Tissue-specific nuclear architecture and gene expression regulated by SATB1

Eukaryotic chromosomes are packaged in nuclei by many orders of folding. Little is known about how higher-order chromatin packaging might affect gene expression. SATB1 is a cell-type specific nuclear protein that recruits chromatin-remodeling factors and regulates numerous genes during thymocyte differentiation. Here we show that in thymocyte nuclei, SATB1 has a cage-like 'network' distribution circumscribing heterochromatin and selectively tethers specialized DNA sequences onto its network. This was shown by fluorescence in situ hybridization on wild-type and Satb1-null thymocytes using in vivo SATB1-bound sequences as probes. Many gene loci, including that of Myc and a brain-specific gene, are anchored by the SATB1 network at specific genomic sites, and this phenomenon is precisely correlated with proper regulation of distant genes. Histone-modification analyses across a gene-enriched genomic region of 70 kb showed that acetylation of histone H3 at Lys9 and Lys14 peaks at the SATB1-binding site and extends over a region of roughly 10 kb covering genes regulated by SATB1. By contrast, in Satb1-null thymocytes, this site is marked by methylation at H3 Lys9. We propose SATB1 as a new type of gene regulator with a novel nuclear architecture, providing sites for tissue-specific organization of DNA sequences and regulating region-specific histone modification.     

Stochastic switching as a survival strategy in fluctuating environments

A classic problem in population and evolutionary biology is to understand how a population optimizes its fitness in fluctuating environments. A population might enhance its fitness by allowing individual cells to stochastically transition among multiple phenotypes, thus ensuring that some cells are always prepared for an unforeseen environmental fluctuation. Here we experimentally explore how switching affects population growth by using the galactose utilization network of Saccharomyces cerevisiae. We engineered a strain that randomly transitions between two phenotypes as a result of stochastic gene expression. Each phenotype was designed to confer a growth advantage over the other phenotype in a certain environment. When we compared the growth of two populations with different switching rates, we found that fast-switching populations outgrow slow switchers when the environment fluctuates rapidly, whereas slow-switching phenotypes outgrow fast switchers when the environment changes rarely. These results suggest that cells may tune inter-phenotype switching rates to the frequency of environmental changes.     

PDGF signaling specificity is mediated through multiple immediate early genes

Growth factor signaling leads to the induction or repression of immediate early genes, but how these genes act collectively as effectors of downstream processes remains unresolved. We have used gene trap-coupled microarray analysis to identify and mutate multiple platelet-derived growth factor (PDGF) intermediate early genes in mice. Mutations in these genes lead to a high frequency of phenotypes that affect the same cell types and processes as those controlled by the PDGF pathway. We conclude that these genes form a network that controls specific processes downstream of PDGF signaling.     

Screening a large reference sample to identify very low frequency sequence variants: comparisons between two genes

Most human sequence variation is in the form of single-nucleotide polymorphisms (SNPs). It has been proposed that coding-region SNPs (cSNPs) be used for direct association studies to determine the genetic basis of complex traits. The success of such studies depends on the frequency of disease-associated alleles, and their distribution in different ethnic populations. If disease-associated alleles are frequent in most populations, then direct genotyping of candidate variants could show robust associations in manageable study samples. This approach is less feasible if the genetic risk from a given candidate gene is due to many infrequent alleles. Previous studies of several genes demonstrated that most variants are relatively infrequent (<0.05). These surveys genotyped small samples (n<75) and thus had limited ability to identify rare alleles. Here we evaluate the prevalence and distribution of such rare alleles by genotyping an ethnically diverse reference sample that is more than six times larger than those used in previous studies (n=450). We screened for variants in the complete coding sequence and intron-exon junctions of two candidate genes for neuropsychiatric phenotypes: SLC6A4, encoding the serotonin transporter; and SLC18A2, encoding the vesicular monoamine transporter. Both genes have unique roles in neuronal transmission, and variants in either gene might be associated with neurobehavioral phenotypes.     

Network modeling links breast cancer susceptibility and centrosome dysfunction

Many cancer-associated genes remain to be identified to clarify the underlying molecular mechanisms of cancer susceptibility and progression. Better understanding is also required of how mutations in cancer genes affect their products in the context of complex cellular networks. Here we have used a network modeling strategy to identify genes potentially associated with higher risk of breast cancer. Starting with four known genes encoding tumor suppressors of breast cancer, we combined gene expression profiling with functional genomic and proteomic (or 'omic') data from various species to generate a network containing 118 genes linked by 866 potential functional associations. This network shows higher connectivity than expected by chance, suggesting that its components function in biologically related pathways. One of the components of the network is HMMR, encoding a centrosome subunit, for which we demonstrate previously unknown functional associations with the breast cancer-associated gene BRCA1. Two case-control studies of incident breast cancer indicate that the HMMR locus is associated with higher risk of breast cancer in humans. Our network modeling strategy should be useful for the discovery of additional cancer-associated genes.     

The gene mutated in bare patches and striated mice encodes a novel 3beta-hydroxysteroid dehydrogenase.

X-linked dominant disorders that are exclusively lethal prenatally in hemizygous males have been described in human and mouse. None of the genes responsible has been isolated in either species. The bare patches (Bpa) and striated (Str) mouse mutations were originally identified in female offspring of X-irradiated males. Subsequently, additional independent alleles were described. We have previously mapped these X-linked dominant, male-lethal mutations to an overlapping region of 600 kb that is homologous to human Xq28 (ref. 4) and identified several candidate genes in this interval. Here we report mutations in one of these genes, Nsdhl, encoding an NAD(P)H steroid dehydrogenase-like protein, in two independent Bpa and three independent Str alleles. Quantitative analysis of sterols from tissues of affected Bpa mice support a role for Nsdhl in cholesterol biosynthesis. Our results demonstrate that Bpa and Str are allelic mutations and identify the first mammalian locus associated with an X-linked dominant, male-lethal phenotype. They also expand the spectrum of phenotypes associated with abnormalities of cholesterol metabolism.     

Epistasis analysis with global transcriptional phenotypes

Classical epistasis analysis can determine the order of function of genes in pathways using morphological, biochemical and other phenotypes. It requires knowledge of the pathway's phenotypic output and a variety of experimental expertise and so is unsuitable for genome-scale analysis. Here we used microarray profiles of mutants as phenotypes for epistasis analysis. Considering genes that regulate activity of protein kinase A in Dictyostelium, we identified known and unknown epistatic relationships and reconstructed a genetic network with microarray phenotypes alone. This work shows that microarray data can provide a uniform, quantitative tool for large-scale genetic network analysis.     

Predicting phenotypic variation in yeast from individual genome sequences

A central challenge in genetics is to predict phenotypic variation from individual genome sequences. Here we construct and evaluate phenotypic predictions for 19 strains of Saccharomyces cerevisiae. We use conservation-based methods to predict the impact of protein-coding variation within genes on protein function. We then rank strains using a prediction score that measures the total sum of function-altering changes in different sets of genes reported to influence over 100 phenotypes in genome-wide loss-of-function screens. We evaluate our predictions by comparing them with the observed growth rate and efficiency of 15 strains tested across 20 conditions in quantitative experiments. The median predictive performance, as measured by ROC AUC, was 0.76, and predictions were more accurate when the genes reported to influence a trait were highly connected in a functional gene network.     

Identification of acquired somatic mutations in the gene encoding chromatin-remodeling factor ATRX in the alpha-thalassemia myelodysplasia syndrome (ATMDS)

Inherited mutations of specific genes have elucidated the normal roles of the proteins they encode by relating specific mutations to particular phenotypes. But many potentially informative mutations in such genes are lethal early in development. Consequently, inherited mutations may not reflect all the functional roles of such proteins. Acquired, somatic defects should reflect a wider spectrum of mutations because they are not prone to negative selection in development. It has been difficult to identify such mutations so far, but microarray analysis provides a new opportunity to do so. Using this approach, we have shown that in individuals with myelodysplasia associated with alpha-thalassemia (ATMDS), somatic mutations of the gene encoding the chromatin remodeling factor ATRX cause an unexpectedly severe hematological phenotype compared with the wide spectrum of inherited mutations affecting this gene. These findings cast new light on this pleiotropic cofactor, which appears to be an essential component rather than a mere facilitator of globin gene expression.     

Functional screening of an asthma QTL in YAC transgenic mice.

Many quantitative trait loci (QTLs) contributing to genetically complex conditions have been discovered, but few causative genes have been identified. This is mainly due to the large size of QTLs and the subtle connection between genotype and quantitative phenotype associated with these conditions. Transgenic mice have been successfully used to analyse well-characterized genes suspected of contributing to quantitative traits. Although this approach is powerful for examining one gene at a time, it can be impractical for surveying the large genomic intervals containing many genes that are typically associated with QTLs. To screen for genes contributing to an asthma QTL mapped to human chromosome 5q3 (refs 6,7), we characterized a panel of large-insert 5q31 transgenics based on studies demonstrating that altering gene dosage frequently affects quantitative phenotypes normally influenced by that gene. This panel of human YAC transgenics, propagating a 1-Mb interval of chromosome 5q31 containing 6 cytokine genes and 17 partially characterized genes, was screened for quantitative changes in several asthma-associated phenotypes. Multiple independent transgenic lines with altered IgE response to antigen treatment shared a 180-kb region containing 5 genes, including those encoding human interleukin 4 (IL4) and interleukin 13 (IL13 ), which induce IgE class switching in B cells. Further analysis of these mice and mice transgenic for mouse Il4 and Il13 demonstrated that moderate changes in Il4 and Il13 expression affect asthma-associated phenotypes in vivo. This functional screen of large-insert transgenics enabled us to identify genes that influence the QTL phenotype in vivo.     

Reverse engineering of regulatory networks in human B cells

Cellular phenotypes are determined by the differential activity of networks linking coregulated genes. Available methods for the reverse engineering of such networks from genome-wide expression profiles have been successful only in the analysis of lower eukaryotes with simple genomes. Using a new method called ARACNe (algorithm for the reconstruction of accurate cellular networks), we report the reconstruction of regulatory networks from expression profiles of human B cells. The results are suggestive a hierarchical, scale-free network, where a few highly interconnected genes (hubs) account for most of the interactions. Validation of the network against available data led to the identification of MYC as a major hub, which controls a network comprising known target genes as well as new ones, which were biochemically validated. The newly identified MYC targets include some major hubs. This approach can be generally useful for the analysis of normal and pathologic networks in mammalian cells.     

Clustering of housekeeping genes provides a unified model of gene order in the human genome

It is often supposed that, except for tandem duplicates, genes are randomly distributed throughout the human genome. However, recent analyses suggest that when all the genes expressed in a given tissue (notably placenta and skeletal muscle) are examined, these genes do not map to random locations but instead resolve to clusters. We have asked three questions: (i) is this clustering true for most tissues, or are these the exceptions; (ii) is any clustering simply the result of the expression of tandem duplicates and (iii) how, if at all, does this relate to the observed clustering of genes with high expression rates? We provide a unified model of gene clustering that explains the previous observations. We examined Serial Analysis of Gene Expression (SAGE) data for 14 tissues and found significant clustering, in each tissue, that persists even after the removal of tandem duplicates. We confirmed clustering by analysis of independent expressed-sequence tag (EST) data. We then tested the possibility that the human genome is organized into subregions, each specializing in genes needed in a given tissue. By comparing genes expressed in different tissues, we show that this is not the case: those genes that seem to be tissue-specific in their expression do not, as a rule, cluster. We report that genes that are expressed in most tissues (housekeeping genes) show strong clustering. In addition, we show that the apparent clustering of genes with high expression rates is a consequence of the clustering of housekeeping genes.     

Deletion in a gene associated with grain size increased yields during rice domestication

The domestication of crops involves a complex process of selection in plant evolution and is associated with changes in the DNA regulating agronomically important traits. Here we report the cloning of a newly identified QTL, qSW5 (QTL for seed width on chromosome 5), involved in the determination of grain width in rice. Through fine mapping, complementation testing and association analysis, we found that a deletion in qSW5 resulted in a significant increase in sink size owing to an increase in cell number in the outer glume of the rice flower; this trait might have been selected by ancient humans to increase yield of rice grains. In addition, we mapped two other defective functional nucleotide polymorphisms of rice domestication-related genes with genome-wide RFLP polymorphisms of various rice landraces. These analyses show that the qSW5 deletion had an important historical role in artificial selection, propagation of cultivation and natural crossings in rice domestication, and shed light on how the rice genome was domesticated.     

Evidence for co-evolution of gene order and recombination rate

There is increasing evidence in eukaryotic genomes that gene order is not random, even allowing for tandem duplication. Notably, in numerous genomes, genes of similar expression tend to be clustered. Are there other reasons for clustering of functionally similar genes? If genes are linked to enable genetic, rather than physical clustering, then we also expect that clusters of certain genes might be associated with blocks of reduced recombination rates. Here we show that, in yeast, essential genes are highly clustered and this clustering is independent of clustering of co-expressed genes and of tandem duplications. Adjacent pairs of essential genes are preferentially conserved through evolution. Notably, we also find that clusters of essential genes are in regions of low recombination and that larger clusters have lower recombination rates. These results suggest that selection acts to modify both the fine-scale intragenomic variation in the recombination rate and the distribution of genes and provide evidence for co-evolution of gene order and recombination rate.     

INTERMEDIUM-C, a modifier of lateral spikelet fertility in barley, is an ortholog of the maize domestication gene TEOSINTE BRANCHED 1

The domestication of cereals has involved common changes in morphological features, such as seed size, seed retention and modification of vegetative and inflorescence architecture that ultimately contributed to an increase in harvested yield. In barley, this process has resulted in two different cultivated types, two-rowed and six-rowed forms, both derived from the wild two-rowed ancestor, with archaeo-botanical evidence indicating the origin of six-rowed barley early in the domestication of the species, some 8,600-8,000 years ago. Variation at SIX-ROWED SPIKE 1 (VRS1) is sufficient to control this phenotype. However, phenotypes imposed by VRS1 alleles are modified by alleles at the INTERMEDIUM-C (INT-C) locus. Here we show that INT-C is an ortholog of the maize domestication gene TEOSINTE BRANCHED 1 (TB1) and identify 17 coding mutations in barley TB1 correlated with lateral spikelet fertility phenotypes.