Search

International audience; Leveraging systems biology approaches, we illustrate how metabolically distinct species of Clostridia protect against or worsen Clostridioides difficile infection in mice by modulating the pathogen's colonization, growth, and virulence to impact host survival. Gnotobiotic mice colonized with the amino acid fermenter Paraclostridium bifermentans survive infection with reduced disease severity, while mice colonized with the butyrate-producer, Clostridium sardiniense, succumb more rapidly. Systematic in vivo analyses revealed how each commensal alters the gut-nutrient environment to modulate the pathogen's metabolism, gene regulatory networks, and toxin production. Oral administration of P. bifermentans rescues conventional, clindamycin-treated mice from lethal C. difficile infection in a manner similar to that of monocolonized animals, thereby supporting the therapeutic potential of this commensal species. Our findings lay the foundation for mechanistically informed therapies to counter C. difficile disease using systems biology approaches to define host-commensal-pathogen interactions in vivo.

International audience; Pathogen-imposed selection pressures have been paramount during human evolution. Detecting such selection signatures in ancient and modern human genomes can thus help us to identify genes of temporal and spatial immunological relevance. Admixture with ancient hominins and between human populations has been a source of genetic diversity open to selection by infections. Furthermore, cultural transitions, such as the advent of agriculture, have exposed humans to new microbial threats, with impacts on host defense mechanisms. The integration of population genetics and systems immunology holds great promise for the increased understanding of the factors driving immune response variation between individuals and populations.

International audience; The field of system biology poses many challenges to computer scientists. Regulatory networks, intra- and extra-cellular signalling pathways, metabolic networks and whole cells models suffer both from high conceptual difficulties and from a huge combinatorial complexity. From the description of these models, by organising and collecting facts based on knowledge representation databases, to the execution of these models, through the abstraction of their main properties, formal methods offer promising potential approaches. While existing formal frameworks can be adapted to cope with the specific kinds of problems, new formal methods emerge from the confrontation with this new field of application. Besides the intrinsic combinatorial complexity, biological systems are characterised by the fact that they have not been humanly designed. Their dynamics are driven by competition for shared resources, time- and concentration-scale separation, and non-linear feedback loops. Many components have antagonistic effects according to the context. It is impossible to understand from which mechanisms the collective behaviour of these models come from.Mainstream quantitative approaches (based on ODEs) suffer from various limitations. To cope with the combinatorial complexity, series of simplifications are usually required. This results in several drawbacks: not only does it usually restrict the context of the application of the models, but also, it is likely to skew the model in the direction of what is intended by the modeller. Moreover, their results are specific for particular sets of parameters and they are difficult to generalize. As a consequence, they usually offer an unsatisfying level of confidence.Formal methods provide various means to increase confidence in models and to compute their properties. Formal languages provide hierarchical constructions to collect information about mechanistic interactions, in a more goal agnostic manner. Formal methods also assist the modeller to gradually refine/abstract information about a given process by including more/fewer details from the literature. Symbolic static analyses not only enhance confidence in models, but also provide new perspectives on models by efficiently computing informative invariants. Causal analyses provide explanations for the potential scenarios for events of interest to occur. The behaviour of models can be captured while avoiding the exploration of the different interleaving of these massively distributed agent-centric models. Lastly, model reduction provides new systems while identifying the elements that mainly drive the dynamics of these models.This Special Issue is dedicated to formal methods developed for the analysis of biological systems and their application to current open questions in biology. Its scope covers the design of specific language and modelling tools, the design of static analyses, causal analyses, model reduction techniques. We are also interested in feedback on the usage of formal approaches in the design and the analysis of models to address biological issues.

Gene expression is a noisy process: in constant environment and genotype, cell to cell variability occurs because of randomness of biochemical reactions..., Biochemical reactions within individual cells result from the interactions of molecules, typically in small numbers. Consequently, the inherent stochasticity of binding and diffusion processes generates noise along the cascade that leads to the synthesis of a protein from its encoding gene. As a result, isogenic cell populations display phenotypic variability even in homogeneous environments. The extent and consequences of this stochastic gene expression have only recently been assessed on a genome-wide scale, owing, in particular, to the advent of single-cell transcriptomics. However, the evolutionary forces shaping this stochasticity have yet to be unraveled. Here, we take advantage of two recently published data sets for the single-cell transcriptome of the domestic mouse Mus musculus to characterize the effect of natural selection on gene-specific transcriptional stochasticity. We show that noise levels in the mRNA distributions (also known as transcriptional noise) significantly correlate with three-dimensional nuclear domain organization, evolutionary constraints on the encoded protein, and gene age. However, the position of the encoded protein in a biological pathway is the main factor that explains observed levels of transcriptional noise, in agreement with models of noise propagation within gene networks. Because transcriptional noise is under widespread selection, we argue that it constitutes an important component of the phenotype and that variance of expression is a potential target of adaptation. Stochastic gene expression should therefore be considered together with the mean expression level in functional and evolutionary studies of gene expression.

International audience; Metabolic control in Escherichia coli is a complex process involving multilevel regulatory systems but the involvement of post-transcriptional regulation is uncertain. The post-transcriptional factor CsrA is stated as being the only regulator essential for the use of glycolytic substrates. A dozen enzymes in the central carbon metabolism (CCM) have been reported as potentially controlled by CsrA, but its impact on the CCM functioning has not been demonstrated. Here, a multiscale analysis was performed in a wild-type strain and its isogenic mutant attenuated for CsrA (including growth parameters, gene expression levels, metabolite pools, abundance of enzymes and fluxes). Data integration and regulation analysis showed a coordinated control of the expression of glycolytic enzymes. This also revealed the imbalance of metabolite pools in the csrA mutant upper glycolysis, before the phosphofructokinase PfkA step. This imbalance is associated with a glucose-phosphate stress. Restoring PfkA activity in the csrA mutant strain suppressed this stress and increased the mutant growth rate on glucose. Thus, the carbon storage regulator system is essential for the effective functioning of the upper glycolysis mainly through its control of PfkA. This work demonstrates the pivotal role of post-transcriptional regulation to shape the carbon metabolism.

International audience; Given the complexity of biological neural circuits and of their component cells and synapses, building and simulating robust, well-validated, detailed models increasingly surpasses the resources of an individual researcher or small research group. In this article, I will briefly review possible solutions to this problem, argue for open, collaborative modelling as the optimal solution for advancing neuroscience knowledge, and identify potential bottlenecks and possible solutions.

Many protein interactions are mediated by small linear motifs interacting specifically with defined families of globular domains. Quantifying the specificity of a motif requires measuring and comparing its binding affinities to all its putative target domains. To this aim, we developed the high-throughput holdup assay, a chromatographic approach that can measure up to a thousand domain-motif equilibrium binding affinities per day. Extracts of overexpressed domains are incubated with peptide-coated resins and subjected to filtration. Binding affinities are deduced from microfluidic capillary electrophoresis of flow-throughs. After benchmarking the approach on 210 PDZ-peptide pairs with known affinities, we determined the affinities of two viral PDZ-binding motifs derived from Human Papillomavirus E6 oncoproteins for 209 PDZ domains covering 79% of the human PDZome. We obtained exquisite sequence-dependent binding profiles, describing quantitatively the PDZome recognition specificity of each motif. This approach, applicable to many categories of domain-ligand interactions, has a wide potential for quantifying the specificities of interactomes.

International audience; Gravity profoundly influences plant growth and development. Plants respond to changes in orientation by using gravitropic responses to modify their growth. Cholodny and Went hypothesized over 80 years ago that plants bend in response to a gravity stimulus by generating a lateral gradient of a growth regulator at an organ's apex, later found to be auxin. Auxin regulates root growth by targeting Aux/IAA repressor proteins for degradation. We used an Aux/IAA-based reporter, domain II (DII)-VENUS, in conjunction with a mathematical model to quantify auxin redistribution following a gravity stimulus. Our multidisciplinary approach revealed that auxin is rapidly redistributed to the lower side of the root within minutes of a 90° gravity stimulus. Unexpectedly, auxin asymmetry was rapidly lost as bending root tips reached an angle of 40° to the horizontal. We hypothesize roots use a "tipping point" mechanism that operates to reverse the asymmetric auxin flow at the midpoint of root bending. These mechanistic insights illustrate the scientific value of developing quantitative reporters such as DII-VENUS in conjunction with parameterized mathematical models to provide high-resolution kinetics of hormone redistribution.

Fragile X syndrome (FXS) is the first cause of inherited intellectual disability, due to the silencing of the X-linked Fragile X Mental Retardation 1 gene encoding the RNA-binding protein FMRP. While extensive studies have focused on the cellular and molecular basis of FXS, neither human Fragile X patients nor the mouse model of FXS—the Fmr1-null mouse—have been profiled systematically at the metabolic and neurochemical level to provide a complementary perspective on the current, yet scattered, knowledge of FXS. Using proton high-resolution magic angle spinning nuclear magnetic resonance (1H HR-MAS NMR)-based metabolic profiling, we have identified a metabolic signature and biomarkers associated with FXS in various brain regions of Fmr1-deficient mice. Our study highlights for the first time that Fmr1 gene inactivation has profound, albeit coordinated consequences in brain metabolism leading to alterations in: (1) neurotransmitter levels, (2) osmoregulation, (3) energy metabolism, and (4) oxidative stress response. To functionally connect Fmr1-deficiency to its metabolic biomarkers, we derived a functional interaction network based on the existing knowledge (literature and databases) and show that the FXS metabolic response is initiated by distinct mRNA targets and proteins interacting with FMRP, and then relayed by numerous regulatory proteins. This novel “integrated metabolome and interactome mapping” (iMIM) approach advantageously unifies novel metabolic findings with previously unrelated knowledge and highlights the contribution of novel cellular pathways to the pathophysiology of FXS. These metabolomic and integrative systems biology strategies will contribute to the development of potential drug targets and novel therapeutic interventions, which will eventually benefit FXS patients.

This paper proposes an abstract mathematical frame for describing some features of cognitive and biological time. We focus here on the so called "extended present" as a result of protentional and retentional activities (memory and anticipation). Memory, as retention, is treated in some physical theories (relaxation phenomena, which will inspire our approach), while protention (or anticipation) seems outside the scope of physics. We then suggest a simple functional representation of biological protention. This allows us to introduce the abstract notion of "biological inertia"., Comment: This paper was made possible only as part of an extended collaboration with Francis Bailly (see references), a dear friend and "ma\^itre \'a penser", who contributed to the key ideas. Francis passed away in february 2008: we continue here our inspiring discussions and joint work

Background Reverse engineering in systems biology entails inference of gene regulatory networks from observational data. This data typically include gene expression measurements of wild type and mutant cells in response to a given stimulus. It has been shown that when more than one type of experiment is used in the network inference process the accuracy is higher. Therefore the development of generally applicable and effective methodologies that embed multiple sources of information in a single computational framework is a worthwhile objective. Results This paper presents a new method for network inference, which uses multi-objective optimisation (MOO) to integrate multiple inference methods and experiments. We illustrate the potential of the methodology by combining ODE and correlation-based network inference procedures as well as time course and gene inactivation experiments. Here we show that our methodology is effective for a wide spectrum of data sets and method integration strategies. Conclusions The approach we present in this paper is flexible and can be used in any scenario that benefits from integration of multiple sources of information and modelling procedures in the inference process. Moreover, the application of this method to two case studies representative of bacteria and vertebrate systems has shown potential in identifying key regulators of important biological processes.

International audience; Quantitative microscopy relies on imaging of large cell numbers but is often hampered by time-consuming manual selection of specific cells. The 'Micropilot' software automatically detects cells of interest and launches complex imaging experiments including three-dimensional multicolor time-lapse or fluorescence recovery after photobleaching in live cells. In three independent experimental setups this allowed us to statistically analyze biological processes in detail and is thus a powerful tool for systems biology.

International audience; Characterizing the behavior and robustness of enzymatic networks with numerous variables and unknown parameter values is a major challenge in biology, especially when some enzymes have counter-intuitive properties or switch-like behavior between activation and inhibition. In this paper, we propose new methodological and tool-supported contributions, based on the intuitive formalism of temporal logic, to express in a rigorous manner arbitrarily complex dynamical properties. Our multi-step analysis allows efficient sampling of the parameter space in order to define feasible regions in which the model exhibits imposed or experimentally observed behaviors. In a first step, an algorithmic methodology involving sensitivity analysis is conducted to determine bifurcation thresholds for a limited number of model parameters or initial conditions. In a second step, this boundary detection is supplemented by a global robustness analysis, based on quasi-Monte Carlo approach that takes into account all model parameters. We apply this method to a well-documented enzymatic reaction network describing collagen proteolysis by matrix metalloproteinase MMP2 and membrane type 1 metalloproteinase (MT1-MMP) in the presence of tissue inhibitor of metalloproteinase TIMP2. For this model, our method provides an extended analysis and quantification of network robustness toward paradoxical TIMP2 switching activity between activation or inhibition of MMP2 production. Further implication of our approach is illustrated by demonstrating and analyzing the possible existence of oscillatory behaviors when considering an extended open configuration of the enzymatic network. Notably, we construct bifurcation diagrams that specify key parameters values controlling the co-existence of stable steady and non-steady oscillatory proteolytic dynamics.

The present article discusses the possibility that catalysed chemical networks can evolve. Even simple enzyme-catalysed chemical reactions can display this property. The example studied is that of a two-substrate proteinoid, or enzyme, reaction displaying random binding of its substrates A and B. The fundamental property of such a system is to display either emergence or integration depending on the respective values of the probabilities that the enzyme has bound one of its substrate regardless it has bound the other substrate, or, specifically, after it has bound the other substrate. There is emergence of information if p(A)>p(AB) and p(B)>p(BA). Conversely, if p(A)

This paper proposes an abstract mathematical frame for describing some features of biological time. The key point is that usual physical (linear) representation of time is insufficient, in our view, for the understanding key phenomena of life, such as rhythms, both physical (circadian, seasonal ...) and properly biological (heart beating, respiration, metabolic ...). In particular, the role of biological rhythms do not seem to have any counterpart in mathematical formalization of physical clocks, which are based on frequencies along the usual (possibly thermodynamical, thus oriented) time. We then suggest a functional representation of biological time by a 2-dimensional manifold as a mathematical frame for accommodating autonomous biological rhythms. The "visual" representation of rhythms so obtained, in particular heart beatings, will provide, by a few examples, hints towards possible applications of our approach to the understanding of interspecific differences or intraspecific pathologies. The 3-dimensional embedding space, needed for purely mathematical reasons, allows to introduce a suitable extra-dimension for "representation time", with a cognitive significance., Presented in an invited Lecture, conference "Biologie e selezioni naturali", Florence, December 4-8, 2009

Background Comparison of metabolic networks across species is a key to understanding how evolutionary pressures shape these networks. By selecting taxa representative of different lineages or lifestyles and using a comprehensive set of descriptors of the structure and complexity of their metabolic networks, one can highlight both qualitative and quantitative differences in the metabolic organization of species subject to distinct evolutionary paths or environmental constraints. Results We used a novel representation of metabolic networks, termed network of interacting pathways or NIP, to focus on the modular, high-level organization of the metabolic capabilities of the cell. Using machine learning techniques we identified the most relevant aspects of cellular organization that change under evolutionary pressures. We considered the transitions from prokarya to eukarya (with a focus on the transitions among the archaea, bacteria and eukarya), from unicellular to multicellular eukarya, from free living to host-associated bacteria, from anaerobic to aerobic, as well as the acquisition of cell motility or growth in an environment of various levels of salinity or temperature. Intuitively, we expect organisms with more complex lifestyles to have more complex and robust metabolic networks. Here we demonstrate for the first time that such organisms are not only characterized by larger, denser networks of metabolic pathways but also have more efficiently organized cross communications, as revealed by subtle changes in network topology. These changes are unevenly distributed among metabolic pathways, with specific categories of pathways being promoted to more central locations as an answer to environmental constraints. Conclusions Combining methods from graph theory and machine learning, we have shown here that evolutionary pressures not only affects gene and protein sequences, but also specific details of the complex wiring of functional modules in the cell. This approach allows the identification and quantification of those changes, and provides an overview of the evolution of intracellular systems.

Background A systems biology interpretation of genome-scale RNA interference (RNAi) experiments is complicated by scope, experimental variability and network signaling robustness. Over representation approaches (ORA), such as the Hypergeometric or z-score, are an established statistical framework used to associate RNA interference effectors to biologically annotated gene sets or pathways. These methods, however, do not directly take advantage of our growing understanding of the interactome. Furthermore, these methods can miss partial pathway activation and may be biased by protein complexes. Here we present a novel ORA, protein interaction permutation analysis (PIPA), that takes advantage of canonical pathways and established protein interactions to identify pathways enriched for protein interactions connecting RNAi hits. Results We use PIPA to analyze genome-scale siRNA cell growth screens performed in HeLa and TOV cell lines. First we show that interacting gene pair siRNA hits are more reproducible than single gene hits. Using protein interactions, PIPA identifies enriched pathways not found using the standard Hypergeometric analysis including the FAK cytoskeletal remodeling pathway. Different branches of the FAK pathway are distinctly essential in HeLa versus TOV cell lines while other portions are uneffected by siRNA perturbations. Enriched hits belong to protein interactions associated with cell cycle regulation, anti-apoptosis, and signal transduction. Conclusion PIPA provides an analytical framework to interpret siRNA screen data by merging biologically annotated gene sets with the human interactome. As a result we identify pathways and signaling hypotheses that are statistically enriched to effect cell growth in human cell lines. This method provides a complementary approach to standard gene set enrichment that utilizes the additional knowledge of specific interactions within biological gene sets.

International audience; Diseases are complex systems. Modelling them, i.e. systems physiopathology, is a quite demanding, complicated, multidimensional, multiscale process. As such, in order to achieve the goal of the model and further to optimise a rather-time and resource-consuming process, a relevant and easy to practice methodology is required. It includes guidance for validation. Also, the model development should be managed as a complicated process, along a strategy which has been elaborated in the beginning. It should be flexible enough to meet every case. A model is a representation of the available knowledge. All available knowledge does not have the same level of evidence and, further, there is a large variability of the values of all parameters (e.g. affinity constant or ionic current) across the literature. In addition, in a complex biological system there are always values lacking for a few or sometimes many parameters. All these three aspects are sources of uncertainty on the range of validity of the models and raise unsolved problems for designing a relevant model. Tools and techniques for integrating the parameter range of experimental values, level of evidence and missing data are needed.

We present the current state of the development of the SAPHIR project (a Systems Approach for PHysiological Integration of Renal, cardiac and respiratory function). The aim is to provide an open-source multi-resolution modelling environment that will permit, at a practical level, a plug-and-play construction of integrated systems models using lumped-parameter components at the organ/tissue level while also allowing focus on cellular- or molecular-level detailed sub-models embedded in the larger core model. Thus, an in silico exploration of gene-to-organ-to-organism scenarios will be possible, while keeping computation time manageable. As a first prototype implementation in this environment, we describe a core model of human physiology targeting the short- and long-term regulation of blood pressure, body fluids and homeostasis of the major solutes. In tandem with the development of the core models, the project involves database implementation and ontology development.

Background High-throughput genome biological experiments yield large and multifaceted datasets that require flexible and user-friendly analysis tools to facilitate their interpretation by life scientists. Many solutions currently exist, but they are often limited to specific steps in the complex process of data management and analysis and some require extensive informatics skills to be installed and run efficiently. Results We developed the Annotation, Mapping, Expression and Network (AMEN) software as a stand-alone, unified suite of tools that enables biological and medical researchers with basic bioinformatics training to manage and explore genome annotation, chromosomal mapping, protein-protein interaction, expression profiling and proteomics data. The current version provides modules for (i) uploading and pre-processing data from microarray expression profiling experiments, (ii) detecting groups of significantly co-expressed genes, and (iii) searching for enrichment of functional annotations within those groups. Moreover, the user interface is designed to simultaneously visualize several types of data such as protein-protein interaction networks in conjunction with expression profiles and cellular co-localization patterns. We have successfully applied the program to interpret expression profiling data from budding yeast, rodents and human. Conclusion AMEN is an innovative solution for molecular systems biological data analysis freely available under the GNU license. The program is available via a website at the Sourceforge portal which includes a user guide with concrete examples, links to external databases and helpful comments to implement additional functionalities. We emphasize that AMEN will continue to be developed and maintained by our laboratory because it has proven to be extremely useful for our genome biological research program.

International audience; BACKGROUND: Much recent work in bioinformatics has focused on the inference of various types of biological networks, representing gene regulation, metabolic processes, protein-protein interactions, etc. A common setting involves inferring network edges in a supervised fashion from a set of high-confidence edges, possibly characterized by multiple, heterogeneous data sets (protein sequence, gene expression, etc.). RESULTS: Here, we distinguish between two modes of inference in this setting: direct inference based upon similarities between nodes joined by an edge, and indirect inference based upon similarities between one pair of nodes and another pair of nodes. We propose a supervised approach for the direct case by translating it into a distance metric learning problem. A relaxation of the resulting convex optimization problem leads to the support vector machine (SVM) algorithm with a particular kernel for pairs, which we call the metric learning pairwise kernel. This new kernel for pairs can easily be used by most SVM implementations to solve problems of supervised classification and inference of pairwise relationships from heterogeneous data. We demonstrate, using several real biological networks and genomic datasets, that this approach often improves upon the state-of-the-art SVM for indirect inference with another pairwise kernel, and that the combination of both kernels always improves upon each individual kernel. CONCLUSION: The metric learning pairwise kernel is a new formulation to infer pairwise relationships with SVM, which provides state-of-the-art results for the inference of several biological networks from heterogeneous genomic data.

Motivation: The modeling and simulation of genetic regulatory networks have created the need for tools for model validation. The main challenges of model validation are the achievement of a match between the precision of model predictions and experimental data, as well as the efficient and reliable comparison of the predictions and observations. Results: We present an approach towards the validation of models of genetic regulatory networks addressing the above challenges. It combines a method for qualitative modeling and simulation with techniques for model checking, and is supported by a new version of the computer tool Genetic Network Analyzer (GNA). The model-validation approach has been applied to the analysis of the network controlling the nutritional stress response in Escherichia coli. Availability: GNA and the model of the stress response network are available at http://www-helix.inrialpes.fr/gna Contact: Hidde.de-Jong@inrialpes.fr

With the completion of the many genomes, genetics is positioned to meet physiology. In this review, we summarize the coming of "systems metabolism" in mammals through the use of the mouse, as a model system to study metabolism. Building on mouse genetics with increasingly sophisticated clinical and molecular phenotyping strategies has enabled scientists to now tackle complex biomedical questions, such as those related to the pathogenesis of the common metabolic disorders. The ultimate goal of such strategies will be to mimic human metabolism with the click of a mouse.

The definition of theoretical biology may be controversial. Indeed, the basic definition of theoretical biology classifies it as biomathematics. It assumes that theoretical biology is divided into, at least, four subfields: Biological mathematical modeling, Relational biology/Complex systems biology, Bioinformatics and Biomodeling Computation. In epistemology, theoretical biology is defined as the action of ‘‘theorizing’’ life sciences. It aims to introduce laws; it highlights some biological phenomena and makes possible the identification of new biological principles. The latter definition does not imply a direct relationship with mathematics. In both cases, theoretical biology is a strongly multidisciplinary science, for the tools imported from other science and for the shared ideas. This controversy was one of the main concerns of the French-speaking Society for Theoretical Biology (SFBT). In 1985, it succeeded the Association for the development of Theoretical Methods for Biology (AMTB). The initial purpose of AMTB was to organize Schools; rapidly it took a more general objective: ‘‘the development of methods and theoretical formalisms useful for fundamental biological research and its practical applications’’. The origin of SFBT Conferences comes from 1975, when Jean-Henri Maresquelle, used to invite colleagues with different scientific backgrounds to round tables, twice a year. These meetings lead to scientifically fruitful multidisciplinary

Motivation: The rational design of biological networks with prescribed functions is limited to gene circuits of a few genes. Larger networks involve complex interactions with many parameters and the use of automated computational tools can be very valuable. We propose a new tool to design transcriptional networks with targeted behavior that could be used to better understand the design principles of genetic circuits.Results: We have implemented a Simulated Annealing optimization algorithm that explores throughout the space of transcription networks to obtain a specific behavior. The software outputs a transcriptional network with all the corresponding kinetic parameters in SBML format. We provide examples of transcriptional circuits with logical and oscillatory behaviors. Our tool can also be applied to design networks with multiple external input and output genes.Availability: The software, a tutorial manual, parameter sets and examples are freely available at http://synth-bio.yi.org/genetdes.htmlContact: Alfonso.Jaramillo@polytechnique.edu

Alternative cell differentiation pathways are believed to arise from the concerted action of signalling pathways and transcriptional regulatory networks. However, the prediction of mammalian cell differentiation from the knowledge of the presence of specific signals and transcriptional factors is still a daunting challenge. In this respect, the vertebrate hematopoietic system, with its many branching differentiation pathways and cell types, is a compelling case study. In this paper, we propose an integrated, comprehensive model of the regulatory network and signalling pathways controlling Th cell differentiation. As most available data are qualitative, we rely on a logical formalism to perform extensive dynamical analyses. To cope with the size and complexity of the resulting network, we use an original model reduction approach together with a stable state identification algorithm. To assess the effects of heterogeneous environments on Th cell differentiation, we have performed a systematic series of simulations considering various prototypic environments. Consequently, we have identified stable states corresponding to canonical Th1, Th2, Th17 and Treg subtypes, but these were found to coexist with other transient hybrid cell types that co-express combinations of Th1, Th2, Treg and Th17 markers in an environment-dependent fashion. In the process, our logical analysis highlights the nature of these cell types and their relationships with canonical Th subtypes. Finally, our logical model can be used to explore novel differentiation pathways in silico., Author Summary T lymphocytes play a key role in the regulation of the immune response in mammals. Various T-helper subtypes (Th1, Th2, Th17, Treg,…) have been identified over the years, characterised by the expression of specific transcription factors and cytokines, which have a critical influence on the selection of different immune responses, driving pro-inflammatory or allergic responses, promoting alternative antibody classes, or preventing (auto)immunity by inhibiting the activation and proliferation of other cells. To gain insight into the heterogeneity and the plasticity of late T-helper lineages, we have built an integrated model of the regulatory network and signalling pathways controlling Th cell differentiation. Relying on a logical modelling framework, we have performed a systematic series of simulations to assess the effects of heterogeneous environments on Th cell differentiation. We have identified stable states corresponding to canonical Th1, Th2, Th17 and Treg subtypes, but also to hybrid cell types co-expressing combinations of Th1, Th2, Treg and Th17 markers in an environment-dependent fashion. Our analysis highlights the nature of these cell types and their relationships with canonical Th subtypes.