3D cell cultures are rapidly becoming the method of choice for the physiologically relevant modeling of many aspects of non-malignant and malignant cell behavior ex vivo. Nevertheless, only a limited number of distinct cell types have been evaluated in this assay to date. Here we report the first large scale comparison of the transcriptional profiles and 3D cell culture phenotypes of a substantial panel of human breast cancer cell lines. Each cell line adopts a colony morphology of one of four main classes in 3D culture. These morphologies reflect, at least in part, the underlying gene expression profile and protein expression patterns of the cell lines, and distinct morphologies were also associated with tumor cell invasiveness and with cell lines originating from metastases. We further demonstrate that consistent differences in genes encoding signal transduction proteins emerge when even tumor cells are cultured in 3D microenvironments.

We show that comprehensive sequence-function maps obtained by deep sequencing can be used to reprogram interaction specificity and to leapfrog over bottlenecks in affinity maturation by combining many individually small contributions not detectable in conventional approaches. We use this approach to optimize two computationally designed inhibitors against H1N1 influenza hemagglutinin and, in both cases, obtain variants with subnanomolar binding affinity. The most potent of these, a 51-residue protein, is broadly cross-reactive against all influenza group 1 hemagglutinins, including human H2, and neutralizes H1N1 viruses with a potency that rivals that of several human monoclonal antibodies, demonstrating that computational design followed by comprehensive energy landscape mapping can generate proteins with potential therapeutic utility.

Approximately 98% of mammalian DNA is noncoding, yet we understand relatively little about the function of this enigmatic portion of the genome. The cis-regulatory elements that control gene expression reside in noncoding regions and can be identified by mapping the binding sites of tissue-specific transcription factors. Cone-rod homeobox (CRX) is a key transcription factor in photoreceptor differentiation and survival, but its in vivo targets are largely unknown. Here, we used chromatin immunoprecipitation with massively parallel sequencing (ChIP-seq) on CRX to identify thousands of cis-regulatory regions around photoreceptor genes in adult mouse retina. CRX directly regulates downstream photoreceptor transcription factors and their target genes via a network of spatially distributed regulatory elements around each locus. CRX-bound regions act in a synergistic fashion to activate transcription and contain multiple CRX binding sites which interact in a spacing-and orientation-dependent manner to fine-tune transcript levels. CRX ChIP-seq was also performed on Nrl -/-retinas, which represent an enriched source of cone photoreceptors. Comparison with the wild-type ChIP-seq data set identified numerous rod-and cone-specific CRX-bound regions as well as many shared elements. Thus, CRX combinatorially orchestrates the transcriptional networks of both rods and cones by coordinating the expression of photoreceptor genes including most retinal disease genes. In addition, this study pinpoints thousands of noncoding regions of relevance to both Mendelian and complex retinal disease.

Cis-regulatory elements (CREs) control gene expression by recruiting transcription factors (TFs) and other DNA binding proteins. We aim to understand how individual nucleotides contribute to the function of CREs. Here we introduce CRE analysis by sequencing (CRE-seq), a high-throughput method for producing and testing large numbers of reporter genes in mammalian cells. We used CRE-seq to assay >1,000 single and double nucleotide mutations in a 52-bp CRE in the Rhodopsin promoter that drives strong and specific expression in mammalian photoreceptors. We find that this particular CRE is remarkably complex. The majority (86%) of single nucleotide substitutions in this sequence exert significant effects on regulatory activity. Although changes in the affinity of known TF binding sites explain some of these expression changes, we present evidence for complex phenomena, including binding site turnover and TF competition. Analysis of double mutants revealed complex, nucleotide-specific interactions between residues in different TF binding sites. We conclude that some mammalian CREs are finely tuned by evolution and function through complex, nonadditive interactions between bound TFs. CRE-seq will be an important tool to uncover the rules that govern these interactions.utations in cis-regulatory elements (CREs) often have unexpected effects on gene regulation. We lack models with the predictive power to accurately interpret the functional consequences of noncoding polymorphisms. More generally, we do not understand the nucleotide-level architecture that distinguishes true CREs from nonfunctional groupings of transcription factor (TF) binding sites (TFBS). Although consortium-driven efforts continue to predict that large numbers of mammalian sequences are CREs (1, 2), we lack a corresponding high-throughput method for functionally analyzing the consequences of variants in these elements. Addressing these problems requires fine structure mutational analysis of mammalian CREs on a large scale-experiments that are difficult to perform using traditional assays. To facilitate such experiments, we developed CRE analysis by sequencing (CREseq), a high-throughput reporter gene assay for mammalian cells.CRE-seq leverages recent advances in oligonucleotide (oligo) synthesis (3) and high-throughput sequencing (4). Using arraybased oligo synthesis, we construct large numbers of reporter genes with unique sequence barcodes in their 3′ UTRs. These libraries of barcoded reporter genes are then transfected, en masse, into mammalian cells and quantified by performing RNA sequencing (RNA-Seq) (5) on the sequence barcodes. Here we present a study using CRE-seq to dissect a CRE in mouse Rhodopsin (Rho), a gene that is expressed strongly and specifically in the mammalian retina.Tight control of Rho expression is critical for the function of mammalian retinas (6, 7). Rho expression is regulated in mice by multiple CREs located at varying distances from the transcription start site (TSS) (8, 9). These elements are occupied in vivo by CRX, a re...

Transcription factors (TFs) recognize short sequence motifs that are present in millions of copies in large eukaryotic genomes. TFsmust distinguish their target binding sites from a vast genomic excess of spurious motif occurrences; however, it is unclear whether functional sites are distinguished from nonfunctional motifs by local primary sequence features or by the larger genomic context in which motifs reside. We used a massively parallel enhancer assay in living mouse retinas to compare 1,300 sequences bound in the genome by the photoreceptor transcription factor Cone-rod homeobox (Crx), to 3,000 control sequences. We found that very short sequences bound in the genome by Crx activated transcription at high levels, whereas unbound genomic regions with equal numbers of Crx motifs did not activate above background levels, even when liberated from their larger genomic context. High local GC content strongly distinguishes bound motifs from unbound motifs across the entire genome. Our results show that the cis-regulatory potential of TF-bound DNA is determined largely by highly local sequence features and not by genomic context. (1), yet the sequence features that distinguish functional cis-regulatory sites from the millions of spurious motif occurrences in large eukaryotic genomes are poorly understood (2-6). Several models have been proposed to explain how TFs distinguish between functional cis-regulatory elements (CREs) and nonfunctional motif occurrences (3, 6, 7). In one model, large-scale chromatin context directs TF binding to target sites while limiting TF access to spurious motif occurrences (7-9). This model is supported by recent analyses of genomic DNaseI hypersensitivity, which show that only 1% of the genome typically resides in open chromatin in any given cell type (1, 10), suggesting that most spurious motif occurrences are inaccessible. A second model states that target sites are recognized through cooperative TF binding to highly specific combinations of sequence motifs, which are unlikely to occur by chance in nonregulatory regions of the genome (6, 11). This model is supported by evidence that the binding specificity of many TFs is affected by cooperative interactions with cofactors (12). A third model states that most TF binding is promiscuous, low occupancy, and nonfunctional, whereas functional CREs are characterized by high TF occupancy, achieved through either a permissive chromatin context or high affinity for TFs (3,6). This model is motivated by recent genomewide binding studies demonstrating that binding locations of functionally diverse TFs overlap substantially (13, 14), a result that suggests binding is unlikely to be primarily determined by rare, specific combinations of cooperative interactions. Regardless of the mechanisms by which TFs select functional CREs, the distinction between functional and nonfunctional motif occurrences must ultimately depend on information encoded either locally or within the larger sequence context surrounding functional CREs.To distinguish between t...

The photoreceptor cells of the retina are subject to a greater number of genetic diseases than any other cell type in the human body. The majority of more than 120 cloned human blindness genes are highly expressed in photoreceptors. In order to establish an integrative framework in which to understand these diseases, we have undertaken an experimental and computational analysis of the network controlled by the mammalian photoreceptor transcription factors, Crx, Nrl, and Nr2e3. Using microarray and in situ hybridization datasets we have produced a model of this network which contains over 600 genes, including numerous retinal disease loci as well as previously uncharacterized photoreceptor transcription factors. To elucidate the connectivity of this network, we devised a computational algorithm to identify the photoreceptor-specific cis-regulatory elements (CREs) mediating the interactions between these transcription factors and their target genes. In vivo validation of our computational predictions resulted in the discovery of 19 novel photoreceptor-specific CREs near retinal disease genes. Examination of these CREs permitted the definition of a simple cis-regulatory grammar rule associated with high-level expression. To test the generality of this rule, we used an expanded form of it as a selection filter to evolve photoreceptor CREs from random DNA sequences in silico. When fused to fluorescent reporters, these evolved CREs drove strong, photoreceptor-specific expression in vivo. This study represents the first systematic identification and in vivo validation of CREs in a mammalian neuronal cell type and lays the groundwork for a systems biology of photoreceptor transcriptional regulation.

SummaryOpportunistic infections, such as aspergillosis, are among the most serious complications suffered by immunocompromised patients. Aspergillus fumigatus and other pathogenic fungi synthesize a toxic epipolythiodioxopiperazine metabolite called gliotoxin. Gliotoxin exhibits profound immunosuppressive activity in vivo. It induces apoptosis in thymocytes, splenocytes, and mesenteric lymph node cells and can selectively deplete bone marrow of mature lymphocytes. The molecular mechanism by which gliotoxin exerts these effects remains unknown. Here, we report that nanomolar concentrations of gliotoxin inhibited the activation of transcription factor NF-~B in response to a variety of stimuli in T and B cells. The effect ofgliotoxin was specific because, at the same concentrations, the toxin did not affect activation of the transcription factor NF-AT or of interferon-responsive signal transducers and activators of transcription. Likewise, the activity of the constitutively DNA-binding transcription factors Oct-1 and cyclic AMP response element binding protein (CREB), as well as the activation of protein tyrosine kinases p56 lck and p59 fyn, was not altered by gliotoxin. Very high concentrations of gliotoxin prevented NF-~B DNA binding in vitro. However, in intact cells, inhibition of NF-~13 did not occur at the level of DNA binding; rather, the toxin appeared to prevent degradation of IKB-ci, NF-KB's inhibitory subunit. Our data raise the possibility that the immunosuppression observed during aspergillosis results in part from gliotoxin-mediated NF-KB inhibition.

A crucial step in human breast cancer progression is the acquisition of invasiveness. There is a distinct lack of human cell culture models to study the transition from preinvasive to invasive phenotype as it may occur ''spontaneously'' in vivo. To delineate molecular alterations important for this transition, we isolated human breast epithelial cell lines that showed partial loss of tissue polarity in three-dimensional reconstituted basement membrane cultures. These cells remained noninvasive; however, unlike their nonmalignant counterparts, they exhibited a high propensity to acquire invasiveness through basement membrane in culture. The genomic aberrations and gene expression profiles of the cells in this model showed a high degree of similarity to primary breast tumor profiles. The xenograft tumors formed by the cell lines in three different microenvironments in nude mice displayed metaplastic phenotypes, including squamous and basal characteristics, with invasive cells exhibiting features of higher-grade tumors. To find functionally significant changes in transition from preinvasive to invasive phenotype, we performed attribute profile clustering analysis on the list of genes differentially expressed between preinvasive and invasive cells. We found integral membrane proteins, transcription factors, kinases, transport molecules, and chemokines to be highly represented. In addition, expression of matrix metalloproteinases MMP9, MMP13, MMP15, and MMP17 was up-regulated in the invasive cells. Using small interfering RNA-based approaches, we found these MMPs to be required for the invasive phenotype. This model provides a new tool for dissection of mechanisms by which preinvasive breast cells could acquire invasiveness in a metaplastic context. [Cancer Res 2008;68(5):1378-87]