Session K51: Systems Biology

Translation Bottlenecks and Bacterial Growth Laws Predict Combined Antibiotic Action

Fast growth of bacteria requires a well-orchestrated translation machinery which is modulated internally by translation factors and perturbed by certain antibiotics (translation inhibitors). Coupling of the growth rate and ribosome synthesis results in ribosomal growth laws - relations connecting ribosome concentration to the growth rate. Joint effects of antibiotic combinations range from synergistic to antagonistic and are hard to predict, as the mechanisms of these interactions remain unknown. We hypothesize that the interactions between translation inhibitors arise from kinetic properties of the antibiotics together with the interplay of different stages in which ribosomes are halted. To test how halting of the ribosomes affects the efficacy of antibiotics, we constructed bacterial strains in which we imposed artificial bottlenecks in translation by controlling the abundance of different translation factors. Measuring the change in the antibiotic efficacy for a given limiting step, we used the growth law-based mathematical model to predict the interactions between various translation inhibitors. These findings offer new insights into the mechanisms of antibiotic interactions and translation itself and suggest a novel way of designing new antibiotic therapies.   

Optimal Structure of Odorant-Receptor Interaction for Effective Olfactory Discrimination

An important task of olfactory sensing is the discrimination of different odors. An odor captures the chemical state of the environment in a mixture of smell molecules, called odorants. Olfactory sensing is realized by the selective binding of odorants to a set of olfactory receptors (ORs), which in turn activates the corresponding olfactory sensory neurons, constructing the brain's first representation of the odor. Despite the high-dimensional nature of olfactory sensing, recent measurements with human ORs suggest that the odorant-receptor interaction is sparse; only a small fraction of all available pairs interact, while the interaction strengths vary. Why did the OR system evolve this way? More generally, what are the optimal interaction structures for effective olfactory discrimination? We investigate these questions by combining studies of model systems and analyses of experimental data. In particular, we discuss how the graph-theoretic properties of the interaction should reflect the environmental odor statistics. We also point out that the human OR system adopts such optimal design.   

Universal properties of estimating many ligand species concentrations by many cellular receptor types

All organisms, and specifically cells, are faced with the task to sense the concentrations of a large variety of ligands using a limited set of receptors. As the number of receptor types is significantly smaller than ligands, the inference is impossible through deterministic steady-state chemical kinetics. Recently (Singh and Nemenman, 2017, PLOS Comp Biol), we have shown that the sequence of bound-unbound states of a receptor carries more information than its mean occupancy. This can be used to infer concentrations of several ligands from the activity of a single receptor, as long as the unbinding rates of the ligands are sufficiently different. Here, extending these ideas to a network of multiple receptors that bind multiple ligands, we show that the temporal sequences of binding-unbinding events again carry more information than the mean occupancy. The analysis of the Fisher information matrix shows that, for a random distribution of unbinding rates, one can estimate concentration of 3-10 times as many ligands as there are receptors, in realistic time. The spectrum of the Fisher matrix shows interesting scaling regimes with the ratio of the number of ligands to receptors, and the scaling is universal for different probability distribution of the unbinding rates.   

Design principles for the optimal arrangement of consecutive enzymes

The spatial organization of enzymes is important in the regulation and the efficiency of metabolic pathways. In previous studies, we showed how to optimize the production rate of the last metabolite of a cascade of enzymatic reactions. We observed a transition in the optimal enzyme arrangement from a clustered configuration into a more extended profile. This transition resulted to be independent of the reaction kinetics, spatial dimensions and loss mechanisms considered. However this generality was not fully understood. Here we derive a condition defining the boundary between the two different regimes. The condition compares two measurable quantities: the reaction flux at the cluster and the diffusive flux away from the cluster. Further, we describe the optimal arrangements as the result of a diversification strategy for enzymes investment. We analyze the marginal returns of the product flux per extra enzyme added at different positions in the system and we find that, for the optimal profiles, different positions generate the same marginal returns. Starting from this strategy, we derive an algorithm for building the optimal arrangements as the amount of enzymes in the system increases, thereby suggesting possible instructions for the realization of optimal bio-reactors.   

Modeling metabolic exchange between microalgae and bacteria to test mechanisms of bacteria-mediated biomass enhancement

Experiments have demonstrated that the rate of biomass production in laboratory batch cultures of the green algae Chlamydomonas reinhardtii is increased when it is co-cultured with the bacterium Arthrobacter strain P2b compared to when cultured on its own. The algal/bacterial co-culture shows an up to 2.4 times higher chlorophyll concentration compared to the axenic culture. This suggests a mutualistic interaction between the two organisms potentially mediated by metabolite exchange. Here we report the results of in silico investigations of the metabolic interaction between these two organisms with the aim of elucidating the underpinnings for this phenomenon. We have created the first genome-scale, constraint-based model (GSM) of strain P2b and used it to examine robustness of the organism’s metabolism to genomic and environmental perturbations. Additionally, we have also discovered a pathway for phytohormone production in P2b and have used the model to study the effect on the rest of its metabolism. Finally, by pairing the P2b model with a well-curated GSM of C. reinhardtii, and using in silico analysis methods like dynamic FBA and multi-objective flux analysis, we have begun to examine interactions between C. reinhardtii and P2b.   

The Resource Allocation Strategies of E.coli under Different Limitations of Ribosome Synthesis

Facing constant selection pressure, microorganisms have to wisely allocate their limited resources. Building ribosomes takes a significant amount of cellular resources and their abundance is found to be tightly linked to cell growth rate. However, contrary to the prevailing theory that ribosomes always work at maximal efficiency, we found that phosphorus-limited E. coli produce protein at the same rate but contain significantly fewer ribosomes than cells limited for carbon or nitrogen. Why and how do ribosomes perform differently under different nutrient limitations? By modeling ribosome dynamics on both the macroscopic and microscopic scales, with input from detailed experimental measurements, we discovered that C/N-limitation results in high rates of aborted translation. As predicted by the model and validated by experiment, this mechanism, while not optimized for steady-state translation efficiency, promotes rapid acceleration of growth rate upon nutrient repletion. This system provides a thought provoking example of an evolutionary trade-off between current growth and future expectation. Our findings also indicate a previously unappreciated role for premature termination, and add an unexpected nutrient-dependent sub-optimality to the bacterial growth laws.   

A Quantitative, Genome-Wide Study of Translation Efficiency in E. coli

There has been much interest in how living cells control protein synthesis, both at the single-gene and genome-wide levels. Here we present a comprehensive study of translation in E. coli, combining quantitative proteomic and transcriptomic methods in a wide variety of growth conditions. We find that most mRNAs are transcribed with similarly high efficiencies across the growth conditions examined. The high average efficiency corresponds to a high density of translating ribosomes on the mRNAs, about 8 Rb/kb, not far from the range of maximal packing. Variations in translation efficiencies, due e.g. to post-transcriptional regulation, are negligible for most genes. A simple model of translation initiation is introduced to discuss the coordination between translation initiation and elongation processes, and to explore what may be optimal for protein synthesis in E. coli.   

Target Control in Logical Models using Domain of Influence of nodes

Dynamical models of biological networks are used to understand the underlying mechanisms of complex diseases and to design therapeutic strategies. Network control problems, and especially target control, are a promising avenue toward developing disease therapies. In target control it is assumed that a small subset of nodes is most relevant to the system’s state and the goal is to drive the target nodes into their desired states. An example of target control would be driving a cell to commit to apoptosis. From the experimental perspective, gene knockout, pharmacological inhibition of proteins and providing sustained external signals are among practical intervention techniques. We propose to use the stabilizing effect of sustained interventions in logical models to solve this problem in biological networks, especially signal transduction networks. Specifically, we define the domain of influence of a node to be the nodes which will be ultimately stabilized by a sustained state of this node regardless of the initial state of the system. Thus a solution to the target control problem is the set of nodes whose domain of influence can cover the desired target node states. We apply our strategy to several biological networks to demonstrate its effectiveness.   

The dark matter of the human gut microbiome follows neutral ecological assembly rules

The discrepancy between the number of microbes that can be observed directly vs those that can be cultured has vexed scientists for decades. Here, we investigate whether yet-to-be-cultured microbes of the human gut microbiome are inherently different from cultured microbes. Using metagenomics from 22 fecal samples, we assembled 85 complete, previously unsequenced genomes from uncultured microbes. We then explore the ecological assembly processes and functional characteristics among these organisms. Specifically, we test the hypothesis that culturing selects for microbes that occupy a stable metabolic niche, whereas microbes that have yet to be cultured lack such a niche and assemble under stochastic, neutral processes (e.g. births, deaths, and migration). We find that neutral processes dominate in uncultured microbes and that functional predictions based on gene content fail to explain why they that have not been cultured. The findings from this study—which apply the tool of metagenomics to the problem of ecological assembly rules—may help lay the groundwork for understanding how the human gut microbiome is assembled, with important implications for health.   

In Silico Analysis of Antibiotic-Induced C. difficile Infection

We study antibiotic-induced C. difficile infection (CDI), caused by the toxin-producing C. difficile (CD), and implement clinically-inspired simulated treatments in a computational framework that synthesizes a generalized Lotka-Volterra (gLV) model with SIR modeling techniques. The gLV model uses parameters derived from an experimental mouse model, in which the mice are administered antibiotics and subsequently dosed with CD. We numerically identify which of the experimentally measured initial conditions are vulnerable to CD colonization, then formalize the notion of CD susceptibility analytically. We simulate fecal transplantation, a clinically successful treatment for CDI, and discover that both the transplant timing and transplant donor are relevant to the the efficacy of the treatment. We incorporate two nongeneric yet dangerous attributes of CD into the gLV model, sporulation and antibiotic-resistant mutation, and for each identify relevant SIR techniques that describe the desired attribute. Finally, we rely on the results of our framework to analyze an experimental study of fecal transplants in mice.   

Statistical physics of (meta)genomes

Despite the contingency of evolution, the gene content of microbes shows remarkable scaling relationships. For example, the number of transcription factors scales quadratically with the total number of genes, and similar scaling laws emerge for groups of gene sharing evolutionary history or function. On an ecological scale, microbial communities are also characterized by strong regularities in their species composition. At both the genomic and the ecological scales, statistical physics approaches help us understand the emergence of such regularities from simple ingredients, such as birth-death of individuals, duplication, transfer, and loss of genes. However, we are only in the early days of a unified description. Metagenome sequencing gives access to both the gene content and the taxonomic composition of microbial communities. Here we show how the combination of within-genomes laws and ecological patterns leads to regularities in the gene content of whole microbial communities. Exploiting one of these regularities, we introduce a robust method to estimate the average genome size of the community.   

Dynamics and Optimal Behavioral Strategies of Motile Networks

Network models play a key role in understanding connectivity and flow in diverse systems such as gene regulation, transportation, and ecosystems. Far less explored are networks that move through space. Here we investigate an understudied class of network models, i.e., trees where the network morphology can dynamically change while the overall mass (sum of all edge lengths) is conserved. We developed a simple mathematical model accounting for the branching, growth, and retraction rates, leading to a compact yet insightful phase space that reveals different optimal strategies for system behavior under different external constraints. We successfully apply this model to various natural and artificial systems, e.g., cellular chemotaxis, slime mold behavior, and businesses evolution – suggesting universal optimal behavioral strategies in motile network systems.   

Long-term growth rate of nonlinear autocatalytic networks

A flux network is a fundamental structure of many biochemical, ecological, and economical systems. Yet, the long-term behavior of general flux systems remains to be characterized, especially when nonlinear functions are involved. Here, using ergodic theory, we prove the convergence of the growth rate for a large class of nonlinear flux system. This class of flux systems exhibits not only steady-state growth but also allows complex dynamics such as limit cycles to coexist with robust growth.

Our mathematical framework provides explicit formulae to analyze the following types of questions: (Q1) Given a flux network, how much will growth rate vary if the underlying flux function is modified? (e.g., by reducing enzymatic activity or by changing the Hill coefficient) (Q2) if the system state is transiently altered (e.g., by introducing/removing a specific amount of molecules, or through noise fluctuations), what would be the long-term impact on the system size? We envision that this framework will stimulate future understanding on fundamental principles of general autocatalytic networks, including cellular systems and ecosystems.