rPAC: Route based pathway analysis for cohorts of gene expression data sets.

Pathway analysis is a popular method aiming to derive biological interpretation from high-throughput gene expression studies. However, existing methods focus mostly on identifying which pathway or pathways could have been perturbed, given differential gene expression patterns. In this paper, we present a novel pathway analysis framework, namely rPAC, which decomposes each signaling pathway route into two parts, the upstream portion of a transcription factor (TF) block and the downstream portion from the TF block and generates a pathway route perturbation analysis scheme examining disturbance scores assigned to both parts together. This rPAC scoring is further applied to a cohort of gene expression data sets which produces two summary metrics, "Proportion of Significance" (PS) and "Average Route Score" (ARS), as quantitative measures discerning perturbed pathway routes within and/or between cohorts. To demonstrate rPAC's scoring competency, we first used a large amount of simulated data and compared the method's performance against those by conventional methods in terms of power curve. Next, we performed a case study involving three epithelial cancer data sets from The Cancer Genome Atlas (TCGA). The rPAC method revealed specific pathway routes as potential cancer type signatures. A deeper pathway analysis of sub-groups (i.e., age groups in COAD or cancer sub-types in BRCA) resulted in pathway routes that are known to be associated with the sub-groups. In addition, multiple previously uncharacterized pathways routes were identified, potentially suggesting that rPAC is better in deciphering etiology of a disease than conventional methods particularly in isolating routes and sections of perturbed pathways in a finer granularity.

Single-cell Transcriptome Landscape of DNA Methylome Regulators Associated with Orofacial Clefts in the Mouse Dental Pulp.

OBJECTIVE: Significant evidence links epigenetic processes governing the dynamics of DNA methylation and demethylation to an increased risk of syndromic and nonsyndromic cleft lip and/or cleft palate (CL/P). Previously, we characterized mesenchymal stem/stromal cells (MSCs) at different stages of osteogenic differentiation in the mouse incisor dental pulp. The main objective of this research was to characterize the transcriptional landscape of regulatory genes associated with DNA methylation and demethylation at a single-cell resolution. DESIGN: We used single-cell RNA sequencing (scRNA-seq) data to characterize transcriptome in individual subpopulations of MSCs in the mouse incisor dental pulp. SETTINGS: The biomedical research institution. PATIENTS/PARTICIPANTS: This study did not include patients. INTERVENTIONS: This study collected and analyzed data on the single-cell RNA expssion in the mouse incisor dental pulp. MAIN OUTCOME MEASURE(S): Molecular regulators of DNA methylation/demethylation exhibit differential transcriptional landscape in different subpopulations of osteogenic progenitor cells. RESULTS: scRNA-seq analysis revealed that genes encoding DNA methylation and demethylation enzymes (DNA methyltransferases and members of the ten-eleven translocation family of methylcytosine dioxygenases), methyl-DNA binding domain proteins, as well as transcription factors and chromatin remodeling proteins that cooperate with DNA methylation machinery are differentially expressed within distinct subpopulations of MSCs that undergo different stages of osteogenic differentiation. CONCLUSIONS: These findings suggest some mechanistic insights into a potential link between epigenetic alterations and multifactorial causes of CL/P phenotypes.

Predicting the targets of IRF8 and NFATc1 during osteoclast differentiation using the machine learning method framework cTAP.

BACKGROUND: Interferon regulatory factor-8 (IRF8) and nuclear factor-activated T cells c1 (NFATc1) are two transcription factors that have an important role in osteoclast differentiation. Thanks to ChIP-seq technology, scientists can now estimate potential genome-wide target genes of IRF8 and NFATc1. However, finding target genes that are consistently up-regulated or down-regulated across different studies is hard because it requires analysis of a large number of high-throughput expression studies from a comparable context. METHOD: We have developed a machine learning based method, called, Cohort-based TF target prediction system (cTAP) to overcome this problem. This method assumes that the pathway involving the transcription factors of interest is featured with multiple "functional groups" of marker genes pertaining to the concerned biological process. It uses two notions, Gene-Present Sufficiently (GP) and Gene-Absent Insufficiently (GA), in addition to log2 fold changes of differentially expressed genes for the prediction. Target prediction is made by applying multiple machine-learning models, which learn the patterns of GP and GA from log2 fold changes and four types of Z scores from the normalized cohort's gene expression data. The learned patterns are then associated with the putative transcription factor targets to identify genes that consistently exhibit Up/Down gene regulation patterns within the cohort. We applied this method to 11 publicly available GEO data sets related to osteoclastgenesis. RESULT: Our experiment identified a small number of Up/Down IRF8 and NFATc1 target genes as relevant to osteoclast differentiation. The machine learning models using GP and GA produced NFATc1 and IRF8 target genes different than simply using a log2 fold change alone. Our literature survey revealed that all predicted target genes have known roles in bone remodeling, specifically related to the immune system and osteoclast formation and functions, suggesting confidence and validity in our method. CONCLUSION: cTAP was motivated by recognizing that biologists tend to use Z score values present in data sets for the analysis. However, using cTAP effectively presupposes assembling a sizable cohort of gene expression data sets within a comparable context. As public gene expression data repositories grow, the need to use cohort-based analysis method like cTAP will become increasingly important.

A route-based pathway analysis framework integrating mutation information and gene expression data.

We propose a new way of analyzing biological pathways in which the analysis combines both transcriptome data and mutation information and uses the outcome to identify "routes" of aberrant pathways potentially responsible for the etiology of disease. Each pathway route is encoded as a Bayesian Network which is initialized with a sequence of conditional probabilities which are designed to encode directionality of regulatory relationships encoded in the pathways, i.e. activation and inhibition relationships. First, we demonstrate the effectiveness of our model through simulation in which the model was able to easily separate Test samples from Control samples using fictitiously perturbed pathway routes. Second, we apply our model to analyze the Breast Cancer data set, available from TCGA, against many cancer pathways available from KEGG and rank the significance of identified pathways. The outcome is consistent with what have already been reported in the literature. Third, survival analysis has been carried out on the same data set by using pathway routes as features. Overall, we envision that our model of using pathway routes for analysis can further refine the conventional ways of subtyping cancer patients as it can discover additional characteristics specific to individual's tumor.

The LG/J murine strain exhibits near-normal tendon biomechanical properties following a full-length central patellar tendon defect.

PURPOSE OF THE STUDY: Identifying biological success criteria is needed to improve therapies, and one strategy for identifying them is to analyze the RNA transcriptome for successful and unsuccessful models of tendon healing. We have characterized the MRL/MpJ murine strain and found improved mechanical outcomes following a central patellar tendon (PT) injury. In this study, we evaluate the healing of the LG/J murine strain, which comprises 75% of the MRL/MpJ background, to determine if the LG/J also exhibits improved biomechanical properties following injury and to determine differentially expressed transcription factors across the C57BL/6, MRL/MpJ and the LG/J strains during the early stages of healing. MATERIALS AND METHODS: A full-length, central PT defect was created in 16-20 week old MRL/MpJ, LG/J, and C57BL/6 murine strains. Mechanical properties were assessed at 2, 5, and 8 weeks post surgery. Transcriptomic expression was assessed at 3, 7, and 14 days following injury using a novel clustering software program to evaluate differential expression of transcription factors. RESULTS: Average LG/J structural properties improved to 96.7% and 97.2% of native LG/J PT stiffness and ultimate load by 8 weeks post surgery, respectively. We found the LG/J responded by increasing expression of transcription factors implicated in the inflammatory response and collagen fibril organization. CONCLUSIONS: The LG/J strain returns to normal structural properties by 8 weeks, with steadily increasing properties at each time point. Future work will characterize the cell populations responding to injury and investigate the role of the differentially expressed transcription factors during healing.

GABRA2 Alcohol Dependence Risk Allele is Associated with Reduced Expression of Chromosome 4p12 GABAA Subunit Genes in Human Neural Cultures.

BACKGROUND: Genetic variation in a region of chromosome 4p12 that includes the GABAA subunit gene GABRA2 has been reproducibly associated with alcohol dependence (AD). However, the molecular mechanisms underlying the association are unknown. This study examined correlates of in vitro gene expression of the AD-associated GABRA2 rs279858*C-allele in human neural cells using an induced pluripotent stem cell (iPSC) model system. METHODS: We examined mRNA expression of chromosome 4p12 GABAA subunit genes (GABRG1, GABRA2, GABRA4, and GABRB1) in 36 human neural cell lines differentiated from iPSCs using quantitative polymerase chain reaction and next-generation RNA sequencing. mRNA expression in adult human brain was examined using the BrainCloud and BRAINEAC data sets. RESULTS: We found significantly lower levels of GABRA2 mRNA in neural cell cultures derived from rs279858*C-allele carriers. Levels of GABRA2 RNA were correlated with those of the other 3 chromosome 4p12 GABAA genes, but not other neural genes. Cluster analysis based on the relative RNA levels of the 4 chromosome 4p12 GABAA genes identified 2 distinct clusters of cell lines, a low-expression cluster associated with rs279858*C-allele carriers and a high-expression cluster enriched for the rs279858*T/T genotype. In contrast, there was no association of genotype with chromosome 4p12 GABAA gene expression in postmortem adult cortex in either the BrainCloud or BRAINEAC data sets. CONCLUSIONS: AD-associated variation in GABRA2 is associated with differential expression of the entire cluster of GABAA subunit genes on chromosome 4p12 in human iPSC-derived neural cell cultures. The absence of a parallel effect in postmortem human adult brain samples suggests that AD-associated genotype effects on GABAA expression, although not present in mature cortex, could have effects on regulation of the chromosome 4p12 GABAA cluster during neural development.

Genome-wide analysis of FoxO1 binding in hepatic chromatin: potential involvement of FoxO1 in linking retinoid signaling to hepatic gluconeogenesis.

The forkhead transcription factor FoxO1 is a critical regulator of hepatic glucose and lipid metabolism, and dysregulation of FoxO1 function has been implicated in diabetes and insulin resistance. We globally identified FoxO1 occupancy in mouse hepatic chromatin on a genome-wide level by chromatin immunoprecipitation coupled with high-throughput DNA sequencing (ChIP-seq). To establish the specific functional significance of FoxO1 against other FoxO proteins, ChIP-seq was performed with chromatin from liver-specific FoxO1 knockout and wild-type mice. Here we identified 401 genome-wide FoxO1-binding locations. Motif search reveals a sequence element, 5' GTAAACA 3', consistent with a previously known FoxO1-binding site. Gene set enrichment analysis shows that the data from FoxO1 ChIP-seq are highly correlated with the global expression profiling of genes regulated by FoxO1, demonstrating the functional relevance of our FoxO1 ChIP-seq study. Interestingly, gene ontology analysis reveals the functional significance of FoxO1 in retinoid metabolic processes. We show here that FoxO1 directly binds to the genomic sites for the genes in retinoid metabolism. Notably, deletion of FoxO1 caused a significantly reduced induction of Pck1 and Pdk4 in response to retinoids. As Pck1 and Pdk4 are downstream targets of retinoid signaling, these results suggest that FoxO1 plays a potential role in linking retinoid metabolism to hepatic gluconeogenesis.

The chromatin accessibility landscape in the dental pulp of mouse molars and incisors.

The dental pulp is a promising source of progenitor cells for regenerative medicine. The natural function of dental pulp is to produce odontoblasts to generate reparative dentin. Stem cells within the pulp tissue originate from the migrating neural crest cells and possess mesenchymal stem cell properties with the ability to differentiate into multiple lineages. To elucidate the transcriptional control mechanisms underlying cell fate determination, we compared the transcriptome and chromatin accessibility in primary dental pulp tissue derived from 5-6-day-old mice. Using RNA sequencing and assay for transposase-accessible chromatin using sequencing (ATAC-seq), we correlated gene expression with chromatin accessibility. We found that the majority of ATAC-seq peaks were concentrated at genes associated with development and cell differentiation. Most of these genes were highly expressed in the mouse dental pulp. Surprisingly, we uncovered a group of genes encoding master transcription factors that were not expressed in the dental pulp but retained open chromatin states. Within this group, we identified key developmental genes important for specification of the neural crest, adipocyte, neural, myoblast, osteoblast and hepatocyte lineages. Collectively, our results uncover a complex relationship between gene expression and the chromatin accessibility landscape in the mouse dental pulp.

Genome-wide distribution of 5hmC in the dental pulp of mouse molars and incisors.

The dental pulp is critical for the production of odontoblasts to create reparative dentin. In recent years, dental pulp has become a promising source of mesenchymal stem cells that are capable of differentiating into multiple cell types. To elucidate the transcriptional control mechanisms specifying the early phases of odontoblast differentiation, we analysed the DNA demethylation pattern associated with 5-hydroxymethylcytosine (5hmC) in the primary murine dental pulp. 5hmC plays an important role in chromatin accessibility and transcriptional control by modelling a dynamic equilibrium between DNA methylation and demethylation. Our research revealed 5hmC enrichment along genes and non-coding regulatory regions associated with specific developmental pathways in the genome of mouse incisor and molar dental pulp. Although the overall distribution of 5hmC is similar, the intensity and location of the 5hmC peaks significantly differs between the incisor and molar pulp genome, indicating cell type-specific epigenetic variations. Our study suggests that the differential DNA demethylation pattern could account for the distinct regulatory mechanisms underlying the tooth-specific ontogenetic programs.

Single-Cell Transcriptome Analysis Defines Expression of Kabuki Syndrome-Associated KMT2D Targets and Interacting Partners.

Objectives. Kabuki syndrome (KS) is a rare genetic disorder characterized by developmental delay, retarded growth, and cardiac, gastrointestinal, neurocognitive, renal, craniofacial, dental, and skeletal defects. KS is caused by mutations in the genes encoding histone H3 lysine 4 methyltransferase (KMT2D) and histone H3 lysine 27 demethylase (KDM6A), which are core components of the complex of proteins associated with histone H3 lysine 4 methyltransferase SET1 (SET1/COMPASS). Using single-cell RNA data, we examined the expression profiles of Kmt2d and Kdm6a in the mouse dental pulp. In the incisor pulp, Kmt2d and Kdm6a colocalize with other genes of the SET1/COMPASS complex comprised of the WD-repeat protein 5 gene (Wdr5), the retinoblastoma-binding protein 5 gene (Rbbp5), absent, small, and homeotic 2-like protein-encoding gene (Ash2l), nuclear receptor cofactor 6 gene (Ncoa6), and Pax-interacting protein 1 gene (Ptip1). In addition, we found that Kmt2d and Kdm6a coexpress with the downstream target genes of the Wingless and Integrated (WNT) and sonic hedgehog signaling pathways in mesenchymal stem/stromal cells (MSCs) at different stages of osteogenic differentiation. Taken together, our results suggest an essential role of KMT2D and KDK6A in directing lineage-specific gene expression during differentiation of MSCs.

Single-cell transcriptomics defines Dot1L interacting partners and downstream target genes in the mouse molar dental pulp.

Although histone methyltransferases are implicated in many key developmental processes, the contribution of individual chromatin modifiers in dental tissues is not well understood. Using single-cell RNA sequencing, we examined the expression profiles of the disruptor of telomeric silencing 1-like (Dot1L) gene in the postnatal day 5 mouse molar dental pulp. Dot1L is the only known enzyme that methylates histone 3 on lysine 79, a modification associated with gene expression. Our research revealed 15 distinct clusters representing different populations of mesenchymal stromal cells (MSCs), immune cells, pericytes, ameloblasts and endothelial cells. We documented heterogeneity in gene expression across different subpopulations of MSCs, a good indicator that these stromal progenitors undergo different phases of osteogenic differentiation. Interestingly, although Dot1L was broadly expressed across all cell clusters within the molar dental pulp, our analyses indicated specific enrichment of Dot1L within two clusters of MSCs, as well as cell clusters characterized as ameloblasts and endothelial cells. Moreover, we detected Dot1L co-expression with protein interactors involved in epigenetic activation such as Setd2, Sirt1, Brd4, Isw1, Bptf and Suv39h1. In addition, Dot1L was co-expressed with Eed2, Cbx3 and Dnmt1, which encode epigenetic factors associated with gene silencing and heterochromatin formation. Dot1l was co-expressed with downstream targets of the insulin growth factor and WNT signaling pathways, as well as genes involved in cell cycle progression. Collectively, our results suggest that Dot1L may play key roles in orchestrating lineage-specific gene expression during MSC differentiation.

Genome-Wide Analysis for Identifying FOXO Protein-Binding Sites.

Forkhead box O (FOXO) proteins comprise a superfamily of transcription factors that play important roles in controlling various biological processes. Transcriptional control constitutes a crucial component in regulating complex biological processes. The identification of cis-regulatory elements is essential to understand the regulatory mechanism of gene expression. Chromatin immunoprecipitation followed by high-throughput sequencing (ChIP-seq) is widely used to identify the cis-regulatory elements of transcription factors and other DNA-binding proteins on a genome-wide level. It is a powerful tool to analyze the regulatory networks underlying the biological processes. Here, we describe a detailed protocol for preparing ChIP-seq samples that are used for sequencing and subsequent data analyses.

Astaxanthin attenuates the increase in mitochondrial respiration during the activation of hepatic stellate cells.

Upon liver injury, quiescent hepatic stellate cells (qHSCs) transdifferentiate to myofibroblast-like activated HSCs (aHSCs), which are primarily responsible for the accumulation of extracellular matrix proteins during the development of liver fibrosis. Therefore, aHSCs may exhibit different energy metabolism from that of qHSCs to meet their high energy demand. We previously demonstrated that astaxanthin (ASTX), a xanthophyll carotenoid, prevents the activation of HSCs. The objective of this study was to determine if ASTX can exert its antifibrogenic effect by attenuating any changes in energy metabolism during HSC activation. To characterize the energy metabolism of qHSCs and aHSCs, mouse primary HSCs were cultured on uncoated plastic dishes for 7 days for spontaneous activation in the presence or absence of 25 µM ASTX. qHSCs (1 day after isolation) and aHSCs treated with or without ASTX for 7 days were used to determine parameters related to mitochondrial respiration using a Seahorse XFe24 Extracellular Flux analyzer. aHSCs had significantly higher basal respiration, maximal respiration, ATP production, spare respiratory capacity and proton leak than those of qHSCs. However, ASTX prevented most of the changes occurring during HSC activation and improved mitochondrial cristae structure with decreased cristae junction width, lumen width and the area in primary mouse aHSCs. Furthermore, qHSCs isolated from ASTX-fed mice had lower mitochondrial respiration and glycolysis than control qHSCs. Our findings suggest that ASTX may exert its antifibrogenic effect by attenuating the changes in energy metabolism during HSC activation.

BioTarget: A Computational Framework Identifying Cancer Type Specific Transcriptional Targets of Immune Response Pathways.

Transcriptome data can provide information on signaling pathways active in cancers, but new computational tools are needed to more accurately quantify pathway activity and identify tissue-specific pathway features. We developed a computational method called "BioTarget" that incorporates ChIP-seq data into cellular pathway analysis. This tool relates the expression of transcription factor TF target genes (based on ChIP-seq data) with the status of upstream signaling components for an accurate quantification of pathway activity. This analysis also reveals TF targets expressed in specific contexts/tissues. We applied BioTarget to assess the activity of TBX21 and GATA3 pathways in cancers. TBX21 and GATA3 are TF regulators that control the differentiation of T cells into Th1 and Th2 helper cells that mediate cell-based and humoral immune responses, respectively. Since tumor immune responses can impact cancer progression, the significance of our pathway scores should be revealed by effective patient stratification. We found that low Th1/Th2 activity ratios were associated with a significantly poorer survival of stomach and breast cancer patients, whereas an unbalanced Th1/Th2 response was correlated with poorer survival of colon cancer patients. Lung adenocarcinoma and lung squamous cell carcinoma patients had the lowest survival rates when both Th1 and Th2 responses were high. Our method also identified context-specific target genes for TBX21 and GATA3. Applying the BioTarget tool to BCL6, a TF associated with germinal center lymphocytes, we observed that patients with an active BCL6 pathway had significantly improved survival for breast, colon, and stomach cancer. Our findings support the effectiveness of the BioTarget tool for transcriptome analysis and point to interesting associations between some immune-response pathways and cancer progression.

Chemotherapy-induced oral mucositis is associated with detrimental bacterial dysbiosis.

BACKGROUND: Gastrointestinal mucosal injury (mucositis), commonly affecting the oral cavity, is a clinically significant yet incompletely understood complication of cancer chemotherapy. Although antineoplastic cytotoxicity constitutes the primary injury trigger, the interaction of oral microbial commensals with mucosal tissues could modify the response. It is not clear, however, whether chemotherapy and its associated treatments affect oral microbial communities disrupting the homeostatic balance between resident microorganisms and the adjacent mucosa and if such alterations are associated with mucositis. To gain knowledge on the pathophysiology of oral mucositis, 49 subjects receiving 5-fluorouracil (5-FU) or doxorubicin-based chemotherapy were evaluated longitudinally during one cycle, assessing clinical outcomes, bacterial and fungal oral microbiome changes, and epithelial transcriptome responses. As a control for microbiome stability, 30 non-cancer subjects were longitudinally assessed. Through complementary in vitro assays, we also evaluated the antibacterial potential of 5-FU on oral microorganisms and the interaction of commensals with oral epithelial tissues. RESULTS: Oral mucositis severity was associated with 5-FU, increased salivary flow, and higher oral granulocyte counts. The oral bacteriome was disrupted during chemotherapy and while antibiotic and acid inhibitor intake contributed to these changes, bacteriome disruptions were also correlated with antineoplastics and independently and strongly associated with oral mucositis severity. Mucositis-associated bacteriome shifts included depletion of common health-associated commensals from the genera Streptococcus, Actinomyces, Gemella, Granulicatella, and Veillonella and enrichment of Gram-negative bacteria such as Fusobacterium nucleatum and Prevotella oris. Shifts could not be explained by a direct antibacterial effect of 5-FU, but rather resembled the inflammation-associated dysbiotic shifts seen in other oral conditions. Epithelial transcriptional responses during chemotherapy included upregulation of genes involved in innate immunity and apoptosis. Using a multilayer epithelial construct, we show mucositis-associated dysbiotic shifts may contribute to aggravate mucosal damage since the mucositis-depleted Streptococcus salivarius was tolerated as a commensal, while the mucositis-enriched F. nucleatum displayed pro-inflammatory and pro-apoptotic capacity. CONCLUSIONS: Altogether, our work reveals that chemotherapy-induced oral mucositis is associated with bacterial dysbiosis and demonstrates the potential for dysbiotic shifts to aggravate antineoplastic-induced epithelial injury. These findings suggest that control of oral bacterial dysbiosis could represent a novel preventive approach to ameliorate oral mucositis.

Analysis of αSMA-labeled progenitor cell commitment identifies notch signaling as an important pathway in fracture healing.

Fracture healing is a regenerative process that involves coordinated responses of many cell types, but characterization of the roles of specific cell populations in this process has been limited. We have identified alpha smooth muscle actin (αSMA) as a marker of a population of mesenchymal progenitor cells in the periosteum that contributes to osteochondral elements during fracture healing. Using a lineage tracing approach, we labeled αSMA-expressing cells, and characterized changes in the periosteal population during the early stages of fracture healing by histology, flow cytometry, and gene expression profiling. In response to fracture, the αSMA-labeled population expanded and began to differentiate toward the osteogenic and chondrogenic lineages. The frequency of mesenchymal progenitor cell markers such as Sca1 and PDGFRα increased after fracture. By 6 days after fracture, genes involved in matrix production and remodeling were elevated. In contrast, genes associated with muscle contraction and Notch signaling were downregulated after fracture. We confirmed that activating Notch signaling in αSMA-labeled cells inhibited differentiation into osteogenic and adipogenic lineages in vitro and ectopic bone formation in vivo. By characterizing changes in a selected αSMA-labeled progenitor cell population during fracture callus formation, we have shown that modulation of Notch signaling may determine osteogenic potential of αSMA-expressing progenitor cells during bone healing.

Diversity and complexity in chromatin recognition by TFII-I transcription factors in pluripotent embryonic stem cells and embryonic tissues.

GTF2I and GTF2IRD1 encode a family of closely related transcription factors TFII-I and BEN critical in embryonic development. Both genes are deleted in Williams-Beuren syndrome, a complex genetic disorder associated with neurocognitive, craniofacial, dental and skeletal abnormalities. Although genome-wide promoter analysis has revealed the existence of multiple TFII-I binding sites in embryonic stem cells (ESCs), there was no correlation between TFII-I occupancy and gene expression. Surprisingly, TFII-I recognizes the promoter sequences enriched for H3K4me3/K27me3 bivalent domain, an epigenetic signature of developmentally important genes. Moreover, we discovered significant differences in the association between TFII-I and BEN with the cis-regulatory elements in ESCs and embryonic craniofacial tissues. Our data indicate that in embryonic tissues BEN, but not the highly homologous TFII-I, is primarily recruited to target gene promoters. We propose a "feed-forward model" of gene regulation to explain the specificity of promoter recognition by TFII-I factors in eukaryotic cells.