- Open Access
The contributions of normal variation and genetic background to mammalian gene expression
© Pritchard et al.; licensee BioMed Central Ltd. 2006
- Received: 22 September 2005
- Accepted: 28 February 2006
- Published: 31 March 2006
Qualitative and quantitative variability in gene expression represents the substrate for external conditions to exert selective pressures for natural selection. Current technologies allow for some forms of genetic variation, such as DNA mutations and polymorphisms, to be determined accurately on a comprehensive scale. Other components of variability, such as stochastic events in cellular transcriptional and translational processes, are less well characterized. Although potentially important, the relative contributions of genomic versus epigenetic and stochastic factors to variation in gene expression have not been quantified in mammalian species.
In this study we compared microarray-based measures of hepatic transcript abundance levels within and between five different strains of Mus musculus. Within each strain 23% to 44% of all genes exhibited statistically significant differences in expression between genetically identical individuals (positive false discovery rate of 10%). Genes functionally associated with cell growth, cytokine activity, amine metabolism, and ubiquitination were enriched in this group. Genetic divergence between individuals of different strains also contributed to transcript abundance level differences, but to a lesser extent than intra-strain variation, with approximately 3% of all genes exhibiting inter-strain expression differences.
These results indicate that although DNA sequence fixes boundaries for gene expression variability, there remain considerable latitudes of expression within these genome-defined limits that have the potential to influence phenotypes. The extent of normal or expected natural variability in gene expression may provide an additional level of phenotypic opportunity for natural selection.
- Gene Ontology
- Mouse Strain
- Additional Data File
- Individual Mouse
- Amine Metabolism
Biological entities such as individual cells, organs, and entire organisms display phenotypes that are simultaneously dictated and constrained by the composition of nucleic acids comprising their genomes. Differences in DNA sequence between individuals within the same species may produce qualitative and quantitative alterations in gene expression that influence biochemical processes conferring disease susceptibility and the beneficial or adverse responses to pharmacological intervention [1, 2]. Thus, a critical component of biomedicine centers on establishing the cause, extent and result of gene expression variability with an aim toward establishing pathological associations. To this end, the development of technologies such as DNA microarrays have allowed for quantitative assessments of transcriptional activity for thousands of genes simultaneously . Microarray-based methods have been used to measure transcriptional variance in a variety of organisms, including yeast , flies , fish , mice , and men ; usually in the context of assessing the contribution of gene expression to phenotypic attributes of age, sex, strain, or disease. While the major component of phenotypic diversity within species is thought to be provided by combinations of heritable variations in DNA, it is readily apparent that individuals sharing nearly identical genomes, such as inbred mouse strains and monozygotic twins, may exhibit strikingly different characteristics [9, 10].
To assess the extent and nature of gene expression variability both within populations of genetically identical individuals and between genetically heterogeneous individuals, we selected five strains of commonly used laboratory mice; inbred 129, Balb/c, and FVB, and outbred CD1 and CFW, isolated RNA from the livers of three males from each strain, and quantified transcript abundance levels by comparative hybridizations to cDNA microarrays.
Mice bred for more than 60 generations should fix the vast majority (potentially all) of genetic contribution to variation , and thus individual mice within each inbred strain are considered genetically identical. We studied the liver in view of its important contribution to a wide variety of metabolic processes as well as practical considerations involving sample quantities and the ease of tissue procurement. To account for technical inconsistencies and facilitate comparisons within and between strains, each array hybridization used a common reference consisting of RNA combined from the liver, testes, and kidney of all mice used in the experiments. Two replicate arrays were performed for each individual mouse liver sample with each of two different fluorescent dyes to control for potential dye bias, thereby generating 4 replicate arrays per mouse and a total of 60 arrays.
We anticipated that three major sources of measurable variation in transcript levels would be represented in this dataset. The first involves the technical inconsistencies in experimental procedures and was assessed by the four replicate arrays performed for each mouse sample. The second source of variation is represented predominantly by intrinsic and extrinsic non-genetic factors influencing gene expression. This variance component was measured through the determination of transcript levels between mice of the same strain with identical genomes. All mice were matched for age, and were provided consistent diets and living environments. The third source of gene expression variability was expected to be driven by differences in DNA sequence or genome structure between the different mouse strains. This inter-strain variability was measured by determining transcript abundance levels between mice of different strains.
To identify genes whose transcript levels varied between genetically identical individuals, we first used an ANOVA model with a conservative assessment of significance . This method yielded the following number of variable genes within each strain: 129, 37 genes; Balb/c, 36 genes; CD1, 26 genes; FVB, 21 genes; and CFW, 11 genes. Our previous study of liver gene expression in C57BL/6 mice identified 21 variable genes (0.8% of all genes assessed), indicating that the overall experimental results are quite consistent . While this method identifies variable genes with high confidence, we concluded that the approach has a high rate of false negatives and is unduly restrictive when one is interested in assessing overall levels of variability rather than focusing on any particular gene product.
Variable genes within strains
Total variable genes*
Of the genes exhibiting variable expression levels within strains, 33 were variable within all 5 strains, and 154 were variable in 4 of the 5 strains (see Additional data file 2, supplemental Table 2). To determine how many genes are expected to be in common by chance if genes were chosen randomly, we undertook a simulation study with 50,000 datasets generated by randomly selecting groups of 554, 1,059, 749, 610, and 661 genes from the 5 strains for each data set and determined the number of genes represented in all 5 sets. The greatest number of genes in common, by chance alone, was 19, though typically fewer than 10 genes were found in common. This analysis indicates that the 33 genes identified in our study represent a highly significant level of overlap (p < 0.00002). Searches based on gene ontology (GO) classification indicated that genes associated with cell growth [GO:0008151], cytokine activity [GO:0005125], amine metabolism [GO:0009308], and the ubiquitin ligase complex [GO:0000151] were enriched among the genes with consistent intra-strain variation, when compared to the array as a whole. All genes showing significant inter-individual variability in our previous study of C57BL/6 mice  also varied in at least one strain analyzed in the current study. Moreover, genes previously found to exhibit substantial hepatic intra-strain variability, including CisH, Hhex, Cyp4a14, and Gadd45a, varied in at least three out of the five strains. Interestingly, four genes identified in the current study are involved in the ubiquitination process; Wsb1, Arih1, Cdc27, and Chordc1. This finding suggests the possibility that normal variability in protein degradation pathways could provide an additional level of global gene expression variability either through direct targeting of specific proteins or via a cascade of indirect effects influencing transcriptional regulation.
Comprehensive studies of gene expression in model organisms such as Saccharomyces and Drosophila have delineated the contributions of age, sex, and genotype to corresponding variations in transcript levels. However, the size constraints of these species necessitates the use of sample pools composed of hundreds to millions of discrete organisms, an approach that eliminates the ability to assess variability at the level of the individual. In contrast, assessing the relationships between the genome and gene expression variability in humans is hampered by the inability to precisely control the multitude of environmental influences that profoundly influence gene expression in qualitative and quantitative ways. In this context, the mouse represents a useful model system highly suited for establishing that component of variability that is independent of diversity directly encoded in the genome. Measurements of intra-strain gene expression levels reflect the allowable latitudes of gene expression in any single individual in a fixed environment at a given point in time. The inter-strain measurements reflect the additional contribution of heterogeneity at the level of the genome.
Based on the analyses of transcript levels in individual mice, we found the greatest contribution to overall gene expression variability occurred among genetically identical individuals: 23% to 44% of all genes exhibited measurable variation, depending on strain (see Additional data file 2, supplemental Figure 3). Substantially less variance was attributable to genome differences between strains (about 2.8%). Few studies assessing natural gene expression variability in mammalian species that might provide a context for these findings have been reported. Analyses of transcript levels in skeletal muscle between five mouse strains found greater inter-strain than intra-strain differences. . This suggests that muscle tissue exhibits a narrow range of normal variation relative to liver. However, the study design in which two mice per strain and two microarrays per mouse were compared provides substantially less statistical power to detect differences within strain. Interestingly, concordant with the findings reported here, Balb/c mice demonstrated the greatest level of intra-strain variation. A comparative analysis of mRNA abundance levels in the hippocampus of mice from 8 mouse strains identified more than 200 genes with significant strain differences using very stringent statistical criteria . The experimental design involved tissue pooled from six mice of each strain, rather than individual mice. This pooling strategy was apparently based in part on the results of a prior microarray study indicating that transcript levels of genes expressed in the hippocampus of genetically identical mice were quite similar with only about 0.1% of all transcripts called differentially expressed . It is possible that there is lower inter-individual variability in hippocampus than in liver. However, this previous study directly compared only pairs of mice in a head-to-head fashion, and the criteria for differential expression were based on a 1.7-fold change in abundance level, and not on statistical criteria.
Overall, we found that the expression of most hepatic genes in mice housed in standard 'steady-state' laboratory vivarium conditions is similar between individuals of the same or different strain. However, the transcript levels of a sizeable minority varied substantially. The proportion of genes exhibiting significantly variable expression between individual fish (18%) , yeast strains (24%) , and fly genotypes (25%)  is similar to that observed here between individual mice (23% to 44%). Analyses of gene expression in human tissues have also shown considerable variability between individuals. Importantly, substantial contributions to this variation cannot be attributed to genotypic differences between subjects [8, 21, 22]. Comparisons of transcript and protein levels between humans and non-human primates identified significantly greater variation among the human subjects than between humans and chimpanzees , a finding further supporting the conclusion that a sizeable component of transcript abundance measurements reflects non-genomic variation.
There are several possible contributors to the gene expression variability observed in genetically identical individuals. Technical factors include subclinical disease states, unrecognized differences in environments and diet, or heterogeneity in the cell-type compositions of the analyzed tissues. We attempted to precisely control environmental and handling effects during the design of this study, and we did not observe any histological differences in the cellular composition of livers within or between strains. The ideal experiment would assess temporal variation in tissue transcript levels within an individual mouse, but in the case of liver gene expression these measurements would be confounded by changes resulting from repeated tissue biopsies. Importantly, our analyses of separate liver samples acquired from the same mouse yielded highly concordant transcript measurements.
One component of inter-individual variability could be represented by stochastic events or noise. Recently, gene expression measurements at the level of the single cell have provided direct experimental evidence of quantifiable contributions of stochastic biochemical noise to phenotypic variation in isogenic populations [24, 25]. The end-result of this component of variability has long been appreciated through studies of developmental processes that revealed requirements for feed-back amplifications of initial asymmetrical noise for cell fate determination .
A second potential contributor to individual differences in gene expression centers on epigenetic regulation. Methylation of cytosine residues in the CpG islands of gene promoters and the covalent modifications of histones represent two important epigenetic modifications that influence gene transcription. Recent studies emphasize the importance of these regulatory mechanisms for dictating phenotypes in individuals with minimal divergence in genome sequence. A provocative report by Rakyan et al.  determined that the penetrance of the highly variable kinky-tail phenotype found in the well-studied Axin-fused (Axin Fu ) mouse strain correlated with the differential methylation of a retrotransposon within Axin Fu . Importantly, the methylation state of the retrotransposon was inherited transgenerationally after both maternal and paternal transmission, and was influenced by strain background. Striking differences in DNA methylation and histone acetylation have been observed in identical twins with increasing 'epigenetic drift' associated with advanced age . Similar age-related epigenetic shifts have been reported in mice . In the studies reported here, we found several genes that exhibited high variability in more than one strain, suggesting that certain genomic loci may be prone to imprecise regulatory control.
In the context of complex multicellular organisms, the end-result of phenotypic diversity in the setting of a fixed genome has long been appreciated. Toxicology studies have repeatedly shown differing susceptibilities to drug effects, such as carcinogen-induced tumor promotion within isogenic mouse strains . Genetically identical animals aged under tightly controlled environments exhibit wide ranges in lifespans . Indeed, the seeming incongruity between genetic homogeneity and phenotypic variability was recognized more than 40 years ago . Importantly, the magnitude of gene expression variability measured in this study suggests either a tolerance for wide abundance ranges of certain transcripts, or potentially an organismal advantage for maintaining a state of gene expression variability offering an additional level of phenotypic opportunity for natural selection.
Animal work and RNA preparation
Mice were purchased from Charles River Laboratories (Wilmington, MA, USA), maintained in a barrier facility and cared for in accordance with an approved Animal Care and Use Committee (IACUC) protocol. All mice were between 68 and 73 days old and were housed in identical environments with the same diet (Harlan Teklad 8664), constant temperature (20 to 22°C), and consistent light and dark cycles (controlled photoperiod of 12 hour light/12 hour dark). Water was provided ad libitum. Three male mice were sacrificed from each of the following strains (nomenclature in italics is used throughout this paper): 129S4 (129), Balb/cAnNCrlBR (Balb/c), Crl:CD-1®(ICR)BR (CD1), FVB/NCrlBR (FVB), and Crl:CFW®(SW)BR (CFW); CFW is sometimes referred to as 'Swiss Webster'. Each mouse was brought individually into a separate room for sacrifice and killed in a CO2 chamber. The liver, left kidney, and left testis were removed from each mouse and immediately snap-frozen in liquid nitrogen. Care was taken to ensure that the minimum amount of time elapsed from the sacrifice of the first mouse to the last. Total RNA was extracted from the tissue using the TRIzol reagent (Life Technologies, Grand Island, NY, USA) according to the manufacturer's protocol. For an RNA reference standard, equal quantities of total RNA were combined from all three organs of all the mice. This same reference RNA was used on every array to standardize comparisons between arrays. For confirmation studies, 3 additional male mice of each strain, ages 68 to 72 days, were processed in a similar manner except that the liver was divided into 4 sections before snap-freezing.
Microarray construction, probe generation, and data collection
Each microarray comprised 5,285 mouse cDNAs obtained from the Research Genetics' sequence-verified set of IMAGE clones (Research Genetics, Invitrogen Corporation, Carlsbad, CA, USA). All cDNA clones used for array construction were sequence verified and annotated accordingly. Clone inserts were amplified by PCR, purified, verified by gel electrophoresis and spotted onto polylysine-coated glass microscope slides using a GeneMachines (San Carlos, CA, USA) robotic spotter as described previously . cDNA probes were generated from 50 μg of total RNA in a reaction volume of 30 μl containing oligo(dT) primer/0.2 mM amino acid-dUTP (Sigma-Aldrich, St Louis, MO, USA)/0.3 mM dTTP/0.5 mM each dATP, dGTP, and dCTP/380 units of Superscript II reverse transcriptase (Life Technologies). The purified cDNA was combined with either Cy3 or Cy5 monoreactive fluorophores (GE Healthcare, Piscataway, NJ, USA (formerly Amersham Pharmacia)) that covalently couple to the cDNA-incorporated aminoallyl linker in the presence of 50 mM NaHCO3 (pH 9.0). The experimental and reference probes were combined and competitively hybridized to microarrays under a coverslip in a volume of 24 μl for 16 h at 63°C. Slides were washed in graded sodium chrolide/sodium citrate buffer (SSC, 1× SSC = 0.15 M NaCl/0.015 M sodium citrate, pH 7) and spun dry. Array images were collected for the Cy3 and Cy5 emissions using a GenePix 4000A fluorescent scanner (Axon Instruments, Foster City, CA, USA). The image data were extracted and analyzed using GENEPIX 3.0 microarray analysis software (Axon Instruments).
For each array spot, the intensity levels of the two fluorophores were obtained by subtracting median background intensity from median foreground intensity. A gene was only considered expressed if the fluorescence intensity of the corresponding spot was at least six foreground pixels greater than four standard deviations above background on every array. For each gene, the logarithm base 2 ratios (referred to henceforth as log ratios) of the two channels were calculated to quantify to relative expression levels between the experimental and reference samples. To allow for inter-array comparisons, each array was normalized to remove systemic sources of variation. This normalization was accomplished by means of a print-tip-specific intensity-based normalization method. . A scatter-plot smoother, which uses robust locally linear fits, was applied to capture the dependence of the log ratios on overall log-spot intensities. The log ratios were normalized by subtracting the fitted values based on the print-tip-specific scatter-plot smoother from the log ratios of experimental and control channels. Examination of the spread of the normalized log ratios via boxplots indicated no systemic variation due to any experimental variable such as different batches of arrays or RNA preparations. Therefore, no scale adjustment was performed on the arrays before combining data across samples.
The expression of genes that vary among mice within each strain was evaluated using an ANOVA model (Pritchard et al. ). Here, an F-value with degrees of freedom 2 and 8 was used to assess the variability of mouse variance within each strain.
To identify genes that varied among strains of mice, a nested mixed effects ANOVA model was used. Specifically, the model is written as:
y = overall mean + dye + strain + mouse within strain
where y is the normalized log2 ratio and the mouse within strain is a random effect. Treating the mouse as a random effect basically assumes that the three mice have been randomly selected from an 'infinite' mouse population of that strain and its observed effect for a particular mouse is an observation of a random variable. Specifically, the F test statistic is:
where ij indexes the ith mouse for the jstrain, and ij , j , and are the means of normalized ratios for the ith mouse in the jth strain, all mice in jth strain, and over all strains, respectively. An F-value with degrees of freedom 4 and 10 for each gene is used to assess how variable the gene is among strains. An ANOVA table for this analysis is provided in Additional data file 2 (supplemental Table 1). To examine a possible dependence of statistical significance and signal intensity, we plotted the F-values versus the log2(intensityCy3 + intensityCy5). There was no dependence on intensity for the significant genes either within strain or between strains (see Additional data file 2, supplemental Figure 2). The significance of these F-values was determined through estimating the pFDR, which is the proportion of falsely rejected hypotheses among the rejected hypotheses for pre-selected critical values . As the overall goal here is to assess how genes vary among and within strains, it is natural to control the proportion of falsely rejected hypotheses among the rejected ones while examining the genes that vary among/within strains. In this paper, the pFDR level was set to be 0.10. This means that we expect 10% of our rejected hypotheses ('significant' genes) to have been falsely rejected. The pFDR level of 0.10 is a somewhat liberal cutoff as we are most interested in assessing overall levels of variation rather than defining a small subset of genes that vary with high confidence. A q-value that measures the strength of the F-value with respect to pFDR was also calculated for each gene using the algorithm proposed by Storey and Tibshirani . The q-value is the minimum pFDR that occurs when rejecting a statistic with the observed F-value for the set of nested rejected regions. To avoid the distributional assumption, 1,000 bootstrap samples were used to calculate the pFDRs for a series of critical values and the q-values for all the genes.
To determine which gene ontology terms were enriched among the variable genes we used EASE software. EASE compares the proportion of genes that are assigned a given GO term among the list of variable genes to the proportion of genes with that GO term on the array as a whole. A statistical score similar to a p value is generated based on the upper bound of the distribution of Jackknife Fisher exact probabilities. For genes that varied within strain we performed separate EASE analyses for each strain, and then reported the GO terms that were enriched by >1.5-fold in at least 4 out of 5 strains and had the lowest average EASE score (cell growth, 0.35; amine metabolism, 0.25; cytokine activity, 0.39; ubiquitin ligase complex, 0.43).
Hierarchical clustering was performed using Cluster 3.0 software (Michael Eisen, Stanford University). We used complete linkage clustering for both genes and arrays with a correlation (uncentered) similarity metric with data either unweighted or weighted by F-value.
The normalized log ratios, F-values, q-values, and mean squares for the 2,382 genes assessed in the unpooled analysis are included in Additional data file 1. In addition, information about the microarray used in this study and the unprocessed gpr files may be obtained through the ArrayExpress website at the European Bioinformatics Institute . The accession number is: A-MEXP-320.
Quantitative PCR was performed using SYBR GREEN as a reporter as previously described . Total RNA from each mouse liver was treated with DnaseI, purified using a Rneasy Minikit (Qiagen, Valencia, CA, USA), and 20 μg was used to generate cDNA for PCR reactions. Primers to ribosomal protein S16 were used to normalize for cDNA loading. The sequences of the primers used were: S16 forward, 5'-AGGAGCGATTTGCTGGTGTGGA-3'; S16 reverse, 5'-GCTACCAGGCCTTTGAGATGGA-3' (102 base-pair (bp) amplicon); Pfk2 forward, 5'-AAGAGGCCAAAGCTGGAGG-3'; Pfk2 reverse, 5'-GTCAGCATTCCGGTGGTGTA-3'; Cth forward, 5'-TCTTGCTGCCACCATTACGA-3'; Cth reverse, 5'-GCCTCCATACACTTCATCCAT-3'; Dnase2a forward, 5'-TCCAGGGAAAACTGCTGACC-3'; Dnase2a reverse, 5'-AGGAAAAGGCTGTCGGTGG-3'; Apoa4 forward, 5'-AGACAGGTGGTGGGGCAGGAC-3'; Apoa4 reverse, 5'-GCCCTCAGCCCATCACAGCAG-3'; Akr1e1 forward, 5'-CAAGGAGGGCGTGGTGAAGAG-3'; Akr1e1 reverse, 5'-GCTGGTGTGACTGGGTATGAC-3'; Cish forward, 5'-GGTGGGGCACAACATAGAGA-3'; Cish reverse, 5'-GGTGGCCAGACAGACAGGAG-3'; Socs2 forward, 5'-GGAATGGGACTGTTCACCTG-3'; Socs2 reverse, 5' GCAGAGTGGGTGCTGATGTA-3'.
The following additional data are available with the online version of this paper. Additional data file 1 is a Microsoft Excel file containing the normalized log ratios, F-values, q-values, and mean squares for the 2,382 genes assessed in the unpooled analysis. Additional data file 2 contains two supplemental tables and three supplemental figures. Supplemental Table 1 shows the analysis of variance for the mixed effect model. Supplemental Table 2 shows selected genes with variable expression within mouse strains. Supplemental Figure 1, titled 'F-values are independent of intensity', shows plots of F-values versus intensity for each of the 2,383 genes analyzed both within and between strains. Supplemental Figure 2, titled 'Quantitative RT-PCR analysis on replicate mice confirms the expression variability patterns of ApoA4, Dnase2a and Socs2', shows transcript abundance measurements from independent RNA preparations from the same liver samples compared across different mice of different strains. Supplemental Figure 3, titled 'Comparisons of variances associated with array, mouse, and strain', shows the numbers of variable genes at specific average fold-changes across different mouse strains.
We thank Barbara Trask and Catherine Peichel for critical reviews of this work and for helpful suggestions. We thank the microarray facility at the Fred Hutchinson Cancer Research Center. This work was supported by NIH grant DK65204, CA84294 and CA85859. CP was supported by a Poncin Scholarship and a Molecular Training Program in Cancer Research Fellowship (T32 CA09437).
- Lin MT, Storer B, Martin PJ, Tseng LH, Gooley T, Chen PJ, Hansen JA: Relation of an interleukin-10 promoter polymorphism to graft-versus-host disease and survival after hematopoietic-cell transplantation. N Engl J Med. 2003, 349: 2201-2210. 10.1056/NEJMoa022060.PubMedView ArticleGoogle Scholar
- Sachse C, Brockmoller J, Bauer S, Roots I: Cytochrome P450 2D6 variants in a Caucasian population: allele frequencies and phenotypic consequences [see comments]. Am J Hum Genet. 1997, 60: 284-295.PubMedPubMed CentralGoogle Scholar
- Schena M, Shalon D, Davis RW, Brown PO: Quantitative monitoring of gene expression patterns with a complementary DNA microarray. Science. 1995, 270: 467-470.PubMedView ArticleGoogle Scholar
- Brem RB, Yvert G, Clinton R, Kruglyak L: Genetic dissection of transcriptional regulation in budding yeast. Science. 2002, 296: 752-755. 10.1126/science.1069516.PubMedView ArticleGoogle Scholar
- Jin W, Riley RM, Wolfinger RD, White KP, Passador-Gurgel G, Gibson G: The contributions of sex, genotype and age to transcriptional variance in Drosophila melanogaster. Nat Genet. 2001, 29: 389-395. 10.1038/ng766.PubMedView ArticleGoogle Scholar
- Oleksiak MF, Churchill GA, Crawford DL: Variation in gene expression within and among natural populations. Nat Genet. 2002, 32: 261-266. 10.1038/ng983.PubMedView ArticleGoogle Scholar
- Pritchard CC, Hsu L, Delrow J, Nelson PS: Project normal: defining normal variance in mouse gene expression. Proc Natl Acad Sci USA. 2001, 98: 13266-13271. 10.1073/pnas.221465998.PubMedPubMed CentralView ArticleGoogle Scholar
- Cheung VG, Conlin LK, Weber TM, Arcaro M, Jen KY, Morley M, Spielman RS: Natural variation in human gene expression assessed in lymphoblastoid cells. Nat Genet. 2003, 33: 422-425. 10.1038/ng1094.PubMedView ArticleGoogle Scholar
- Phelan JP, Austad SN: Selecting animal models of human aging: inbred strains often exhibit less biological uniformity than F1 hybrids. J Gerontol. 1994, 49: B1-11.PubMedView ArticleGoogle Scholar
- Healey SC, Kirk KM, Hyland VJ, Munns CF, Henders AK, Batch JA, Heath AC, Martin NG, Glass IA: Height discordance in monozygotic females is not attributable to discordant inactivation of X-linked stature determining genes. Twin Res. 2001, 4: 19-24. 10.1375/1369052012100.PubMedView ArticleGoogle Scholar
- Bailey DW: How pure are inbred strains of mice?. Immunol Today. 1982, 3: 210-214. 10.1016/0167-5699(82)90093-7.PubMedView ArticleGoogle Scholar
- Storey JD, Tibshirani R: Statistical significance for genomewide studies. Proc Natl Acad Sci USA. 2003, 100: 9440-9445. 10.1073/pnas.1530509100.PubMedPubMed CentralView ArticleGoogle Scholar
- Williams SC, Grant SG, Reue K, Carrasquillo B, Lusis AJ, Kinniburgh AJ: cis-acting determinants of basal and lipid-regulated apolipoprotein A-IV expression in mice. J Biol Chem. 1989, 264: 19009-19016.PubMedGoogle Scholar
- Reue K, Purcell-Huynh DA, Leete TH, Doolittle MH, Durstenfeld A, Lusis AJ: Genetic variation in mouse apolipoprotein A-IV expression is determined pre- and post-transcriptionally. J Lipid Res. 1993, 34: 893-903.PubMedGoogle Scholar
- Wong WM, Hawe E, Li LK, Miller GJ, Nicaud V, Pennacchio LA, Humphries SE, Talmud PJ: Apolipoprotein AIV gene variant S347 is associated with increased risk of coronary heart disease and lower plasma apolipoprotein AIV levels. Circ Res. 2003, 92: 969-975. 10.1161/01.RES.0000069688.94567.7A.PubMedView ArticleGoogle Scholar
- Clothier B, Robinson S, Akhtar RA, Francis JE, Peters TJ, Raja K, Smith AG: Genetic variation of basal iron status, ferritin and iron regulatory protein in mice: potential for modulation of oxidative stress. Biochem Pharmacol. 2000, 59: 115-122. 10.1016/S0006-2952(99)00306-8.PubMedView ArticleGoogle Scholar
- Beck JA, Lloyd S, Hafezparast M, Lennon-Pierce M, Eppig JT, Festing MF, Fisher EM: Genealogies of mouse inbred strains. Nat Genet. 2000, 24: 23-25. 10.1038/71641.PubMedView ArticleGoogle Scholar
- Turk R, t Hoen PA, Sterrenburg E, de Menezes RX, de Meijer EJ, Boer JM, van Ommen GJ, den Dunnen JT: Gene expression variation between mouse inbred strains. BMC Genomics. 2004, 5: 57-10.1186/1471-2164-5-57.PubMedPubMed CentralView ArticleGoogle Scholar
- Fernandes C, Paya-Cano JL, Sluyter F, D'Souza U, Plomin R, Schalkwyk LC: Hippocampal gene expression profiling across eight mouse inbred strains: towards understanding the molecular basis for behaviour. Eur J Neurosci. 2004, 19: 2576-2582. 10.1111/j.0953-816X.2004.03358.x.PubMedView ArticleGoogle Scholar
- Carter TA, Del Rio JA, Greenhall JA, Latronica ML, Lockhart DJ, Barlow C: Chipping away at complex behavior: transcriptome/phenotype correlations in the mouse brain. Physiol Behav. 2001, 73: 849-857. 10.1016/S0031-9384(01)00522-4.PubMedView ArticleGoogle Scholar
- Whitney AR, Diehn M, Popper SJ, Alizadeh AA, Boldrick JC, Relman DA, Brown PO: Individuality and variation in gene expression patterns in human blood. Proc Natl Acad Sci USA. 2003, 100: 1896-1901. 10.1073/pnas.252784499.PubMedPubMed CentralView ArticleGoogle Scholar
- Schadt EE, Monks SA, Drake TA, Lusis AJ, Che N, Colinayo V, Ruff TG, Milligan SB, Lamb JR, Cavet G, et al: Genetics of gene expression surveyed in maize, mouse and man. Nature. 2003, 422: 297-302. 10.1038/nature01434.PubMedView ArticleGoogle Scholar
- Enard W, Khaitovich P, Klose J, Zollner S, Heissig F, Giavalisco P, Nieselt-Struwe K, Muchmore E, Varki A, Ravid R, et al: Intra- and interspecific variation in primate gene expression patterns. Science. 2002, 296: 340-343. 10.1126/science.1068996.PubMedView ArticleGoogle Scholar
- Elowitz MB, Levine AJ, Siggia ED, Swain PS: Stochastic gene expression in a single cell. Science. 2002, 297: 1183-1186. 10.1126/science.1070919.PubMedView ArticleGoogle Scholar
- Ozbudak EM, Thattai M, Kurtser I, Grossman AD, van Oudenaarden A: Regulation of noise in the expression of a single gene. Nat Genet. 2002, 31: 69-73. 10.1038/ng869.PubMedView ArticleGoogle Scholar
- Lai EC: Notch signaling: control of cell communication and cell fate. Development. 2004, 131: 965-973. 10.1242/dev.01074.PubMedView ArticleGoogle Scholar
- Rakyan VK, Chong S, Champ ME, Cuthbert PC, Morgan HD, Luu KV, Whitelaw E: Transgenerational inheritance of epigenetic states at the murine Axin(Fu) allele occurs after maternal and paternal transmission. Proc Natl Acad Sci USA. 2003, 100: 2538-2543. 10.1073/pnas.0436776100.PubMedPubMed CentralView ArticleGoogle Scholar
- Fraga MF, Ballestar E, Paz MF, Ropero S, Setien F, Ballestar ML, Heine-Suner D, Cigudosa JC, Urioste M, Benitez J, et al: From the cover: Epigenetic differences arise during the lifetime of monozygotic twins. Proc Natl Acad Sci USA. 2005, 102: 10604-10609. 10.1073/pnas.0500398102.PubMedPubMed CentralView ArticleGoogle Scholar
- Bennett-Baker PE, Wilkowski J, Burke DT: Age-associated activation of epigenetically repressed genes in the mouse. Genetics. 2003, 165: 2055-2062.PubMedPubMed CentralGoogle Scholar
- Wolff GL: Variability in gene expression and tumor formation within genetically homogeneous animal populations in bioassays. Fundam Appl Toxicol. 1996, 29: 176-184. 10.1006/faat.1996.0019.PubMedView ArticleGoogle Scholar
- Martin GM: Epigenetic drift in aging identical twins. Proc Natl Acad Sci USA. 2005, 102: 10413-10414. 10.1073/pnas.0504743102.PubMedPubMed CentralView ArticleGoogle Scholar
- Wolff GL: Some genetic aspects of physiological variability. Cancer Res. 1961, 21: 1119-1123.PubMedGoogle Scholar
- Dudoit S, Yang Y, Callow M, Speed T: Technical report. 2000, Berkeley: Department of Statistics, University of California at BerkeleyGoogle Scholar
- ArrayExpress Database. [http://www.ebi.ac.uk/arrayexpress]
This article is published under license to BioMed Central Ltd. This is an open access article ditributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribtion, and reproduction in any medium, provided the original work is properly cited.