Logo of geneticsGeneticsCurrent IssueInformation for AuthorsEditorial BoardSubscribeSubmit a Manuscript
Genetics. 2007 Nov; 177(3): 1553–1567.
PMCID: PMC2147974

Extensive Additivity of Gene Expression Differentiates Subspecies of the House Mouse


We have studied different subspecies of the house mouse and their reciprocal F1 hybrids to estimate the within-locus mode of inheritance for subspecies differences in gene expression in three tissues (brain, liver, and testis) of male mice. This study investigates the mode of inheritance in crosses at a larger taxonomic distance than has been previously systematically investigated. We found the vast majority of transcripts to be additively expressed with only a few transcripts showing dominance or overdominance in expression, except for one direction of one cross, which showed large mis-expression in the testis. We suggest that, as time passes, more genes come to influence expression, and if there is no directional dominance, additivity becomes increasingly more likely, up to a threshold beyond which there is F1 hybrid breakdown. Some previous studies on different organisms have found a large degree of dominance, commonly at shorter taxonomic differences. We surveyed these findings and show that the most consistent association exists between the amount of additivity detected in a study and the expression analysis method (in particular microarray platform), suggesting that at least some of the differences among studies might be methodological. Most studies agree with ours in that within-locus additivity seems to be general mode of inheritance for transcript expression. Differentially expressed transcripts identified in our screen among subspecies of house mice are candidate genes that could be involved in reproductive isolation between these subspecies.

THE methods of quantitative genetics are now being applied to transcript abundance, as measured using microarrays. Expression quantitative trait locus (eQTL) mapping identifies the genetic basis of expression differences among organisms (Rockman and Kruglyak 2006). Such mapping studies are being used to ask how many loci contribute to a quantitative trait, what is the size of the effect of the loci, and how individual loci interact to generate a quantitative trait (Mackay 2001). Treating gene expression levels as measured by microarrays as quantitative traits enables the simultaneous assessment of thousands of traits in parallel. eQTL studies can potentially also be used to ask whether regulatory variants, especially those that are important in evolution, are located in cis (i.e., directly linked to the gene that shows the differential expression) or in trans (i.e., somewhere else in the genome) (Hoekstra and Coyne 2007; Wray 2007).

Two approaches have been used to study the genetics of expression traits. In the first, DNA sequence polymorphisms are correlated with expression differences (Brem et al. 2002; Schadt et al. 2003; Doss et al. 2005; Storey et al. 2005). A key finding from these studies is that expression traits are often affected by multiple underlying loci and interactions among them.

The second approach makes use of F1 hybrids generated from a cross between divergent taxa to ask if a transcript's expression is intermediate (“additive”) to that of the two parents. These studies have produced varying results. Gibson et al. (2004) crossed two lines of Drosophila melanogaster and found most transcripts to be “nonadditively” (i.e., the hybrids showed expression levels that were closer to either one of the parents) expressed in F1 hybrids. Widespread nonadditivity was also identified in the Pacific oyster (Hedgecock et al. 2007) and, to a lesser extent, in Arabidopsis (Vuylsteke et al. 2005). However, later studies on Drosophila (Hughes et al. 2006), maize (Stupar and Springer 2006; Swanson-Wagner et al. 2006), and laboratory strains of the house mouse (Cui et al. 2006) were consistent with mostly additive expression. Hughes et al. (2006) suggested that the discrepancy between the different Drosophila studies might be explained in two ways. First, inbreeding might affect the results. Much within-locus additivity was observed using natural populations of D. melanogaster (Hughes et al. 2006) while nonadditivity was observed with strongly inbred lines (Gibson et al. 2004). It is well known that crosses between inbred strains produce offspring that exhibit greater biomass, speed of development, and fertility than both parents (heterosis) (Comings and MacMurray 2000). Heterosis is attributed to overdominance (superiority of heterozygotes at genes affecting fitness) or dominance (masking of recessive deleterious mutations) and plausibly could result in overdominant or dominant expression at the transcript level. However, the inbreeding explanation seems to be an unlikely general explanation because additional recent studies have found strong evidence for additivity despite the use of inbred parental lines (Cui et al. 2006; Stupar and Springer 2006; Swanson-Wagner et al. 2006). A second explanation for the Drosophila differences is that genetic architecture of an expression trait might depend critically on the taxonomic level at which the variation is investigated (i.e., between populations, as in the Gibson et al. 2004 study, as opposed to within populations, as in the Hughes et al. 2006 study). These authors assume that the more highly diverged the parents are, the greater the nonadditivity in gene expression that will be observed. However, we might expect additivity of expression to become more likely as taxa diverge and the trait becomes influenced by more and more genes, provided that they do not show directional dominance. This would apply up to some certain threshold beyond which incompatible genes lead to massive mis-expression in F1 hybrids.

Here we investigate patterns of expression in the house mouse subspecies complex of Mus musculus. We used parental mouse strains that were collected in the wild and inbred in the laboratory for several generations. The strains belong to the three different subspecies of house mice, M. m. musculus, M. m. domesticus, and M. m. castaneus, with estimated divergence times of between 300,000 and 1 million years (Boursot et al. 1993). Previous systematic studies focused on comparisons within a single species or used house mouse laboratory strains. In contrast, our study employs naturally derived lines that are rather highly diverged. We found predominantly additive gene expression differences among these subspecies.


Mouse strains:

Wild-derived strains of domesticus (STRA from Germany, Piálek et al. 2007), musculus (STUS from the Czech Republic), and castaneus (CIM from India) were used for the microarray experiment. The domesticus and musculus strains were provided by J. Pialek. These strains were collected in the wild and inbred by brother–sister matings for ∼13 generations in the laboratory of J. Pialek in the Department of Population Biology, Institute of Vertebrate Biology, Academy of Sciences of the Czech Republic. The castaneus strain was provided by A. Orth and F. Bonhomme. This strain has been kept in the Laboratory of Genome, Populations, Interactions, and Adaptation in Montpellier, France, for >30 generations in a closed colony. We set up reciprocal crosses between domesticus and musculus and between musculus and castaneus to obtain F1 hybrids. For each parental strain and each reciprocal cross, two male individuals were analyzed. Altogether, we performed 42 microarray experiments, 14 in each of three tissues (brain, liver, and testis). An overview of the samples used in this study is given in Table 1. All mice were raised under identical standard laboratory conditions and were sacrificed at the age of 6–8 weeks.

Overview of the samples hybridized separately on the microarray

RNA extraction:

We extracted RNA from three different tissues (brain, liver, and testis) using Trizol (Invitrogen, Carlsbad, CA) following the manufacturer's protocol. Quality and integrity of the total RNA was controlled by using the Agilent Technologies 2100 Bioanalyzer and the RNA 6000 Nano LabChip kit (Agilent Technologies, Waldbronn, Germany).


Expression profiles were determined for >39,000 mouse transcripts using the Mouse Genome 430 2.0 Affymetrix (Santa Clara, CA) GeneChip. For biotin-labeled target synthesis starting from 1 μg of total RNA we used standard protocols supplied by the manufacturer. After hybridization, the GeneChips were washed, stained, and read using an Affymetrix GeneChip fluidic station and scanner.

Data analysis:

All data processing and statistical analyses were performed using the statistical language R. Raw signal intensities were normalized and summarized according to the standard Affymetrix MA Suite 5.0 algorithm using the program Bioconductor (http://www.bioconductor.org/). Signal intensities were ln-transformed prior to statistical analyses. MA Suite 5.0 expression values have been submitted to the Gene Expression Omnibus (accession no. GSE9338). We define a “group” as one of domesticus, musculus, castaneus, F1 hybrids from one direction of the cross or F1 hybrids from the other direction of the cross. We called a transcript “expressed” in this group if the average expression level among the replicate samples within this group was >500. Applying a threshold of >500 yields ∼11,0000 transcripts expressed per group per tissue, a value that corresponds to previous studies (Su et al. 2004).

First, we studied all transcripts that are differentially expressed between the parental subspecies and assessed additivity vs. nonadditivity of expression of these transcripts in the F1 hybrids (see Figure 1, A and B). Second, we investigated those transcripts that do not differ in expression between the parental lines, but which are expressed differently in the F1 hybrids relative to both their parents (Figure 1C):

Figure 1.
Schematic of within-locus modes of gene action in the comparison between house mouse subspecies and their F1 hybrids. Sp, subspecies.

Additivity/nonadditivity of differentially expressed transcripts:

For each parental comparison (i.e., musculus vs. domesticus and musculus vs. castaneus), standard one-way ANOVA was performed for each tissue separately to identify differentially expressed transcripts. We applied the false discovery rate (FDR; Reiner et al. 2003) to the ANOVA P-values and selected only transcripts with an FDR <10%. Applying this stringent criterion to the musculusdomesticus comparison corresponds to a significance threshold of P < 0.00025 in the brain, P < 0.0000053 in the liver, and P < 0.002 in the testis. In the musculuscastaneus comparison, this corresponds to P < 0.002 in the testis; no transcript at all is significant in the brain and liver under a 10% false discovery rate (see results). However, preliminary experiments using quantitative real-time PCR of one gene differentially expressed between musculus and domesticus in the liver (Ces1) suggest that transcripts up to a P-value level of 0.06 on the microarray can be confirmed on independent samples (data not shown). Therefore, we repeated the selection of differentially expressed genes for each tissue using significance values arbitrarily set at P < 0.01 and P < 0.05.

Significant transcripts were considered separately according to whether they were expressed (microarray signal intensity >500) in both subspecies or in just one of the subspecies within a given comparison. Signal intensities <500 likely include transcripts that are not expressed at all. In this case, additivity of transcript expression is meaningless. However, if the transcript is absent in one parental subspecies but present in the other, one can at least ask whether or not the expression status in the hybrid is consistent with dominance. We used three methods to determine the mode of inheritance for the differentially expressed transcripts between the parental subspecies using the F1 hybrids between them. The musculusdomesticus comparison was processed in parallel with the musculuscastaneus comparison using identical methods. We define the average expression level across replicates of the subspecies with lower expression to be x, the average expression level of the subspecies with higher expression to be y, and the expression in the hybrid to be hij, where the subscript i indicates the cross (i.e., either subspecies a female with subspecies b male or subspecies a male with subspecies b female; i = 1, 2) and the subscript j indicates the replicate within each cross (j = 1, 2). First, following Gibson et al. (2004), we defined additivity as y/1.25 > h > x × 1.25, for all h (note that y, x, and h are measured on the untransformed scale). In this model, only transcripts with a difference between the parental subspecies of >2.5 can show additivity, and hence the analysis is restricted to only these transcripts. Dominance is defined as 1.25 × x > h > x/1.25 or 1.25 × y > h > y/1.25 for all h, and overdominance as h < x/1.25 or h > y × 1.25 for all h. We call this the “fixed threshold method.”

The problem with the fixed threshold method is that as the fold difference between two subspecies becomes greater, the zone where additivity will be accepted becomes larger. Accordingly, we devised a second measure, which we term the “fractional threshold method,” based on the relative difference in expression levels between the parental subspecies. We defined additivity to be (x + y)/2 − ɛ < h < (x + y)/2 + ɛ, for all h, where ɛ was set to 5, 10, 15, and 20% of the difference between x and y [e.g., ɛ = (yx) × 0.15]. Dominance was set to be x − ɛ/2 < h < x + ɛ/2 or y − ɛ/2 < h < y + ɛ/2 for all h, so that for each transcript the interval over which either dominance or additivity is accepted is equivalent. Overdominance was set to be outside of the range for dominance, i.e., h < x − ɛ/2 or h > y + ɛ/2 for all h. To be comparable to the analysis using a fixed threshold of 1.25, we included only genes where the expression difference (fold change) between the parents was >2.5-fold. For both methods we determined the mode of inheritance by (1) treating F1 hybrids from both directions of the cross together (i.e., all four replicates had to be consistent with a certain mode of inheritance) and (2) treating both directions of the cross separately to account for possible maternal effects, sex linkage, and directional dominance.

In addition to classifying transcripts into arbitrary mode-of-inheritance classes, we analyzed the distribution of dominance effects, which was calculated as d/a, where a is half the difference in expression level between the parental strains and d is the difference in expression level between the F1 hybrid (the average across replicates independent of the direction of cross) and the average of the parental strains (Falconer and Mackay 1996; Gibson et al. 2004). If the expression level of the transcript in the hybrid is exactly intermediate between the parental strains, d = 0. A d/a value of 0 corresponds to perfect within-locus additivity, |d/a| = 1 to complete dominance and |d/a| > 1 to overdominance. We performed this analysis separately for each tissue (brain, liver, and testis) and both comparisons (i.e., musculusdomesticus and musculuscastaneus) including all differentially expressed transcripts between parental strains (P < 0.01 and P < 0.05), with average expression levels >500 in at least one of the parental subspecies. Note that, in contrast to the analyses where transcripts are assigned to discrete additive and nonadditive classes, we do not require that the differentially expressed transcript shows a fold change >2.5 between the parental subspecies, thus allowing us to more generally assess dominance effects.

Nonadditive expression of nondifferentially expressed transcripts:

To assay parent–F1 hybrid offspring expression differences at the large number of transcripts not differentially expressed between the parental subspecies, we used pairwise Tukey post-hoc tests. We selected transcripts in which F1 hybrids have significant (P < 0.05) and 1.25-fold higher or lower expression compared to both parents. We performed this analysis for F1 hybrids from both directions of the cross together as well as from each direction of the cross separately. For those transcripts that had expression levels <500 in both the parental subspecies, we analyzed only those crosses in which the F1 hybrids from at least one direction of the cross had average expression levels >500.

Determination of subspecies origin of each probe set:

The probe sets of the Mouse Genome 430 2.0 Affymetrix GeneChip were designed on the basis of the sequence of the laboratory inbred strain C57BL/6J. The genome of this strain is a mixture of genetic contributions from all three subspecies—domesticus, musculus, and castaneus (Wade et al. 2002; Wade and Daly 2005; Yang et al. 2007). According to the most recent analyses (Yang et al. 2007), up to 92% of the genome of laboratory inbred strains (and thus also the genome of C57BL/6J) is derived from domesticus, ∼7% is from musculus, and <1% is from castaneus. Whenever DNA sequences show differences between these subspecies, we might expect probe sets to hybridize more efficiently to the subspecies from which the particular genomic region was derived. While sequence divergence between the subspecies is generally small [typically 1% for noncoding DNA at autosomal loci (Harr 2006)], it could potentially enhance or alleviate expression differences. This problem is likely to be more prevalent when other subspecies are compared to domesticus, the subspecies that contributed most of the genetic material to C57BL/6J. Thus, we specifically address this problem for the comparison between domesticus and musculus. We used >8 million single nucleotide polymorphism (SNP) loci distributed all over the genome and downloaded from the Perlegene website (http://mouse.perlegen.com/mouse/download.html). These SNPs have been typed in 15 commonly used inbred strains of the laboratory house mouse. The data from these strains can be combined with the known sequence of strain C57BL/6J (Waterston et al. 2002) at the respective SNP position. Of the 15 strains, 2 are so-called “wild-derived” strains, 1 of which belongs to musculus (PWD/PhJ) and 1 to domesticus (WSB/EiJ). We first used the SNPs that distinguished the musculus from the domesticus strain to assign the subspecies origin of each SNP in the C57BL/6J strain. We then calculated the frequency of musculus-like or domesticus-like SNPs in a 20-kb region surrounding the location of the transcript that is targeted by a specific probe set. A probe set was called musculus-like if >60% of the SNPs in the region matched the musculus strain and domesticus-like if >60% of the SNPs matched the domesticus strain. We then partitioned the data into two sets: transcripts that are located in “musculus-like” regions but show higher expression in domesticus and transcripts that are located in “domesticus-like” regions but show higher expression in musculus. For these transcripts, sequence divergence is an unlikely explanation of the expression differences. Due to the absence of a wild-derived castaneus strain in the Perlegen data set, we cannot assign regions in the genome of C57BL/6J that were contributed by this subspecies. Given that only a very small proportion of the genome derives from castaneus, this is probably not a very serious limitation.

Chromosomal location:

The chromosomal location (according to NCBI Build 36) of each additively expressed transcript [identified by the fractional (15%) method in F1 hybrids from both directions of the cross (P < 0.05, fold change between the parental subspecies >2.5)] was obtained from the annotation file (“Mouse430_2.na21.annot.csv”) downloaded from the Affymetrix website in January 2007 (http://www.affymetrix.com) and transformed into a MySQL database. We counted the number of additively expressed transcripts on each chromosome. If two alternative transcripts of the same gene were additively expressed, we counted that transcript only once. For each chromosome we also obtained the total number of known protein-coding genes from Ensembl (http://www.ensembl.org). To test for differential representation of additively expressed transcripts on any particular autosome, we first calculated the ratio of additive transcripts to the total number of protein-coding genes on this autosome. We then compared this to the ratio of the total number of additively expressed transcripts on all autosomes relative to the total number of protein-coding genes on all autosomes, using a Fisher's exact test. To test for an overrepresentation of additively expressed transcripts on the X chromosome, we counted the number of additive transcripts on the X chromosome relative to the total number of protein-coding genes on the X chromosome. This number is compared to the number of additive transcripts on all autosomes relative to the total number of protein-coding genes on all autosomes. All analyses were repeated for each tissue separately.

Clustering of additively expressed transcripts along a chromosome:

For each chromosome, we determined the chromosomal location of each additively expressed gene and sorted them according to chromosomal position (NCBI Build 36). Next we calculated the distance between each gene and its nearest downstream neighbor and expressed this distance relative to the total length of that particular chromosome. We then asked whether this distance is smaller than would be expected by chance. The expectation was derived by randomly drawing two values from a uniform distribution. The absolute distance between these two numbers was calculated and the lower 2.5 percentile was derived from repeating this procedure 100,000 times. We asked whether the observed distance was smaller than the 2.5 percentile from the simulation.

Functional annotation:

We used PANTHER (http://www.pantherdb.org/tools/genexAnalysis.jsp) to analyze the additively expressed transcripts in each tissue for overrepresentation (relative to the full gene content of the mouse genome) of certain biological functions. We performed this analysis for additively expressed transcripts identified by the fractional (15%) method in F1 hybrids from both directions of the cross (setting P < 0.05, >2.5-fold difference between the parental subspecies). In addition, we analyzed the transcripts differentially expressed between the parental subspecies independently of the mode of inheritance in F1 hybrids (again setting the same criteria of P < 0.05, fold change between parental subspecies >2.5).


About one-fourth of all the transcripts represented on the Affymetrix GeneChip have signal intensities >500 (i.e., are expressed) in at least one of the three tissues from at least one of the three subspecies. The number of expressed transcripts did not differ greatly among tissues (brain: 13,137 expressed transcripts; liver: 11,155 expressed transcripts; testis: 11,608 expressed transcripts). In Table 2 we show the pattern of differential transcript expression between musculus and domesticus and between musculus and castaneus, respectively. At a significance level of P < 0.05, we find roughly similar numbers of transcripts in each of the three organs. Given the ∼12,000 transcripts expressed in at least one subspecies and tissue, 600 should be assigned “significantly different” by chance at P < 0.05. Instead, we found roughly three times as many significant transcripts in each tissue. The number of transcripts differentially expressed between the subspecies drops considerably when only transcripts with a larger-than-twofold change are considered [i.e., musculusdomesticus: 445 (brain), 683 (liver), and 847 (testis); musculuscastaneus: 435 (brain), 691 (liver), and 827 (testis)]. However, even when parental expression differences are constrained to differ by fivefold, we find a relatively large number of significantly different transcripts for each tissue (between 96 and 218 transcripts, depending on tissue and subspecies comparison; Table 2).

Transcripts differentially expressed in an ANOVA analysis between the parental subspecies M. m. musculus and M. m. domesticus and between M. m. musculus and M. m. castaneus at different ANOVA P-value thresholds and at different magnitudes of change

If the significance is set to a more stringent level, a difference between the tissues becomes apparent (Figure 2). At the extreme, when the P-value level corresponds to a false discovery rate of 10%, we find in the musculusdomesticus comparison 9 transcripts differentially expressed in the liver, 1 transcript differentially expressed in the brain, and 218 transcripts differentially expressed in the testis. In the musculuscastaneus comparison, significant transcripts were found only in the testis (213 transcripts). It is clear that there is an excess of differential transcript expression in the testis compared to the other organs (Figure 2).

Figure 2.
Number of differentially expressed transcripts between M. m. musculus and M. m. domesticus and between M. m. musculus and M. m. castaneus in different tissues and at different P-value thresholds.

For both comparisons, we also determined the overlap of differentially expressed transcripts between subspecies among the three different tissues. The vast majority of transcripts are differentially expressed in just one tissue within a comparison (Figure 3A), even though about one-half of the differentially expressed transcripts in each tissue are detectably expressed in at least one additional tissue (i.e., a large proportion of differentially expressed genes are not tissue specific; data not shown). Between one-third and one-half (depending on the tissue) of the transcripts that are differentially expressed in the musculusdomesticus comparison are also differentially expressed in the musculuscastaneus comparison (Figure 3B), suggesting that the expression phenotype at these transcripts is specific to musculus.

Figure 3. Figure 3.
(A) Venn diagram showing the overlap of transcripts differentially expressed between parental subspecies (P < 0.01, fold change >2.5) among the different tissues. (B) Venn diagram showing the overlap of differentially expressed transcripts ...

Proportion of differentially expressed genes showing additivity, dominance, and overdominance:

As described in materials and methods, we used three different methods to assay the mode of inheritance for transcript expression, two of which assign transcripts into discrete classes (i.e., additivity, dominance, and overdominance): the fixed threshold method and the fractional threshold method. In Tables 3 and and4,4, we show the number of transcripts identified for each mode of inheritance according to these two methods, using an ɛ corresponding to 15% of the parental difference in expression for the fractional method.

Additively, dominantly, and overdominantly expressed transcripts among the differentially (fold change >2.5, ANOVA P < 0.01) expressed transcripts between M. m. musculus and M. m. domesticus in different tissues
Additively, dominantly, and overdominantly expressed transcripts among the differentially (fold change >2.5, ANOVA P < 0.01) expressed transcripts between M. m. musculus and M. m. castaneus in different tissues

We scored these numbers separately for the transcripts with signal intensities >500 in both parental strains and those with signal intensity >500 in only one of the two parents. For the latter group, we cannot unequivocally determine whether a transcript is additively expressed in the F1 hybrid because it may have no expression in one of the parental subspecies. For this reason, we place numbers for additively expressed transcripts in parentheses in Tables 3 and and44.

When compared to the fixed threshold method (Gibson et al. 2004), the fractional threshold method generally yields fewer transcripts that can be classified into one or the other mode of inheritance. Nevertheless, for both the musculusdomesticus and musculuscastaneus comparisons, the fraction of additive vs. nonadditive effects is almost identical for both methods and changes little with varying values of ɛ in the fractional method. In all cases, transcripts were classified into the same mode-of-inheritance categories by both methods.

The relative frequency of additive effects is generally independent of whether additivity was identified in F1 hybrids from only one direction of the cross or from both directions of the cross combined. The percentage of transcripts showing additivity was also similar across tissues. In all but one comparison, >80% of the transcripts are additively expressed, with ∼10% dominantly and 10% overdominantly expressed. The one exception to this pattern is in the musculus mother–castaneus father cross. In this cross, the testis (but not the other tissues) shows a much higher fraction of nonadditivity (∼50% of all the assigned transcripts).

Similar results come from the analysis of transcripts with expression levels >500 in only one of the parental subspecies. Dominance and overdominance among these transcripts are rare, and expression levels of most transcripts are consistent with additivity. The fact that we find such a large fraction of transcripts consistent with additivity suggests that those transcripts with expression levels <500 may not be absent, but rather are expressed at a low level.

The distribution of dominance values is shown in Figure 4. Unlike the previous tests that included only genes showing a >2.5-fold difference in expression between the parental subspecies, this test includes all transcripts that are differentially expressed (at the given P-value threshold). The vast majority of d/a values fall within the −0.5 to +0.5 interval, again suggesting that additivity is predominant (recall that a value of d = 0 indicates the exact intermediacy of expression).

Figure 4.
Distribution of dominance values in F1 hybrids for differentially expressed transcripts between M. m. musculus and M. m. domesticus and M. m. musculus and M. m. domesticus for two different P-value thresholds.

Contribution of sequence divergence to expression change:

Divergence in the sequences that represent the probes on the microarray may influence the results. The Affymetrix probe sequences are based on predominantly domesticus DNA [i.e., the lab strain C57BL/6J (Wade et al. 2002; Wade and Daly 2005; Yang et al. 2007)] and we used this microarray to compare domesticus to musculus. In a direct comparison between domesticus and musculus, probe sets should hybridize more efficiently in domesticus than in musculus in all regions of the genome where these two subspecies have accumulated sequence changes. To quantify the effect of sequence divergence of probe sets for the musculusdomesticus comparison, we obtained counts of the number of transcripts differentially expressed separately for the cases where musculus shows higher expression than domesticus and where domesticus shows higher expression than musculus (Table 5). At low significance, there is no difference between these numbers but, at high significance, about twice as many significant transcripts show higher expression in domesticus than in musculus. This suggests that sequence divergence might contribute to some extent to the detection of differentially expressed transcripts.

Transcripts differentially expressed between the parental subspecies M. m. musculus and M. m. domesticus at different P-value thresholds and at different magnitudes of change

We performed two tests to determine whether such an effect could contribute to our estimate of the proportion of additively expressed transcripts. First, as described in materials and methods, we assigned each probe set on the microarray as “musculus–like” or as “domesticus–like.” Then, we restricted the analysis to two sets of transcripts: (1) differentially expressed transcripts where the probe set is musculus-like but the expression change is in the direction of higher expression in domesticus and (2) differentially expressed transcripts where the probe set is domesticus-like but the expression change is in the direction of higher expression in musculus. In both cases, sequence divergence in the probe is not likely to explain the measured expression divergence. As shown in Table 6, the proportions of additively expressed transcripts among the differentially expressed transcripts in this subset of genes do not differ from the analysis including all probe sets.

Additively, dominantly, and overdominantly expressed transcripts among the differentially (fold change >2.5, ANOVA P < 0.01) expressed transcripts between M. m. musculus and M. m. domesticus in different tissues

As a second test, we considered the reciprocal crosses between musculus and castaneus. These two subspecies are more closely related to each other than either is to domesticus (Prager et al. 1998). Thus, this comparison should suffer much less from an ascertainment bias because both of these subspecies are equally diverged from the subspecies that contributes most sequence to the probe (i.e., domesticus). As shown above (Table 4), with one exception (one direction of the musculuscastaneus cross, hybrid testis), we can confirm the large surplus of additively expressed transcripts. Thus, we conclude that the excess of additive transcript expression is not substantially affected by divergence in the sequences that correspond to the probes on the Affymetrix microarray.

We also analyzed all differentially expressed transcripts between the parental subspecies with respect to maternal or paternal effects. To identify these transcripts, we analyzed both directions of the cross separately and looked for transcripts that were dominantly expressed in both directions of the cross. We classified transcripts as maternal-effect transcripts if the expression level in the F1 hybrids always matched that of the mother, while for paternal-effect transcripts F1 hybrids always matched that of the father. This analysis was restricted to transcripts that are not sex linked. Interestingly, we found one region on chromosome 7 (∼59 Mb, NCBI Build 36, gene Snrnp) that seems to show a paternal effect in the brain of F1 hybrids from the musculusdomesticus cross as well as in F1 hybrids from the musculuscastaneus cross. The former cross does show the same paternal effect also in the liver. No maternal-effect transcripts were found.

Nonadditive expression in F1 hybrids at transcripts that are not differentially expressed between the parents:

We included in this analysis all transcripts that did not show a significant difference in expression between the parental subspecies (i.e., P > 0.05; see Figure 1C). However, if the transcript had an expression level <500 in both parental subspecies for a given comparison, the transcript had to have an expression level >500 in the respective F1 hybrid to be included. Among the ∼10,000 transcripts meeting the criteria for this analysis (i.e., ∼12,000 minus the ∼2000 transcripts that are differentially expressed between two subspecies at P < 0.05) we found a few (between 0.01 and 0.5%) that differed significantly in expression between parents and F1 hybrids (Table 7). There is again the familiar exception, i.e., the musculus mother–castaneus father cross in the testis, where about one-third of all transcripts show significant nonadditivity. Thus, apart from the one exception, also in this analysis, nonadditivity of transcript expression is rare.

Transcripts nonadditively expressed in F1 hybrids compared to the parental subspecies

Chromosomal distribution of additively expressed transcripts:

In supplemental Table 1 (http://www.genetics.org/supplemental/) we list the chromosomal location (based on NCBI Build 36) of additively expressed transcripts in F1 hybrids that are found in both directions of the cross. For musculusdomesticus, a significant overrepresentation of additively expressed transcripts was found for chromosome 13 in the testis (Fisher's exact test, P = 0.01) and for chromosome 12 in the liver (P = 0.014). In the brain, no chromosome was significantly overrepresented. The musculuscastaneus comparison yielded a significant overrepresentation of additively expressed genes only for chromosome 10 in the brain (P = 0.03). The significant effect of all chromosomes disappeared when P-values were Bonferroni corrected. We found either no transcripts (musculusdomesticus) or only one transcript (musculuscastaneus) additively expressed in F1 hybrids on the X chromosome; although these are small numbers, they are not significantly different from expectation (P > 0.05). Due to small sample sizes, we did not test transcripts showing dominance or overdominance.

Clustering of additively expressed transcripts:

We analyzed the additively expressed transcripts for chromosomal clustering. Significant clustering might be expected if genes that are involved in the same pathway colocalize along the chromosome (Hurst et al. 2004) or, alternatively, if differential expression is caused by deletion or duplication of a region encompassing more than one gene. We found a few instances of significant clustering of differentially expressed transcripts that show additivity. For the musculusdomesticus comparison, we found chromosome 17 (Glo1, Glpr; P = 0.008) and chromosome 5 (Camkk2, Tmed2; P = 0.024) in the brain, chromosome 13 (Akr1c18, Akr1c12; P = 0.0006) and chromosome 7 (Aldoa, Hirip3; P = 0.0006) in the liver, and chromosome 2 (Rims2, 4930429F24Rik; P = 0.001) and chromosome 3 (Gm440, Tm2d2, P = 0.0012) in the testis. For the musculuscastaneus comparison, only the brain [chromosome 10 (unannotated gene and Fuca1; P = 0.005)] and the liver [chromosome 6 (1700019G17Rik and 2810474O19Rik; P = 0.0005)] showed evidence for clustering of additively expressed transcripts.

Functional categories of transcripts:

For brain and testis, no overrepresentation of any biological process was detected among the additively expressed transcripts. In the liver, metabolic functions were significantly overrepresented among the additively expressed transcripts in musculusdomesticus hybrids and oxygenase and acetyltransferase functions were overrepresented in musculuscastaneus hybrids (both assessed after Bonferroni correction).

We also performed this analysis for all transcripts differentially expressed between the parental subspecies, ignoring the expression status in F1 hybrids. As shown in supplemental Table 2 (http://www.genetics.org/supplemental/), we found a significant overrepresentation of certain biological processes in all tissues and in both comparisons of subspecies. Comparing musculus with domesticus, similar biological processes were overrepresented in the brain and testis (these included transcripts involved in intracellular protein traffic), and in the liver we found a strong and highly significant overrepresentation of transcripts involved in lipid, fatty acid, and steroid metabolism. The musculuscastaneus comparison showed similar patterns in the liver, but brain and testis showed different and distinct patterns of overrepresentation [i.e., carbohydrate metabolism (brain) and detoxification (testis); supplemental Table 3 at http://www.genetics.org/supplemental/].

Finally, we functionally annotated the large number of nonadditively expressed transcripts that did not show an expression difference between the parental subspecies in the testis from the musculus mother–castaneus father direction of the cross (see supplemental Table 4 at http://www.genetics.org/supplemental/). Various processes seem to be affected. “Spermatogenesis and motility” and “gametogenesis” rank fifth and sixth in the list with an overrepresentation P-value of 5 × 10−6.


We assayed expression levels in subspecies of the house mouse and their reciprocal F1 hybrids in three different tissues: brain, liver, and testis. These tissues were chosen as representative of different aspects of the phenotype of an organism. Previously, we suggested that expression changes in the brain between the subspecies could reflect behavioral differences and expression changes in the liver could indicate general metabolic differences, while changes in the testis pertain to changes in reproduction (Voolstra et al. 2007). We found a strikingly high preponderance of transcripts additively expressed in F1 hybrids between the subspecies in all three tissues. Additively expressed transcripts were found more or less evenly distributed among all the chromosomes of the house mouse, and functional annotation of the transcripts did not reveal an overrepresentation of specific functional categories. While we did see an overrepresentation of metabolic functions for additively expressed transcripts in the liver, the same overrepresentation is also seen when all transcripts differentially expressed between the parents are functionally annotated, suggesting that this is liver specific rather than a phenomenon associated with additive expression.

We used three different methods to identify additivity and nonadditivity in transcript expression, two of which classify transcript expression into discrete mode-of-inheritance categories, i.e., the fixed and the fractional threshold methods. Both these methods have different forms of bias. The fixed threshold method biases the result toward detecting more additivity as the difference in expression between the parental strains rises, whereas the fractional method is biased toward more overdominance as the differences between the parents become smaller. Because both methods affect the result in different directions, yet both of them yield the same result, we find strong support for the notion that additive transcript expression is common. The third method that assesses additivity/nonadditivity in transcript expression level, using a continuous distribution of dominance effects, was also consistent with our interpretation of most expression changes being additively expressed (Figure 4).

Among the seven studies in addition to ours that so far have addressed this question rigorously, two found evidence for substantial nonadditivity in gene expression (Gibson et al. 2004; Hedgecock et al. 2007) while the findings of four studies matched ours, showing a strong preponderance of additivity [Drosophila (Hughes et al. 2006), maize (Stupar and Springer 2006; Swanson-Wagner et al. 2006), and the laboratory mouse (Cui et al. 2006)]. One study in Arabidopsis (Vuylsteke et al. 2005) found relatively similar numbers of additively and nonadditively expressed transcripts. Several factors could explain the discrepancy among studies:

  1. Degree of divergence might play a critical role. Hughes et al. (2006) suggested that more recently diverged lines could show a higher fraction of additivity. However, one might expect that, as more and more genes come to influence a trait, provided those genes have no net directional dominance, additivity of expression will increase with time, at least up to a certain threshold.
  2. The fraction of additivity might depend on whether or not the parental lines are inbred. Given the fact that many crosses from highly inbred lines exhibit heterosis (i.e., F1 hybrid fitness-related traits lie outside the range of the two parents), one might expect that nonadditively expressed transcripts are causally related to the heterosis phenotype in studies using inbred lines.
  3. There could be methodological reasons for the discrepancy among these studies. We pursue this possibility in more detail in the next paragraphs.

Several methodological differences between studies are relevant. First, the definition of additivity depends on the measurement scale; i.e., on some scales an expression level might be called additive and on others not. In our study of house mouse subspecies, we have tried to minimize such an effect by using three different analysis methods, all of which gave very similar results and supported the view of pervasive additivity in gene expression in F1 hybrid mice. In addition, while we used only a small number of replicates in our study, our result is very robust with respect to changes in P-value and fold-change levels (data not shown) and thus is unlikely to be explained by the measurement scale.

Second, the outcome of a study might depend on the biological material that is used for RNA extractions. Some studies use single organs (such as the study of laboratory strains of the house mouse and our study) while other studies pool different tissues (i.e., a whole plant or a whole fly is processed). Third, studies might differ due to technical reasons, most notably in the microarray platform used. Recently, there has been an interest in comparing the performance of different methods to measure gene expression level on the genomewide scale. While some studies did not find differences in performance among platforms (Patterson et al. 2006), other studies concluded that generally one-color microarrays (such as Affmetrix or ABI) outperform two-color microarrays (such as homemade cDNA arrays or Agilent arrays) (De Reynies et al. 2006; Kuo et al. 2006).

In Table 8, we have summarized all studies that have systematically assayed patterns of inheritance in gene expression. We have also included four additional studies that have less rigorously treated the issue, but these studies allow a qualitative evaluation. As shown in Table 8, neither the level of inbreeding nor differences in divergence time are associated with the frequency of additivity. This is, for example, illustrated by the study by Ranz et al. (2004) who compared D. melanogaster to D. simulans and their female F1 hybrids. Using heads as the source for the RNA, Ranz et al. found a high proportion of additivity, but the whole body of these flies showed a surplus of nonadditivity. In this example, complexity of the tissue used as the RNA source might explain the difference, although this cannot apply in other cases, such as the study by Hughes et al. (2006) who also used whole flies as the RNA source and found a strong preponderance of additivity (Table 8). The most striking correlate of additivity is the platform used. Among the eight studies that support additivity as the major mode of inheritance, five used Affymetrix arrays, two were based on cDNA arrays, and one study used differential display. None of the four studies that found a preponderance of nonadditive expression used the Affymetrix platform but instead used either a two-color microarray or sequencing-based expression measures. Thus, the combined evidence seems to suggest that methodological issues are a main cause of different results. This implies that additivity may be a general feature of most (but not all) genes differentially expressed among taxa, independent of divergence level or inbreeding and most likely also independent of the complexity of the tissue.

Literature review of studies on differentiating additivity and nonadditivity in gene expression

House mice are increasingly becoming a model system for studying the genetics of incipient speciation (Guenet and Bonhomme 2003; Galtier et al. 2004; Harr 2006). Under the biological species concept (Mayr 1963), the different taxa of house mice have been considered subspecies (Boursot et al. 1993) because they are only partially reproductively isolated from each other (i.e., male and female hybrids are fully viable and only some crosses between subspecies show hybrid male sterility, while females are fully fertile). Voolstra et al. (2007) have previously tested whether the evolution of reproductive isolation between house mouse subspecies could be explained by an exceptionally fast rate of divergence in gene expression between the subspecies in reproductive organs (i.e., testis), as predicted by some models of sexual selection and sexual conflict (Rice 1998). While Voolstra et al. (2007) found a large number of genes differing in expression in the testis between two species of Mus (M. spretus and M. musculus), comparisons between house mouse subspecies showed most differences in expression in the combined liver and kidney tissue and only very few genes differentially expressed in the testis. In contrast, we find here that most gene expression changes between subspecies of M. musculus are in the testis, especially at a high significance threshold. Some reasons that could explain the discrepancy between this study and Voolstra et al. (2007) are methodological. Voolstra et al. (2007) used two-color microarrays while this study uses Affymetrix microarrays. Voolstra et al. (2007) also used individuals that were collected directly in the wild and standardized to laboratory conditions for a few days only, while all mice used in this study have been born in the laboratory. Finally, Voolstra et al. (2007) used multiple unrelated individuals within each subspecies as replicates in the statistical tests, whereas we used only a single (and to a large extent inbred) strain within each subspecies. It is possible that genes affecting hybrid sterility are segregating within populations. If different individuals in Voolstra et al.'s sample were of different genotypes, the consequent within-subspecies variability would obscure between-subspecies differences. In support of this explanation are experiments by Vyskocilova et al. (2005), who crossed laboratory strain C57BL/6J to several different musculus strains from various locations in the Czech Republic and found several hybrid sterility loci segregating within natural populations. Interestingly, the only other study of which we are aware that has mapped hybrid sterility loci in recently diverged taxa (in two species of Mimulus) with some precision and determined the frequency of these alleles in natural populations also found several hybrid sterility loci segregating within natural populations but no fixed differences between the species (Sweigart et al. 2007).

We have not measured the fertility status of the F1 hybrid males that we studied. Some may be sterile. In one cross, between a musculus mother and a castaneus father, an extraordinarily large number of transcripts were “mis-expressed” in F1 hybrids relative to both parents in the testis. The fact that this high number of mis-expressed transcripts is restricted to the testis and is not observed in brain or liver rules out methodological/technical problems as causes of this mis-expression. If this is linked to sterility, it indicates that a very large number of genes are affected, perhaps a result of a few master control genes influencing many downstream targets. The observed case of pervasive mis-expression in the testis could represent the crossing of a threshold whereby genetic incompatibilities between taxa result in low hybrid fitness. Such incompatibilities are often thought to affect fertility before viability (Coyne and Orr 2004).

It is less clear how the relatively large number of differentially expressed transcripts between the parental strains that are additively expressed in F1 hybrids would relate to F1 hybrid fitness. While intermediate expression level might already be deleterious in F1 hybrids, these genes might also be important isolating genes in F2 or later generation backcrosses, a phenomenon that has been observed in some between-subspecies crossing schemes (Oka et al. 2007). While the contribution of the identified genes to reproductive isolation at this stage is highly speculative, our study highlights valuable candidate genes (both the additive and the nonadditively expressed transcripts) on which future expression, phenotypic, and population genetic studies using larger samples of multiple unrelated individuals within each subspecies could be targeted. Ultimately, these efforts will help in determining the genetic basis of reproductive isolation in the house mouse in nature.


We are especially grateful to J. Pialek for providing the mouse strains of M. m. musculus and M. m. domesticus and A. Orth and F. Bonhomme for providing M. m. castaneus. B. Schmitz contributed expert technical help and C. Voolstra helpful experimental advice. P. Nuernberg and C. Becker from the Center for Functional Genomics in Cologne, Germany, performed the microarray experiments. We thank T. Price for valuable discussion and J. Baines and R. Scavetta and two anonymous reviewers for helpful comments on the manuscript. The work has been supported by Emmy-Noether and SFB680 funds by the Deutsche Forschungsgemeinschaft to B. Harr.


  • Boursot, P., J. C. Auffray, J. Brittondavidian and F. Bonhomme, 1993. The evolution of house mice. Annu. Rev. Ecol. Syst. 24: 119–152.
  • Brem, R. B., G. Yvert, R. Clinton and L. Kruglyak, 2002. Genetic dissection of transcriptional regulation in budding yeast. Science 296: 752–755. [PubMed]
  • Comings, D. E., and J. P. MacMurray, 2000. Molecular heterosis: a review. Mol. Genet. Metab. 71: 19–31. [PubMed]
  • Coyne, J. A., and H. A. Orr, 2004. Speciation. Sinauer Associates, Sunderland, MA.
  • Cui, X., J. Affourtit, K. R. Shockley, Y. Woo and G. A. Churchill, 2006. Inheritance patterns of transcript levels in F1 hybrid mice. Genetics 174: 627–637. [PMC free article] [PubMed]
  • de Reynies, A., D. Geromin, J. M. Cayuela, F. Petel, P. Dessen et al., 2006. Comparison of the latest commercial short and long oligonucleotide microarray technologies. BMC Genomics 7: 51. [PMC free article] [PubMed]
  • Doss, S., E. E. Schadt, T. A. Drake and A. J. Lusis, 2005. Cis-acting expression quantitative trait loci in mice. Genome Res. 15: 681–691. [PMC free article] [PubMed]
  • Falconer, D. S., and T. F. C. Mackay, 1996. Introduction to Quantitative Genetics. Longman, Essex, UK.
  • Galtier, N., F. Bonhomme, C. Moulia, K. Belkhir, P. Caminade et al., 2004. Mouse biodiversity in the genomic era. Cytogenet. Genome Res. 105: 385–394. [PubMed]
  • Gibson, G., R. Riley-Berger, L. Harshman, A. Kopp, S. Vacha et al., 2004. Extensive sex-specific nonadditivity of gene expression in Drosophila melanogaster. Genetics 167: 1791–1799. [PMC free article] [PubMed]
  • Guenet, J. L., and F. Bonhomme, 2003. Wild mice: an ever-increasing contribution to a popular mammalian model. Trends Genet. 19: 24–31. [PubMed]
  • Harr, B., 2006. Genomic islands of differentiation between house mouse subspecies. Genome Res. 16: 730–737. [PMC free article] [PubMed]
  • Hedgecock, D., J. Z. Lin, S. Decola, C. D. Haudenschild, E. Meyer et al., 2007. Transcriptomic analysis of growth heterosis in larval Pacific oysters (Crassostrea gigas). Proc. Natl. Acad. Sci. USA 104: 2313–2318. [PMC free article] [PubMed]
  • Hey, J., and R. M. Kliman, 1993. Population genetics and phylogenetics of DNA sequence variation at multiple loci within the Drosophila melanogaster species complex. Mol. Biol. Evol. 10: 804–822. [PubMed]
  • Hoekstra, H. E., and J. A. Coyne, 2007. The locus of evolution: evo devo and the genetics of adaptation. Evolution Int. J. Org. Evolution 61: 995–1016. [PubMed]
  • Hughes, K. A., J. F. Ayroles, M. M. Reedy, J. M. Drnevich, K. C. Rowe et al., 2006. Segregating variation in the transcriptome: cis regulation and additivity of effects. Genetics 173: 1347–1355. [PMC free article] [PubMed]
  • Hurst, L. D., C. Pal and M. J. Lercher, 2004. The evolutionary dynamics of eukaryotic gene order. Nat. Rev. Genet. 5: 299–310. [PubMed]
  • Karn, R. C., A. Orth, F. Bonhomme and P. Boursot, 2002. The complex history of a gene proposed to participate in a sexual isolation mechanism in house mice. Mol. Biol. Evol. 19: 462–471. [PubMed]
  • Kuo, W. P., F. Liu, J. Trimarchi, C. Punzo, M. Lombardi et al., 2006. A sequence-oriented comparison of gene expression measurements across different hybridization-based technologies. Nat. Biotechnol. 24: 832–840. [PubMed]
  • Mackay, T. F., 2001. The genetic architecture of quantitative traits. Annu. Rev. Genet. 35: 303–339. [PubMed]
  • Mayr, E., 1963. Animal Species and Evolution. The Belknap Press, Cambridge, MA.
  • Michalak, P., and M. A. Noor, 2003. Genome-wide patterns of expression in Drosophila pure species and hybrid males. Mol. Biol. Evol. 20: 1070–1076. [PubMed]
  • Oka, A., T. Aoto, Y. Totsuka, R. Takahashi, M. Ueda et al., 2007. Disruption of genetic interaction between two autosomal regions and the X chromosome causes reproductive isolation between mouse strains derived from different subspecies. Genetics 175: 185–197. [PMC free article] [PubMed]
  • Patterson, T. A., E. K. Lobenhofer, S. B. Fulmer-Smentek, P. J. Collins, T. M. Chu et al., 2006. Performance comparison of one-color and two-color platforms within the MicroArray Quality Control (MAQC) project. Nat. Biotechnol. 24: 1140–1150. [PubMed]
  • Prager, E. M., C. Orrego and R. D. Sage, 1998. Genetic variation and phylogeography of central Asian and other house mice, including a major new mitochondrial lineage in Yemen. Genetics 150: 835–861. [PMC free article] [PubMed]
  • Piálek, J., M. Vyskocilová, B. Bímová, D. Havelková, J. Piálková et al., 2007. Development of unique house mouse resources suitable for evolutionary studies of speciation. J. Hered. (in press). [PubMed]
  • Ranz, J. M., K. Namgyal, G. Gibson and D. L. Hartl, 2004. Anomalies in the expression profile of interspecific hybrids of Drosophila melanogaster and Drosophila simulans. Genome Res. 14: 373–379. [PMC free article] [PubMed]
  • Reiland, J., and M. A. Noor, 2002. Little qualitative RNA misexpression in sterile male F1 hybrids of Drosophila pseudoobscura and D. persimilis. BMC Evol. Biol. 2: 16. [PMC free article] [PubMed]
  • Reiner, A., D. Yekutieli and Y. Benjamini, 2003. Identifying differentially expressed genes using false discovery rate controlling procedures. Bioinformatics 19: 368–375. [PubMed]
  • Rice, W. R., 1998. Intergenomic conflict, interlocus antagonistic coevolution, and the evolution of reproductive isolation, pp. 261–270 in Endless Forms: Species and Speciation, edited by S. Berlocher. Oxford University Press, London/New York//Oxford.
  • Rockman, M. V., and L. Kruglyak, 2006. Genetics of global gene expression. Nat. Rev. Genet. 7: 862–872. [PubMed]
  • Schadt, E. E., S. A. Monks, T. A. Drake, A. J. Lusis, N. Che et al., 2003. Genetics of gene expression surveyed in maize, mouse and man. Nature 422: 297–302. [PubMed]
  • Storey, J. D., J. M. Akey and L. Kruglyak, 2005. Multiple locus linkage analysis of genomewide expression in yeast. PLoS Biol. 3: e267. [PMC free article] [PubMed]
  • Stupar, R. M., and N. M. Springer, 2006. Cis-transcriptional variation in maize inbred lines B73 and Mo17 leads to additive expression patterns in the F1 hybrid. Genetics 173: 2199–2210. [PMC free article] [PubMed]
  • Su, A. I., T. Wiltshire, S. Batalov, H. Lapp, K. A. Ching et al., 2004. A gene atlas of the mouse and human protein-encoding transcriptomes. Proc. Natl. Acad. Sci. USA 101: 6062–6067. [PMC free article] [PubMed]
  • Swanson-Wagner, R. A., Y. Jia, R. DeCook, L. A. Borsuk, D. Nettleton et al., 2006. All possible modes of gene action are observed in a global comparison of gene expression in a maize F1 hybrid and its inbred parents. Proc. Natl. Acad. Sci. USA 103: 6805–6810. [PMC free article] [PubMed]
  • Sweigart, A. L., A. R. Mason and J. H. Willis, 2007. Natural variation for a hybrid incompatibility between two species of Mimulus. Evolution Int. J. Org. Evolution 61: 141–151. [PubMed]
  • Voolstra, C., D. Tautz, P. Farbrother, L. Eichinger and B. Harr, 2007. Contrasting evolution of expression differences in the testis between species and subspecies of the house mouse. Genome Res. 17: 42–49. [PMC free article] [PubMed]
  • Vuylsteke, M., F. van Eeuwijk, P. Van Hummelen, M. Kuiper and M. Zabeau, 2005. Genetic analysis of variation in gene expression in Arabidopsis thaliana. Genetics 171: 1267–1275. [PMC free article] [PubMed]
  • Vyskocilova, M., Z. Trachtulec, J. Forejt and J. Pialek, 2005. Does geography matter in hybrid sterility in house mice? Biol. J. Linn. Soc. 84: 663–674.
  • Wade, C. M., and M. J. Daly, 2005. Genetic variation in laboratory mice. Nat. Genet. 37: 1175–1180. [PubMed]
  • Wade, C. M., E. J. Kulbokas, III, A. W. Kirby, M. C. Zody, J. C. Mullikin et al., 2002. The mosaic structure of variation in the laboratory mouse genome. Nature 420: 574–578. [PubMed]
  • Waterston, R. H., K. Lindblad-Toh, E. Birney, J. Rogers, J. F. Abril et al., 2002. Initial sequencing and comparative analysis of the mouse genome. Nature 420: 520–562. [PubMed]
  • Wray, G. A., 2007. The evolutionary significance of cis-regulatory mutations. Nat. Rev. Genet. 8: 206–216. [PubMed]
  • Yang, H., T. A. Bell, G. A. Churchill and F. Pardo-Manuel de Villena, 2007. On the subspecific origin of the laboratory mouse. Nat. Genet. 39: 1100–1107. [PubMed]

Articles from Genetics are provided here courtesy of Genetics Society of America
PubReader format: click here to try


Save items

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


  • BioProject
    BioProject links
  • GEO DataSets
    GEO DataSets
    Gene expression and molecular abundance data reported in the current articles that are also included in the curated Gene Expression Omnibus (GEO) DataSets.
  • MedGen
    Related information in MedGen
  • PubMed
    PubMed citations for these articles
  • Taxonomy
    Taxonomy records associated with the current articles through taxonomic information on related molecular database records (Nucleotide, Protein, Gene, SNP, Structure).
  • Taxonomy Tree
    Taxonomy Tree

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...