Research

Earth harbors countless extreme environments — the desert interior, the deep sea at ten thousand meters, nuclear reactor cooling water — yet life persists in all of them. Understanding what makes living things so resilient is one of the central questions in modern biology.

Bacteria are the most striking example. They inhabit virtually every environment on Earth, from boiling hot springs and Antarctic ice to acidic waters and the human gut, and their total number is estimated to exceed the stars in the universe by a factor of one million. Dividing roughly every twenty minutes, bacteria accumulate in days what would take humans tens of thousands of years in generations, continuously generating new variants and diversifying at extraordinary speed.

What drives this resilience? One key insight is that adaptive change in bacteria is not random: specific patterns and directionalities govern which genes mutate and how traits such as metabolism shift. These patterns might seem to constrain evolution, yet they are thought to play a crucial role in ensuring that functional variants arise at high frequency. Uncovering these patterns and the molecular mechanisms behind them could inform the design of antibiotics that resist evolutionary escape, enable more stable genetic engineering, and help explain why complex biological systems evolve so efficiently. A second dimension concerns the population level — how bacteria interact collectively to prevent group collapse, a question central to understanding microbial ecology.

My research aims to uncover these mechanisms of adaptation, from the molecular to the population level.

One reason bacteria can adapt to their environment with remarkable speed is their ability to generate genetic changes rapidly. But how exactly do bacteria alter their genes? To obtain a coherent picture, I'm trying to reveal multiple distinct mechanisms, from single-letter substitutions to sweeping structural rearrangements of the genome.

We focused on the simplest kind of DNA change: the substitution of a single “letter” in the genetic code, known as a point mutation. Cells have multiple systems for correcting errors that arise during DNA replication, and disabling these repair genes increases the overall mutation rate (PLOS Genet., 2015). Importantly, however, not all types of mutations increase equally — each disrupted repair pathway produces its own characteristic pattern of mutations (Genes Cells, 2015). Our research revealed that some types of mutations that cannot be generated by disabling repair genes alone can be produced by ultraviolet (UV) irradiation. In other words, “regulation of repair functions” and “responses to UV-induced DNA damage” are complementary mechanisms, each generating a distinct spectrum of mutations Sci. Rep., 2017; SLAS Tech., 2019; G3, 2023). By combining multiple such pathways, bacteria can explore a far broader range of genetic changes than any single mechanism alone would allow.

A key driver of large-scale structural changes is a class of mobile DNA elements called insertion sequences (ISs). When an IS moves within the genome, it can cause deletions or duplications of nearby sequences, substantially reorganizing the arrangement of surrounding genes. Our research showed that this reorganization is not mere “breakage” — it can carry adaptive significance. For example, we experimentally demonstrated that two genes initially located far apart can be brought into physical proximity through IS activity, enabling them to be regulated together as a single unit called an operon (Nucleic Acids Res., 2022). In a further experiment, introducing high-activity ISs into E. coli produced structural changes affecting more than 5% of the entire genome within just ten weeks, illustrating how rapidly ISs can reshape the genetic landscape (Nucleic Acids Res., 2025).

How do genetic changes alter the behavior of a cell?

Changes in DNA ultimately manifest as changes in the observable behavior of a cell. We use the term trait to refer to a measurable property such as growth rate or metabolic activity, and phenotype to refer to the specific state a trait takes in a given organism. Cells possess many traits simultaneously, and if these traits were to change independently and at random, the finely tuned systems inside the cell would quickly break down. In practice, traits change in a coordinated fashion: when one shifts, others shift in concert, allowing cells to adapt efficiently without losing function. The molecular mechanisms that give rise to this coordination, however, remain largely unknown.

Directionality in gene expression change

To ask whether mutations tend to shift gene expression in particular directions, we analyzed expression patterns in strains that had accumulated random mutations. Regardless of which gene was mutated, specific groups of genes consistently increased or decreased together, revealing preferred directions of expression change. These directions aligned with the expression shifts induced by environmental perturbations. Mapping these directions onto the transcriptional regulatory network, we found that a small set of core hub genes — capable of sensing a wide range of intracellular and extracellular signals — were strongly associated with these preferred directions. This suggests that such core regulatory structures provide the molecular basis for coordinated trait change (Nature Commun., 2025; Nature Commun., 2015).

Directionality in metabolic change

A similar coordination appears in metabolic capacity. To test whether mutations shift metabolism in particular directions, we measured the growth rates of randomly evolved E. coli strains on 32 distinct carbon sources. Across strains, changes in metabolic capacity were coordinated in discrete modules: glycolytic carbon sources such as glucose and fructose changed together, while gluconeogenic carbon sources formed a separate module. This modular structure was consistent across both natural isolates and laboratory-evolved strains. Because these modules map directly onto the pathway architecture of the metabolic network, the results suggest that the structure of the metabolic network itself constrains the directions available to evolution (preprint).

How do populations avoid collapse?

No matter how rapidly individual bacteria adapt, the collapse of the population as a whole renders that adaptation meaningless. Bacterial populations face constant threats of collapse — through resource exhaustion, interspecies competition, and other pressures. How, then, do bacteria interact with one another to keep their populations intact? This question is central both to understanding the stability of microbial communities in nature and to explaining what makes bacteria so resilient.

Recycling of dead cells sustains long-term population survival

Analysis of population survival dynamics under resource exhaustion revealed an unexpected mechanism. When nutrients run out, most bacteria die — but surviving cells persist by scavenging and reusing materials released from dead cells. What is striking is that the survivors do not consume all of the resources available to them. Once viable cell density reaches a certain threshold, cells restrain their own growth and refrain from depleting the remaining nutrients. We demonstrated, through both experiments and mathematical modeling, that this self-restraint prevents premature resource exhaustion and enables long-term population survival. The balance between cell death and nutrient recycling thus functions as a population-level survival strategy, operating independently of adaptation at the individual level (mBio, 2017; Sci. Rep., 2023).

Interspecies cooperation under resource-poor conditions

In communities where multiple bacterial species coexist, interspecies interactions strongly influence community stability. We analysed 896 pairwise interactions among eight bacterial species isolated from diverse natural environments, grown under 32 conditions varying in carbon source diversity. We found that synergistic interactions — in which species promote each other’s growth — were prevalent in environments with few carbon sources, and that interactions shifted toward competitive as carbon source diversity increased. This shift was explained by the intensification of resource competition in more complex environments. These findings show that a single environmental factor, the diversity of available carbon sources, is a key determinant of the nature of interspecies relationships in bacterial communities (The ISME Journal, 2025).

Cells can change their state far more rapidly than mutation or sensor-based environmental responses allow, through stochastic fluctuations in gene expression — “noise” arising from the small number of molecules inside a cell. Even genetically identical cells in the same environment show cell-to-cell variability in expression levels, and this variability is thought to help cell populations explore diverse states in response to environmental shifts too rapid or too subtle for mutation or sensing to handle. Yet noise that is too large or too fast to be matched to environmental dynamics offers no adaptive benefit. This project combined single-cell quantification of expression noise with synthetic biology to test whether such fluctuations are genuinely used as a functional adaptive strategy.

Tracing stochastic noise in gene expression

Gene expression noise is observable in living cells, but the complexity of cellular metabolism, growth, and division makes it difficult to determine whether this noise arises intrinsically from the gene expression process itself (Phys. Biol., 2009). To address this, we enclosed a cell-free transcription and translation system inside lipid vesicles (artificial cells), reconstituting gene expression under conditions free from the confounding effects of living cells. We measured the expression levels of two fluorescent proteins driven by the same promoter across individual vesicles and quantified stochasticity from the bias in fluorescence intensities. The resulting cell-to-cell variation in expression was substantial, and its magnitude matched that observed in living E. coli — demonstrating that gene expression noise is an intrinsic property of the transcription and translation process itself, not a product of the complexity of living cells (ACS Synth. Biol., 2015).

Adaptation by stochastic noise

To test whether gene expression noise can directly support adaptation, we engineered an E. coli strain in which hisC — a gene essential for histidine biosynthesis — was removed from its native regulatory context and placed under the control of a foreign promoter, severing its connection to the cell’s endogenous histidine-starvation sensor. In this strain, survival under histidine depletion depends entirely on whether stochastic fluctuations happen to produce cells with sufficiently high hisC expression. Single-cell imaging confirmed that only those cells in which noise had transiently elevated expression could grow, allowing the population to escape extinction. Critically, the timescale of the fluctuations was slow enough to influence cellular metabolism and growth, showing that the noise was functionally coupled to growth fitness rather than being merely transient. These results demonstrate that gene expression noise alone — with no evolved sensing mechanism — can be sufficient for a population to adapt to environmental challenge (Mol. Syst. Biol., 2011; PLOS One., 2011; Nucleic Acids Res., 2014).