ABSTRACT. Although cell cycle control is a conserved and essential process, some core animal and fungal cell cycle regulators are not homologous (e.g. E2F and SBF). We recently showed that evolution along the fungal lineage was punctuated by the early acquisition and entrainment of the SBF transcription factor, a protein with homology to a domain commonly found in DNA viruses (Medina et al, 2016).  Ancestral SBF likely hijacked cell cycle control by binding cis-regulatory elements targeted by E2F and activating transcription of G1/S genes to drive cell proliferation.  Cell cycle evolution in the fungal ancestor proceeded through a hybrid network containing both SBF and its ancestral animal counterpart E2F, which is still maintained in early-diverging Fungi such as Spizellomyces punctatus (a Chytrid). Chytrids are unique in that they exhibit both fungal and animal-like features, including zoospores that swim using a posterior flagellum nucleated from centrioles, or crawl on surfaces using amoeboid movement.

To address the question of redundancy and specificity of E2F and SBF, we measured the binding specificity of E2F and SBF from several early-diverging Fungi using a protein-binding microarray (PBM) assay. PBM assays measure, in a single experiment, the binding of recombinant proteins to tens of thousands of synthetic DNA sequences, guaranteed to cover all possible 10-bp DNA sequences in a maximally compact representation. We showed that E2F and SBF from early-diverging Fungi have nearly identical DNA-binding specificity to their human and yeast counterparts. We further showed that E2F and SBF can bind a common set of motifs, which supports the hijacking hypothesis and binding redundancy between E2F and ancestral SBF. Last, we showed that there are specific motifs that can be bound only by E2F or only by SBF. This suggests that certain classes of genes could be under E2F-only or SBF-only control, which may explain why both transcription factors are still maintained in some species, such as Spizellomyces punctatus.

ABSTRACT. Maintaining robust circadian rhythms has been linked to longevity and metabolic health. Because these rhythms are disturbed by factors such as jet lag, shift work, and high-fat diets, there is interest in developing pharmacological control strategies to modulate circadian function. The design of therapeutic strategies is currently limited by the lack of a clear mechanistic understanding of interactions between posttranslational regulators, as efficient control of clock behavior will likely require several simultaneous modulations. Although small molecules that modulate clock function might offer therapeutic approaches to such diseases, only a few compounds have been identified that selectively target core clock proteins. Using mathematical modeling and systems biology approaches, we provide a mechanistic interpretation for the relationship between candidate regulators, lending insight into circadian regulation and potential pharmacological control. This study provides further insight into the molecular clock machinery responsible for maintaining robust circadian rhythms.

ABSTRACT. In natural environment bacteria seldom encounter optimal conditions that could sustain its growth exponentially. Thus, bacteria have adapted distinctive strategies to survive during famine period; for example, most species of gram-positive bacteria will form dormant spores in response to starvation. Gram-negative bacteria are able to survive this period without entering dormancy. Also, it was shown that Escherichia coli can survive for long periods of time (months or even years) under starvation condition and this phase has been termed as long-term stationary phase (LTSP) (Finkel S. E., 2006; PMID 16415927). Two interesting phenomena associated with LTSP; growth advantage in the stationary phase (GASP) and viable but nonculturable (VBNC) phenotypes were observed in cells survival during the stationary phase. GASP phenotype has been characterized as mutant cells that exhibit fitness advantage against wild-type and eventually dominates the culture, which was proven through competition experiments between aged cells and young cells populations. To date, mutations in rpoS, lrp, gltIJKL and subsequently sgaABC genes were characterized with GASP phenotype. In addition, VBNC phenotype are seen in many bacteria as a response to a variety of stress though the molecular mechanism of VBNC phenotype have not been cleared. We are interested to study this phenomenon from a systems perspective to understand the function that allows gram-negative bacteria to survive prolonged periods of starvation. For this purpose, we have constructed a new E. coli single-gene deletion library, with 20 nucleotides molecular barcode which allows us to monitor population dynamics in mixed culture by high-throughput sequencing. We had grown the mutant library in Luria-Bertani (LB) medium and sampled for three weeks (LTSP) under batch culture condition to monitor population changes of each deletion strains. Serial passage was carried out at each sampling to differentiate between the viable and the non-viable populations.

ABSTRACT. Genome-scale metabolic models allow researchers to investigate the metabolism of a given organism in various growth conditions. In addition, they provide a means to calculate yields, to predict consumption and production rates, and to study the effect of genetic modifications, without running resource-intensive experiments. While metabolic models have become an invaluable tool for optimizing industrial production hosts like E. coli and S. cerevisiae, few such models exist for C1 metabolizers. Here we present a genome-scale metabolic model for Methylococcus capsulatus, a well-studied obligate methanotroph, which, since the 70s has been the industry's focus as a production strain of single cell protein (SCP). The model was manually curated, and spans a total of 783 metabolites connected via 840 reactions. The inclusion of 730 genes and a host of annotations, make this model not only a useful tool for knock-out studies, but also a centralized knowledge base for M. capsulatus. We are confident that our contribution will serve the ongoing fundamental research of C1 metabolism, and pave the way for rational strain design strategies towards improved SCP production in M. capsulatus.

Paper in preparation

ABSTRACT. A major challenge in biology is fully understanding the relationship between genotype and phenotype. Decades after genome sequencing emerged, we are far from comprehensive models that can predict phenotypic outcomes from environmental conditions and genetic perturbations. While genomic, proteomic, and metabolomics technologies have greatly advanced in the past decade, phenotyping has not reached the same level of refinement. In this work, we develop tools to extract quantitative phenotypes at multiple levels. We use the multicellular organism C. elegans to characterize in vivo intermediate and downstream phenotypic states such as gene expression, cellular and subcellular morphology, and behavior. By incorporating experimental platforms that enable high-throughput imaging, controlled environmental conditions, and on-line image analysis, we are able to extract quantitative phenotypic data sets that enable identifying underlying biological functions and genetic relationships via statistical and mathematical analysis tools. This is made possible by the integration of customized microfluidic platforms and computer vision which allow fast animal handling, controlled environmental conditions, and quantitative image analysis. We have used these tools to identify alleles that produce subtle phenotypes in synaptic patterning. Through statistical analyses, we characterized a spectrum of phenotypic severity exhibited by a library of mutants, rather than a binary outcome. Through mathematical models, genetic relationships can be predicted from phenotypes that are hidden to human vision. In addition, we have developed microfluidic platforms that allow longitudinal lifelong monitoring of C. elegans populations, while performing high-resolution imaging. Through these platforms, lifelong spatiotemporal gene expression patterns can be extracted, and correlated with subcellular, cellular and physiological outcomes. These deep-phenotyping tools allow quantitative characterization of intermediate and downstream outcomes (such as spatiotemporal gene expression, cellular and subcellular phenotypes), and thus enable building predictive models that link phenotype and genotype.

ABSTRACT. Abstract A petri net representation and simulation of biosynthesis of menthol from Geranyl diphosphate (GPP) has been performed with the objective of understating new insights of the structure of this pathway affecting the synthesis of menthol. The model has been validated for its structural and behavioural properties. This understanding is expected to identified reactions that could be experimentally manipulated to enhance the productivity of this medically and commercially important material. Petri nets are a special class of networks, introduced in 1962 by Carl Adam Petri, that provide a convenient language and graphical representation for many kinds of processes in a variety of areas of science and engineering. The petri net model is generated on the basis of literature survey and from the biological databases such as KEGG (Kyoto Encyclopaedia of Genes and Genomes). Petri net is designed through employing the Petri net tool “Hybrid Petri net ICSI simulator”(hisim-1.0)1. The model is simulated and validated against the known experimental data obtained from extensive literature searches. The model posses the basic structural properties of Petri net like PUR, ORD and SC which are necessary for preliminary consistency of the net and its correctness. P-invariant and T-invariant properties were computed with the help of software “Platform Independent Petri net Editor” (Pipe-v4.3)2. The net is utilized to simulate the time (pt) with concentrations of GPP, (-)-Limonene, (+)-pulegone, (-)-Menthone and (-)-Menthol. Dimethylallyl diphosphate (DMAPP) and Isopentenyl diphosphate (IPP) are main precursors for this biosynthesis. Biological data needed for simulation where obtained from extensive survey of literature. The results are shown graphically, the nature of graphs represent the variation of concentrations with time (pt). Although this metabolic model is basic, it will facilitate a platform for analysing high- throughput data, and it should lead to a more generous understanding of menthol biosynthesis 1 https://sourceforge.net/projects/hisim/files/hisim/hisim-1.0/ 2 https://sourceforge.net/projects/pipe2/

ABSTRACT. Understanding the development of tissues and organs at a single cell level remains a challenge. Here we show that Polylox – an artificial DNA recombination locus based on the loxP-Cre recombination system developed recently by Hans-Reimer Rodewald and colleagues – allows for the endogenous, non-invasive barcoding of single cells. Based on a Markov model for barcode generation, we find that more than 10^6 different barcodes can be generated. However, individual barcodes have different probabilities of generation (as also noticed for alternative barcoding approaches using CRISPR/Cas9). Our model, calibrated against experimental barcoding data, allows the assignment of barcode generation probabilities and thus the selection of low-probability, informative barcodes for clonal analyses. We have used this mathematical framework to analyze data on the formation of hematopoietic stem cells (HSC) clones during development and on the differentiation of mature cell lineages from HSC during development and adult hematopoiesis. Many HSC realize multipotency in vivo, yet the spreading of barcodes from such cells reveals a fundamental split between myelo-erythroid and lymphoid lineage development. These findings support the long-held, but currently contested, view of a tree-like hematopoietic structure with few major branches.

ABSTRACT. The CRISPR-based gene-editing tool has revolutionized molecular cell biology research. It makes efficient genome editing possible, and is fundamentally changing our practice in biological research, synthetic biology, biomedical treatment, and other technology development such as biofuels.

CRISPR-based gene knock-in requires synthesis of DNA constructs containing the knock-in sequence as a template. Some constructs have high G/C content and high sequence complementation, and sometimes include a drug selectable marker and associated LoxP sites with palindromic structure. These properties impose severe challenges to any DNA synthesis approach. Specifically, the widely-used Gibson assembly procedure, which here is designed to assemble four DNA segments into a construct, often leads to misassembly. We developed a procedure that greatly improves the efficiency of making the DNA constructs. Experimental tests on multiple constructs confirmed that our new procedure leads to at least 3-8 folds of increase of the assembly efficiency. More impressively we achieved high assembly efficiency on DNA constructs we failed to make otherwise. The modularized procedures save time and cost while making multiple constructs, and it may accelerate applications of the CRISPR technique in synthetic and systems biology studies.

ABSTRACT. Signaling systems control many pivotal processes in a cell, which need to be strictly regulated. The corresponding signaling networks become big and complex with many cross-talks. The automatic detection of all possible signal transduction pathways from the receptor binding to the cell response would be useful for profound network analysis. Even if signaling pathways, which lead, for example to cell death or apoptosis, will not be repeated by the same cell again, the same signaling pathway may take place in the next generation of cells. Thus, a steady-state behavior can be assumed. Elementary mode (EM) analysis [1], which takes into account steady-state conditions, has not only been applied to model metabolic systems, but also to model signaling networks [2]. There are some specific properties in signaling pathways. For example, feedback loops may lead to EMs that do not capture the entire signal flow. To get the complete signaling pathways, we have to combine the EMs in a special way. We expressed EMs as transition invariants (TIs) in the Petri net formalism [3]. To combine the EMs in a proper way, we define Manatee invariants [4] based on feasible TIs [2]. We determine the cause of the disruptions of signal flows as internal place invariants in the subnetwork of a TI and generate the linear combination of TI to link interrelated processes. In the talk, we explain the concept of Manatee invariants, giving the necessary definitions. For illustration, we introduce a Petri net model of mitophagy processes [5] edited and analyzed using the software tool MonaLisa [6]. We will show that at least one EM did not cover a signaling pathway from the receptor to the cell response, while the corresponding Manatee invariant recovers this pathway. 1. Schuster et al. (1993) Proc Second Gauss Symp, 101 2. Sackmann et al. (2006) BMC Bioinformatics, 7:482 (2006), doi:10.1186/1471-2105-7-482 3. Koch, Reisig, Schreiber (Eds.) (2011) Springer, Modeling in Systems Biology: The Petri Net Approach 4. Amstein et al. (2017) in revision 5. Lass (2016) Bachelor’s Thesis, Goethe-University Frankfurt/Main 6. Einloft et al. (2013) Bioinformatics, 29: 1469

ABSTRACT. Metabolic models generated by automated reconstruction pipelines are widely used for high-throughput prediction of microbial phenotypes. However, the generation of accurate in-silico phenotype predictions based solely on genomic data continues to be a challenge as metabolic models often require extensive gapfilling in order to produce biomass. As a result, the true physiological profile of an organism can be altered by the addition of non-native biochemical pathways or reactions during the gapfilling process. In this study, we constructed draft genome-scale metabolic models for ~1000 diverse set of reference microbial genomes currently available in GenBank, and we decomposed these models into a set of classical biochemical pathways. We then determine the extent to which each pathway is either consistently present or absent in each region of the phylogenetic tree, and we study the degree of conservation in the specific steps where gaps exist in each pathway across a phylogenetic neighborhood. Based on this analysis, we improved the reliability of our gapfilling algorithms, which in turn, improved the reliability of our models in predicting auxotrophy. This also resulted in improvements to the genome annotations underlying our models. We validated our improved auxotrophy predictions using growth condition data collected for a diverse set of organisms. Our improved gapfilling algorithm will be available for use within the DOE Knowledgebase (KBase) platform (https://kbase.us).

ABSTRACT. Learning of gene regulatory networks from transcriptomic data is challenging due to the large space of potential gene regulatory interactions to explore, and the relative sparsity of data points. Further challenges are presented by RNA-seq data sets where the data cannot be assumed to follow a multivariate normal distribution, and so many previously applied graphical modelling approaches are no longer valid. Building on previous methods from the literature exploiting Gaussian process regression in a variable selection framework, we present a Sequential Monte Carlo approach to learning gene regulatory network structures. Such methods utilise a population of particles to more thoroughly explore the parameter space, and are easily amenable to parallelisation on multicore CPUs or GPGPU systems to accelerate the inference procedure. Each particle represents a potential network structure, and particles are propagated through a sequence of probability distributions, starting as a sample from the prior distribution on network edges, and finishing as a sample from the posterior distribution of the model. We benchmark the performance of our approach when compared to traditional Markov Chain Monte Carlo approaches using synthetic RNA-seq data with known network structure, and apply the method to real world RNA-seq data.

ABSTRACT. Mechanism-based mathematical models are the foundation for diverse applications in science and engineering. It is often critical to explore the massive parametric space for each model. For certain models, e.g., agent-based, PDEs, and SDEs, this practice can impose a prohibitive barrier for practical applications even when computer clusters are used. To overcome this limitation, we develop a fundamentally new framework to improve the computational efficiency by orders of magnitude. The key concept is to train an artificial neural network (ANN) using a limited number of simulations generated by well-defined mechanism-based models. The number of simulations is small enough such that the simulations are still manageable, but large enough to train the ANN sufficiently well to make reliable predictions. Then, the trained ANN will be used to explore the system dynamics in a much larger parametric space. In addition to the framework, we also illustrate the application of this approach using several hand-on examples in synthetic biology design and in exploring stochastic dynamics of complex networks. Our work can potentially be a platform for faster pattern screening, cell strain identification as well as new drug development.

ABSTRACT. We develop and offer integrated data management support for research in the fields of systems biology and systems medicine within and across research consortia. This support is applied and offered to geographically dispersed, interdisciplinary and large-scale research initiatives in which we are part of, like the German systems biology network Virtual Liver and its successor the research initiative ‘Systems Medicine of the Liver’ (LiSyM: http://www.lisym.org), as well as European research networks like ERASysAPP, the former SysMO network (Systems Biology of Microorganisms) or NMTrypI (New Medicines for Trypanosomatidic Infections). Parts of these solutions are also applied to projects with a local focus as the Synthetic Biology Centres at Manchester (SynBioChem) and Edinburgh (SynthSys).

Our data management concept aims at bundling, storing and integrating research assets like data, models and description of processes and biological samples in a Findable, Accessible, Interoperable and Reusable (FAIR) manner (http://fair-dom.org) and consists of 4 major pillars:

1) Infrastructure backbone: The SEEK software as registry and a commons for data, models, samples, processes and resulting publications or presentations, at the same time yellow pages for projects, people and events. SEEK is either implemented as data management platform that is maintained by the research project itself (e.g. LiSyM SEEK: http://seek.lisym.org) or as hub service maintained by us and spanning different consortia (FAIRDOMhub: https://www.fairdomhub.org).

2) Standardized data description: Data spreadsheet templates and tailored use of controlled vocabularies and ontologies to describe data and metadata.

3) Modelling support: Seamless handling and simulation of models by integrated modelling platforms (JWS-Online, SYCAMORE, Cytoscape).

4) Social support: Facilitators (PALs) in the projects for gathering requirements and dissemination

Unlike the majority of data management systems, we specifically support the interaction between modelling and experimentation. Datasets can be associated with models and/or workflows or biological samples, and model simulations can be compared with experimental data.

ABSTRACT. Advances in high-throughput sequencing technologies have led to a high volume of public RNA-seq data, enabling assembly of large data sets to search for novel biological patterns not visible to individual studies, although methods for doing so remain a significant challenge.

The use of clusters and bi-clusters is a popular unsupervised machine learning approach for discovering co-expressed, and hence functionally related, gene sets. Different notions of clustering have been used, including graph-theoretical methods based on density and hierarchical clustering. Expression data can be viewed as a signed dataset, with up or down regulation captured by a positive or negative quantity, respectively. However, most of these prior approaches tend to ignore the signs and works on unsigned data. This is partly because the analysis of signed data tends to be much more challenging.

We develop a novel approach for finding coordinated motifs of expression by formalizing them as quasicliques in signed networks. This is computationally much harder than the problem in unsigned networks, and we use a convex optimization approach, combined with pruning, to find the top k quasicliques, in terms of their objective values. We incorporate functional similarity measures on nodes in quasicliques, e.g., the fraction of genes within each cluster that have high scores of semantic similarity as annotated on the Gene Ontology. Clusters with low known functional similarity can be indicators of new biological patterns in such data, and might help guide further experiments. We also study a new approach that involves finding quasicliques with given constraints on the level of functional similarity within the nodes. We evaluate these methods and present findings from analysis of a large compilation of RNA-seq expression data from humans.

ABSTRACT. We developed a computational model of mouse iron physiology to gain insights into its complex hormonal regulations. Model calibration revealed an essential role of non-transferrin bound iron (NTBI) uptake by the liver under high iron diet condition, without this the model failed to explain iron distribution in the liver and red blood cells . The model was validated by its ability to simulate the pathophysiology of several iron disorders such as hemochromatosis, β-thalassemia and anemia of inflammation. We also tested various other experimental observations under normal and pathological states which not only further validated our model but also provided better understanding of underlying mechanisms. Moreover, we show that how such a model can be used for optimal therapies of various iron disorders. This physiological model paves the way for a more comprehensive multiscale model across organs, cells, and molecules and also provides a prototype for human iron metabolism. The present model contributes to a deeper understanding of iron physiology and can be used for predictive exploration of therapeutic interventions in iron disorders.

ABSTRACT. Animals, including humans, can cope with environmental stress by adapting their physiology and behavior. The free-living nematode Caenorhabditis elegans can adapt to harsh environments by undergoing a whole-animal change, which involves exiting reproductive development and entering the stress-resistant dauer larval stage. The dauer is a dispersal stage with dauer-specific behaviors that allow C. elegans to find and stow onto carrier animals for transportation to improved environments, but how the dauer acquires these behaviors, despite having a physically limited nervous system of 302 neurons, is poorly understood. We compared dauer and reproductive development using whole-animal RNA-seq at fine time points, and at sufficient depth to measure transcriptional changes within single cells. We detected 8,042 differentially expressed genes (39% of the protein-coding genome) during dauer and reproductive development, and observed striking up-regulation of 60 of the 118 C. elegans neuropeptide genes during dauer entry. We knocked down a large set of neuropeptides using sbt-1 mutants, which are defective in neuropeptide processing, and demonstrated that neuropeptide signaling promotes the dauer entry decision over reproductive development. We then demonstrated that neuropeptide signaling in dauers promotes the dauer-specific nictation (a carrier animal-hitchhiking) behavior, and is necessary for switching from CO2 (a carrier animal cue) repulsion in non-dauers to CO2 attraction in dauers. We then tested individual neuropeptides using CRISPR knockouts and existing strains, and revealed that the combined effects of two neuropeptide genes, flp-10 and flp-17, strongly explain the sbt-1 effects on nictation and CO2 attraction. Through a meta-analysis, we discovered a shared up-regulation of neuropeptides in dauers and the dauer-like infective juveniles of diverse parasitic nematodes, suggesting the anti-parasitic potential of targeting SBT-1 in these species. Our findings reveal that C. elegans animals can adapt to stress by using neuropeptides to enhance their decision-making accuracy, and to expand their behavioral repertoire.

ABSTRACT. The human immune system is well characterized for its critical role in host defense. It has evolved into a complex defense network that recognizes and protect against a range of pathogens that threaten the host viability. Computational models of immune system dynamics may contribute to a better understanding of the relationship between complex phenomena and immune response with respect to various pathogens. While computational models have been developed to study the dynamics of individual functions of the immune system (e.g., immune-receptor signaling, immune response to a tumor antigen, HIV infection of macrophage cells, etc.), recent efforts lack comprehensive computational models capable of capturing the multi-cellular, system-wide complexity of the immune system. We developed a multi-cellular computational network model that integrates various components of the immune system. Specifically, the model includes immune cells (antigen-presenting cells, monocytes, erythrocytes, lymphocytes, granulocytes etc.), non-immune cells (epithelial cells, endothelial cells, keratinocytes, hepatocytes, etc.), cytokines, chemokines, and pathogens (Influenza A virus, Human immunodeficiency virus, Human papillomavirus, Ebolavirus, Mycobacterium tuberculosis, Plasmodium falciparum, Leishmania donovani, Ascaris lumbricoides, Candida albicans). The network model was manually curated and annotated and consists of 152 components and 456 interactions, which enable the understanding of how the different complex phenomena interact with structures and elements during an immune response. The comprehensive nature of the model simulated a mounted response to nine different pathogens (and/or any combination thereof). Analyses of the dynamical model under a range of simulated conditions validated the existing immune responses and revealed complex signatures of the immune system during single infection and co-infections. The intuitive model construction allowed us to make novel predictions and validations to gain insights into different infectious diseases. The computational model would be very useful for the biomedical community to generate new hypotheses through iterative interactions with the model and its simulations.

ABSTRACT. The cell division cycle is the process by which eukaryotic cells replicate their chromosomes and partition the sister chromatids to two daughter cells. To maintain the integrity of the genome, proliferating cells must be able to block progression through the division cycle at key transition points (called ‘checkpoints’), if there have been problems in the replication of the chromosomes or their biorientation on the mitotic spindle. These checkpoints are governed by protein-interaction networks, composed of phase-specific cell-cycle activators and inhibitors. Examples include: Cdk1:Clb5 and its inhibitor Sic1 at the G1/S checkpoint in budding yeast, APC:Cdc20 and its inhibitor MCC at the metaphase checkpoint, and PP2A:B55 and its inhibitor ENSA at the mitotic-exit checkpoint. Each of these inhibitors (I) is a substrate as well as a stoichiometric inhibitor of the cell-cycle activator (A). Because the production (or actuation) of each inhibitor is promoted by a regulatory protein (R) that is itself inhibited by the cell cycle activator, the A-I-R interaction network presents a regulatory motif characteristic of a ‘feedback-amplified domineering substrate’. In this short talk/poster, we describe how the FADS motif responds to signals in the fashion of a bistable toggle switch, and then we discuss in detail how this toggle switch accounts for the abrupt and irreversible nature of three specific cell-cycle checkpoints: at the G1/S transition, at metaphase, and at mitotic exit.

ABSTRACT. Great strides have been made in the understanding of complex networks; however, our understanding of natural microecologies is limited. Modeling of complex natural ecological systems has allowed for new findings, but these models typically ignore the constant evolution of species. Due to the complexity of natural systems, unanticipated interactions may lead to erroneous conclusions concerning the role of specific molecular components. Modeling of synthetic microecologies has allowed researchers to explore specific questions (e.g. evolution and maintenance of coexistence) using simplified models, the findings of which can be used to infer results about natural systems. Using synthetic systems, researchers have been able to engineer better-defined cellular interactions and thus shed light on how these interactions lead to particular collective cell behaviors. Most of the microbial association studies involve co-cultures and fail to reflect the spatial relationship which is important to study pattern formation and evolution. The first synthetic predator-prey ecosystem showed the oscillatory population dynamics arising from the interaction of quorum-sensing modules in a spatiotemporal fashion. Recently, a large step forward in the field was the analysis of the microbial evolution and growth arena (MEGA)-plate where the spatiotemporal dynamics of microbial evolution of a single type of motile Escherichia coli was studied on an antibiotic background. We use a synthetic system to understand the spatiotemporal dynamics of growth and to study acquired resistance in vivo. Our system differs from previous systems in that it focuses on the evolution of a microecology from a killer-prey relationship to coexistence using two different non-motile E. coli strains. Using empirical data, we developed the first ecological model emphasizing the concept of the constant evolution of species, where the survival of the prey species is dependent on location (distance from the killer) or on the evolution of resistance. Our simple model, when expanded to complex microecological association studies under varied spatial and nutrient backgrounds may help to understand the complex associations between multiple species in intricate natural ecological networks.

ABSTRACT. Transcriptional output scales genome-wide with cell-size and growth rates. As a result, cells of different size and physiological states contain different numbers of mRNAs and proteins. The molecular mechanisms underlying this remarkably coordinated regulation remain largely mysterious. We have used single molecule FISH (smFISH) to investigate the role of cell size in regulation of transcription at the single cell level. Using fission yeast size mutants, we find that mRNA numbers of a series of diagnostic genes increase proportionally with cell size. Interestingly, we observe that scaling of mRNA numbers with cell size is linear across size mutants and as single cells progress through the cell cycle. We use mathematical models of stochastic gene expression in growing and dividing cells and Bayesian statistical inference to shed light on the source of this linear relationship between transcription and cell size. Our models suggest that for mRNAs with lifetimes shorter than a cell cycle duration, linear scaling is evidence for direct coordination of parameters of gene expression such as transcription or mRNA degradation with cell size. Moreover, we find that noise in mRNA numbers is mainly explained by the cell cycle and random birth-deaths of mRNA molecules with no evidence of transcriptional bursting occurring in constitutively expressed genes. This indicates that either transcription rate or mRNA lifetime, but not the frequency of promoter activation, are scaling with cell size. Using smFISH to quantify nascent mRNA we provide direct evidence of transcription rate being cell size dependent. Finally, we use the power of yeast genetics to explore further the mechanistic origin of the global coupling of gene expression and cell size.

ABSTRACT. In order to maintain genome integrity and an effective nucleocytoplasmic ratio from one generation to the next, cells carefully monitor progression through their replication-division cycle and fix any errors before they jeopardize the progeny of the cellular reproduction process. These error surveillance and correction mechanisms operate at distinct ‘checkpoints’ in the cell division cycle, where a growing cell must ‘decide’ whether it must wait for errors to be corrected or it may proceed to the next phase of the cell cycle. Once a decision is made to proceed, the cell unequivocally enters into a qualitatively different biochemical state, which makes cell cycle transitions switch-like and irreversible. These characteristics of cell cycle transitions are best explained by bistable switches with different activation and inactivation thresholds, resulting in a hysteresis effect. Almost 25 years ago, John Tyson and I proposed that the activity of the mitosis-inducing protein kinase, Cdk1:CycB, is controlled by an underlying bistable switch generated by positive feedbacks involving inhibitory phosphorylations of the kinase subunit. Numerous predictions of this model were experimentally verified by different groups, and bistability has become a paradigm of cell cycle transitions. The phosphorylation of mitotic proteins by Cdk1:CycB is counteracted by a protein phosphatase, PP2A:B55, which is inhibited during mitosis by a stoichiometric binding partner, ENSA-P, which is itself activated by Greatwall-kinase. Using mathematical modelling guided by biochemical reconstitution experiments, we showed recently that the BEG (B55-ENSA-Greatwall) pathway also represents a bistable, hysteretic switch controlled by the activity of Cdk1:CycB. Bistable regulation of the kinase (Cdk1:CycB) and the phosphatase (PP2A:B55) makes hysteresis a robust property of mitotic control, with suppression of futile cycling of protein phosphorylation and dephosphorylation during M phase. These considerations show that both entry into and exit from mitosis are controlled by bistable switches intimately connected to the activities of the major mitotic kinase, Cdk1:CycB, and phosphatase, PP2A:B55. Intriguingly, the ‘design principle’ of the BEG pathway is operative as well at two other cell cycle checkpoints, as will be discussed.

ABSTRACT. The detailed molecular mechanisms underlying the regulation of sleep duration in mammals are still elusive. To address this challenge, we constructed a simple computational model, which recapitulates the electrophysiological characteristics of the slow-wave sleep and awake states. Comprehensive bifurcation analysis predicted that a Ca2+-dependent hyperpolarization pathway may play a role in slow-wave sleep and hence in the regulation of sleep duration. To experimentally validate the prediction, we generate and analyze 26 KO mice. Here we found that impaired Ca2+-dependent K+ channels (Kcnn2 and Kcnn3), voltage-gated Ca2+ channels (Cacna1g and Cacna1h), or Ca2+/calmodulin-dependent kinases (Camk2a and Camk2b) decrease sleep duration, while impaired plasma membrane Ca2+ ATPase (Atp2b3) increases sleep duration. Genetical (Nr3a) and pharmacological intervention (PCP, MK-801) and whole-brain imaging validated that impaired NMDA receptors reduce sleep duration and directly increase the excitability of cells. Based on these results, we propose a hypothesis that a Ca2+-dependent hyperpolarization pathway underlies the regulation of sleep duration in mammals.

ABSTRACT. Actin is the most abundant protein in eukaryotic cells and is responsible for their dynamic structures. It accomplishes its function through interaction with a multitude of binding partners and signaling proteins, which control the assembly and branching of actin polymers. Thus signaling, polymerization and mechanics all play important roles in actin function. But accounting for this complex interplay of physics and chemistry presents modeling challenges that require multiple approaches. I will illustrate several such approaches with models and experiments aimed at elucidating actin dynamics in: motile dendritic filopodia, invasive pathogen motility and the leading lamelipodium of a migrating cell,. All these are deterministic models require the solution of partial differential equations in 1, 2, and 3 dimensions respectively. I will also introduce results on upstream signaling to the actin cytoskeleton within the foot processes of kidney podocytes in which the multivalent interactions between nephrin, Nck and NWAsp can produce multivalent clusters. This last system is modeled with stochastic Langevin dynamics simulations. (Supported through NIH Grant Number P41 GM103313 from the National Institute for General Medical Sciences.)

ABSTRACT. In order to maintain chromosome number between mitotic cell cycles, it is essential that the transition from interphase to M phase is abrupt and irreversible. This transition requires the switch-like phosphorylation of hundreds of proteins by the cyclin-dependent kinase 1 (Cdk1):cyclin B (CycB) complex. Previous studies have ascribed these switch-like phosphorylations to the self-activation of Cdk1:CycB through the removal of inhibitory phosphorylations on Cdk1-Tyr15, which creates a bistable switch that makes mitotic commitment irreversible.

Cdk1 self-activation, however, is dispensable for irreversible, switch-like mitotic entry due to a second mechanism that has recently been discovered: Cdk1:CycB inhibits one of its major counteracting phosphatases (PP2A:B55) via Greatwall kinase, which is phosphorylated and activated by Cdk1:CycB and then inhibits PP2A:B55 by phosphorylating the small, heat-stable protein ENSA. PP2A:B55 in turn dephosphorylates and inactivates Greatwall, forming a double negative feedback loop that gives a bistable PP2A:B55 activity profile with respect to Cdk1:CycB activity (1). PP2A:B55 also dephosphorylates Wee1 and Cdc25, generating crosstalk between the two modules. The resulting network that we are investigating – of two bistable mechanisms that mutually inhibit each other, is one that operates with maximum theoretical switching efficiency and is therefore ideal for the transitions between interphase and M phase.

Based on this theoretical framework we have explored these bistable mechanisms and their crosstalk combining experiments in HeLa cells with mathematical modelling. Our data suggest that two interlinked bistable mechanisms provide a robust solution for irreversible and switch-like mitotic entry and that either mechanism can maintain a bistable system response. When either of the bistable mechanisms are removed experimentally, hysteresis is maintained but reduced; when they are both removed, hysteresis is lost. In summary, we show how two mutually inhibiting bistable mechanisms generate robust separation of interphase and M phase.

Reference: 1. Mochida et al. (2016): Curr. Biol. 26: 3361–3367

ABSTRACT. CD4+ T-helper cells direct the cell-based and antibody-based arms of the adaptive immune system via the secretion of cytokines. The classical view has been that naïve T-helper (Th) cells differentiate into a small number of distinct stable states that express certain cytokine profiles (Th1, Th2 etc.). Recently this view has been challenged by experimental findings that suggest a higher complexity and point towards a long-lived tunable continuum of cell states between the well-known extremes. These hybrid states stably co-express graded levels of lineage-specifying transcription factors, such as T-bet and GATA-3. The mechanistic basis of such a stable continuum of cell states is unknown. To interrogate the underlying gene-regulatory mechanisms, we integrated data on histone modification patterns, RNA expression and transcription factor binding to systematically identify enhancers and repressors involved in Th cell lineage specification, map these regulatory elements to their target genes and identify their control by transcription factors, thus obtaining a bipartite graph linking enhancers and genes. To this end, we developed a novel correlation model, via a data-driven multivariable histone correlation measure, for inferring enhancer/repressor-gene interactions on topologically associated domains. This approach recovered well-known cis-regulatory elements and predicted new ones with comparable statistical confidence. We then used machine learning approaches to classify the large number of individual enhancers into functional classes according to their regulation by the lineage-specifying transcription factors T-bet and GATA-3 as well as external differentiation signals. This comprehensive topological analysis provides the basis for understanding the multistable dynamics of the Th cell differentiation network.

ABSTRACT. A growing number of studies have elucidated the essential role of ceramides and sphingolipids in the glucose homeostasis and insulin signaling. However, the mechanistic interplay between various components of ceramide metabolism remains to be quantified. To this end, we have resorted to dynamical modeling to gain insights into the sphingolipid metabolism and their role in the development of the insulin resistance. In particular, we have focused on the C16 ceramides family. Our model extends and refines a previously published model by including those reactions that connect sphingolipids de-novo synthesis with the salvage pathway. The latter recycles complex sphingolipids by transforming them in ceramides and it accounts for a significant part of the total ceramide production. We estimated unknown parameters of the model using mice macrophage cell line data. For the parameter estimation, we have used a multi-start approach with a least squares method. We have validated this extended model on an independent dataset for the same tissue in mice. We have integrated the model with transcriptomic data from a different experiment in obese/diabetic murine macrophages at 5 and 16 weeks. Our in silico experiments on the behavior of ceramide and related bioactive lipids, in accordance with the observed transcriptomic changes, support the observation on insulin resistance at the later phase. Furthermore, it provides a mechanistic explanation of its development. Our model suggests the key role of ceramide, glucosylceramide, and S1P in the development of insulin resistance. In addition, sensitivity analysis on the model allowed us to quantify the effect of the availability of each enzyme involved in the metabolism on each sphingolipid. We have visualized such effects using an interaction network. These visualizations should guide wet lab scientist in identifying new potential drug targets. In addition, the visualizations may help in identifying collateral effects by highlighting the nontrivial interactions among all the metabolites, supporting the development of more precise drugs. This article is published in January 2017, at the Nature Publishing Group journal of Scientific Reports.

ABSTRACT. The eukaryotic cell cycle is robustly designed, with molecules interacting and organized within definite network topologies that ensure its precise timing. This is governed by a transcriptional oscillator interlocked with waves of dedicated enzymatic activities, called cyclin-dependent kinases (cyclin/Cdk). These guarantee execution of definite phases throughout cell cycle progression. Although details about transcription of cyclins, the regulatory subunits of these enzymes, are available, a lack of understanding exists about network motifs responsible for the precise timing of waves of cyclin activation. We investigate the robustness of molecular designs interlocking the transcriptional oscillator with waves of cyclin/Cdk1 kinase activities. We have recently identified a transcriptional cascade that regulates the relative timing of waves of mitotic (Clb) cyclin expression in budding yeast. This cascade involves the Forkhead (Fkh) transcription factors (TF). Here we aim to unravel the network motif(s) responsible for timely cyclin/Cdk1 dynamics that interlock Clb waves through Fkh-mediated signaling. An integrated computational and experimental framework is presented. A kinetic model of the cyclin/Cdk1 network is simulated under a quasi-steady state assumption, and fitted to in vivo time course data of Clb dynamics. Robustness analyses are then performed by testing 1024 possible network motifs for their ability to fit Clb oscillations. Biochemical experiments support computational analyses, revealing the Clb/Cdk1-Fkh2 axis to be pivotal for timely transcriptional dynamics. A novel regulatory motif synchronizing Clb waves, coined as Multiplex Phase Interlocker, is unraveled. This motif uniquely describes a molecular timer (TF) that relies on separate inputs (cyclin/Cdk1 complexes) converging on a common target (TF itself). Within the motif, a progressive TF (Fkh2) activation may be realized by the sequential Clb/Cdk1 complexes. Altogether, our integrative approach is able to pinpoint robustness of cell cycle control by revealing a novel and conserved principle of design that ensures timely oscillations of cyclin/Cdk1 activities.

ABSTRACT. The Assay for Transposase Accessible Chromatin (ATAC)-seq provides a unique opportunity for inference of the transcriptional regulatory networks (TRN), especially for cell types and contexts where sample material is limiting and a priori knowledge of transcriptional regulation is scarce. Integration of accessible chromatin regions with TF motif database provides an initial network, where putative interactions are based on TF motif occurrences cis to gene loci. This initial network contributes to network inference in two ways (1) as a network prior for downstream inference and (2) to estimate TF activities (TFAs) from putative target gene expression levels (as opposed to more traditional estimates based on TF mRNA levels). The TFAs, network prior, and target gene expression then serve as input to the Inferelator algorithm, which uses Bayesian best subset regression to learn the entire transcriptional regulatory network. Noisy prior network edges that lack support from the gene expression measurements are removed, while new edges can be learned. We first validate our method in T Helper 17 (Th17) cells, integrating new ATAC-seq data with published RNA-seq data and making use of TF knockout and ChIP-seq to evaluate model performance. Then we infer TRNs for the relatively recently discovered innate lymphoid cells (ILCs), where very few transcriptional regulatory interactions are known. We validate the ILC TRN models both computationally (using gene expression prediction in new tissues) and experimentally through TF perturbation response measurements. We rigorously demonstrate the strength of our method to learn predictive transcriptional regulatory network models from ATAC-seq and RNA-seq experimental designs.

ABSTRACT. When developing mathematical models of biochemical processes, the relative paucity of quantitative data concerning concentrations of molecular species and especially reaction rates makes parametrization a challenge. On the other hand, experimentalists have obtained large amounts of phenotypic data from mutants in model organisms. Although such data do not provide specific values for model parameters, fitting models to these phenotypes can significantly constrain parameter sets. In the past, we have used existing phenotypic data in the literature to develop models of the yeast cell cycle, and then tested model predictions by generating a few dozen new cell cycle mutants. To address the bottle neck of experimental validation, we have scaled up the generation of new yeast cell cycle mutants using the synthetic genetic array (SGA) approach. Using selectable markers for haploid cells and gene deletions, SGA facilitates large-scale genetic crosses performed in arrays of 96 up to 1536 crosses per plate. Instead of only testing pairwise genetic interactions, we will be making strains with combinations of up to 6 deletion mutations. We are currently in the process of generating a set of ~3000 yeast cell mutants that our current model predicts to be synthetic lethal combinations or to rescue synthetic lethal combinations. The observed genetic interactions will either confirm aspects of the model or allow us to infer new molecular interactions and to further tune our model parameters.

ABSTRACT. Cellular morphology and morphodynamics are commonly used for qualitative and quantitative assessments of cellular states, since they are the phenotypic outcome of cellular processes such as differentiation, proliferation, migration, and apoptosis. Here we propose a framework to profile cellular morphodynamics based on an adaptive spectrum decomposition approach, in which time series of local cell boundary motion is decomposed and profiled into analytical instantaneous frequency spectra defined by the Hilbert-Huang transform (HHT), with a spatial and temporal resolution that matches the dimension of localized protrusion events. For an isogenic, clonal population of spontaneously protrusive Cos7 fibroblast, we found that the instantaneous frequency distributions are remarkably consistent in spite of vast cell-to-cell heterogeneity in cell boundary motion magnitude. Meanwhile, by acute silencing of the Vav2 GEF which activates Rac1 signaling, we discovered frequency distribution shifts. From this data we conclude that the frequency spectra encode the state of molecular regulation that governs the local cellular morphodynamic behavior, whereas the cell boundary motion magnitude merely captures the activation level of a given regulatory regime. Next, we adopted the profiled frequency spectra as features to divide the cell periphery into sub-regions with different morphodynamic behavior using a statistical region merging (SRM) algorithm in space and time. To evaluate and verify the clustering accuracy and quality, we then monitored Rac1 activity of Cos7 cells in clustered sub-regions of cell periphery with distinct morphodynamic profiles. The quantification results did identify different levels of local signaling strength and signaling coordination in those sub-regions. We claim that the proposed framework serves as a pre-classifier of cell boundary motion with consistent molecular states, which is of significance to further investigate the spatiotemporal coupling between cellular morphodynamics and underlying molecular regulation.

ABSTRACT. Metabolic networks are typically large and contain many metabolites and reactions. This results in dynamical models consisting of large systems of ordinary differential equations having many kinetic parameters. These parameters, however, are often unknown and must be estimated from experimental data. In our work, we assume the data to be metabolomics measurements of various steady states obtained by different input flux configurations to the network. Under the assumption of first order kinetics and normally distributed error terms, we are able to calculate the Fisher information matrix analytically for the unknown kinetic parameters. This enables us to study the structural identifiability of the kinetic parameters, i.e. if data can provide the information needed for estimation of the parameters or not for given input fluxes. If the input fluxes are controllable, the so-called D-optimality criterion can be used to find the optimal configuration of input fluxes that maximizes the information. In some cases, adding more measurements does not increase the information about the parameters. For smaller test networks this can be shown visually, and we are able to explain and predict the properties of the Fisher information matrix by the use of extreme pathways. Altogether, our work provides a methodology for analyzing the identifiability of kinetic parameters from steady state metabolomics data using various mathematical properties of the Fisher information matrix.

Acknowledgements: This research was supported by the Research Council of Norway through grant 248840.

ABSTRACT. The budding yeast cell cycle is regulated by complex and multi-scale control mechanisms, and is subject to inherent noise, resulted from low copy numbers of species such as critical mRNAs. Conventional deterministic models cannot capture this inherent noise. Although stochastic models can generate simulation results to better represent inherent noise, the stochastic approach is often computationally too expensive for complex systems, which exhibit multi-scale features in two aspects: species with different scales of abundances and reactions with different scales of firing frequencies. Moreover, it often requires detailed information of chemical kinetics, which is either not available or costly. To address this challenge, one promising solution is to adopt a hybrid approach. Hybrid stochastic method replaces the single mathematical representation of either discrete-stochastic or continuous deterministic formulation with an integration of both approaches, so that the corresponding advantageous features are well kept to achieve a trade-off between accuracy and efficiency. In this work, a comprehensive deterministic model is augmented first by adding mRNAs as the source of intrinsic noise in cell cycle. Second, we propose a hybrid stochastic model that represents a gene-protein regulatory network of the budding yeast cell cycle control mechanism, respectively, by Gillespie’s stochastic simulation algorithm (SSA) and ordinary differential equations (ODEs). Simulation results of our model are compared with published experimental measurement on the budding yeast cell cycle. The comparison demonstrates that our hybrid model well represents many critical characteristics of the budding yeast cell cycle, and reproduces phenotypes of more than 100 mutant cases. Moreover, the model accounts for partial viability of certain mutant strains. The last but not the least, the proposed scheme is shown to be considerably faster in both modeling and simulation than the equivalent stochastic simulation.

ABSTRACT. Model based design of biochemical networks of microorganisms with improved target metabolite yield or flux is a typical kinetic model optimization task. Selecting efficient small set of adjustable parameters to improve metabolic features of an organism is important to reduce implementation costs and risks of unpredicted side effects. The feasibility of optimization result can be improved by integration of biologically and biochemically relevant constraints during the optimization of kinetic models. Implementation of constraints reduces the impact of suggested changes in processes that are out of the scope of the model. This increases the probability that the result of kinetic model optimization can be carried out by an organism after its introduction in vivo. A case study was carried out to determine the impact of total enzyme activity and homeostatic constraints on the objective function values and the following ranking of adjustable parameter combinations. The total enzyme activity constraint limits the overexpression of enzymes by taking into account the limits of enzyme production resources. The homeostatic constraint limits large changes in metabolite concentrations in the model to avoid their potential impact on other reactions not included in the kinetic model but present in the living organism. The constraints were applied and optimizations were carried out using COPASI software. Several kinetic models were used. Experiments revealed that a homeostatic constraint caused heavier objective function value decrease than the total enzyme activity constraint. Both constraints changed the ranks of adjustable parameter combinations in the list of best combinations. No “universal” constraint-independent combinations of adjustable parameters were found suggesting that the mentioned constraints heavily influence the rank of best adjustable parameter combinations and should be applied in the optimization.

ABSTRACT. As computational tools became more affordable and powerful, there has been an extensive research on the role of stochasticity in biochemical systems, such as in bacteriophage lambda, phenotypical variations in cloned cells, and cell mutations in tumor formation. For these systems, stochastic modeling is necessary to capture the inherent variability of the underlying mechanism. However, many interesting phenomena that require stochastic modeling also involve rarity in that they are not a part of typical system behavior. The number of trajectories required to accurately estimate the probability of a rare event, which is inversely proportional to the probability, can be prohibitively expensive with conventional Monte Carlo simulation techniques. The doubly weighted stochastic simulation algorithm was developed by Daigle et al. to efficiently estimate a rare event probability by combining Rubinstein’s multilevel cross-entropy method into Gillespie’s stochastic simulation algorithm (SSA). The main contribution of this method was that the importance sampling parameters required to propagate the original system toward the event of interest were selected automatically without requiring user insight. The authors demonstrated a significant computational gain over using SSA alone. However, convergence rate of the multilevel cross-entropy method can be slow or non-existent if the system reaches a parametric plateau, where the cross-entropy method is unable to detect sufficient signal due to lack of stochasticity. Locations or existence of such parametric plateau is not known in advance, and the dwSSA does not provide a way to “leap” through the plateau. In this talk we introduce dwSSA++ that significantly improves the convergence rate of the original dwSSA. The speed up is achieved by integrating a modified version of the cross-entropy leaping that was first introduced in Stochastic Parameter Search for Events++ (SParSE++) by Roh et al. We compare the performance of dwSSA and dwSSA++ using a Susceptible-Infectious-Recovered-Susceptible (SIRS) disease dynamics model as well as a yeast polarization model.

ABSTRACT. Correlation matrices describing variable interdependency in biological systems provide valuable information about systems’ structures and functions. Sometimes, the interdependence is even more critical for differentiating biological system subtypes than commonly used characteristics, such as levels of omics, images, or other biological markers and attributes. There is an increasing need to use correlation matrices representing interdependence as signals for identifying patterns within biological systems. However, cluster analysis methods using correlation matrices have received little attention. We propose to use the upper-triangle elements of a correlation matrix to form a “snake vector” representing the entire correlation matrix. Then we cluster “snake vectors” by using existing clustering methods. We benchmark our method with existing methods for comparing correlation matrices, such as random skewers, T statistics, and S statistics, which focus on calculating the distances among the correlation matrices. Importantly, these existing methods only allow the use of hierarchical clustering based on pairwise distances between matrices. One advantage of our proposed method, which transforms correlation matrices into vectors, is the ability to apply a variety of standard clustering algorithms (e.g., K-means, K-medoids and hierarchical). Furthermore, a “snake vector” can be concatenated with the mean and variance of each attribute to form a “dragon vector” that synchronously leverages both levels of attributes and their interdependencies. We tested our method by clustering fMRI brain images from young and old individuals using connectivity matrices (correlations of brain signals across brain regions). The misclassification error of 2.7% using our proposed “snake vector” method outperformed the three existing methods (each with >40% misclassification error). In a simulation study, we derived two groups of matrices from two prototypes representing substantially different brain connectivity matrices. We also obtained smaller misclassification errors using our method compared to using existing methods, which demonstrated the superiority of “snake vectors” method for clustering correlation matrices.

ABSTRACT. Stochastic processes become more and more popular to model the mesoscopic nonequilibrium biophysical dynamics, which well fit the recent development of advanced experimental techniques at single-cell level.

Here I will take about two short stories. One is the molecular mechanism of transcriptional burst, which is uncovered by combining single-molecule in vitro experiments and stochastic models. The other is a new rate formula for phenotype transition at the intermediate region of gene-state switching for single cells, which is more general and more close to the reality of living cells. The new rate formula can explain a "noise enhancer" therapy for HIV reported recently, which motivated a future project of us.

ABSTRACT. Temporally ordered progression through a series of molecular events is essential for the successful completion of the cell cycle. Particularly, the splitting of chromosomes during anaphase needs to be coordinated with changes in the spindle apparatus, and with reversal from the mitotic to the interphasic state [1-4]. Here we address the question of how cellular heterogeneities (such as cell-to-cell variation in the abundance of anaphase regulatory proteins like securin and cyclin B) affect the relative timing of chromosome splitting and mitotic exit, with the aim of predicting mechanisms of temporal robustness. To this end, we extended an existing cell population average model, which describes the dynamics of anaphase regulatory proteins [5], to an ensemble of single cell models by sampling the protein concentrations from a lognormal distribution. This allows us to adequately represent cell-to-cell variation. This model is then calibrated by fitting to single cell data of wild type and perturbed cells. The single cell model reproduces the experimentally observed robustness of temporal order in wild type cells. For certain perturbation conditions, it predicts subpopulations of cells that exhibit problems in the temporal coordination of anaphase events, with possible implications for genomic instability. These predictions could be tested by assessing cell cycle completion under novel perturbation experiments, in which anaphase regulatory proteins are downregulated or overexpressed. Analysis of our model together with single cell data will aid in determining how genomically stable and unstable subpopulations differ in their protein content. Taken together, our results provide insights into the buffering of heterogeneity that ensures genome stability.

References: 1. Funabiki, H., Yamano, H., Kumada, K., Nagao, K., Hunt, T. & Yanagida, M. Cut2 proteolysis required for sister-chromatid separation in fission yeast. Nature 381, 438-441, doi:10.1038/381438a0 (1996). 2. Vazquez-Novelle, M. D., Mirchenko, L., Uhlmann, F. & Petronczki, M. The 'anaphase problem': how to disable the mitotic checkpoint when sisters split. Biochemical Society transactions 38, 1660-1666, doi:10.1042/BST0381660 (2010). 3. Sullivan, M. & Morgan, D. O. Finishing mitosis, one step at a time. Nature reviews. Molecular cell biology 8, 894-903, doi:10.1038/nrm2276 (2007). 4. Higuchi, T. Uhlmann, F. Stabilization of microtubule dynamics at anaphase onset promotes chromosome segregation. Nature 433, 171-176, doi:13.7020/15650742 (2005)

5. Kamenz, J., Mihaljev, T., Kubis, A., Legewie, S. & Hauf, S. Robust ordering of anaphase events by adaptive thresholds and competing degradation pathways. Molecular Cell 60, 446-459, doi:10.1016/j.molcel.2015.09.022 (2015).

ABSTRACT. An important component of the transcription process in eukaryotic cells is RNA splicing accomplished by the spliceosome. The Spliceosome is a complicated and highly dynamic machine that removes introns from precursor mRNA (pre-mRNA) and produces the mRNA. It consists of five protein–RNA complexes in the form of small nuclear ribonucleoproteins (snRNP).

We study the sequential formation of each snRNP, the assembly of the spliceosome, the removal of introns and the production of mRNAs in spatially-resolved models of a yeast and human cells. Specifically, using whole-cell simulations, we evaluate the noise associated with each of these processes under variable physiological conditions which ultimately leads to mRNA production noise. In addition, because determining the exact geometry of cellular components using experiments is a challenging task, we explicitly estimate the noise produced by the varying geometry in a population of eukaryotic cells. We start with the formation of a snRNP from its components. We have derived a series of kinetic equations that describe the assembly process of snRNPs. Preliminary results from this kinetic model in our S. cerevisiae and Hella cell geometries show that we can successfully generate the most abundant component of spliceosome, and observe the noisiness of its formation at different cellular RNA and protein (as reacting species) counts. The simulated time scale of snRNP particle formation compares well with experimental results, serving to validate our model.

These simulations can lay the foundation for studying eukaryotic systems with spatial resolution and pathogen-related processes such as alternative splicing in detailed models.

ABSTRACT. Signal transduction networks allow eukaryotic cells to make decisions based on information about intracellular state and the environment. Biochemical noise significantly diminishes the fidelity of signaling: networks examined to date appear to transmit less than 1 bit of information. It is unclear how networks that control critical cell fate decisions (e.g. cell division and apoptosis) can function with such low levels of information transfer. Here, we employ theory, experiments and numerical analysis to demonstrate an inherent trade-off between the information transferred in individual cells and the information available to control population-level responses. Noise in receptor-mediated apoptosis reduces information transfer to approximately one bit at the single-cell level but allows 3-4 bits of information to be transmitted at the population level. For processes such as eukaryotic chemotaxis, in which single cells are the functional unit, we find high levels of information transmission at a single-cell level. Thus, low levels of information transfer are unlikely to represent a physical limit. Instead, we propose that signaling networks exploit noise at the single-cell level to increase population-level information transfer, allowing extracellular ligands, whose levels are also subject to noise, to incrementally regulate phenotypic changes. This is particularly critical for discrete changes in fate (e.g. life vs. death) for which the key variable is the fraction of cells engaged. Our findings provide a framework for rationalizing the high levels of noise in metazoan signaling networks and have implications for the development of drugs that target these networks in the treatment of cancer and other diseases.