Bacterial expression of human small GTPases induces stochastic predatory responses and motility interference in Physarum polycephalum
Click through the PLOS taxonomy to find articles in your field.
For more information about PLOS Subject Areas, click here.
We investigated the behavioral responses of the slime mold Physarum polycephalum to Escherichia coli expressing human small GTPases to elucidate how internalized proteins affect its actomyosin-based motility machinery. We found that prey bacteria expressing hKRAS WT or hRhoA triggered a deterministic “conflict” behavior, in which P. polycephalum failed to complete ingestion despite physical contact. Notably, the oncogenic hKRAS G12C mutation induced a transition to stochastic switching among predation, conflict, and sampling-mediated rejection, whereas hRac1 expression showed a tendency toward diversified behavioral outcomes. To explain this, we propose a working model centered on molecular jamming or local perturbation of intracellular oscillatory dynamics, in which non-prenylated G-domains act as decoys that decouple the contractile machinery from leading-edge actin dynamics. This behavioral stochasticity likely reflects physical noise arising from the specific conformational states and functional heterogeneity of the expressed proteins. Our results suggest that P. polycephalum functions as a sensitive biological sensor of the biophysical properties of prey proteins, providing a unique platform for exploring the co-evolutionary arms race and molecular survival strategies between bacteria and amoeboid predators.
Citation: Saiki R, Iimura R, Ichimura M, Osada Y, Sano K-I (2026) Bacterial expression of human small GTPases induces stochastic predatory responses and motility interference in Physarum polycephalum. PLoS One 21(9): e0359275. https://doi.org/10.1371/journal.pone.0359275
Editor: Yosuke Senju, Okayama University, JAPAN
Received: June 12, 2026; Accepted: September 11, 2026; Published: September 24, 2026
Copyright: © 2026 Saiki et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the manuscript and its Supporting Information files.
Funding: JSPS KAKENHI grant awarded to K. S. (24K08605) The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
The plasmodium of Physarum polycephalum is a multinucleated giant amoeboid organism that exhibits dynamic motility driven by the active formation of lamellipodia and filopodia and by rhythmic actomyosin contractions [1–3]. Despite lacking a central nervous system, this organism demonstrates primitive biological intelligence, as evidenced by its ability to solve optimization problems [4–10]. For instance, the plasmodium can find the shortest path in a maze to connect food sources and solve the traveling salesman problem through neurocomputing based on photophobic responses [5,9]. These intelligent behaviours are realized through the chaotic fluctuations of its oscillatory protoplasmic streaming, which allow the organism to search for optimal solutions and spontaneously destabilize stable states to explore alternatives [5,6]. Thus, P. polycephalum represents a unique biological system where complex behavioural patterns and decision-making are governed by its intrinsic rhythmic dynamics [6,7,11–13].
As a multinucleate organism, the plasmodium grows to a macroscopic scale through repeated nuclear divisions without cytokinesis, resulting in a single cell that behaves as a single individual [1]. In this organism, morphological deformation is synonymous with behaviour, driven by actin cytoskeletal reorganization, acto-myosin cytoskeletal reorganization, and internal pressure generated through rhythmic contraction [3,12]. This unique organization enables the simultaneous analysis of molecular information flow, cytoskeletal remodelling, and macroscopic behavioural outputs. By perturbing these pathways at the molecular level, we aim to elucidate the direct link between cytoskeletal dynamics and biological intelligence. Our approach—integrating micro-level signalling with macro-level behavioural analysis—represents a novel fusion of nanotechnology, biotechnology, and information technology.
A cornerstone of our research is the reconstruction of such biological dynamics through an in vitro reconstituted approach [14,15]. We have previously developed a functional actin-based gel cross-linked with polyethylene glycol (PEG), which exhibits sustained autonomous oscillation of its storage modulus with a periodicity of 15–25 min [16]. This underscores that the rhythmic motility and decision-making of P. polycephalum emerge from the autonomous, multiscale rheological properties of its internal actin machinery.
Complementing this bottom-up study, we further explore how exogenous molecular interference can directly modulate these behavioural patterns through the cytoskeletal machinery. Although various physical and chemical stimuli have been reported to modulate the plasmodium motility, direct regulation through exogenous molecular interference with cytoskeletal dynamics remains an underexplored frontier. One potential method for such manipulation is feeding-based RNA interference (RNAi), in which double-stranded RNA (dsRNA) expressed in prey Escherichia coli is ingested by the plasmodium [17,18]. However, establishing an effective feeding RNAi protocol for P. polycephalum has remained challenging because of its acute chemosensory system and selective predatory nature [19,20].
In this study, we initially sought to suppress the expression of ras1, a member of the small guanosine triphosphatase (small GTPase) superfamily, involved in cytoskeletal regulation, via feeding of RNAi [21]. Surprisingly, we discovered that the presence of Ras1 protein in prey bacteria, even as a result of leaky expression, triggered a distinct “conflict” behaviour in the plasmodium. In this state, the plasmodium approached the bacterial spot but failed to complete ingestion, exhibiting local wandering and stagnation. This interference was consistently observed with human KRAS, RhoA, and Rac1; notably, the oncogenic KRAS G12C mutation G12C mutation induced stochastic switching among conflict, predation, and sampling-mediated avoidance, whereas Rac1 expression showed a tendency toward diversified behavioral outcomes [22–24]. To interpret this behavioral stochasticity, we propose a working model centered on physical noise and molecular jamming or localized signaling perturbation within the motility apparatus, highlighting how P. polycephalum acts as a sensitive biological sensor for the physical and conformational states of intracellular proteins. While the amoeba-bacteria interaction is widely recognized as an “evolutionary training ground” for infection and defence, P. polycephalum—with its more complex signalling repertoire than Dictyostelium discoideum—offers a unique platform to explore how early eukaryotes navigate the fine line between nutrient acquisition and infectious risk [25–27]. This study proposes a novel intersection between protein biophysics and the behavioural determination of microbial predation.
The myxomycete Physarum polycephalum strain HU195X200 (kindly provided by Dr. Masashi Aono, Keio University) was used throughout this study. Plasmodia were revived from sclerotia stored at −80 °C on filter paper. Sclerotia were cut into approximately 1 cm² pieces and inoculated onto 1% (w/v) plain agar medium (Eiken Chemical Co., Ltd.). Cultures were maintained at 25 °C in the dark and supplemented with moisturized oatmeal. For genetic manipulations, Escherichia coli strains XL1-Blue or DH5α were employed. BL21(DE3) and HT115(DE3) strains were utilized for protein and double-stranded RNA (dsRNA) expression, respectively. E. coli cells were cultured in LB medium supplemented with 100 μg/mL carbenicillin at 37 °C.
cDNAs of ras1 and hemagglutinin (HA) were cloned into L4440 and pET-3a/3d vectors. Human KRAS, KRAS G12C, hRhoA, and hRac1 sequences were codon-optimized for E. coli expression, synthesized (Eurofins Genomics), and cloned into pET-3a and 3d vectors. Protein and dsRNA expression was induced with 0.5 μM IPTG at the mid-log phase for 3 h at 37 °C. Harvested cells were washed twice with PBS (–). For sotorasib treatment, hKRAS G12C-expressing cells were incubated in PBS (–) containing 5 μM sotorasib (MedChemExpress) for 30 min at 25 °C, followed by two washes to remove residual sotorasib. The final bacterial paste was prepared by complete removal of the supernatant.
Bacterial paste was inoculated onto the centre of a 1% (w/v) agar plate, and a 1 cm² block of P. polycephalum plasmodium was placed adjacent to it. Time-lapse imaging was performed at 10 min intervals for 24 h using a digital camera (FinePix XP140; Fujifilm) inside a humidified incubator at 25 °C. Due to the 10 min frame interval, high-frequency oscillatory dynamics (1–2 min period) were not resolved; thus, responses were evaluated based on macroscopic behavioral phenotypes observed over 24 h. Behavioural responses were categorized into three defined phenotypes: Predation (defined by the clear consumption by the end of the 24 h period), Conflict (prolonged contact typically lasting 6-12h without sustained feeding), and Sampling-mediated avoidance (initial contact followed by reversal and retreat). Trials that could not be classified within 24 h were designated as “Indeterminable.” For each experimental condition, at least 12 independent trials were performed to calculate the occurrence rate of each phenotype.
Harvested E. coli cells were denatured in SDS-PAGE sample buffer at 95 °C for 10 min. Proteins were separated on 5–20% gradient gels and transferred to PVDF membranes. Following blocking with PBS-BT, membranes were incubated with primary anti-Ras antibody (1:1,000; #3965, Cell Signalling Technology) and an alkaline phosphatase-conjugated secondary antibody (S3731; Promega). Visualization was performed using SIGMA FAST BCIP/NBT.
The occurrence rate of each phenotype was calculated as the percentage of the total number of experiments for each condition. To evaluate the statistical significance of behavioural shifts between groups (e.g., deterministic vs. stochastic behaviour), frequency distributions were compared using the Chi-square test (or Fisher's exact test, P < 0.05 being considered statistically significant). All calculations and visualizations were performed using Microsoft Excel, Adobe Photoshop and Illustrator.
We initially attempted to establish a feeding RNAi protocol for P. polycephalum using E. coli expressing dsRNA targeting the endogenous ras1 gene. However, the plasmodium exhibited an unexpected “conflict” behaviour toward the ras1 dsRNA-expressing strain, in which it is willing to eat but unable to ingest the bacteria. While the plasmodium approached the bacterial colony, it failed to complete ingestion, manifesting as localized wandering or stagnation (Fig 1B). This response stood in sharp contrast to the typical predatory expansion observed on control bacteria (Fig 1A). Western blotting revealed leaky expression of Ras1 protein in the dsRNA-expressing strain (Fig 1C). To verify whether the protein itself caused this interference, we overexpressed Ras1 protein in both BL21(DE3) and HT115(DE3) strains. Both strains consistently elicited the same conflict response (Figs 1D & E). Furthermore, the elimination of Ras1 protein expression via a nonsense mutation (M67*) completely rescued the predatory behaviour, allowing the plasmodium to successfully consume the bacteria within 24 h (Fig 1F). These results indicate that the presence of the Ras1 protein, rather than the dsRNA, induces predatory interference.
(A) Normal predatory behaviour. P. polycephalum plasmodium completely ingests control E. coli cells expressing hemagglutinin (HA) dsRNA. (B) Predatory interference (conflict behaviour) when provided with E. coli targeting the endogenous ras1 gene via dsRNA expression. (C) Detection of leaky Ras1 protein expression. CBB-stained SDS-PAGE gel (left) and western blot analysis (right) of HT115(DE3) cells carrying the ras1 (residues 7–627) L4440 plasmid. Arrowheads indicate the leaky expression of recombinant Ras1 protein. (D, E) Robustness of the conflict behaviour. The conflict response was consistently observed in both (D) BL21(DE3) and (E) HT115(DE3) strains intentionally overexpressing the Ras1 protein, confirming that the protein itself, rather than the dsRNA, triggers the interference. (F) Rescue of predatory behaviour via elimination of Ras1 protein. E. coli expressing a ras1 variant with a nonsense mutation at M67 (M67*).
https://doi.org/10.1371/journal.pone.0359275.g001
To investigate the generality of this phenomenon, we examined behavioural responses to human small GTPases [21,28,29]. Among the various human Ras isoforms, KRAS is of particular clinical importance; notably, molecular targeted drugs have been developed to specifically inhibit the KRAS G12C mutant [23,24]. When provided with E. coli expressing wild-type human KRAS (hKRAS WT), the plasmodia exhibited a deterministic conflict behaviour in 100% of trials (12/12, Table 1, Fig 2; S1 Video). In contrast, the introduction of the oncogenic hKRAS G12C mutation shifted this deterministic response to a stochastic one. The behavioural outcomes branched into conflict (12/18, 66.7%), predation (4/18, 22.2%), and sampling-mediated avoidance (2/18, 11.1%) (n = 18, Table 1, Fig 3; S2-S4 Video). Statistical analysis confirmed that the behavioural distribution of the hKRAS G12C group was significantly different from the hKRAS WT group (p = 0.025, Fisher’s exact test). Treatment with the covalent inhibitor sotorasib, which targets the Cys12 residue, yielded a behavioral distribution similar to that of the hKRAS G12C group (p = 0.731) and remained significantly different from the hKRAS WT group (p = 0.009) (n = 18, Table 1) [23].
https://doi.org/10.1371/journal.pone.0359275.t001
Time-lapse images showing the behavioural response of P. polycephalum over 24 h.
https://doi.org/10.1371/journal.pone.0359275.g002
Time-lapse images illustrating three distinct behavioural patterns induced by the hKRAS G12C mutant. (A) Conflict behaviour, (B) predation, and (C) sampling-mediated avoidance.
https://doi.org/10.1371/journal.pone.0359275.g003
We further examined the Rho-family GTPases, RhoA and Rac1, which are key regulators of the actin cytoskeleton [30,31]. Human RhoA (hRhoA) induced a deterministic conflict response in 91.7% (11/12) of cases, similar to hKRAS WT (n = 12, Table 1, Fig 4). However, human Rac1 (hRac1) resulted in a more stochastic distribution, with 75.0% (9/12) conflict, 16.7% (2/12) predation, and 8.3% (1/12) sampling-mediated avoidance behaviours (Table 1, Fig 5). Although a tendency toward diversified behavioral outcomes in the hRac1 group was observed (25.0% total of predation and avoidance), the difference compared to the hKRAS WT group did not reach statistical significance (p = 0.232, Fisher’s exact test). In all trials, protoplasmic streaming remained active, suggesting that the interference was localized to the leading edge of the plasmodium.
(A) Consistent conflict behaviour of P. polycephalum towards E. coli expressing human RhoA. Time-lapse images showing the behavioural response of P. polycephalum over 24 h. (B) Indeterminable behaviour of P. polycephalum. The plasmodium initially approached the E. coli (8 h) but subsequently bypassed the colony between 12 and 16 h. It then migrated along the periphery of the Petri dish (20 h) and eventually encircled the E. coli by 24 h.
https://doi.org/10.1371/journal.pone.0359275.g004
Time-lapse images illustrating three distinct behavioural patterns induced by hRac1. (A) Conflict behaviour, (B) predation, and (C) sampling-mediated avoidance.
https://doi.org/10.1371/journal.pone.0359275.g005
Fig 6A summarizes the macroscopic modes of behavioural decision making in the P. polycephalum plasmodium. Our results demonstrate that the presence of hKRAS WT or hRhoA in prey E. coli disrupts the normal predatory behaviour of P. polycephalum, shifting it into a deterministic “conflict” state, whereas hKRAS G12C induced stochastic behavioral outcomes, and hRac1 showed a tendency toward diversified outcomes.
(A) Macroscopic modes of behavioural decision-making in Physarum polycephalum during predatory interactions. (B) Working model for localized movement arrest at the leading edge, illustrating potential mechanisms including localized molecular jamming and phase desynchronization of rhythmic contractions.
https://doi.org/10.1371/journal.pone.0359275.g006
Our behavioural observations reveal that P. polycephalum initially approaches the E. coli colony through positive chemotaxis and initiates predation. However, when encountering bacteria expressing hKRAS WT or hRhoA, the leading edge macroscopically appeared locally arrested specifically at the contact site. While these contact points appeared locally restricted, the rest of the plasmodium—driven by continued chemotactic attraction—attempts to surround the bacterial colony. This spatial mismatch between localized arrest and global attraction leads to a state we termed “conflict”: a sustained behavioural tension in which the organism is willing to eat but unable to ingest, consistent with our proposed model of localized molecular jamming (Fig 6B). This phenomenon suggests that such localized mechanical disruption prevents the coordination required for complete engulfment.
One plausible interpretation is a “molecular jamming” model at the leading edge to explain this interference (Fig 6B). Small GTPases expressed in E. coli lack C-terminal lipid modifications (e.g., prenylation), which are essential for membrane anchoring and functional signalling in their native eukaryotic context [32]. When ingested, these non-prenylated G-domains may act as dominant-negative decoys, sequestering endogenous guanosine nucleotide exchange factors GEFs or downstream effectors. This could induce a critical mismatch: while the global contractile engine continues pumping protoplasm forward, the leading edge becomes physically immobilized, forcing internal pressure to redistribute laterally into unproductive localized oscillations [33,34].
Alternatively, these observations can be interpreted through the framework of classical Physarum oscillatory dynamics. Pioneering work by Wohlfarth-Bottermann, Matsumoto and co-workers, and Ueda and co-workers established that local perturbations of contraction rhythm, phase wave propagation, and shuttle streaming velocity can directly alter directional migration and taxis without requiring complete mechanical arrest [35–37]. In the present study, parameters such as local contraction period, phase synchronization, and shuttle streaming velocity were not directly measured. Therefore, an equally compelling explanation is that internalized GTPases locally perturb intracellular signaling, resulting in phase desynchronization of rhythmic contractions at the contact zone. Crucially, molecular jamming and oscillatory phase desynchronization are not mutually exclusive; initial molecular jamming at the leading edge could perturb local signal integration, which subsequently disrupts oscillatory phase synchronization and manifests as the macroscopically observed conflict behavior.
A key finding of this study is the shift from deterministic to stochastic behaviour induced by the hKRAS G12C mutation. This stochasticity of hKRAS G12C likely reflects the physical heterogeneity of the interfering proteins within the bacterial cells. The Cys12 residue of the hKRAS G12C mutant is highly reactive and prone to forming intermolecular disulfide-bonded aggregates during protein overexpression in the bacterial cytoplasm, which we hypothesize introduces “physical noise” into the leading-edge machinery. Consistent with this view, sotorasib treatment—which targets the specific -SH group of the G12C mutation—which specifically blocks Cys12—did not alter the stochastic distribution [23,24]. Rather than executing a distinct biochemical pathway, we propose as a working model that the non-uniform distribution of G12C aggregates may destabilize the local leading-edge dynamics, allowing intrinsic oscillatory fluctuations to tip decision-making toward either predation or retreat. It should be noted, however, that protein aggregation and soluble/insoluble fractions were not directly measured in this study, and this mechanism remains a proposed explanation.
The contrast between hRhoA and hRac1 provides additional insight into the spatial regulation of plasmodial motility. RhoA typically governs actomyosin contractility and the maturation of focal adhesions, and its interference likely induces a “molecular anchoring” effect that rigidly fixes the leading edge. In contrast, Rac1 is the primary orchestrator of membrane protrusion, coordinating dynamic actin polymerization at the lamellipodia [38,39]. While the behavioural shift for hRac1 did not reach statistical significance in this study, the observed stochasticity may stem from its distinct regulatory network. Exogenous, non-prenylated Rac1 is hypothesized to act as a decoy for Rac1-specific GEFs or GAPs that are spatio-temporally distinct from those targeting KRAS or RhoA [29,31,39,40]. Given Rac1's central role in driving the leading-edge protrusion, such interference might destabilize the steering of the plasmodial front rather than imposing a complete arrest. This “unstable steering” could lead to the observed behavioural bifurcations, where local fluctuations in the leading-edge dynamics may help determine whether the organism proceeds with predation or retreats.
The ability of P. polycephalum to discriminate between different protein states carries significant evolutionary implications. The long-standing predator-prey relationship between amoebae and bacteria serves as an “evolutionary training ground” for both bacterial infection strategies and eukaryotic defence mechanisms [25,26]. Although these pathogens primarily utilize these effectors to invade host cells, early amoeboid predators must have faced strong evolutionary pressure to detect and counter such molecular mimicry during ingestion. Pathogenic bacteria, such as Salmonella, utilize specialised effectors that mimic or modulate host signalling molecules to evade digestion [27,41,42]. For instance, Salmonella delivers effectors like SopE (a GEF mimic) and SptP (a GAP mimic) to directly manipulate host Rho-family GTPases, thereby controlling actin dynamics and membrane trafficking to facilitate its survival within the host [27,41,42].
Our results raise the possibility that P. polycephalum has evolved to sense such molecular perturbations with high sensitivity. In this context, P. polycephalum may offer a superior experimental platform compared to the widely used model organism Dictyostelium discoideum. While D. discoideum possesses a streamlined and well-characterised signalling system, its repertoire of sensory receptors and signal integration pathways is relatively limited. In contrast, the P. polycephalum genome reveals a vastly expanded array of G-protein-coupled receptors (GPCRs) and ancestral two-component systems, reflecting a more complex evolutionary history of environmental sensing [43]. This increased molecular complexity likely functions as a high-resolution biological sensor, enabling the plasmodium to integrate subtle conformational and physical perturbations. To directly test the proposed hypotheses, future studies incorporating direct optical measurements of local contraction rhythms, soluble/insoluble protein fractionations, and quantitative motility tracking will be valuable.
Further investigating the specific behavioural outputs triggered by these bacterial-mimetic proteins, will provide new insights into the co-evolutionary arms race and the strategic competition between bacteria and amoeboid predators. Such investigations will clarify how early eukaryotes navigate the fine line between nutrient acquisition and infectious risk, positioning P. polycephalum as a powerful organismal platform for decoding the molecular logic of microbial warfare at the interface of protein biophysics and ethology.
Time-lapse movie showing the behavioural response of P. polycephalum over 24 h (frames captured at 10 min intervals).
https://doi.org/10.1371/journal.pone.0359275.s001
Time-lapse movie illustrating the first of three distinct behavioural patterns induced by the hKRAS G12C mutant over 24 h (1 frame = 10 min), representing the conflict behaviour.
https://doi.org/10.1371/journal.pone.0359275.s002
Time-lapse movie illustrating the second of three distinct behavioural patterns induced by the hKRAS G12C mutant over 24 h (1 frame = 10 min), representing the predation behaviour.
https://doi.org/10.1371/journal.pone.0359275.s003
Time-lapse movie illustrating the third of three distinct behavioural patterns induced by the hKRAS G12C mutant over 24 h (1 frame = 10 min), representing the sampling-mediated avoidance.
https://doi.org/10.1371/journal.pone.0359275.s004
Complete list of independent experimental trials recorded for each bacterial variant, showing the experiment date, sample condition, and classified behavioral phenotype over 24 h of observation.
https://doi.org/10.1371/journal.pone.0359275.s005
We would like to thank Prof. Masashi Aono for helpful discussions and Mr. Yuta Arai for technical assistance. We are grateful to Prof. Asako Sugimoto for generously providing the L4440 plasmid and E. coli HT115(DE3). We also thank Editage for editing the manuscript.