Jose Davila-Velderrain

Modeling the epigenetic attractors landscape: toward a post-genomic mechanistic understanding of development

Robust temporal and spatial patterns of cell types emerge in the course of normal development in multicellular organisms. The onset of degenerative diseases may result from altered cell fate decisions that give rise to pathological phenotypes. Complex networks of genetic and non-genetic components underlie such normal and altered morphogenetic patterns. Here we focus on the networks of regulatory interactions involved in cell-fate decisions. Such networks modeled as dynamical non-linear systems attain particular stable configurations on gene activity that have been interpreted as cell-fate states. The network structure also restricts the most probable transition patterns among such states. The so-called Epigenetic Landscape (EL), originally proposed by C. H. Waddington, was an early attempt to conceptually explain the emergence of developmental choices as the result of intrinsic constraints (regulatory interactions) shaped during evolution. Thanks to the wealth of molecular genetic and genomic studies, we are now able to postulate gene regulatory networks (GRN) grounded on experimental data, and to derive EL models for specific cases. This, in turn, has motivated several mathematical and computational modeling approaches inspired by the EL concept, that may be useful tools to understand and predict cell-fate decisions and emerging patterns. In order to distinguish between the classical metaphorical EL proposal of Waddington, we refer to the Epigenetic Attractors Landscape (EAL), a proposal that is formally framed in the context of GRNs and dynamical systems theory. In this review we discuss recent EAL modeling strategies, their conceptual basis and their application in studying the emergence of both normal and pathological developmental processes. In addition, we discuss how model predictions can shed light into rational strategies for cell fate regulation, and we point to challenges ahead.

Molecular Evolution Constraints in the Floral Organ Specification Gene Regulatory Network Module Across 18 Angiosperm Genomes

The gene regulatory network of floral organ cell fate specification of Arabidopsis thaliana is a robust developmental regulatory module. Although such finding was proposed to explain the overall conservation of floral organ types and organization among angiosperms, it has not been confirmed that the network components are conserved at the molecular level among flowering plants. Using the genomic data that have accumulated, we address the conservation of the genes involved in this network and the forces that have shaped its evolution during the divergence of angiosperms. We recovered the network gene homologs for 18 species of flowering plants spanning nine families. We found that all the genes are highly conserved with no evidence of positive selection. We studied the sequence conservation features of the genes in the context of their known biological function and the strength of the purifying selection acting upon them in relation to their placement within the network. Our results suggest an association between protein length and sequence conservation, evolutionary rates, and functional category. On the other hand, we found no significant correlation between the strength of purifying selection and gene placement. Our results confirm that the studied robust developmental regulatory module has been subjected to strong functional constraints. However, unlike previous studies, our results do not support the notion that network topology plays a major role in constraining evolutionary rates. We speculate that the dynamical functional role of genes within the network and not just its connectivity could play an important role in constraining evolution.

XAANTAL2 (AGL14) Is an Important Component of the Complex Gene Regulatory Network that Underlies Arabidopsis Shoot Apical Meristem Transitions

In Arabidopsis thaliana, multiple genes involved in shoot apical meristem (SAM) transitions have been characterized, but the mechanisms required for the dynamic attainment of vegetative, inflorescence, and floral meristem (VM, IM, FM) cell fates during SAM transitions are not well understood. Here we show that a MADS-box gene, XAANTAL2 (XAL2/AGL14), is necessary and sufficient to induce flowering, and its regulation is important in FM maintenance and determinacy. xal2 mutants are late flowering, particularly under short-day (SD) condition, while XAL2 overexpressing plants are early flowering, but their flowers have vegetative traits. Interestingly, inflorescences of the latter plants have higher expression levels of LFY, AP1, and TFL1 than wild-type plants. In addition we found that XAL2 is able to bind the TFL1 regulatory regions. On the other hand, the basipetal carpels of the 35S::XAL2 lines lose determinacy and maintain high levels of WUS expression under SD condition. To provide a mechanistic explanation for the complex roles of XAL2 in SAM transitions and the apparently paradoxical phenotypes of XAL2 and other MADS-box (SOC1, AGL24) overexpressors, we conducted dynamic gene regulatory network (GRN) and epigenetic landscape modeling. We uncovered a GRN module that underlies VM, IM, and FM gene configurations and transition patterns in wild-type plants as well as loss and gain of function lines characterized here and previously. Our approach thus provides a novel mechanistic framework for understanding the complex basis of SAM development.

Gene Regulatory Network Models for Floral Organ Determination

Understanding how genotypes map unto phenotypes implies an integrative understanding of the processes regulating cell differentiation and morphogenesis, which comprise development. Such a task requires the use of theoretical and computational approaches to integrate and follow the concerted action of multiple genetic and nongenetic components that hold highly nonlinear interactions. Gene regulatory network (GRN) models have been proposed to approach such task. GRN models have become very useful to understand how such types of interactions restrict the multi-gene expression patterns that characterize different cell-fates. More recently, such temporal single-cell models have been extended to recover the temporal and spatial components of morphogenesis. Since the complete genomic GRN is still unknown and intractable for any organism, and some clear developmental modules have been identified, we focus here on the analysis of well-curated and experimentally grounded small GRN modules. One of the first experimentally grounded GRN that was proposed and validated corresponds to the regulatory module involved in floral organ determination. In this chapter we use this GRN as an example of the methodologies involved in: (1) formalizing and integrating molecular genetic data into the logical functions (Boolean functions) that rule gene interactions and dynamics in a Boolean GRN; (2) the algorithms and computational approaches used to recover the steady-states that correspond to each cell type, as well as the set of initial GRN configurations that lead to each one of such states (i.e., basins of attraction); (3) the approaches used to validate a GRN model using wild type and mutant or overexpression data, or to test the robustness of the GRN being proposed; (4) some of the methods that have been used to incorporate random fluctuations in the GRN Boolean functions and enable stochastic GRN models to address the temporal sequence with which gene configurations and cell fates are attained; (5) the methodologies used to approximate discrete Boolean GRN to continuous systems and their use in further dynamic analyses. The methodologies explained for the GRN of floral organ determination developed here in detail can be applied to any other functional developmental module.

Reshaping the epigenetic landscape during early flower development: induction of attractor transitions by relative differences in gene decay rates

BackgroundGene regulatory network (GRN) dynamical models are standard systems biology tools for the mechanistic understanding of developmental processes and are enabling the formalization of the epigenetic landscape (EL) model.MethodsIn this work we propose a modeling framework which integrates standard mathematical analyses to extend the simple GRN Boolean model in order to address questions regarding the impact of gene specific perturbations in cell-fate decisions during development.ResultsWe systematically tested the propensity of individual genes to produce qualitative changes to the EL induced by modification of gene characteristic decay rates reflecting the temporal dynamics of differentiation stimuli. By applying this approach to the flower specification GRN (FOS-GRN) we uncovered differences in the functional (dynamical) role of their genes. The observed dynamical behavior correlates with biological observables. We found a relationship between the propensity of undergoing attractor transitions between attraction basins in the EL and the direction of differentiation during early flower development - being less likely to induce up-stream attractor transitions as the course of development progresses. Our model also uncovered a potential mechanism at play during the transition from EL basins defining inflorescence meristem to those associated to flower organs meristem. Additionally, our analysis provided a mechanistic interpretation of the homeotic property of the ABC genes, being more likely to produce both an induced inter-attractor transition and to specify a novel attractor. Finally, we found that there is a close relationship between a gene’s topological features and its propensity to produce attractor transitions.ConclusionsThe study of how the state-space associated with a dynamical model of a GRN can be restructured by modulation of genes’ characteristic expression times is an important aid for understanding underlying mechanisms occurring during development. Our contribution offers a simple framework to approach such problem, as exemplified here by the case of flower development. Different GRN models and the effect of diverse inductive signals can be explored within the same framework. We speculate that the dynamical role of specific genes within a GRN, as uncovered here, might give information about which genes are more likely to link a module to other regulatory circuits and signaling transduction pathways.

Dynamic network modelling to understand flowering transition and floral patterning

Differentiation and morphogenetic processes during plant development are particularly robust. At the cellular level, however, plants also show great plasticity in response to environmental conditions, and can even reverse apparently terminal differentiated states with remarkable ease. Can we understand and predict both robust and plastic systemic responses as a general consequence of the non-trivial interplay between intracellular regulatory networks, extrinsic environmental signalling, and tissue-level mechanical constraints? Flower development has become an ideal model system to study these general questions of developmental biology, which are especially relevant to understanding stem cell patterning in plants, animals, and human disease. Decades of detailed study of molecular developmental genetics, as well as novel experimental techniques for in vivo assays in both wild-type and mutant plants, enable the postulation and testing of experimentally grounded mathematical and computational network dynamical models. Research in our group aims to explain the emergence of robust transitions that occur at the shoot apical meristem, as well as flower development, as the result of the collective action of key molecular components in regulatory networks subjected to intra-organismal signalling and extracellular constraints. Here we present a brief overview of recent work from our group, and that of others, focusing on the use of simple dynamical models to address cell-fate specification and cell-state stochastic dynamics during flowering transition and cell-state transitions at the shoot apical meristem of Arabidopsis thaliana. We also focus on how our work fits within the general field of plant developmental modelling, which is being developed by many others.

Descriptive vs. Mechanistic Network Models in Plant Development in the Post-Genomic Era

Network modeling is now a widespread practice in systems biology, as well as in integrative genomics, and it constitutes a rich and diverse scientific research field. A conceptually clear understanding of the reasoning behind the main existing modeling approaches, and their associated technical terminologies, is required to avoid confusions and accelerate the transition towards an undeniable necessary more quantitative, multidisciplinary approach to biology. Herein, we focus on two main network-based modeling approaches that are commonly used depending on the information available and the intended goals: inference-based methods and system dynamics approaches. As far as data-based network inference methods are concerned, they enable the discovery of potential functional influences among molecular components. On the other hand, experimentally grounded network dynamical models have been shown to be perfectly suited for the mechanistic study of developmental processes. How do these two perspectives relate to each other? In this chapter, we describe and compare both approaches and then apply them to a given specific developmental module. Along with the step-by-step practical implementation of each approach, we also focus on discussing their respective goals, utility, assumptions, and associated limitations. We use the gene regulatory network (GRN) involved in Arabidopsis thaliana Root Stem Cell Niche patterning as our illustrative example. We show that descriptive models based on functional genomics data can provide important background information consistent with experimentally supported functional relationships integrated in mechanistic GRN models. The rationale of analysis and modeling can be applied to any other well-characterized functional developmental module in multicellular organisms, like plants and animals.

Gene regulatory network underlying the immortalization of epithelial cells

BackgroundTumorigenic transformation of human epithelial cells in vitro has been described experimentally as the potential result of spontaneous immortalization. This process is characterized by a series of cell–state transitions, in which normal epithelial cells acquire first a senescent state which is later surpassed to attain a mesenchymal stem–like phenotype with a potentially tumorigenic behavior. In this paper we aim to provide a system–level mechanistic explanation to the emergence of these cell types, and to the time–ordered transition patterns that are common to neoplasias of epithelial origin. To this end, we first integrate published functional and well–curated molecular data of the components and interactions that have been found to be involved in such cell states and transitions into a network of 41 molecular components. We then reduce this initial network by removing simple mediators (i.e., linear pathways), and formalize the resulting regulatory core into logical rules that govern the dynamics of each of the network components as a function of the states of its regulators.ResultsComputational dynamic analysis shows that our proposed Gene Regulatory Network model recovers exactly three attractors, each of them defined by a specific gene expression profile that corresponds to the epithelial, senescent, and mesenchymal stem–like cellular phenotypes, respectively. We show that although a mesenchymal stem–like state can be attained even under unperturbed physiological conditions, the likelihood of converging to this state is increased when pro–inflammatory conditions are simulated, providing a systems–level mechanistic explanation for the carcinogenic role of chronic inflammatory conditions observed in the clinic. We also found that the regulatory core yields an epigenetic landscape that restricts temporal patterns of progression between the steady states, such that recovered patterns resemble the time–ordered transitions observed during the spontaneous immortalization of epithelial cells, both in vivo and in vitro.ConclusionOur study strongly suggests that the in vitro tumorigenic transformation of epithelial cells, which strongly correlates with the patterns observed during the pathological progression of epithelial carcinogenesis in vivo, emerges from underlying regulatory networks involved in epithelial trans–differentiation during development.

Epigenetic Landscape Models: The Post-Genomic Era

Complex networks of regulatory interactions orchestrate developmental processes in multicellular organisms. Such a complex internal structure intrinsically constrains cellular behavior allowing only a reduced set of attainable and observable cellular states or cell types. Thus, a multicellular system undergoes cell fate decisions in a robust manner in the course of its normal development. The epigenetic landscape (EL) model originally proposed by C.H. Waddington was an early attempt to integrate these processes in a universal conceptual model of development. Since then, a wealth of experimental data has accumulated, the general mechanisms of gene regulation have been uncovered, and the placement of specific molecular components within modular gene regulatory networks (GRN) has become a common practice. This has motivated the development of mathematical and computational models inspired by the EL aiming to integrate molecular data and gain a better understanding of development, and hopefully predict cell differentiation and reprogramming events. Both deterministic and stochastic dynamical models have been used to described cell state transitions. In this review, we describe recent EL models, emphasising that the construction of an explicit landscape from a GRN is not the only way to implement theoretical models consistent with the conceptual basis of the EL. Moreover, models based on the EL have been shown to be useful in the study of morphogenic processes and not just cell differentiation. Here we describe the distinct approaches, comparing their strengths and weaknesses and the kind of biological questions that they have been able to address. We also point to challenges ahead.

Comparison of Gene Expression Profiles Between Pansensitive and Multidrug-Resistant Strains of Mycobacterium tuberculosis

Mycobacterium tuberculosis has developed resistance to anti-tuberculosis first-line drugs. Multidrug-resistant strains complicate the control of tuberculosis and have converted it into a worldwide public health problem. Mutational studies of target genes have tried to envisage the resistance in clinical isolates; however, detection of these mutations in some cases is not sufficient to identify drug resistance, suggesting that other mechanisms are involved. Therefore, the identification of new markers of susceptibility or resistance to first-line drugs could contribute (1) to specifically diagnose the type of M. tuberculosis strain and prescribe an appropriate therapy, and (2) to elucidate the mechanisms of resistance in multidrug-resistant strains. In order to identify specific genes related to resistance in M. tuberculosis, we compared the gene expression profiles between the pansensitive H37Rv strain and a clinical CIBIN:UMF:15:99 multidrug-resistant isolate using microarray analysis. Quantitative real-time PCR confirmed that in the clinical multidrug-resistant isolate, the esxG, esxH, rpsA, esxI, and rpmI genes were upregulated, while the lipF, groES, and narG genes were downregulated. The modified genes could be involved in the mechanisms of resistance to first-line drugs in M. tuberculosis and could contribute to increased efficiency in molecular diagnosis approaches of infections with drug-resistant strains.