While most of the genes that influence the segmentation of the fruit fly embryo act only transiently, the segment polarity genes have a stable expression pattern that defines and maintains the borders between different parasegments. The segment polarity genes refine and maintain their expression through a network of intra- and intercellular regulatory interactions between gene products. This talk will present a family of qualitative (logical) models of these interactions and of how they lead to stable gene expression patterns. We investigated three modeling frameworks: synchronous Boolean, asynchronous Boolean and piece-wise linear ODE-based models, collectively spanning the range between discrete and continuous modeling. All models are able to reproduce the wild type expression pattern of the segment polarity genes, as well as the ectopic expressions obtained for gene mutation experiments. We find that a separation between the timescales of posttranslational and transcription/translation processes is necessary for establishing the regular gene expression pattern in the segment polarity network. All our algorithms concur in suggesting that the divergence from wild type can be attributed to an imbalance between the two opposing Cubitus Interruptus transcription factors (CIA, CIR) in the posterior half of the parasegment. We find that the system is vulnerable to large delays in expression of any gene - except for ci - and, in such delayed conditions, the mutant state characteristic to that gene knockout is generated. Interestingly, cell division increases the robustness of the segment polarity network with respect to perturbations in biological processes. Taken together, the results of the synchronous, asynchronous Boolean and hybrid models convincingly demonstrate the Boolean models' capability for effectively describing the basic structure and functioning of gene control networks when detailed kinetic information is unavailable.
All development is ultimately encoded in gene regulatory interactions. Transcription factor (TF) perturbations (e.g. knock downs and knock outs) have been used widely to predict candidate TF targets. But practical constraints limit their applicability in many species, and particularly in studies of later embryonic development. To date identification of causal regulatory interactions between transcription factors and target genes without TF-specific perturbation data has been difficult, costly, time-consuming, and error-prone. I will describe computational approaches that we are developing to address these challenges as we attempt to identify the network of gene regulatory interactions that underlie the development of T-cells in mice.
Chromatin boundaries, or insulators, can block enhancer-promoter interactions and/or limit the spread of silent chromatin. Recent studies indicate that boundary elements are widely present in animal genomes, especially between closely apposed gene promoters, further supporting their roles in maintaining regulatory independence between neighboring genes. We have previously identified SF1, a chromatin boundary in the Drosophila Antennapedia Hox cluster. It is located between the divergently transcribed Hox gene Scr and a non-Hox gene ftz. SF1 exhibits strong enhancer-blocking activity in embryos and protects the miniwhite reporter from the influences of surrounding chromatin. Our recent studies further show that SF1 interacts with neighboring genomic elements to form DNA/chromatin loop domains. We propose that SF1 facilitates the formation of independent gene regulatory domains to modulate stage- and tissue- specific enhancer-promoter interactions.
The cells that lie along the midline of the Drosophila CNS are few in number (22/ganglion) but are represented by a variety of neuronal and glial cell types. These include motorneurons, local interneurons, projection neurons, and glia. The midline cells represent an excellent system to study the regulatory circuitry that controls the generation of distinct neuronal and glial cell types, their migration, axon guidance, and glial-axonal interactions. To this end, we have employed in situ hybridization to describe the spatial and temporal expression of 278 midline-expressed genes -the data is accessible via a searchable, web-base database. Methods were developed for imaging midline cells by confocal microscopy of sim-Gal4 UAS-\tauGFP embryos, and expression of 70 genes that include many transcription factor and neural function genes were examined at multiple stages of development. Thus, each midline precursor and mature cell type can be uniquely identified at each stage of CNS development in both wild-type and mutant embryos.
The single-minded (sim), Notch, and lethal of scute (l(1)sc) genes all play major roles in midline cell development. The sim gene is a master regulator of midline development and plays later roles in midline glial and neuronal development. Notch signaling plays multiple roles in midline development including the neuron-glia switch, neuronal precursor formation, and H-cell sib and iVUM neuronal cell fates. One major goal of our research is to understand how sim and Notch signaling work together to control midline cell development. The l(1)sc gene acts to control neural precursor formation as well as H-cell and mVUM gene expression. Current work is involved with identifying and studying the regulatory proteins that are downstream of Notch and l(1)sc that control the differentiated properties of each neuronal cell type.
The midline glia form a scaffold that ensheaths the commissural axons that cross the midline. We used our imaging methods to visualize midline glial migration, ensheathment, and subdivision of axon commissures, and showed that these events are mediated by the Wrapper (midline glial-expressed) and Neurexin IV (neuronal and axonal-expressed) heterophilic adhesion proteins. We have identified 52 genes expressed in midline glia, including 11 transcription factors, and these are being genetically analyzed to understand how they control the complex morphogenetic and functional properties of midline glia. The overall goal is a comprehensive understanding of the regulatory circuitry involved in CNS developmental decision-making and how specific CNS cell types acquire their differentiated properties.
Work done in collabortaion with Scott R. Wheeler, Stephanie B. Stagg, and Joseph C. Pearson
Many developmental phenomena involve processes that lead to determinate outputs. Lineage information and signaling cues at specific time points, though varied at certain levels, are integrated to yield robust cell fate executions. For sensory systems, fine regulation of gene expression is critical so that information is assessed, relayed, and processed in a specific logical manner. Typically, one molecular receptor type is expressed in a single sensory neuron to prevent sensory confusion.
The fly eye is an example of a sensory system that integrates developmental inputs to yield specific robust cell fate determination. The fly eye is composed of 800 ommatidia (unit eyes) which contain six outer photoreceptors (PRs), R1-6, arranged in a trapezoidal shape surrounding two inner PRs, R7 and R8. The outer PRs express the Rhodopsin1 (Rh1) protein and are used for motion detection. The inner PRs, used for color vision, are organized into two coordinated subtypes. In the pale subtype, R7 expresses Rh3 and Rh8 expresses Rh5 whereas in the yellow subtype, R7 expresses Rh4 and R8 expresses Rh5. Though the distribution of these ommatidial subtypes is spatially randomized throughout the eye, subtype fate determination is robust such that each R7 and R8 expresses a particular rhodopsin in a stable manner and conserved ratio.
How does the fly eye ensure robustness? Here, we describe two distinct roles for the K50 homeodomain transcription factor, Defective proventriculus (Dve). In yellow R7s, Dve specifically represses expression of Rh3. In /dve /mutants, Rh3 is de-repressed in all yellow R7s yielding R7s that express both Rh3 and Rh4. In outer PRs, Dve plays a very different role, repressing noisy expression of Rh3, Rh5, and Rh6. In /dve /null mutants, these rhodopsins are de-repressed in random outer PRs. Dve expression itself is robustly controlled by a complex transcriptional regulatory network. Our analysis suggests that the fly eye utilizes transcriptional repression to mask inherently noisy gene expression and ensure robustness.
Biological cells behave in complex ways by producing different RNA molecules in response to diverse conditions. These RNA molecules fold into specific functional RNA structures, or are translated into peptide sequences, which fold into proteins. RNA transcripts are encoded in and transcribed from DNA segments called genes, in a process called "gene expression".
Gene expression is accomplished by the presence of regulatory DNA sequences present at each gene locus, where they instruct the cell as to the conditions under which that gene should be expressed. Thus a gene encodes a potential RNA transcript as well as several instructions for when to produce the transcript. Regulatory DNAs therefore are critical for specifying the number of different gene expression states available to a cell, and the situations in which a cell transitions between these states. Regulatory DNAs are vastly more numerous and complex than the easily identifiable protein-coding DNAs that they regulate. Regulatory DNAs represent the latest frontier in biology.
In my talk, I will focus on the structure of an equivalence class of regulatory DNAs and how they have been evolving across different Drosophila lineages. I will also discuss how such an example corpus can help guide a unified computational approach to the study of the native computational infrastructure of living cells.
The Bone Morphogenetic Protein (BMP) family member decapentaplegic (dpp) is transcribed uniformly over the dorsal 40% of the blastoderm embryo. In contrast, BMP signaling, which mirrors Dpp - receptor interactions, is present in only a subset of the dpp expression domain. By the onset of gastrulation, BMP signaling refines to become restricted to a sharp stripe comprising the dorsal-most 10% of embryonic cells, which form the extraembryonic amnioserosa. Two processes cooperate to produce this spatially bistable pattern of BMP signaling. First, an extracellular transport system concentrates Dpp into a shallow dorsal gradient. Second, an intracellular positive feedback system promotes Dpp - receptor interactions dependent on previous signaling strength.
We are currently analyzing the mechanisms underlying positive feedback. We propose that an increase in receptor availability in regions of previous signaling both increases capacity for future signaling and acts as a sink for BMP ligands, thereby narrowing the signaling domain. We have shown that the BMP target gene eiger is a component of feedback. Eiger, a Tumor Necrosis alpha homolog, signals through the Jun N-terminal kinase pathway to promote Dpp - receptor interactions. While the effects of eiger mutations on BMP signaling can be visualized in multiple sensitized genetic backgrounds, loss of eiger in an otherwise wild-type embryo has no phenotype. These data indicate that feedback is genetically redundant, and we are currently investigating other possible feedback components.
Work done with Jackie Gavin-Smyth and Yu-Chiun Wang.
A dynamic regulatory network among transcription factors and inductive signals leads to the progressive delineation of cell fates of the developing heart and other muscular tissues. Most of the known regulatory factors exert different functions during consecutive steps of in this regulatory cascade. Of note, the NK homeodomain factor Tinman acts in the early mesoderm in combination with Dpp signals to promote the development of all dorsal mesodermal tissue derivatives, whereas the T-box factors Dorsocross are required specifically for the formation of myocardial cells. Upon heart formation, these cardiogenic factors are then required within the dorsal vessel, where they regulate proper diversification of myocardial cell identities and, in the case of Tinman, cardiac remodeling. A current model of the regulatory interactions in the Drosophila embryonic mesoderm with a focus on cardiogenesis and our present approaches to identify additional components will be presented.
One of the central challenges in biology is to understand how the genome is utilized to orchestrate the development of complex tissues and organisms. While genetic studies have identified a number of essential transcription factors required for cell fate specification, little is known about the molecular mechanisms by which these regulators function. Few of their direct target genes or effector molecules are known. Moreover, the architecture of the underlying transcriptional network in which they operate remains elusive.
Our work attempts to bridge this gap, by integrating genetic, genomic and computational approaches to understand the transcriptional network that drives the selection of cell fates within the mesoderm. By combining ChIP-on-chip through a time-course of Drosophila development we are systematically identifying cis-regulatory module occupancy during developmental progression. These data are enriched by expression profiling of mutant embryos for each transcription factor. The topology of the network was unexpected, showing extensive combinatorial regulation and temporal enhancer occupancy. Current work is focused on understanding how these diverse combinatorial binding 'codes' give rise to specific patterns of enhancer expression.
A systems-level understanding of gene regulation in animal genomes requires the comprehensive characterization of functional regulatory regions, the sequence motifs within them, and the regulatory logic guiding their spatial and temporal activity. Our group at MIT is developing computational methods to address these challenges in Drosophila melanogaster, in collaboration with large-scale experimental efforts. We have used comparative genomics of 12 Drosophila genomes to recognize characteristic patterns of change, or evolutionary signatures, associated with genes and regulatory elements. We have also developed methods for the de novo discovery of recurring combinations of chromatin marks, or chromatin signatures, revealing a small number of distinct chromatin states associated with distinct functional roles, such as enhancer, promoter, insulator, and other regions. Using evolutionary signatures and chromatin signatures together, we have defined a global map of regions of regulatory importance in the Drosophila genome, and a complete map of high-confidence instances of conserved regulatory motifs and motif combinations within them. In parallel, we have studied spatial and temporal patterns of gene expression from in situ images at varying stages of embryonic development, in order to define recurrent patterns of gene expression, or expression primitives, likely to correspond to regulatory signals established by combinations of transcriptional regulators. In this talk, I will describe our progress in each of these areas, and the computational challenge of defining a coherent map between genome sequence and gene expression patterns in development.
This is work by: Chris Bristow, Pouya Kheradpour, Jason Ernst, Rachel Sealfon. Experimental collaborators: Kevin White, Bing Ren, Gary Karpen, Sue Celniker.
In recent years, much research on morphogen gradients has shifted from purely mechanistic questions -how gradients form and how morphogens signa l-to strategic ones- how gradients perform well in the face of various kinds of constraints and perturbations. For example, quite a few cellular and molecular processes have been described as contributing to robustness and precision. Do these processes constitute true strategies of control? Why are there so many of them? Why are some used in certain gradients but not others? Drawing on examples from Drosophila development, I will argue that the constraints imposed by the need to meet multiple performance objectives drives the diversification of strategic approaches, and provides a context within which to understand the perplexing complexity of patterning systems.
Regulation of gene expression along the dorso-ventral axis of the Drosophila embryo is one of the best understood systems of pattern formation. It is especially interesting because of the immediate translation of the fate determination events into morphogenetic processes. In particular the first steps in the establishment of the mesoderm, the formation of the ventral furrow, present a system in which to trace the steps from a fate-determining transcription factor, the transcriptional activator Twist, to the target genes responsible for morphogenetic activity. Six zygotically active Twist target genes are necessary to direct furrow formation. Five directly affect cell shape changes, the sixth is the transcription factor Snail. For the complete understanding of how the dorso-ventral patterning cascade controls morphogenesis via Twist, it will now be necessary to establish the transcriptional events downstream of Snail.
In many developing systems the outcome is buffered to numerous perturbations, ranging from major ones such as separation of the cells at the 2-cell stage in Xenopus (which can lead to one smaller, but normal adult, and an amorphous mass of tissue), to less severe ones such as changes in the ambient temperature or the loss of one copy of a gene. The general question is how systems are buffered against variations in such factors. We address this question in the specific context of scale-invariance: how different size embryos lead to normally-proportioned adults, both in Drosophila and in Xenopus.
I will talk about how we are performing genome alignments of Drosophila at the nucleotide level, and how the alignments can be leveraged to study the functional drivers of genome evolution. The focus of the talk will be on the mathematical questions and issues, with a view towards large scale alignment of thousands of Drosophila genomes, and their study in conjunction with data from high-throughput sequencing based assays in the near future.
The prediction of expression patterns from genomic sequence is an important unsolved problem in modern molecular genetics. Its solution requires an understanding of the transcriptional consequences of particular configurations of bound factors. An important aspect of the problem is to understand how modular enhancers arise from binding sites. We are currently using the eve gene of Drosophila as a testbed for finding the general rules by which sequence controls gene expression in metazoa. We believe that the most informative experimental materials for such studies are instances where the usual additive behavior of enhancers breaks down. Such instances can reveal underlying rules, but the complexity of the experimental phenomena require precise quantitative models for their interpretation. We consider two experimental situations in which modularity breaks down. In one case, a modular enhancer for stripe 2 fused to proximal sequences that do not drive any expression results in a fragment that expresses stripe 7, demonstrating nonadditive behavior. In another case, placing enhancers for stripes 2 and 3 adjacent to one another give rise to a novel expression pattern, an example of another type of nonadditive behavior. I will show how both types of nonadditive behavior can be understood using a quantitative model in conjunction with quantitative expression data from promoter-reporter constructs.
Embryonic development is first controlled by maternal gene products deposited in the egg. Some time after fertilization, this control is transferred to the zygotic genome in a process called the maternal-zygotic transition (MZT). During this time, maternal components are degraded and zygotic genes are activated. In Drosophila, zygotic gene activation starts about one hour after fertilization with a small set of genes activated during cycle 8 to cycle 13. These genes are referred to as the precellular blastoderm genes (pre-CB genes or primary zygotic genes; ten Bosch et al., 2006), while a major burst of zygotic gene activity occurs during and after cellularization. We have identified the zinc-finger protein, Zelda (Zinc-finger early Drosophila activator) that binds specifically to cis-regulatory heptamer motifs called the TAGteam sites, which have been shown to be overrepresented in the upstream regions of many pre-CB genes (ten Bosch et al., 2006; de Renzis et al., 2007). Mutant embryos lacking Zelda are defective in cellular blastoderm formation, and fail to activate many TAGteam containing genes essential for cellularization, sex determination, and pattern formation. Global expression profiling confirmed that Zelda plays a key role in the activation of the early zygotic genome, and suggests that Zelda may also regulate maternal RNA degradation during the MZT (Liang et al., 2008). The discovery of Zelda has provided opportunities to reveal the underline mechanisms of the MZT. We propose that the biological role of Zelda in the preblastoderm embryo is to set the stage for key processes such as cellular blastoderm formation and gastrulation, counting of X chromosomes for dosage compensation and sex determination, and pattern formation, by ensuring the coordinated accumulation of batteries of gene products during the MZT. This early preparedness should allow sufficient time for the formation of molecular machines involved in these processes, and so are ready to spring into action during the prolonged interphase of cycle 14.
We investigate the mechanisms of canalization and embryonic regulation in the morphogenetic field which controls the segment determination in Drosophila. The data used for this characterization are quantitative with cellular resolution in space and about 6 minutes in time. At cycle 13 and the early time classes of cycle 14A the patterns of zygotic segmentation genes show considerable variation in amplitude, the way, time and sequence of domain formation, as well as significant positional variability. Nevertheless, this variation is dynamically reduced, or canalized by the onset of gastrulation. We characterize the epigenetic mechanism of canalization by means of dynamical systems theory supported by quantitative gene expression data.
BMPs play a prominent role in early dorso-ventral patterning in vertebrate and invertebrate embryos. At early stages of embryogenesis, BMPs are produced in a broad domain, abutting a region expressing the extracellular inhibitor Chordin/Sog. How is a morphogen gradient generated within the broad domain of uniform BMP expression? We have used the observation that in Drosophila embryos this gradient is robust to fluctuations in the dose of pathway components, as a basis for a quantitative description of the system, with a focus on the numerical solutions which provide robustness. These solutions present the mechanistic basis for the model, which relies on shuttling of BMP ligands towards the region containing the lowest level of inhibitor, to generate a sharp and robust morphogen gradient. Extrapolation of the findings from flies to Xenopus took into account the presence of an additional ligand in Xenopus (termed ADMP), which behaves in the opposite manner to BMPs: It is expressed on the opposite side of the embryo, at the dorsal side, and its expression is repressed by BMP signaling. Furthermore, in Xenopus the dorso-ventral system is able to scale pattern with size, as demonstrated in the classical Spemann experiments and the manipulations of J. Cooke, providing a further restriction to the numerical solutions. Based on computational and experimental analyses, we have postulated a shuttling mechanism similar to the one identified in Drosophila, which is also able to scale pattern with size by turning off the expression of ADMP according to the size of the embryo.
Work done in collaboration with Danny Ben-Zvi, Avigdor Eldar, Abraham Fainsod, and Naama Barkai.
Eldar A., Dorfman, R., Weiss, D., Ashe, H., Shilo B-Z. and Barkai N. Robustness of the BMP morphogen gradient in Drosophila embryonic patterning. Nature 419, 304-308 (2002).
Ben-Zvi, D., Shilo, B-Z., Fainsod, A. and Barkai, N. Scaling of the BMP activation gradient in Xenopus embryos. Nature 453, 1205-1211 (2008).
Developmental patterning relies on combinatorial action of inductive cues and employs a number of strategies for signal integration. These include regulation of a single gene by multiple transcription factors and biochemical modification of a single transcription factor by multiple signaling pathways. Using the early Drosophila embryo as a model, we show that signal integration can also be mediated by a simple enzymatic network. The anterior structures of Drosophila embryo are specified by two inductive signals. One of them, a homeodomain protein Bicoid, establishes the anteroposterior morphogen gradient. The second (terminal) signal is provided by the localized activation of the MAPK pathway at both anterior and posterior poles. Activated MAPK phosphorylates the uniformly distributed transcriptional repressors Capicua and Groucho, relieving their repression of the terminal gap genes. At the anterior pole, MAPK phosphorylates Bicoid, potentiating its transcriptional effects. Using a combination of biochemical, imaging, and genetic approaches, we demonstrate that modification of Bicoid by MAPK has a reverse effect on MAPK phosphorylation and signaling. In the resulting model, MAPK substrates compete for access to this kinase, establishing an enzyme-substrate competition network that integrates the anterior and terminal signals.
Work done in collaboration with Yoosik Kim, Mathieu Coppey, Leiore Ajuria, Gerardo Jiménez, and Ze'ev Paroush
Bicoid is a homeodomain-containing transcription factor that is expressed in a long-range anterior gradient in the early embryo. Loss of Bicoid function leads to a mutant embryo that lacks all head and thoracic structures. Previous studies have identified approximately 20 target genes that are directly activated by Bicoid. Activation of each target gene involves direct binding of Bicoid to one or more enhancers that appear as rather tightly linked clusters of Bicoid-binding sites. These enhancers direct expression patterns at different positions along the anterior posterior axis (Figure 3), and a major goal is to understand the cis-regulatory logic that controls the differential positioning of different target genes. We are using an integrated approach in pursuit of this goal. First, we use bio-informatics methods and published ChIP-Chip data to identify all clusters of Bicoid-binding sites that are similar to those in the known target genes. Candidate clusters are cloned into reporter genes, transformed into the genome, and tested for in vivo activity by in situ hybridization experiments. While collecting an ever-growing number of Bcd-dependent elements, we are using data mining techniques to identify sequence motifs or binding site arrangements that correlate with target gene positioning. This will lead to specific hypotheses that can be tested by in vitro mutagenesis of binding sites in the context of the reporter genes.
The expression of genes is tightly regulated in spatial and temporal patterns during development. Regulation occurs at several levels both pre- and post-transcriptionally and relies on subtle DNA and RNA sequence signals, forming several classes of regulatory motifs. A systematic understanding of gene regulation relies on the global knowledge of these motifs and their targets. Using whole genome alignments of 12 Drosophila species, we developed a robust framework to discover novel regulatory motifs and predict their individual instances. The predicted targets show highly significant overlap with experimentally derived targets of transcription factors and miRNAs, and recover motif-instances with high sensitivity. The top scoring novel motifs in promoters, introns, 5'UTRs, and intergenic regions correspond to known Drosophila transcription factors. The majority of the novel elements show an enrichment near genes, related by common functions or expression patterns and - similar to known transcription factors - are depleted near ubiquitously expressed genes. The most highly conserved motifs in 3'UTRs and coding sequence are complementary to known miRNA 5'ends, suggesting that coding exons contain physiologically relevant miRNA target sites.
We also predicted and validated Drosophila miRNA genes. We found that the miRNA miR-10 produces functional miRNAs from both hairpin arms and showed that a novel miRNA is produced from the DNA strand opposite of miR-iab-4. Our finding that a single miRNA locus can produce up to 4 functional miRNAs greatly expands the versatility of miRNA-mediated regulation.
The comparative analysis of miRNAs and regulatory motifs in flies reveals principles of gene regulation and provides a framework for future approaches to understand tissue formation and development.
Work done in collaboration with Pouya Kheradpour and Manolis Kellis.
Mathematical models of embryonic development are formulated to illuminate how the spatio-temporal expression of genes that presages the adult body plan of an organism is controlled, but many have limited utility because they oversimplify crucial aspects such as the geometry, the molecular mechanisms, and other components in the system being modeled. To circumvent these limitations we developed a data-driven, 3D, organism-scale model of bone morphogenetic protein (BMP)-mediated embryonic patterning in Drosophila. We tested 7 different receptor/feedback mechanisms and 8 different geometry/gene expression scenarios for their ability to reproduce the mean distributions of pMad signaling in both wild-type and more than twenty different mutant embryos. We found that positive feedback of a secreted BMP binding protein, coupled with the measured embryo geometry, provides the best agreement between model and experiment. The inclusion of all important factors in a 3D model represents a significant step forward in the systems biology of development.
Work done with Hans G. Othmer and Michael B. O'Connor.
Changes in cis-regulatory elements for transcription are thought be an important driving force for the evolution of species. To investigate how changes in cis-regulatory sequence affect a transcriptional regulatory network during development, we study the dorso-ventral (DV) patterning network in four closely related Drosophila species, D. melanogaster, D. simulans, D. erecta, and D. yakuba. Chromatin immunoprecipitation combined with high throughput-sequencing (ChIP-seq) is used to compare the genome-wide distribution of the transcriptional activator Twist and repressor Snail.
This is work by Qiye He and Brianne Patton, in collaboration with Alex Stark and Manolis Kellis.