Abstract :
[en] The fundamental characteristics of living organisms, including growth, adaptation, homeostasis, organization, metabolism, and reproduction, are encoded in their DNA. Despite the crucial control exerted on genes, clonal cells exhibit heterogeneous behavior even when exposed to homogeneous environments. This phenomenon is justified by the theory of bet-hedging, which suggests that it allows populations to spread risks and opportunities. However, this heterogeneity poses significant challenges in biotechnology, reducing bioprocess performance and robustness, and has far-reaching implications for human health, including antibiotic resistance, cancer treatment, and potentially even its onset. The causes of this heterogeneity are multifaceted, including the diversity and uneven distribution of essential intracellular macromolecules necessary for gene expression. Our study reveals that, despite the complexity of these factors, intracellular dynamics alone are insufficient to explain the observed heterogeneity at the population level. We quantified population heterogeneity using Shannon entropy, a metric borrowed from information theory, and identified three diversification regimes characterized by increasing heterogeneity. A key predictor of these regimes is the impact of gene activation on cell growth, the switching cost. We found that as this switching cost increases, also known as the burden or production load, heterogeneity within the population rises accordingly. Interestingly, when the switching cost becomes significant, cells can synchronize their gene expression into periodic bursts, oscillations.
From these observations, the rest of the work is structured around two primary questions: I. Why are burdensome genes
associated with greater population entropy and; II. How can a gene circuit that does not have the properties of an oscillator nevertheless exhibit an oscillatory behavior?
To understand the relationship between population heterogeneity and the burden imposed by gene activation, we introduced the concept of burden entropy compensation. This mechanism suggests that the burden associated with gene activation is offset by the overgrowth of cells with lower activation of the burdensome gene. This overgrowth spreads the population thus increasing entropy, but safeguards the population from being washed out in continuous cultivation
devices.
We focused on the T7 expression system in E. coli BL21 and found that periodic addition of the inducer can reduce entropy, but this control method is countered by the emergence of mutants with weakened promoters, lower burden, and more homogeneous induction. We then demonstrated that reducing the switching cost by lowering the quality of the main carbon source, and thus the maximum growth rate, homogenized gene expression across the population
without sacrificing induction strength.
This counterintuitive outcome highlights the importance of considering phenotypic heterogeneity using a system biology approach, rather than solely focusing on intracellular sources of noise. The significance of this system approach was further confirmed when we attempted to explain why highly burdensome gene circuits can exhibit bursty expression. We developed a mathematical model that incorporates the interplay between cellular stress response and environmental changes. The model revealed that, for highly burdensome gene circuits under certain dilution rate conditions, expression can be bursty and exhibit oscillatory behavior. The model predictions were validated experimentally in chemostat cultures of Bacillus subtilis, where sporulation synchronized across the population and appeared as oscillations. These oscillations result from a feedback loop where if cells are growing then the concentration in glucose drops below a threshold that triggers sporulation, and if cells sporulate, then the concentration
rises above this threshold thus preventing sporulation.
Crucially, our results show that this unwanted instability can be eliminated, and entropy reduced, by carefully choosing process parameters that align with the biological characteristics of the cultivated organism. Our findings establish a clear link between gene expression burden and cellular entropy, demonstrating that entropy can be reduced to benefit overall population productivity, potentially unlocking the wider utilization of continuous cultivation.
