In Chapter 3, we investigated the response of target geneexpression to direct recruitment and de-recruitment of four different CRs. Remarkably, all four CRs demonstrated similar all-or-none event profiles during silencing and reactivation, while displaying a diverse range of silencing and reactivation rates, and types of epigenetic memory. How may these differ- ences be utilized in developmental contexts? During early embryonic development, when cells in the epiblast differentiate into cells in the three germ layers, pluripotency-associated genes are down-regulated, while lineage-associated genes begin to express. In particular, genes that are associated with self-renewal and can induce dedifferentiation, such as Oct4, needs to be permanently silenced and safeguarded against re-expression in somatic cells. Recent research in ES cell differentiation, in response to retinoic acid, showed that this safeguard may be achieved with a multi-step mechanism . This process involves a series of regulations in the following sequence: transcriptional repression, G9a-dependent methylation of H3K9, removal of H3K4 methylation and histone acetylations, and, finally, Dnmt3a/3b-dependent methylation of DNA. Interestingly, our results demonstrated that KRAB-mediated silencing (associated with H3K9 methylation) is rapid yet only partially permanent, while Dnmt3b-mediated silencing (associated with DNA methylation) is perma- nent but slow. Combining different chromatin modifications and regulators may therefore represent a unique strategy for rapid and permanent gene silencing.
Although the genes of S100 proteins are located in a cluster, there is no evidence that their expression is by any means synchronized either in a cell-specific or develop- mental manner. Quite the opposite—there are many reports showing that in a given cell type, a certain S100 protein may be abundant while the one encoded by a neighboring gene is expressed at a low level or absent. Therefore, studies which compared the expression of a panel of S100 proteins in a given cell type or tissue, or in a set of normal versus cancerous tissues, led to the conclusion that, despite structural similarities and clustered genes, each S100 protein has a very specific expression pattern (Pedrocchi et al. 1994; Elder and Zhao 2002; Cross et al. 2005). Another interesting feature of S100 proteins is that expression of an individual protein may be completely different between cell lines, even those derived from related sources. Attempts aimed at identifying cell-specific tran- scription factors that would underlie this phenomenon have failed because exogenously introduced promoter constructs appeared to be equally active in cells differing in endogenous expression of a given S100 protein (Tulchinsky et al. 1992; Wicki et al. 1997; Lesniak et al. 2000). These observations turned the attention to epigenetic factors that could be involved in the control of S100 protein expression.
Among others, one objective of RNA-seq experiments is to characterize transcriptional differences between pre- specified populations of cells (given by experimental con- ditions or cell types). This is a key step for understanding a cell’s fate and functionality. In the context of bulk RNA- seq, two popular methods for this purpose are edgeR  and DESeq2 . However, these are not designed to cap- ture features that are specific to scRNA-seq data sets. In contrast, SCDE  has been specifically developed to deal with scRNA-seq data sets. All of these meth- ods target the detection of differentially expressed genes based on log-fold changes (LFCs) of overall expression between the populations. However, restricting the anal- ysis to changes in overall expression does not take full advantage of the rich information provided by scRNA-seq. In particular – and unlike bulk RNA-seq – scRNA-seq can also reveal information about cell-to-cellexpressionheterogeneity. Critically, traditional approaches will fail to highlight genes whose expression is less stable in any given population but whose overall expression remains unchanged between populations.
levels of these markers, NPCb and NPCd had lost CITED1 almost completely, while retaining some SIX2 expression. In an immunostaining of a w15 kidney, CITED1 and SIX2 appeared overlapping in a subset of cells (Fig 5B–5D). Quantification of the fluorescence signal (Meth- ods) revealed clear differences between their expression patterns. Whereas SIX2 expression was approximately constant throughout the CM, CITED1 expression decreased, relative to SIX2, with increasing (radial) distance from the UB (Fig 5D). A marked drop of CITED1 was visible between 10 and 20 μm from the UB, which approximately corresponds to the first layer of cells. To exclude that the observed difference between SIX2 and CITED1 expression was due to the different fluorophores on the secondary antibodies, we repeated the experiment with swapped fluorophores. This measurement produced a very similar expression gradient (S9A Fig). To exclude that the observed effect was influenced by PTA found in the CM towards the stalk of the UB, the analysis was repeated, taking only the 20% of CM cells closest to the edge of the cortex into account. A similar expression gradient was observed (S9C Fig). This result implies the existence of a CITED1 low/SIX2 high subpopulation of cells, which are not in contact with the UB. Secondly, we observed that CITED1 decreased relative to SIX2 towards the interface with the PTA and the stalk of the UB (Fig 5D). A similar observation was made when the experiment was repeated with swapped fluorophores (S9B Fig). Taken together, these results suggested that NPCa and NPCc were located closer to the surface of the UB and closer to the tip of the UB compared to the other NPC subtypes. Additionally, we also observed differences in subcellular localization of CITED1 protein within the CITED1 high compart- ment. Whereas for the majority of cells CITED1 was found in the cytoplasm, in several cells it was concentrated in the nucleus (right inset in Fig 5C). In contrast, SIX2 was always found restricted to the nucleus (left inset in Fig 5C). This observation might indicate that CITED1 was only active in a small population of cells, which would constitute another layer of cell–cellheterogeneity.
Although at least 300 genes are considered to be gametocyte speciﬁc in P. falci- parum, their roles in male and female development have not yet been fully deﬁned (13–18). Plasmodium berghei and P. falciparum gender-speciﬁc ﬂow sorting studies have revealed late-stage markers for male and female gametocytes, but these studies are based on speciﬁc reporter genes and are therefore biased for late stages (19–21). In particular, the recent ﬂow sorting study using P. falciparum represents the ﬁrst tran- scriptome analysis of male and female gametocytes (20). However, the gender-speciﬁc expression of some genes is still debated (22). We hypothesize that some of the controversies about gender-speciﬁc expression may result from reliance on population analyses of mixed gametocytes that include multiple differentiation stages and tem- poral changes in geneexpression. These issues can best be resolved using single-cell isolation and expression analyses at distinct time points to unequivocally decipher sex-speciﬁc transcripts that may ultimately determine male or female fate. Recently, single-cell RNA sequencing revealed that sexually committed schizonts have a distinct program of geneexpression (23). Here, we describe our efforts using a single-cell approach to deﬁne male and female gametocyte geneexpression in an unbiased manner. Our study incorporates the ﬁrst use of the Fluidigm C1 system for microﬂuidic capture of single gametocytes, followed by real-time PCR (RT-PCR) quantitation of their sex-speciﬁc expression of gametocyte genes on the Biomark HD system. The analysis of stage III through stage V gametocytes separates parasites by gender rather than stage and reveals a number of new candidate genes for male and female development. Additionally, a large female population reveals unexpected cellular heterogeneity among single cells, previously undetected on a population level. Therefore, our study highlights the power of single-cell transcriptome analysis in dissecting the sex-speciﬁc geneexpression of P. falciparum.
The near future will undoubtedly witness the applica- tion of single-cell epigenomic approaches in vivo. For example, mouse zygotes fertilized in vitro and embryos resulting from natural matings will be studied to under- stand epigenome dynamics during this critical stage of development. Due to the low cell numbers associated with these samples, FACS isolation of single cells is in- feasible, so single cells will be manually picked after em- bryo dissociation. For the early stages of development, it should be possible to study every cell isolated from an embryo, while at later time points (E6.5 onwards) the in- creasing cell number may necessitate focused studies on specific cell lineages or on representative subpopulations of each lineage. A limitation to these studies will be the loss of spatial information upon embryo dissociation. Complementary studies including in vivo imaging of lineage-specific genes will be used to map cell types identified by single-cell sequencing back to the three- dimensional embryo . By employing single-cell multi-omics, these studies will reveal the fundamental processes of cell-fate specification and establish an atlas of differentiation in which every tissue type can be traced back to its embryonic origins. This information will bring light to one of the most fascinating processes of biology, clarifying key questions such as whether cell- type-specific epigenetic marks are established during lineage priming prior to cell-fate commitment.
Po is an integral membrane glycoprotein o f molecular weight 28 kDa. The human Po gene, MPZ, was shown to map to chromosome Iq22-q23 (Hayasaka et at., 1993) and the rat Po gene has been isolated, sequenced and analyzed (Lemke and Axel, 1985; Lemke et al., 1988). In rats and mice, this gene consists o f six exons distributed over 7kb o f DNA. The topographic arrangement o f these exons in the genome is consistent with the functional segregation o f the Po protein into extracellular, membrane- spanning, and cytoplasmic domains (Lemke et al., 1988). The primary amino acid struture o f Po has been deduced from cloned cDNAs (Lemke and Axel, 1985) and bovine Po has been directly determined by protein sequencing (Sakamoto et al., 1987). It consists o f a single membrane spanning region, a large extracellular hydrophobic region and small basic cytoplasmic region (Figure 1-1). Being a primitive member o f the immunoblobulin-related protein family, the extracellular domain o f Po is highly glycosylated (see below) and is responsible for the homophilic adhesion properties o f Po and the formation o f the intraperiod line o f compacted myelin (Figure 1-2). In vitro the recombinant extracellular domain is capable o f forming tetramers and dimers, and the same interaction is believed to play an essential role in forming a network to secure the adhesion between apposing Schwann cell membranes (Shapiro et al., 1996). Po molecules are believed to interact via electrostatic charges with the negatively charged lipid bilayer o f the opposing memebrane via its basic intracellular domain in order to form the major dense line in myelin (Ding and Bruden, 1994) (Figure 1-2). Apart from the secondary structure, various post-translational modifications are also believed to contribute to Po function. The Po molecule undergoes many post-translational modifications, including glycosylation (Kitamura et al., 1976; Matthieu et al., 1975; Quarles, 1980), phosphorylation (Brunden and Poduslo, 1987b; Singh and Spritz 1976), sulphation (Matthieu et al., 1975), and acylation (Agrawal et al., 1983). Among these modifications, HNK-1-reacting glycans and mannose-rich N-glycans are potentially involved in cell adhesion (Bollensen et al., 1987; Griffith et al., 1992; Brunden, 1992).
We have performed an extensive analysis of genotype robustness and evolvability using mathe- matical models of two common gene regulatory networks involving a singlegene under auto- (or transcription factor) mediated regulation. We have used the expression levels of this gene as a phenotype and the system parameters controlling expressionlevel as the genotype. Defin- ing several complementary measures for genotype robustness and evolvability under mutations of different size, we have evaluated these properties for several million genotypes for each net- work architecture. This analysis revealed that for most genotypes, robustness and evolvability display a negative correlation, but there exist a significant number of genotypes for which this trade-off can be broken. This observation holds for all the combinations of the different mea- sures utilized. Furthermore, the identified robust and evolvable genotypes using these fitness- independent measures are also found to emerge under in silico evolution when selection schemes that are shown to facilitate adaptation time are used. This suggests that our fitness- independent measures applied to a genotype-phenotype map are then able to identify geno- types that are evolvable in a population dynamics context and using fitness functions based on that same phenotype (such as adaptation time, or performing well in fluctuating environments, section 4 of the Results). Thus we conclude that the fitness-dependent and the fitness-indepen- dent view on evolvability need not be mutually exclusive.
Transcriptional heterogeneity. The better-understood aspect of noisy gene ex- pression is transcriptional heterogeneity. Experimental evidence for geneexpression noise within a population was ﬁrst revealed in bacterial cells (42, 43). Ozbudak et al. showed that the expression levels of a ﬂuorescent protein differ from cell to cell within a population of genetically identical Bacillus subtilis cells (43). Using two ﬂuorescence reporters controlled by identical promoters in E. coli, Elowitz et al. found that promoter activity is heterogeneous among cells and is stochastic within the cell, particularly when the transcription level is low (42). Both of those studies used protein ﬂuorescence as the readout for geneexpression, and the overall heterogeneity of ﬂuorescence intensity reﬂected the cumulative noise from transcription, mRNA degradation, translation, protein degradation, and ﬂuorophore maturation. To speciﬁcally study transcriptional noise, a breakthrough came from the use of MS2-green ﬂuorescent protein (MS2-GFP) to directly count the number of stable mRNA molecules carrying the MS2 binding sites in E. coli (44). Subsequent studies revealed that transcription initiation does not occur continuously but rather as bursts (45, 46). Variations in promoter activity are large contributors to variations in single-cellgeneexpression. In 2012, a study characterized the heterogeneity of every known promoter in E. coli and found that different promot- ers show different levels of heterogeneity in a population (47). Some categories of promoters, such as stress response promoters, are noisier than others (47). Heteroge- neity of geneexpression was initially thought to be a consequence of the stochastic nature of molecular interactions (42). However, recent analyses of the evolution of synthetic promoters de novo revealed that the heterogeneity of promoter expression is low by default (48). This ﬁnding indicates that the high levels of heterogeneity seen in some promoters may have evolved as a beneﬁcial mechanism. Future investigations into the regulation of promoter heterogeneity and evolution of these systems may provide insights into the role and beneﬁts of transcriptional heterogeneity in bacterial populations.
Studies in rodent models pointed out that the tran- scriptional factor forkhead box protein N1 (FOXN1) is both necessary and seemingly sufficient to induce differentiation of functional TEC [17, 18]. FOXN1 ap- pears on day 11 during mouse embryonic development, the sixth week of gestation in humans, and induces the thymic organogenesis program presumably under the control of WNT family of glycoproteins, namely, by WNT-4 [2, 19, 20]. In a model with inducible Cre medi- ated deletion of an SV40 driven transgenic hypomorphic Foxn1 allele, it has been demonstrated, that FOXN1 in TEC induces the expression of MHC II, CD40, PAX1, cathepsin-L, the chemokine CCL25 and the NOTCH ligand Delta-like 4 (DLL4), thus highlighting its orches- trating role in T cell maturation . The lack of FOXN1 in mice and rats results in the absence or the incomplete development of TEC and the thymic epithelial mesh, com- bined with severe immunodeficiency known as the nude phenotype . Nude mice carry a single base pair dele- tion at exon 3 of the Foxn1 gene, which results in aberrant protein production, lacking the DNA-binding and the transcription activation domains, necessary for FOXN1 protein function [23, 24]. Similar phenotype was found in human, carrying a rare non-sense mutation at the residue 255 of the FOXN1 protein, resulting from a single base substitution in exon 5 of the FOXN1 gene . Recently, Bredenkamp and co-workers showed that mouse embry- onic fibroblasts transfected with inducible Foxn1 transdif- ferentiated to functional TECs upon induction . These Foxn1 induced TECs support T cell development in vitro and in vivo. The data clearly demonstrate the central role
The “central dogma” of developmental biology holds that development is underwritten by regulated changes in geneexpression. In order to study this process, we need techniques for measuring how gene activity patterns vary within the developing organism. Since most genes encode messenger RNAs which are translated into proteins, expression can be assessed at either the RNA or the protein level. In principle, protein quantitation is more informative, because only a gene’s peptide end product directly affects cell function. As translation and protein degradation can be loci for gene-specific regulation, RNA assays cannot capture all the expression changes which attend development. This is mitigated by the fact that post- transcriptional control is often effected by transcriptional modulation of regulatory genes. Moreover, it has been discovered recently that eukaryotic genomes harbor large, previously unrecognized classes of untranslated genes, such as microRNAs and shRNAs [1-3]. These genes encode short RNAs which modulate translation, mRNA degradation, epigenetic silencing, and perhaps still other processes; many seem to be involved in developmental processes. It seems likely, then, that most of the gene-expression changes involved in development are associated with measurable changes in the level of cellular RNAs.
concentration of these enzymes to survive. Interestingly, when controlling for the basic cellular environment, by normalizing copy numbers of the glycolysis genes to β-actin geneexpression, noise is seen to increase for the gene ratios as compared to the raw copy number values. However, the noise in the ratio groups decreases in G1 as compared to G2 and the heterogenous sample (Fig. 5.5). Furthermore in all three groups, the ratios’ coefficients of variation are higher than the noise measured for the glycolysis genes alone. Although the total noise in all three genes goes down in G2, the ratios of geneexpression increase most likely due to the genes being active in distinct cellular processes, with intracellular variability non-deleterious if copy numbers stay above a certain threshold. The cell probably makes more than enough of the three genes studied and employs regulation before divison. This mechanism could decrease noise at the protein level (e.g. high copy numbers are the limiting case for translational noise) for the housekeeping genes, with excess transcripts always available for translation, thus
associated with worse prognosis. Authors also found that DNA methylation data could predict various types of im- mune cells in the primary and recurring tumors . De- veloping a novel single-cell technology, Pi-ATAC, which simultaneously measures protein epitopes and active DNA regulatory elements of the same individual cell, Chen et al. found epigenetic variability of tumor cells is linked to the hypoxic tumor microenvironment . By genome-wide methylotyping analysis, Tanas et al. divided breast cancer into six breast cancer methylotypes, and found that the ma- jority of CpG islands appeared to be more densely hyper- methylated in breast cancer cell lines than in primary tumors . Using an epigenome-wide sequencing ap- proach, Grasse et al. observed that aberrantly methylated regions in the PDX tumors were reflected in the corre- sponding primary NSCLC tumors, albeit the levels of differ- ential methylation of the PDX samples were much higher compared to the levels within the primary tumors . Mutations in epigenetic modifier genes, such as SETD2 and DNMT3A, are strongest determinants of ITH amongst a panel of 17 distinct cellular pathways . Epigenetic reg- ulators such as histone modifying enzymes are critical for the establishment of cell-type-specific geneexpression pat- terns, thus, they are also likely to play a role in modulating cell-to-cell variability in transcription. The distinct epigen- etic state of the cells could determine cellular response to treatment . Lysine demethylase 5 (KDM5) was found to be a regulator of cellular transcriptomic heterogeneity in ER + luminal breast cancer, and inhibiting KDM5 activity could decrease resistance to cancer therapies . Pastore et al. suggested that intratumoral epigenetic diversity may permit leukemic cells to stochastically activate alternate gene regulatory programs, facilitating the emergence of novel cell sates, ultimately fostering CLL’s ability to effi- ciently explore the fitness landscape for superior evolution- ary trajectories during tumorigenesis and in response to therapy .
plasmid into pseudo attP site in mammalian genomes (2). PhiC31 integrase system is considered as a specific tool for gene therapy (3, 4) and transgenic research (2, 5). The efficiency of phiC31-integrase has been indicated to be comparable with that of the widely used Cre/loxP system. Furthermore, flippase (FLP) recombinase shows only 10% recombination activity on chromosomal targets in comparison with Cre recombinase (6). Cre and FLP cause deletion of the gene after integration (7) whereas phiC31 integrase can catalyze unidirectional and irreversible recombination between attB and pseudo attP sites (3). Development of phiC31 integrase-based vectors for prolonged therapeutic geneexpression, demonstrated that it is a robust and reliable gene delivery system (4, 8). Sodium butyrate (NaBut) treatment increases the specific productivity of recombinant proteins in mammalian cells; but, it declines cell growth and can provoke apoptosis (9). NaBut inhibits the activity of many histone deacetylases, induces hyperacetylation of histones. Histone acetylation could modify chromatin structure, lead to transcription factors and polymerases binding as well as improving geneexpression (10). Due to its impact on epigenetic mechanisms, NaBut has attracted many interest for the prevention and treatment of different diseases such as genetic/metabolic conditions and neurological degenerative disorders (11). Valproic acid (VPA), a histone deacetylase inhibitor (HDACi), can cause impaired epigenetic modification and suppress cell growth (12). It can increase the expression of genes that are regulated by transcription factors (13). It has been indicated that the HDACi increases both the specific productivity and mRNA transcription level in stable CHO cell lines. Furthermore, no cellular toxicity was reported with VPA compared with other widely used HDACi such as NaBut (14). Blinatumumab, the most advanced bispecific T-cell engager (BiTE) with dual binding specificities (15),
For single-cell RNA-seq data, adapters were trimmed from reads using Trim Galore! [27–29], using default settings. Trimmed reads were mapped to the human reference genome build 37 using STAR  (version: 020201) in two-pass alignment mode, using the defaults proposed by the ENCODE consortium (STAR manual). Expression quantification was performed separately using Salmon  (version: 0.8.2), using the “--seqBias,” “--gcBias,” and “VBOpt” options on transcripts derived from ENSEMBL 75. Transcript-levelexpression values were summarized at the genelevel (estimated counts) and quality control of scRNA-seq data was performed using scater . Cells with the following features were retained for analysis: (i) at least 50,000 counts from en- dogenous genes, (ii) at least 5000 genes with non-zero expression, (iii) less than 90% of counts are assigned to the top 100 expressed genes per cell, (iv) less than 20% of counts are assigned to ERCC spike-in sequences, and (v) a Salmon mapping rate of at least 40%. These filters jointly removed 9 iPS cells and 36 endoderm cells from our analysis.
Many genes are expressed in bursts, which can contribute to cell-to-cellheterogeneity. It is now possible to measure this heterogeneity with high throughput singlecellgene expres- sion assays (singlecell qPCR and RNA-seq). These experimental approaches generate geneexpression distributions which can be used to estimate the kinetic parameters of geneexpression bursting, namely the rate that genes turn on, the rate that genes turn off, and the rate of transcription. We construct a complete pipeline for the analysis of singlecell qPCR data that uses the mathematics behind bursty expression to develop more accurate and robust algorithms for analyzing the origin of heterogeneity in experimental samples, specifi- cally an algorithm for clustering cells by their bursting behavior (Simulated Annealing for Bursty Expression Clustering, SABEC) and a statistical tool for comparing the kinetic parameters of bursty expression across populations of cells (Estimation of Parameter changes in Kinetics, EPiK). We applied these methods to hematopoiesis, including a new singlecell dataset in which transcription factors (TFs) involved in the earliest branchpoint of blood differentiation were individually up- and down-regulated. We could identify two unique sub-populations within a seemingly homogenous group of hematopoietic stem cells. In addition, we could predict regulatory mechanisms controlling the expression levels of eigh- teen key hematopoietic transcription factors throughout differentiation. Detailed information about gene regulatory mechanisms can therefore be obtained simply from high throughput singlecellgeneexpression data, which should be widely applicable given the rapid expan- sion of singlecell genomics.
exclusive expression profile of IFN- γ and IL-4 in Th1 and Th2 cells, respectively. As naïve T cells differentiate to Th1 and Th2 cells, Th1 cells slowly lose their ability to express IL-10 after 1 week of differentiation, while Th2 cells show increased IL-10 production and maintain high IL-10 levels (Lee et al., unpublished result). These results suggest a dynamic chromatin remodeling on IL-10 locus during T cell differentiation. Although the biological function of IL- 10 in immune system has extensively been studied for decades, little information is available on the molecular mechanism of its transcriptional regulation, especially at the chromatin level. Attempts to identify cis-regulatory elements have been made using DNase I HSS mapping. In our previous work, we described the HSS and identified 6 HSS in CD4 + T cells. 89 Later, Wang et al. 90 showed that the CNS
Network modeling is increasingly recognized as a powerful tool for understanding complex biological sys- tems, including the hematopoietic system [55,56]. Efforts are underway to apply network-modeling approaches for the computational elucidation and analysis of single- cell data [20,25,30]. Here, we employed a co-expression network-based method (WGCNA, [50,51]) to analyze single-cellgeneexpression data, using the identified cellular hierarchy as a guide. Our analysis identified a core module that is common between GMPs and leuke- mia networks, and suggested that much of the gene ex- pression level changes between these two cell types can be viewed as a switch of allowable states within a common network module. On the other hand, we also identified significant differences between the networks. For example, Pbx1, which cooperates with Meis1 in leukemogenesis , is regulated by a separate module. As such, our ana- lysis demonstrates that network modeling provides mech- anistic insights into organizing principles of leukemia.
prototype of a gene with typical CpG features, with a large CpG island that spans the promoter and exon 1, encompassing the start site for transcription. CST6 is a member of a family of proteins that represent physiological inhibitors of lysosomal cysteine proteases that are expressed in normal and premalignant breast epithelium, but not in metastatic breast cancer cell lines (103). Ectopic expression of CST6 suppresses the neoplastic phenotype of MDA-MD-435S breast cancer cells, reducing their cell proliferation, migration, and invasion in vitro (113), and delaying tumor growth and reducing metastatic tumor burden in vivo (116). CST6 expression is significantly diminished in primary human breast cancers (116), which is unrelated to gene deletion (103) but may be due to transcriptional silencing through methylation of its CpG island (134). Our methylation analysis of the CST6 promoter shows that this gene is subject to DNA methylation in MCF-7 cells, and that there is an inverse correlation between CST6 expression and methylation of its promoter CpG island (Figure 9). These results strongly suggest that CST6, a putative breast cancer tumor suppressor gene (116), is sensitive to DNA methylation and that methylation-dependent epigenetic silencing may represent an important mechanism for loss of this gene during breast carcinogenesis and/or tumor progression.
The liver is the primary site for metabolism of nutrients, drugs and chemical agents. While metabolic pathways are complex and tightly regulated, genetic variation among individuals, reflected in variation in geneexpression levels, introduces complexity into research on liver disease. This study aimed to dissect genetic networks that control liver geneexpression by combining large-scale quantitative mRNA expression analysis with genetic mapping in a reference population of BXD recombinant inbred mouse strains for which extensive SNP, haplotype and phenotypic data is publicly available. We profiled geneexpression in livers of naive mice of both sexes from C57BL/6J, DBA/2J, B6D2F1, and 37 BXD strains using Agilent oligonucleotide microarrays. This data was used to map quantitative trait loci (QTLs) responsible for variation in expression of about 19,000 transcripts. We identified polymorphic cis- and trans-acting loci, including several loci that control expression of large numbers of genes in liver.