No description available.

The drug discovery process is challenging, time-consuming, and costly, with drug target identification being an essential step in developing effective therapies. Drug repurposing offers a strategy for identifying new uses for existing drugs, aiming to simplify the process. Machine learning models and network analysis methods have demonstrated promise in both drug target identification and repurposing, providing powerful tools for analyzing complex biological data. This thesis will explore the applications of neural networks and multilayer biological networks for drug repurposing opportunities and network inference problems applied to signaling pathways. A novel machine learning and network-based workflow is presented for identifying drug targets for cystinosis, a rare disease that causes progressive kidney disease, currently lacking effective therapies to prevent the kidney failure. This approach permits to recapitulate the disease mechanisms in the context of renal tubular physiology and identify candidate drug targets for further validation using a cross-species workflow and disease-relevant screening technologies. While machine learning approaches have shown promise, they often need more mechanistic understanding, which is necessary for robust drug target identification and repurposing strategies. Mechanistic models provide crucial insights into the underlying biological mechanisms, complementing machine learning techniques. However, inferring mechanistic signaling networks from omics data poses challenges due to non-identifiability, resulting in multiple valid solutions consistent with the data. After that, the focus shifts towards quantifying signaling model diversity through solver-agnostic solution sampling with CORNETO, an ongoing effort that aims to unify network inference problems via constrained optimization. Mechanistic signaling networks can be inferred from omics data and prior knowledge using combinatorial optimization and mathematical solvers to find the optimal network. However, this problem is in general, non-identifiable, and several solutions may be equally valid. Ignoring the existence of these alternative solutions leads to an incomplete picture of the hypothesis space of consistent mechanistic signaling networks. To alleviate this issue, an algorithm to explore the space of alternative solutions and to conduct sensitivity analysis on the optimal solution is implemented and presented. These algorithms are applied to data from pancreatic cancer cell lines treated with kinase inhibitors to study cellular responses to drug perturbations by inferring mechanistic signaling networks from omics data.

Perturbation experiments for example using RNA interference (RNAi) offer an attractive way to elucidate gene function in a high throughput fashion. The placement of hit genes in their functional context and the inference of underlying networks from such data, however, are challenging tasks. One of the problems in network inference is the exponential number of possible network topologies for a given number of genes. Here, we introduce a novel mathematical approach to address this question. We formulate network inference as a linear optimization problem, which can be solved efficiently even for large-scale systems. We use simulated data to evaluate our approach, and show improved performance in particular on larger networks over state-of-the art methods. We achieve increased sensitivity and specificity, as well as a significant reduction in computing time. Furthermore, we show superior performance on noisy data. We then apply our approach to study the intracellular signaling of human primary nave CD4+ T-cells, as well as ErbB signaling in trastuzumab resistant breast cancer cells. In both cases, our approach recovers known interactions and points to additional relevant processes. In ErbB signaling, our results predict an important role of negative and positive feedback in controlling the cell cycle progression.

Perturbation experiments for example using RNA interference (RNAi) offer an attractive way to elucidate gene function in a high throughput fashion. The placement of hit genes in their functional context and the inference of underlying networks from such data, however, are challenging tasks. One of the problems in network inference is the exponential number of possible network topologies for a given number of genes. Here, we introduce a novel mathematical approach to address this question. We formulate network inference as a linear optimization problem, which can be solved efficiently even for large-scale systems. We use simulated data to evaluate our approach, and show improved performance in particular on larger networks over state-of-the art methods. We achieve increased sensitivity and specificity, as well as a significant reduction in computing time. Furthermore, we show superior performance on noisy data. We then apply our approach to study the intracellular signaling of human primary nave CD4+ T-cells, as well as ErbB signaling in trastuzumab resistant breast cancer cells. In both cases, our approach recovers known interactions and points to additional relevant processes. In ErbB signaling, our results predict an important role of negative and positive feedback in controlling the cell cycle progression.

Mouse embryonic stem cells (mESCs) can be maintained in a proliferative and undifferentiated state over many passages (self-renewal) while retaining the potential to give rise to every cell type of the organism (pluripotency). Autocrine FGF4/Erk signalling has been identified as a major stimulus for fate decisions and lineage commitment in these cells. Recent findings on serum-free culture conditions with specific inhibitors (known as 2i) demonstrate that the inhibition of this pathway reduces transcription factor heterogeneity and is vital to maintain ground state pluripotency of mESCs. We suggest a novel mathematical model to explicitly integrate FGF4/Erk signalling into an interaction network of key pluripotency factors (namely Oct4, Sox2, Nanog and Rex1). The envisaged model allows to explore whether and how proposed mechanisms and feedback regulations can account for different expression patterns in mESC cultures. We demonstrate that an FGF4/Erk-mediated negative feedback is sufficient to induce molecular heterogeneity with respect to Nanog and Rex1 expression and thus critically regulates the propensity for differentiation and the loss of pluripotency. Furthermore, we compare simulation results on the transcription factor dynamics in different self-renewing states and during differentiation with experimental data on a Rex1GFPd2 reporter cell line using flow cytometry and qRT-PCR measurements. Concluding from our results we argue that interaction between FGF4/Erk signalling and Nanog expression qualifies as a key mechanism to manipulate mESC pluripotency. In particular, we infer that ground state pluripotency under 2i is achieved by shifting stable expression pattern of Nanog from a bistable into a monostable regulation impeding stochastic state transitions. Furthermore, we derive testable predictions on altering the degree of Nanog heterogeneity and on the frequency of state transitions in LIF/serum conditions to challenge our model assumptions.

Mycobacterium tuberculosis, the causative organism of the disease tuberculosis (TB) in humans, leads to nearly two million deaths each year. This versatile pathogen can exist in highly distinct physiological states such as asymptomatic latent TB infection where bacilli lie dormant or as active TB disease in which the bacilli replicate in macrophages. The pathogenic lifestyle requires the tubercle bacillus to sense and respond to multiple environmental cues to ensure its survival. Such stimuli include hypoxia, nutrient limitation, presence of reactive oxygen and reactive nitrogen intermediates, pH alterations, and cell wall/ membrane stress. Two component systems (TCSs) form the primary apparatus for sensing and responding to environmental cues in bacteria. A prototypical TCS is composed of a sensory protein called sensor kinase (SK) and a response generating protein called response regulator (RR). M. tuberculosis encodes 11 genetically paired TCSs, 2 orphan sensor kinases and six orphan response regulator proteins. Studies of the TB bacilli using transcriptional profiling and gene knockouts have revealed that TCSs play an important role in facilitating successful adaptation to diverse environmental conditions encountered within the host. The mtrAB and prrAB genes encoding corresponding TCSs have been shown to be essential for survival, mprAB for persistence and devRS for hypoxic adaptation. Further, inactivation of the TCSs regX3-senX3, tcrXY, trcRS, phoPR or kdpDE was shown to affect the growth and/or virulence of M. tuberculosis in animal infection models. The SK and RR proteins of TCSs are modular and contain variable input and output domains coupled to conserved ‘transmitter’ and ‘receiver’ domains. Despite the modular nature and extensive homology of SK and RR proteins across TCSs, which may allow non-cognate interactions, it is believed that crosstalk across different TCSs is not favored and that individual pathways are generally well insulated. The existing profiling studies have been performed on the TCSs of bacterial species containing a relatively large number of TCSs. In those studies, specificity and insulation have been the norm and thus become the prevalent paradigm of TCS signaling. In vitro genome wide phosphotransfer profiling has revealed only a few cross- communication nodes in the TCSs of Escherichia coli (~3%), while none in Caulobacter crescentus (in 352 interactions tested, in short time duration) and Myxococcus xanthus (in 250 interactions tested). Yet, many instances of cross talk have been reported in literature. For example, E. coli TCSs PmrAB and EnvZ-OmpR show cross-communication with QseBC and ArcBA, and many more. In M. tuberculosis, indirect evidence of the existence of such cross regulation has originated from studies where mutations in phoPR have been shown to affect the expression of the TCS devRS and its regulon. It is thus interesting to examine the extent of crosstalk in the TCSs of M. tuberculosis, which has an exceptionally small number of TCS proteins compared to E. coli. As mentioned earlier, M. tuberculosis H37Rv has 11 cognate pairs of TCSs, 2 orphan sensor kinases and 6 orphan response regulators. To study the entire landscape, we aimed to study all 221 connections between SK and RR proteins including 12 cognate interactions. While 10 of the cognate TCS interactions were established in the literature, two putative systems KdpDE and NarSL and 5 orphan response regulators were still uncharacterized, therefore we initiated our work with the characterization of these TCSs. At the biochemical level, the KdpDE two component system of M. tuberculosis is not well studied, though one report showed interaction of the C-terminal domain of KdpD SK and KdpE RR using yeast two hybrid assay and another reported the interaction of the SK with LRP protein. Besides these associations, there is no evidence for the functionality of KdpDE system. Similarly, NarSL system also has not been characterized and it not known whether these putative two component proteins are functional. The initial part of the study includes the characterization of these two TCSs, NarS-NarL and KdpD-KdpE, at biochemical and physiological levels. In our studies we demonstrated that KdpDE system is a bonafide two component system of M. tuberculosis, and KdpD SK undergoes autophosphorylation at His642 residue in presence of Mg+2 ions and then it transfers phosphoryl group to a conserved Asp52 residue on the KdpE RR protein. The acid-base stability analysis revealed the nature of chemical bonds present in the KdpD and KdpE proteins, and further confirmed that KdpD and KdpE are typical SK and RR respectively. SPR analysis demonstrated that KdpD and KdpE proteins interact under basal non-phosphorylated conditions and the interaction affinity reduced when SK was phosphorylated. The reduction in the interaction affinity indicated towards a possible dissociation of SK and RR protein during phosphotransfer, which allows RRs to exert their regulatory effect. On the similar line, the phosphorylation defective SK (KdpDH642Q) had least affinity with KdpE suggesting that perhaps this mutant SK, fails to interact with the RR. We have also shown that both the kdpD and kdpE genes are in the same operon and are up regulated in potassium ions limitation and osmotic stress conditions. Overall, using the biochemical approaches, we have established that Rv1027c–Rv1028c operon of M. tuberculosis encodes a functional and a typical KdpDE two component signal transduction system. Using the similar biochemical and biophysical approaches, we have demonstrated that NarS-NarL proteins constitute a functional TCS and His241 and Asp61 are the phosphorylatable residues. In contrast to KdpDE which shows typical behaviour of TCS, NarSL TCS showed atypical behaviour. Malhotra and group’s work on NarSL suggested that there is cross-regulation between NarS/NarL and DevS/DosT/DevR systems. We addressed this possibility on three separate levels, by examining (i) the cross-phosphorylation of DevR and NarL RRs by non-cognate sensor kinases NarS and DevS/DosT respectively, (ii) the interaction between DevR and NarL RR proteins, and (iii) examining the effect of DevR-NarL interactions on their DNA binding properties. Our studies ruled out the presence of any physiologically relevant phosphorylation mediated cross-talk between NarS/NarL and DevS/DosT/DevR. We identified that the cross talk between these TCSs could be explained on the basis of interaction between NarL and DevR RRs and their subsequent binding to the target gene promoter regions for concerted regulation of gene expression. We also identified that DevR activation is critical for cooperative action with NarL. This process comes out as a novel mechanism of gene regulation via heteromerization of RRs. We hypothesized that formation of NarL-DevR heteromers may arise because of high sequence similarities. Conclusively, our study provides insights into the functionality of M. tuberculosis NarL/NarS TCS and regulatory function of NarL protein which acts in concert with another RR, DevR. Overall, NarS-NarL system showed an atypical, novel mode of gene regulation involving RR heteromerization. Subsequent to the basic biochemical characterization of NarSL and KdpDE system, the genome wide phosphotransfer profiling was done to identify the cross-connections between TCSs. Remarkably, we found that specificity was the exception rather than the rule. While only three of the TCS pairs were completely specific, all the other nine TCS pairs exhibited crosstalk, including a few that were highly promiscuous. We classified the interactions as specific, one-to-many, and many-to-one signaling circuits. We also profiled all the RRs including the orphans for their ability to accept phosphoryl group from a low molecular weight donor, acetyl phosphate, and interestingly found that only two RRs DevR and NarL were capable of accepting phosphoryl group from such a donor. Interestingly, none of the orphan RRs accepted phosphoryl group from any donor, neither SKs nor low molecular weight phospho donors, warranting further analysis of their roles and presence in the M. tuberculosis genome. Our exhaustive map of the crosstalk between the TCSs of M. tuberculosis sets the stage for a renewed view of TCS signaling and proposes a dispersive-integrative landscape for TCS signaling rather than one of insulation. As an extension of our basic characterization work of NarSL TCS, we also attempted to understand the localization pattern of NarS sensor kinase in M. smegmatis cells using fluorescence approaches. It is known that many bacterial receptors including sensor kinases form clusters or show specific localization patterns inside the cell. We found that NarS shows distinct cellular localization pattern. However, the functional significance of this localization pattern is not obvious yet and warrants further investigations. We also developed a few non-radioactive methods to study interaction between two component systems to overcome the limitations associated with radioactive experiments in studying TCSs. We developed fluorescence resonance energy transfer (FRET) to study in vitro interaction between two component proteins which was sensitive to the phosphorylation status of the proteins. Using fluorescently tagged SKs and RRs, we determined a change in FRET for KdpDE and NarSL TCS pairs in vitro. Our study thus also provides an alternative approach to study TCS signaling, using an easier, non-radioactive and high throughput approach. In summary, our study presents the evidence of an alternative paradigm of bacterial signaling, where significant crosstalk between the underlying TCSs prevails. The new paradigm is expected to have important implications in our understanding of the virulence and pathogenesis of bacterial infections. Overall, our studies (i) allowed the establishment of functionality of all paired TCSs encoded in the genome of M. tuberculosis including NarSL and KdpDE TCSs, (ii) identified the novel mechanism of gene regulation by NarL RR and DevR, (iii) demonstrated the existence of TCS signaling which is contrary to the existing notion of specificity (iv) showed the distinct localization pattern of NarS and (v) developed non-radioactive approaches to study two component interactions.

Les herbicides sont des polluants suscitant de grandes inquiétudes en raison de leur ubiquité environnementale résultant de leur usage intensif dans l’agriculture moderne et de leur persistance dans les sols et les eaux. Les herbicides peuvent être dégradés par des microorganismes, des plantes ou d’autres processus naturels, produisant alors une vaste gamme de métabolites dont l’impact sur les écosystèmes reste méconnu. Dans un contexte d’évaluation des risques environnementaux, l’étude de la réponse des plantes à des mélanges complexes de xénobiotiques est importante pour estimer les effets des contaminations, notamment dans le cas de pollution résiduelle. Afin d’étudier l’impact de cette diversité de polluants, les mécanismes de réponse et les cibles impliquées, la plante modèle Arabidopsis thaliana a été confrontée à des doses variables de molécules de la famille des triazines constituant une série chimique cohérente : atrazine, herbicide encore largement utilisé au niveau mondial, déséthylatrazine, métabolite chloré de l’atrazine, et hydroxyatrazine, métabolite de déchloration de l’atrazine. Ce travail montre que l’exposition de courte durée à des doses variables d’atrazine, de déséthylatrazine et d’hydroxyatrazine, au niveau racinaire, affecte de manière spécifique et dose-dépendante la croissance précoce et le développement de la plante. La caractérisation d’effets directs et multiples sur la respiration et la croissance racinaire a permis de révéler des mécanismes d’action non-canoniques, distincts de l’action classiquement décrite des triazines sur le photosystème II. Afin d’identifier ces mécanismes, activés en absence de dommages cellulaires, une analyse transcriptomique au niveau du génome entier a été effectuée. Les trois triazines induisent des changements coordonnés et spécifiques dans l’expression des gènes. L’analyse fonctionnelle des gènes différentiellement exprimés et de leur promoteur révèle que les voies de signalisation liées à la fois aux hormones végétales, à la perception de faibles niveaux d’énergie, aux stress environnementaux ainsi qu’aux interactions biotiques sont impliquées dans la réponse aux faibles doses de triazines. Les triazines affectent, en particulier, l’expression de gènes connus pour être régulés par les cytokinines. De manière intéressante, cette famille d’hormones végétales montre des caractéristiques chimiques similaires à celles des triazines. Des études développementales utilisant différentes modalités d’exposition aux triazines et aux cytokinines ont alors été effectuées sur des génotypes sauvages et sur des mutants de la voie de signalisation des cytokinines. L’identification d’interactions spécifiques entre les triazines et les composants de la signalisation des cytokinines a alors mis en évidence des mécanismes potentiels de compétition et/ou d’antagonisme. La caractérisation de ces perturbations au niveau de la transduction du signal pourra permettre à terme d’évaluer l’efficacité des herbicides sur les cultures ainsi que l’impact des contaminations xénobiotiques sur les communautés végétales naturelles. Enfin, l’identification des interactions entre stress xénobiotique, biotique et abiotique approfondira les connaissances sur les effets croisés de la pollution chimique et des stress liés au changement climatique. / Herbicides are pollutants of high concern due to their environmental ubiquity resulting from extensive use in modern agriculture and persistence in soil and water. Degradation events on active molecules mediated by microorganisms, by plants and by natural processes give rise to a plethora of herbicide metabolites of unknown impact on ecosystems. Study of plant behavior toward such complex mixtures of xenobiotic structures is important to evaluate the effects of contaminations, especially in the context of residual pollution. In order to understand the mechanisms underlying the action of this diversity of compounds, the model plant Arabidopsis thaliana was confronted to variable doses of the widely-used triazine herbicide atrazine, and of two of its metabolites, desethylatrazine and hydroxyatrazine. Short exposure to varying concentrations of atrazine, desethylatrazine and hydroxyatrazine was found to affect early growth and development in various dose-dependent and distinct manners. These differential effects pointed out to the multiple involvement of non-canonical mechanisms, directly affecting respiration and root development. In order to identify these mechanisms, which are activated in the absence of major adverse physiological effects, a genome-wide transcriptomic analysis was carried out. All of the triazines under study induced coordinated and specific changes in gene expression. Functional analysis of differentially expressed genes and of their promoters revealed that signaling pathways related to plant hormones, low energy sensing, environmental stresses and biotic interactions were involved in low-dose triazine responses. In particular, triazines affected the expression of genes known to be regulated by cytokinins. Interestingly, this family of plant hormones shares similar chemical features with triazine compounds. Developmental studies on plants bearing mutations in cytokinin sensing and signaling pathways were then carried out under variable triazine exposures. The identification of specific interactions between triazine compounds and cytokinin-signaling components highlighted potential mechanisms of competition and/or antagonism. The characterization of such signal transduction modifications and perturbations will be useful to assess herbicide efficiency in crop systems and xenobiotic contamination impact on natural plant communities. Finally, the identification of crosstalk processes between xenobiotic, abiotic and biotic stress signaling gives novel insights into the interplay between chemical pollution and climate change stressors.

Mouse embryonic stem cells (mESCs) can be maintained in a proliferative and undifferentiated state over many passages (self-renewal) while retaining the potential to give rise to every cell type of the organism (pluripotency). Autocrine FGF4/Erk signalling has been identified as a major stimulus for fate decisions and lineage commitment in these cells. Recent findings on serum-free culture conditions with specific inhibitors (known as 2i) demonstrate that the inhibition of this pathway reduces transcription factor heterogeneity and is vital to maintain ground state pluripotency of mESCs. We suggest a novel mathematical model to explicitly integrate FGF4/Erk signalling into an interaction network of key pluripotency factors (namely Oct4, Sox2, Nanog and Rex1). The envisaged model allows to explore whether and how proposed mechanisms and feedback regulations can account for different expression patterns in mESC cultures. We demonstrate that an FGF4/Erk-mediated negative feedback is sufficient to induce molecular heterogeneity with respect to Nanog and Rex1 expression and thus critically regulates the propensity for differentiation and the loss of pluripotency. Furthermore, we compare simulation results on the transcription factor dynamics in different self-renewing states and during differentiation with experimental data on a Rex1GFPd2 reporter cell line using flow cytometry and qRT-PCR measurements. Concluding from our results we argue that interaction between FGF4/Erk signalling and Nanog expression qualifies as a key mechanism to manipulate mESC pluripotency. In particular, we infer that ground state pluripotency under 2i is achieved by shifting stable expression pattern of Nanog from a bistable into a monostable regulation impeding stochastic state transitions. Furthermore, we derive testable predictions on altering the degree of Nanog heterogeneity and on the frequency of state transitions in LIF/serum conditions to challenge our model assumptions.