 ; BULT and LYNCH 2000 ) and there is selection pressure to supply genetic variation for fitness traits, so most of this variation should be due to epistatic and dominance interactions. Thus, epistasis plays an important role in variation for fitness traits (ARMBRUSTER et al. 1997 ). Our results indicate a few individual dominant gene effects that account for 10% of the variation in maternal performance, but a large number of epistatic interactions. Taken together, all QTL and interactions explain almost 35% of the phenotypic variation for this trait.

This study revealed QTL affecting maternal performance in mice of the F₂ generation of the SM/J and LG/J cross. To ratify the present results, future QTL studies for maternal performance should include fine-scale mapping and the investigation of RI lines. We expect that if these QTL indeed affect maternal performance, they should show different frequencies in successful and unsuccessful RI strains.

Maternal performance intrigues researchers due to its complexity and its role as an environmental influence on the phenotypes of relatives. Identification of individual genes has been mostly restricted to the use of the knockout gene technology. Nevertheless, multiple genes affect maternal behaviors and QTL analysis provides an appropriate framework not only to determine chromosomal localization of putative candidate genes, with a posteriori gene identification, but more importantly, to investigate how these regions interact to produce success or failure in maternal performance.

	ACKNOWLEDGMENTS

We thank T. Ehrich, S. Cropp, J. Wolf, A. Moore, and C. Boake for comments. This work is supported by National Science Foundation grant DEB-9726433, National Institutes of Health grant DK52514, and Fundação de Amparo à Pesquisa do Estado de São Paulo grant 98/16139-6 to A.C.P. and S.R.M.

Manuscript received November 2, 2001; Accepted for publication August 2, 2002.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

ALSTON-MILLS, B., A. C. PARKER, E. J. EISEN, R. WILSON, and S. FLETCHER, 1999  Factors influencing maternal behavior in the hubb/hubb mutant mouse. Physiol. Behav. 68:3-8.[Medline]

ARMBRUSTER, P., W. E. BRADSHAW, and C. M. HOLZAPFEL, 1997  Evolution of the genetic architecture underlying fitness in the pitcher-plant mosquito, Wyeomyia smithii.. Evolution 51:451-458.

BASTEN, C. J., B. S. WEIR and Z-B. ZENG, 1994 Zmap—a QTL cartographer, pp. 65–66 in Proceedings of the 5th World Congress on Genetics Applied to Livestock Production: Computing Strategies and Software, Vol. 22, edited by C. SMITH, J. S. GAVORA, B. BENKEL, J. CHESNAIS, W. FAIRFULL et al. Organizing Committee, 5th World Congress on Genetics Applied to Livestock Production, Guelph, Ontario, Canada.

BASTEN, C. J., B. S. WEIR and Z-B. ZENG, 2002 QTL Cartographer, Version 1.16. Department of Statistics, North Carolina State University, Raleigh, NC.

BERNARDIS, L. L. and L. L. BELLINGER, 1996  The lateral hypothalamic area revisited: ingestive behavior. Neurosci. Biobehav. Rev. 20:189-287.[Medline]

BOYCE, A. J., 1983  Computation of inbreeding and kinship coefficients on extended pedigrees. J. Hered. 74:400-404.[Abstract/Free Full Text]

BRIDGES, R. S., 1998  The genetics of motherhood. Nat. Genet. 20:108-109.[Medline]

BROWN, J. R., H. YE, R. T. BRONSON, P. DIKKES, and M. D. GREENBERG, 1996  A defect in nurturing in mice lacking the immediate early gene fosB.. Cell 86:297-309.[Medline]

BULT, A. and C. B. LYNCH, 2000  Breaking through artificial selection limits of an adaptive behavior in mice and the consequences for correlated responses. Behav. Genet. 30:193-206.[Medline]

CHEN, C., D. G. RAINNIE, R. W. GREENE, and S. TONEGAWA, 1994  Abnormal fear response and aggressive behavior in mutant mice deficient for alpha-calcium-calmodulin kinase II. Science 14:291-294.

CHEVERUD, J. M., 1984  Evolution by kin selection: a quantitative genetic model illustrated by maternal performance in mice. Evolution 38:766-777.

CHEVERUD, J. M., 2000 Detecting epistasis among quantitative trait loci, pp. 58–81 in Epistasis and the Evolutionary Process, edited by J. WOLF, E. BRODIE, III and M. WADE. Oxford University Press, New York.

CHEVERUD, J. M., 2001  A simple correction for multiple comparisons in interval mapping genome scans. Heredity 87:52-58.[Medline]

CHEVERUD, J. M., and A. J. MOORE, 1994 Quantitative genetics and the role of the environment provided by relatives in behavioral evolution, pp. 67–100 in Quantitative Genetic Studies of Behavioral Evolution, edited by C. R. B. BOAKE. The University of Chicago Press, Chicago.

CHEVERUD, J. M. and E. J. ROUTMAN, 1995  Epistasis and its contribution to genetic variance components. Genetics 139:1455-1461.[Abstract]

CHEVERUD, J. M., E. J. ROUTMAN, F. A. M. DUARTE, B. VAN SWINDEREN, and K. COTHRAN et al., 1996  Quantitative trait loci for murine growth. Genetics 142:1305-1319.[Abstract]

CHEVERUD, J. M., E. J. ROUTMAN, and D. K. IRSCHICK, 1997  Pleiotropic effects of individual gene loci on mandibular morphology. Evolution 51:2004-2014.

CHEVERUD, J. M., T. T. VAUGHN, L. S. PLETSCHER, K. J. KING-ELLISON, and C. ERICKSON et al., 1999  Epistasis and the evolution of additive genetic variance in populations that pass through a bottleneck. Evolution 53:1009-1018.

CHEVERUD, J. M., T. T. VAUGHN, L. S. PLETSCHER, A. C. PERIPATO, and E. S. ADAMS et al., 2001  Genetic architecture of adiposity in the cross of LG/J and SM/J inbred mice. Mamm. Genome 12:3-12.[Medline]

CHURCHILL, G. A. and R. W. DOERGE, 1994  Empirical threshold values for quantitative trait mapping. Genetics 138:963-971.[Abstract]

CLEMENT, Y., K. H. KIA, G. DAVAL, and D. VERGE, 1996  An autoradiographic study of serotonergic receptors in a murine genetic model of anxiety-related behaviors. Brain Res. 709:229-242.[Medline]

COHEN, J., and P. COHEN, 1983 Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences, Ed. 2. Lawrence Erlbaum, Hillsdale, NJ.

COHEN, J., and L. WILKINSON, 1997 Systat 8.0 Statistics, pp. 817–840. SPSS, Chicago.

DICKIE, M. M., 1969 lm. Mouse News Lett. 41: 30–31.

DIETRICH, W., H. KATZ, S. LINCOLN, H.-S. SHIN, and J. FRIEDMAN et al., 1992  A genetic map of the mouse suitable for typing intraspecific crosses. Genetics 131:423-447.[Abstract]

DOERGE, R. W. and G. A. CHURCHILL, 1996  Permutation tests for multiple loci affecting a quantitative character. Genetics 142:285-294.[Abstract]

DYKE, B., 1996 PEDSYS: A Pedigree Data Management System, Version 2.0. Southwest Foundation for Biomedical Research, San Antonio, TX.

EICHER, E. M. and W. G. BEAMER, 1976  Inherited ateliotic dwarfism in mice. Characteristics of the mutation, little, on chromosome 6. J. Hered. 67:87-91.[Free Full Text]

FALCONER, D. S., and T. MACKAY, 1996 Introduction to Quantitative Genetics. Longman Press, New York.

FESTING, M. F. W., 1979 Inbred Strains in Biomedical Research. Oxford University Press, New York.

FLANNELLY, K. J., E. D. KEMBLE, D. C. BLANCHARD, and R. J. BLANCHARD, 1986  Effects of septal-forebrain lesions on maternal aggression and maternal care. Behav. Neural. Biol. 45:17-30.[Medline]

FOWLER, K., C. SEMPLE, N. H. BARTON, and L. PARTRIDGE, 1997  Genetic variation for total fitness in Drosophila melanogaster.. Proc. R. Soc. Lond. Ser. B 264:191-199.[Medline]

FRITZ, J. D., L. D. JAYANTHI, M. A. THORESON, and R. D. BLAKELY, 1998  Cloning and chromosomal mapping of the murine norepinephrine transporter. J. Neurochem. 70:2241-2251.[Medline]

GANDELMAN, R., M. X. ZARROW, V. H. DENENBERG, and M. MYERS, 1971  Olfactory bulb removal eliminates maternal behavior in the mouse. Science 171:210-211.[Abstract/Free Full Text]

HALEY, C. S. and S. A. KNOTT, 1992  A simple regression method for mapping quantitative trait loci in line crosses using flanking markers. Heredity 69:315-324.[Medline]

HILTON, G., 1976 Intermediate Politometrics. Columbia University Press, New York.

HORSEMAN, N. D., W. ZHAO, E. MONTECINO-RODRIGUEZ, M. TANAKA, and K. NAKASHIMA et al., 1997  Defective mammopoiesis, but normal hematopoiesis, in mice with a targeted disruption of prolactin gene. EMBO J. 16:6926-6935.[Medline]

HOULE, D., 1992  Comparing evolvability and variability of quantitative traits. Genetics 130:195-204.[Abstract]

KADARMIDEEN, H. N., L. G. JANSS, and J. C. M. DEKKERS, 2000  Power of quantitative trait locus mapping for polygenic binary traits using generalized and regression interval mapping in multi-family half-sib designs. Genet. Res. 76:305-317.[Medline]

KINSLEY, C. H., L. MADONIA, G. W. GIFFORD, K. TURESKI, and G. R. GRIFFIN et al., 1999  Motherhood improves learning and memory. Nature 402:137-138.[Medline]

KORT, F., 1973  Regression analysis and discriminant analysis. Am. Polit. Sci. Rev. 67:555-559.

KRAMER, M., T. T. VAUGHN, L. S. PLETSCHER, K. KING-ELLISON, and E. ADAMS et al., 1998  Genetic variation for body weight gain and composition in the intercross of Large (LG/J) and Small (SM/J) inbred strains of mice. Genet. Mol. Biol. 21:211-218.

LABOSKY, P. A., G. E. WINNIER, T. L. JETTON, L. HARGETT, and A. K. RYAN et al., 1997  The winged helix gene, Mf3, is required for normal development of the diencephalon and midbrain, postnatal growth and the milk-ejection reflex. Development 124:1263-1274.[Abstract]

LANDER, E. S. and D. BOLSTEIN, 1989  Mapping Mendelian factors underlying quantitative traits using RFLP linkage maps. Genetics 121:185-199.[Abstract/Free Full Text]

LANDER, E. S. and L. KRUGLYAK, 1995  Genetic dissection of complex traits: guidelines for interpreting and reporting linkage results. Nat. Genet. 11:241-247.[Medline]

LANDER, E. S., P. GREEN, J. ABRAHAMSON, A. BARLOW, and M. DALEY et al., 1987  MAPMAKER: an interactive computer package for constructing primary genetic linkage maps of experimental and natural populations. Genomics 1:174-181.[Medline]

LEAMY, L. J., E. J. ROUTMAN, and J. M. CHEVERUD, 1997  A search for quantitative trait loci affecting asymmetry of mandibular characters in mice. Evolution 51:957-969.

LEAMY, L. J., E. J. ROUTMAN, and J. M. CHEVERUD, 1998  Quantitative trait loci for fluctuating asymmetry of quasi-continuous skeletal characters in mice. Heredity 80:509-518.

LEAMY, L., E. ROUTMAN, and J. CHEVERUD, 1999  Quantitative trait loci for early and late developing skull characters in mice: a test of the genetic independence model of morphological integration. Am. Nat. 153:201-214.

LEAMY, L., E. ROUTMAN, and J. CHEVERUD, 2002  An epistatic genetic basis for fluctuating asymmetry of mandible size in mice. Evolution 56:642-653.[Medline]

LEE, P. C., P. MAJLUF, and I. J. GORDON, 1991  Growth, weaning and maternal investment from a comparative perspective. J. Zool. 225:99-114.

LEFEBVRE, L., S. VIVILLE, S. C. BARTON, F. ISHINO, and E. B. KEVERNE et al., 1998  Abnormal maternal behavior and growth retardation associated with loss of the imprinted gene Mest.. Nat. Genet. 20:163-168.[Medline]

LEHMAN, A. L., Y. NAKATSU, A. CHING, R. T. BRONSON, and R. J. OAKEY et al., 1998  A very large protein with diverse functional motifs is deficient in rjs (runty, jerky, sterile) mice. Proc. Natl. Acad. Sci. USA 95:9436-9441.[Abstract/Free Full Text]

LI, L.-L., E. B. KEVERNE, S. A. APARICIO, F. ISHINO, and S. C. BARTON et al., 1999  Regulation of maternal behavior and offspring growth by paternally expressed Peg3.. Science 284:330-333.[Abstract/Free Full Text]

LINCOLN, S., M. DALY and E. LANDER, 1992 Constructing Genetic Maps with MAPMAKER/EXP 3.0, Ed. 3. Whitehead Institute Technical Report, Whitehead Institute, Cambridge, MA.

LONSTEIN, J. S. and J. M. STERN, 1997  Role of the midbrains periaqueductal gray in maternal nurturance and aggression: c-fos and electrolytic lesion studies in lactating rats. J. Neurosci. 17:3364-3378.[Abstract/Free Full Text]

LUCAS, B. K., C. J. ORMANDY, N. BINART, R. S. BRIDGES, and P. A. KELLY, 1998  Null mutation of prolactin receptor gene produces a defect in maternal behavior. Endocrinology 139:4102-4107.[Abstract/Free Full Text]

MCINTYRE, L. M., C. J. COFFMAN, and R. W. DOERGE, 2001  Detection and location of a single binary trait locus in experimental populations. Genet. Res. 78:79-92.[Medline]

MOUSE GENOME DATABASE, 2001 Mouse genome informatics (http://www.informatics.jax.org/). The Jackson Laboratory, Bar Harbor, ME.

MOUSSEAU, T. A., and C. W. FOX, 1998 Maternal Effects as Adaptations. Oxford University Press, New York.

NISHIMORI, K., L. J. YOUNG, Q. GUO, W. ZUOXIN, and T. R. INSEL et al., 1996  Oxytocin is required for nursing but is not essential for parturition or reproductive behavior. Proc. Natl. Acad. Sci. USA 93:11699-11704.[Abstract/Free Full Text]

NUMAN, M., M. J. NUMAN, S. R. MARZELLA, and A. PALUMBO, 1998  Expression of c-fos, fos B, and egr-1 in the medial preoptic area and bed nucleus of the stria terminalis during maternal behavior in rats. Brain Res. 792:348-352.[Medline]

OLAZABAL, D. E. and A. FERREIRA, 1997  Maternal behavior in rats with kainic acid-induced lesions of the hypothalamic paraventricular nucleus. Physiol. Behav. 61:779-784.[Medline]

PEDERSEN, C. A., J. A. ASCHER, Y. L. MONROE, and A. J. PRANGE, JR., 1982  Oxytocin induces maternal behavior in virgin female rats. Science 216:648-650.[Abstract/Free Full Text]

PEDERSEN, C. A., J. D. CADWELL, M. MCGUIRE, and D. L. EVANS, 1991  Corticoprotein-releasing hormone inhibits maternal behavior and induces pup-killing. Life Sci. 49:1537-1546.

POTTI, J., 1999  Maternal effects and the pervasive impact of nestling history on egg size in a passerine bird. Evolution 53:279-285.

RAO, D. C., 1998  CAT scans, PET scans, and genomic scans. Genet. Epidemiol. 15:1-18.[Medline]

REBAI, A., 1997  Comparison of methods for regression interval mapping in QTL analysis with non-normal traits. Genet. Res. 69:69-74.

ROSENBLATT, J. S., 1967  Nonhormonal basis of maternal behavior in the rat. Science 156:1512-1514.[Abstract/Free Full Text]

ROUTMAN, E. J. and J. M. CHEVERUD, 1994  A rapid method of scoring simple sequence repeat polymorphisms with agarose gel electrophoresis. Mamm. Genome 5:187-188.[Medline]

ROUTMAN, E. J. and J. M. CHEVERUD, 1995  Polymorphism for PCR-analyzed microsatellites: data for two additional inbred mouse strains and the utility of agarose gel electrophoresis. Mamm. Genome 6:401-404.[Medline]

ROUTMAN, E. J. and J. M. CHEVERUD, 1997  Gene effects on a quantitative trait: two-locus epistatic effects measured at microsatellite markers and at estimated QTL. Evolution 51:1654-1662.

SAS INSTITUTE, 1998 SAS/STAT User's Guide, Version 6, Vol. 1, Ed. 4. SAS Institute, Cary, NC.

SILVER, L. M., 1995 Mouse Genetics. Oxford University Press, New York.

THOMAS, S. A. and R. D. PALMITER, 1997  Impaired maternal behavior in mice lacking norepinephrine and epinephrine. Cell 91:583-592.[Medline]

VAUGHN, T. T., L. S. PLETSCHER, A. PERIPATO, K. KING-ELLISON, and E. ADAMS et al., 1999  Mapping quantitative trait loci for murine growth: a closer look at genetic architecture. Genet. Res. 74:313-322.[Medline]

VISSCHER, P. M., C. S. HALEY, and S. A. KNOTT, 1996  Mapping QTLs for binary traits in backcross and F2 populations. Genet. Res. 68:55-63.

WELLER, J. I., J. Z. SONG, D. W. HEYEN, H. A. LEWIN, and M. RON, 1998  A new approach to the problem of multiple comparison in the genetic dissection of complex traits. Genetics 150:1699-1706.[Abstract/Free Full Text]

WOLF, J. B., E. D. BRODIE, III, J. M. CHEVERUD, A. J. MOORE, and M. J. WADE, 1998  Evolutionary consequences of indirect genetic effects. Trends Ecol. Evol. 13:64-69.

WOLF, J. B., T. T. VAUGHN, L. S. PLETSCHER, and J. M. CHEVERUD, 2002  Contribution of maternal effect QTL to genetic architecture of early growth in mice. Heredity 89:300-310.[Medline]

YAMAMURO, Y. and N. SENSUI, 1998  Exogenous oxytocin attenuates suckling-induced prolactin release but not maternal or infant behavior in lactating rats. Physiol. Behav. 63:939-943.[Medline]

This article has been cited by other articles:

				S. D Becker and J. L Hurst Female behaviour plays a critical role in controlling murine pregnancy block Proc R Soc B, May 7, 2009; 276(1662): 1723 - 1729. [Abstract] [Full Text] [PDF]

				L. J. Leamy, D. Pomp, and J. T. Lightfoot An Epistatic Genetic Basis for Physical Activity Traits in Mice J. Hered., November 1, 2008; 99(6): 639 - 646. [Abstract] [Full Text] [PDF]

				B. Gutierrez-Gil, N. Ball, D. Burton, M. Haskell, J. L. Williams, and P. Wiener Identification of Quantitative Trait Loci Affecting Cattle Temperament J. Hered., November 1, 2008; 99(6): 629 - 638. [Abstract] [Full Text] [PDF]

				P. Hanlon, W. A. Lorenz, Z. Shao, J. M. Harper, A. T. Galecki, R. A. Miller, and D. T. Burke Three-locus and four-locus QTL interactions influence mouse insulin-like growth factor-I Physiol Genomics, September 14, 2006; 26(1): 46 - 54. [Abstract] [Full Text] [PDF]

				M. Liljander, M.-A. Sallstrom, S. Andersson, P. Wernhoff, A. Andersson, R. Holmdahl, and R. Mattsson Identification of Genetic Regions of Importance for Reproductive Performance in Female Mice Genetics, June 1, 2006; 173(2): 901 - 909. [Abstract] [Full Text] [PDF]

				J. P. Jarvis, J. Kenney-Hunt, T. H. Ehrich, L. S. Pletscher, C. F. Semenkovich, and J. M. Cheverud Maternal genotype affects adult offspring lipid, obesity, and diabetes phenotypes in LGXSM recombinant inbred strains J. Lipid Res., August 1, 2005; 46(8): 1692 - 1702. [Abstract] [Full Text] [PDF]

				L. Bunger, R. M Lewis, M. F Rothschild, A. Blasco, U. Renne, and G. Simm Relationships between quantitative and reproductive fitness traits in animals Phil Trans R Soc B, July 29, 2005; 360(1459): 1489 - 1502. [Abstract] [Full Text] [PDF]

				S. C. Gammie Current Models and Future Directions for Understanding the Neural Circuitries of Maternal Behaviors in Rodents Behav Cogn Neurosci Rev, June 1, 2005; 4(2): 119 - 135. [Abstract] [PDF]

				S. J. Garlow, E. Boone, W. Li, M. J. Owens, and C. B. Nemeroff Genetic Analysis of the Hypothalamic Corticotropin-Releasing Factor System Endocrinology, May 1, 2005; 146(5): 2362 - 2368. [Abstract] [Full Text] [PDF]

				Y. Cui, G. Casella, and R. Wu Mapping Quantitative Trait Loci Interactions From the Maternal and Offspring Genomes Genetics, June 1, 2004; 167(2): 1017 - 1026. [Abstract] [Full Text] [PDF]

				D. Zhong, A. Pai, and G. Yan Quantitative Trait Loci for Susceptibility to Tapeworm Infection in the Red Flour Beetle Genetics, November 1, 2003; 165(3): 1307 - 1315. [Abstract] [Full Text] [PDF]

				J. M. Cheverud Evolution in a genetically heritable social environment PNAS, April 15, 2003; 100(8): 4357 - 4359. [Full Text] [PDF]