(updated Nov 2024)
We are a group of biophysicists employing both experimental and theoretical approaches to investigate various topics related to phase transitions, morphology, and evolution, as well as chromatin physics. Our overarching goal is to address and solve compelling biological problems from unique perspectives, leveraging concepts from physics and mathematics, and establishing new experimental systems when necessary.
See also publications.
The emergence of macroscopic properties, such as water solidifying into ice, is often associated with the intricate collective behavior of constituent entities. These phenomena can be understood using simplified models involving fundamental units like spins or interacting agents.
Similar collective phenomena arise in biological systems but exhibit distinct characteristics compared to the simplified models typically studied in equilibrium statistical mechanics. One key difference is the presence of active entities, capable of self-propelled motion that goes beyond simple thermal fluctuations. Furthermore, cells can self-replicate, increasing their numbers, and also die, with subsequent replenishment from the surrounding tissue. These nonequilibrium properties are central to understanding the behavior of our bodies, which are essentially complex collectives of cells.
In our in vitro experiment work, we culture elongated cells with high migratory activity, which effectively create active nematic systems. These cells, exemplified by neural progenitor cells, form nematic patterns reminiscent of liquid crystals, exhibiting topological defects indexed as +1/2 and -1/2. We previously demonstrated that these defects, arising spontaneously as geometrical features of the nematic ordering, act as sources and sinks of cell flow (Kawaguchi, Kageyama, Sano, Nature 2017). These cells exhibit interesting collective dynamics including chiral edge flows (Yamauchi et al arxiv 2020), which is an example of topological active matter (review: Sone et al arxiv 2024) The accumulation at +1/2 defects has been explained by the coupling of anisotropic friction and active stress, which has been measured and refined in recent work (Uwamichi et al arxiv 2024, Zhao et al biorxiv 2024).
A fundamental question is how nonequilibrium driving influences the macroscopic properties of many-body physical systems. One of the simplest examples of activity-induced phase transitions is motility-induced phase separation (MIPS), observed in systems of self-propelled particles with volume exclusion. Another type of transition occurs when entities possess spin degrees of freedom coupled to their direction of motion, potentially leading to flocking behavior, analogous to a ferromagnetic transition.
We have recently investigated the possibility of observing these activity-induced phase transitions in quantum many-body systems. By exploiting the connection between classical stochastic dynamics and quantum many-body models, we have shown that the quantum model extends the phase diagram of classical models (Adachi, Takasan, Kawaguchi, Phys Rev Research 2022), revealing novel phases. We also identified simple mechanisms driving ferromagnetic transitions in the deeply quantum regime (Takasan, Adachi, Kawaguchi, Phys Rev Research 2024).
The quantum phase transitions in these models can be interpreted as dynamic phase transitions from the classical perspective. We also explored minimal models exhibiting dynamical phase transitions, finding that even single-particle Brownian motion can display such transitions in dimensions higher than four (Kanazawa, Kawaguchi, Adachi, arxiv 2024 (letter), arxiv 2024 (full paper)).
Our in vivo work involves collaborations with experimentalists conducting cutting-edge live imaging experiments to capture the complexities of multicellular dynamics. In a series of studies, we have elucidated how skin stem cells coordinate their fates to maintain tissue homeostasis (Rompolas, Mesa et al., Science 2016, Mesa, Kawaguchi, Cockburn et al., Cell Stem Cell 2018). This problem connects interestingly with interacting particle systems, particularly the voter model (Yamaguchi, Kawaguchi, Sagawa, Phys Rev E 2017), a critical model with two absorbing states exhibiting coarsening dynamics.
While applying simple nonequilibrium models to in vivo data is valuable, we are also interested in moving beyond such models to identify truly biologically relevant factors. To this end, we developed a graph neural network (GNN) model to capture the contributions of multiple cellular variables, including genetic markers and cell interactions, to cell fate decisions (Yamamoto et al., Plos Comp Biol 2022). This method incorporates an attribution strategy to unbiasedly assess how the fates of neighboring cells influence the fate of a given cell, revealing additional cell fate coupling beyond targeted searches.
Since cell positioning and cell-cell interactions are crucial for cell fate determination and kinetics, we are currently applying similar graph-based analyses to decipher the rules governing developing embryos, using live imaging data as input (on-going).
Another distinguishing feature of collective phenomena in biology, compared to simple statistical physics models, is the diversity and heterogeneity of interactions. This is exemplified by biological condensates – spontaneously forming structures within cells, such as nucleoli, nuclear speckles, and stress granules – composed of proteins and RNAs. These condensates form through liquid-liquid phase separation driven by multivalent interactions between intrinsically disordered regions (IDRs) of proteins and RNA molecules. Understanding the principles governing their formation and regulation is crucial for deciphering cellular organization and function.
We are interested in the physics of these heterogeneous interactions and have worked on estimating the interaction strength between two heteropolymers, such as IDRs in proteins (Adachi, Kawaguchi, Phys Rev X 2024). We developed a method to effectively predict the outcome of phase separation and demixing simulations using real IDR sequences in coarse-grained molecular dynamics simulations. This theory allows us to predict which sequences will mix or demix, design sequences that attract or repel specific targets, and reveals that simple amino acid interactions can support only three or four distinct, immiscible condensates, implying that multi-phase coexistence observed in cells requires additional mechanisms. We have also explored how the kinetics of the production of the constituents of phase-separating droplets can affect the localization of the formed droplets in the setup of wetting (Adachi, Kawaguchi, Phys Rev E 2021).
We also recently applied percolation theory to the problem of concept emergence in large language models, using a synthetic language approach (Lubana, Kawaguchi et al., arxiv 2024). Percolation, another type of nonequilibrium phase transition, describes the formation of connected clusters in a random system. In the context of large language models, we consider how the models learn concepts—groups of words with shared properties—by forming percolated clusters on graphs.
Eukaryotic DNA is wrapped around histone proteins to form chromatin, enabling its packaging within the nucleus. Chromatin structure is far from random, and polymer physics concepts have been used to understand its basic properties, such as the contact probability observed in Hi-C experiments.
We have developed simplified models of chromatin state transitions that incorporate both the polymer nature of chromatin and the active turnover of its monomeric components, nucleosomes, under statistical physics laws, accounting for histone modification-dependent interactions (Adachi, Kawaguchi, Phys Rev E (R) 2019). This coupling between monomer interactions and polymer behavior (a polymer Potts model) can naturally lead to switch-like phase transitions, enhancing the sharpness of the transitions compared to standard Potts models.
We are also developing techniques to reconstitute chromatin in vitro at larger scales. We are imaging single-molecule fluctuations and fine structures using atomic force microscopy and chromatin conformation capture (Fukai et al., biorxiv 2024).
One of the most fascinating biological challenges is understanding the diversity of living organisms and developing a theory connecting genomics to the development and evolution of morphological features. The Hox genes, vital for body patterning, particularly within the vertebral column, provide a compelling example of this connection. Manipulating Hox genes can induce homeotic transformations, suggesting that variations in vertebral counts across species arise from such shifts, with a gain in one type often accompanied by a loss in an adjacent type. This intricate balance may be linked to the Hox cluster's chromatin organization, where the sequential opening of chromatin, governed possibly by principles of polymer physics, could play a crucial role in coordinating vertebral development.
To explore the role of Hox-mediated homeotic shifts in vertebral count diversity, we constructed a comprehensive database of tetrapod vertebral formulae, systematically analyzing combinations of vertebral counts across different body regions (Cerbus, Hiratani, Kawaguchi, PNAS 2024, github). This analysis revealed previously unknown homeotic patterns in mammals, extending beyond the thoracolumbar region. Intriguingly, we also discovered unexpected balances between distal vertebral counts in other tetrapods, particularly birds, bats, and certain extinct theropods. These balances defy simple explanations by the traditional Hox-vertebrae homeotic relationship, suggesting the involvement of additional developmental mechanisms.