Supramolecular nanofiber stabilization of microdroplets enables ultra-sensitive virus infection analysis

Abstract

Aqueous two-phase system (ATPS) droplets in cells act as fluidic microreactors by concentrating biomacromolecules. Inspired by this phenomenon, dextran-rich microdroplets formed in an ATPS with polyethylene glycol have been explored as artificial microreactors for sensitive detection and spatiotemporal control of biochemical reactions. However, the rapid fusion of the microdroplets into bulk phase separation has limited practical applications of this approach. Here, we report the stabilization of dextran-rich microdroplets using supramolecular nanofibers of an azobenzene-appending self-assembling peptide (AzSAP). Physicochemical characterization and structural analyses elucidate the mechanism of nanofiber formation and its role in droplet stabilization. The nanofiber network prevents droplet coalescence while maintaining macroscopic fluidity, thereby enabling highly sensitive quantitative virus detection via microfluidic analysis by confining infecting viruses within droplets. Furthermore, the photo-responsive properties of the AzSAP allow dynamic control over droplet size and intra-droplet virus activity, highlighting its exceptional potential as a platform for programmable artificial microreactors.

Artificial aqueous two-phase system (ATPS) droplets offer promise as controllable microreactors but are limited by the low stability of dextran-rich microdroplets in 3D media. Here, reversible nanofibers formed from an azobenzene-appended self-assembling peptide stabilize droplets by suppressing fusion without compromising fluidity. These stabilized systems enable concentration of M13 bacteriophages and Escherichia coli, prevent secondary infections, and support quantitative flow cytometry readouts, while allowing precise, photoresponsive modulation of droplet fusion.

Introduction

Cells construct aqueous two-phase system (ATPS) droplets as fluidic microreactors. Due to their interfacial permeability, these droplets can import solutes, including reactants, and condense biomacromolecules such as nucleic acids and proteins, thereby enabling various biological reactions to proceed efficiently and maintain homeostasis in response to external stressors^1–8. Inspired by such biological droplets, artificial ATPS microdroplets^9–26—typically formed by mixing two immiscible water-soluble polymers such as dextran and polyethylene glycol (PEG)—have attracted considerable attention as controllable microreactors in segregated spaces. Leveraging the capacity to accumulate biomolecules and maintain macroscopic fluidity, dextran-rich ATPS microdroplets (Dex-droplets) have been integrated into two-dimensional microchamber devices²⁷, in which fluidity allows the direct filling of microchambers. The property of accumulation is also useful for green chemistry^28,29, constructing artificial cells^30,31, and culturing cells⁹.

Extending this concept to a three-dimensional (3D) Dex-droplet microreactor system would markedly enhance reaction productivity, and the macroscopic fluidity of these systems is compatible with flow-based detectors (such as fluorescence-activated cell sorters) for highly sensitive and quantitative analyses. However, the inherently low stability and short lifetime of conventional Dex-droplets, owing to rapid fusion leading to bulk phase separation, critically limit their applicability in 3D systems. Inspired by intracellular systems, in which F-actin stabilizes ribonucleoprotein droplets³², synthetic supramolecular polymers have been utilized to construct and stabilize ATPS droplets. Short peptides such as carboxybenzyl-protected diphenylalanine form ATPS droplets in the presence of metal cations, but these droplets deform into fibrillar morphologies after several hours of incubation³³. A synthetic amphiphile UPy-Gly appending ureidopyrimidinone and glycine units self-assembles in water to produce ATPS droplets with a characteristic tactoid morphology and high rigidity after extended incubation over several days¹³. The supramolecular polymers of UPy-Gly are densely generated within Dex-droplets in the dextran and PEG system for stabilization and deformation. The use of Pickering emulsions with liposomes¹⁰ and platelets¹¹, and coating of droplets with triblock copolymers are also reported to be effective for droplet stabilization¹². However, these conventional approaches compromise key microreactor properties of the droplets, such as interfacial permeability and fluidity. In addition, stimuli-responsive control of droplet fusion has remained a challenging task.

Here, we report effective stabilization of Dex-droplets in a 3D medium by suppressing fusion using nanofibers formed by an azobenzene-appending self-assembling peptide (AzSAP, Fig. 1a–c), without loss of macroscopic fluidity. Physicochemical characterization revealed that AzSAP nanofibers retain their structural integrity within the ATPS, thereby remaining permeable to small molecules, and they do not interfere with droplet mobility. Cryo-electron microscopy (cryo-EM) structural analyses further showed that the nanofibers consist of two symmetrical protofilaments connected via cross-β hydrogen bonds and stabilized by azobenzene π-stacking. The nanofibers provide a robust supramolecular scaffold that suppresses droplet coalescence. Functionally, stabilized Dex-droplets efficiently concentrated M13 bacteriophages (M13 phage)^34–41 and Escherichia coli (E. coli) cells, while preventing secondary infections outside the droplets (Fig. 1d, e). The preserved macroscopic fluidity enabled ultra-sensitive, quantitative detection of infection events within droplets using flow cytometry (Fig. 1e). In the previously reported Dex-droplets, no such application for analyzing virus infection has been achieved. Furthermore, the reversible photo-responsiveness of AzSAP nanofibers, arising from the trans–cis isomerization of azobenzene units, allowed precise photo-control over droplet fusion, enabling size tuning and spatiotemporal modulation of viral infection.

Fig. 1. Concept of micro-bioreactors based on aqueous two-phase system (ATPS) droplets stabilized by supramolecular nanofibers for microfluidic analysis of virus infection.

a Molecular structures of AzSAP and rev-AzSAP. b Schematic illustration of dextran-rich droplets (Dex-droplets) in an aqueous two-phase system (ATPS) with a polyethylene glycol (PEG)-rich phase, and spontaneous fusion of Dex-droplets leading to macroscopic phase separation. c Schematic illustration of Dex-droplets formed in the presence of supramolecular AzSAP nanofibers. The droplets are arrested by the nanofibers to suppress fusion. d, e Difference in infection of E. coli cells by M13 bacteriophages in Dex-droplets in the absence (d) and presence (e) of AzSAP nanofibers. In the absence of AzSAP nanofibers, M13 bacteriophages secondarily infect E. coli cells to initiate unrestrained amplification. This secondary infection is suppressed by encapsulating M13 bacteriophages in fusion-suppressed Dex-droplets, which allows for microfluidic analysis of the infected E. coli cells.

Design of peptides that form supramolecular nanofibers

To suppress the fusion of microdroplets while maintaining macroscopic fluidity, we focused on mixing Dex-droplets with supramolecular nanofibers that form 3D network structures in aqueous media and exhibit the dynamic property of collapsing and re-forming due to non-covalent interactions. We chose peptides as the supramolecular nanofiber-forming component because of their high structural designability, based on established synthesis principles that can generate a variety of molecular and self-assembling structures. We recently developed self-assembling peptides with an iterative sequence between hydrophilic and hydrophobic side groups containing an azobenzene group^42,43. These peptides form supramolecular nanofibers via the construction of β-sheet motifs through hydrophobic interactions and hydrogen bonds. By coupling with optical controls, the potential applications of the photoresponsive functionality of the azobenzene group can be expanded. Based on our previous study, herein we designed azobenzene-containing amphiphilic peptides, AzSAP (Ac-RAzRADMR-NH₂, where Az indicates an amino acid residue with an azobenzene group appended at the side chain) and rev-AzSAP (Ac-RMDARAzR-NH₂) consisting of 7 amino acid residues with opposite sequences (Fig. 1a, Supplementary Fig. 1).

AzSAP and rev-AzSAP dissolved in water at 1.0 wt% to provide transparent solutions, suggesting good dispersibility. After incubation at 25 °C for 6 h, AzSAP remained in solution, whereas the fluidity of rev-AzSAP in water decreased (Supplementary Fig. 2). Rheological analysis of rev-AzSAP in water (1.0 wt%) after incubation indicated G’ and G” values of 12.4 Pa and 2.1 Pa, respectively (Supplementary Fig. 2). Even though rev-AzSAP exhibited rather weak viscoelasticity, the G’ > G” relationship suggested the formation of a hydrogel. Interestingly, transmission electron microscopy (TEM) revealed that AzSAP and rev-AzSAP formed well-defined nanofibers with high rectilinearity after incubation (Fig. 2a, c). Circular dichroism (CD) spectroscopic measurements of AzSAP (1.0 wt%) in water showed negative Cotton signals at 219 and 332 nm, indicative of β-sheet formation and stacking at the azobenzene unit in a chiral fashion, respectively (Fig. 2b, Supplementary Fig. 3). Although rev-AzSAP also showed a negative Cotton effect at 219 nm immediately after dissolution, a continuous increase in CD signals was observed during incubation, with a stronger negative signal at 219 nm than AzSAP and a positive signal at 326 nm (Fig. 2d, Supplementary Fig. 3). The opposite signals of AzSAP and rev-AzSAP at around 330 nm suggest different packing structures of the azobenzene moiety within the fibers⁴⁴. The stronger 219-nm negative CD signal was suggestive of a larger population of β-sheet structures of rev-AzSAP in water compared with AzSAP, which was likely the reason for the different macroscopic states between the two peptides. We observed that AzSAP fibers maintained a fibrous structure against external environmental changes such as saline conditions and upon heating, while rev-AzSAP fibers were less stable and disassembled upon heating (Supplementary Fig. 4). In addition, AzSAP fibers showed high stability even against pH changes and in serum and saliva (Supplementary Fig. 4).

Fig. 2. Structures of supramolecular nanofibers.

a, c Transmission electron micrographs of AzSAP (1.0 wt%, a) and rev-AzSAP (1.0 wt%, c) in water. The inset shows a higher magnification image. Scale bars:100 and 20 nm (inset). Stain: uranyl acetate. b, d Circular dichroism (CD) spectra of AzSAP (1.0 wt%, b) and rev-AzSAP (1.0 wt%, d) after incubation for 1 min (green), 2 h (blue), and 5 h (red). e–p Structural representations of supramolecular nanofibers of AzSAP (e-j) and rev-AzSAP (k-p) refined by cryo-electron microscopy (cryo-EM) analysis. Lateral views of one helical pitch of the supramolecular nanofibers (e, k), zoomed-in views highlighting the protofilaments (f, l), cross-sectional views (g, m), and zoomed-in views highlighting the packing structure at the core (h, n) shown in the red squares in (g) and (m). i, j, o, p High-resolution cryo-EM density maps showing amino acid residues surrounding the azobenzene groups in cluster 1 (i, j) and cluster 2 (o, p), resolved at 2.32 Å for AzSAP fibers and 1.91 Å for rev-AzSAP fibers. Source data are provided as a Source Data file.

To investigate the structural mechanism by which different macroscopic states of AzSAP and rev-AzSAP are formed in water, we further conducted structural analyses by cryogenic electron microscopic (cryo-EM) observation. Cryo-EM imaging of AzSAP nanofibers visualized the packing of two symmetric protofilaments with a crossover distance of 72.5 nm (Fig. 2e, f, Supplementary Figs. 5–7). Structural optimization at a resolution of 0.19 nm revealed a C₂-symmetric butterfly-shaped cross-sectional structure composed of 14 molecules of AzSAP, and the horizontal and vertical sizes of the cross-section were 9.0 nm and 7.5 nm, respectively (Fig. 2g–i, Supplementary Figs. 8, 9). Interestingly, the butterfly-type cross-section is constructed by dimerization of two clusters of AzSAP formed by the assembly of three (a/b/c-cluster; AzSAP-a/AzSAP-b/AzSAP-c) and four (d/e/f/g-cluster; AzSAP-d/AzSAP-e/AzSAP-f/AzSAP-g) molecules of AzSAP. In the AzSAP nanofiber, two filaments were assembled by electrostatic interactions between two AzSAP-g molecules and hydrophobic interactions between AzSAP-a and AzSAP-g. The two clusters between a/b/c-cluster and d/e/f/g-cluster were assembled by interactions between AzSAP-c and AzSAP-d via guanidine groups at the terminal arginine residues, where water molecules were arranged to fill the space between the clusters (Supplementary Figs. 10, 11). A similar butterfly-shaped cross-sectional structure was observed with rev-AzSAP nanofibers. In both the AzSAP and rev-AzSAP fibers, each cluster contained aggregations of the azobenzene groups and hydrophobic side chains of Met and Ala in the interior region, and these azobenzene groups formed π-stacks both within the clusters and between the layers along the filaments (Fig. 2g, h, i, l, m, n). The crossover distance of the rev-AzSAP fibers was 94.3 nm (Fig. 2j, k, Supplementary Figs. 5, 9–11). The horizontal and vertical dimensions of the cross-section of rev-AzSAP fibers were 8.5 nm and 6.5 nm, respectively (Fig. 2l). A characteristic difference in the packing structures was observed at the core of the cross-section. Namely, observation of the cross-section of the AzSAP assembly revealed a cavity at the central core. By contrast, the core of the rev-AzSAP assembly is filled by residues R1, D3, and R5 in rev-AzSAP-g fibers, in which rev-AzSAP-g and rev-AzSAP-g’ interact via contact at these residues. This filled-core architecture results in a denser, mechanically more rigid packing that likely enhances nanofiber stability at the cost of dynamic flexibility, correlating with the formation of a self-standing hydrogel and reduced macroscopic fluidity. Due to the difference in the packing structure at the core, rev-AzSAP fibers have a smaller cross-sectional area than AzSAP fibers (i.e., rev-AzSAP fibers are more densely packed than AzSAP fibers). In addition, molecular dynamics (MD) simulation showed that the interactions between AzSAP clusters mediated by hydrophilic peptide residues are not robust (Supplementary Fig. 12). Cryo-EM analyses thus directly reveal how subtle rearrangements in peptide packing determine the balance between stability and fluidity in the resulting supramolecular networks, an essential structure–function relationship that explains why the densely packed rev-AzSAP nanofibers exhibit greater stability and more readily form hydrogels, whereas the more loosely packed AzSAP fibers maintain macroscopic fluidity.

Stabilization of ATPS microdroplets by supramolecular peptide nanofibers

Although vortexing of a mixture of dextran and PEG in water (5.0 wt% dextran; M_w = 450,000–650,000, 5.0 wt% PEG; M_w = 35,000) afforded a cloudy suspension due to the formation of Dex-droplets, the solution returned to a transparent state due to macroscopic phase separation within a 15 min incubation (Fig. 3a). Phase-contrast microscopic observation of the mixture in a 200 μm-thick glass chamber consistently showed rapid fusion of the dextran droplets during the initial 10 min, resulting in the formation of enlarged droplets with a diameter of >100 μm (Fig. 3f). Critically, such rapid fusion limits application of the conventional dextran droplets formed in aqueous PEG solutions to microreactors. Interestingly, the aqueous mixture of dextran and PEG remained turbid even after 24 h of incubation after vortexing in the presence of AzSAP (Dex-droplets/AzSAP, Fig. 3b) and rev-AzSAP (Dex-droplets/rev-AzSAP, Fig. 3c) fibers. The size of the dextran droplets ( < 30 µm) changed only minimally during the incubation, clearly indicating suppression of droplet fusion (Fig. 3g, h). Also, the turbidity of the dispersions due to the formation of Dex-droplets changed very little in the presence of AzSAP and rev-AzSAP, in contrast to the Dex-droplets without peptides, which showed a rapid drop in turbidity due to fusion (Supplementary Fig. 13). Confocal laser scanning microscopy (CLSM) observation of Cy5-labeled peptide nanofibers and high-performance liquid chromatography (HPLC) analysis indicated that AzSAP and rev-AzSAP fibers were distributed in both the dextran-rich and PEG-rich phases (Supplementary Fig. 14). TEM observations showed formation of AzSAP and rev-AzSAP nanofibers in an aqueous medium containing dextran and PEG, with similar morphologies observed in water (Supplementary Fig. 15). In addition, fluorescence recovery after photobleaching (FRAP) analysis showed that molecular diffusivity in the system was maintained even after the stabilization of Dex-droplets by AzSAP fibers (Supplementary Fig. 16). To investigate the structural influences of the peptide self-assemblies on droplet stabilization efficiency, we synthesized several AzSAP derivatives, AzSAP-K (Ac-KAzRADMK-NH₂) and AzSAP-D (Ac-DAzRADMD-NH₂) possessing lysine and aspartic acid residues at the respective termini, respectively, JigSAP (Ac-RIDARMRADAIR-NH₂) having a jigsaw-shaped hydrophobic region without an azobenzene unit⁴², and RADA16 (Ac-RADARADARADARADA-NH₂) having a planar hydrophobic region composed solely of alanine residues⁴⁵. Importantly, AzSAP-K, AzSAP-D, and RADA16 formed particulate, plate-like, and flexible fibrous assemblies in water, respectively, as observed by TEM, CD measurement, and Fourier transform infrared spectroscopy (FTIR) analyses (Supplementary Figs. 17, 18), and their stabilization effect on Dex-droplets was limited (Fig. 3d, e, i, j, Supplementary Fig. 19). Although JigSAP self-assembled into continuous nanofibers stabilizing Dex-droplets, the dispersion formed a hydrogel with low fluidity (Supplementary Fig. 20). These findings, in light of the results of cryo-EM structural analyses, suggest that the nanofibers with a long persistence length, consisting of AzSAP, rev-AzSAP, and JigSAP fibers that are amenable to the construction of homogeneous networks, are critical for arresting droplet fusion. In contrast, non-fibrous self-assembling morphologies observed in AzSAP-K (particle-like assembly) and AzSAP-D (bundling) lack sufficient capacity to suppress droplet fusion; therefore, the terminal residues play important roles to develop the nanofibers with high aspect ratio.

Fig. 3. Stabilization of Dex-droplets by supramolecular peptide nanofibers.

a–e Photographs of aqueous dispersions of Dex-droplets (5.0 wt% dextran, 5.0 wt% PEG) without additive (a), with AzSAP (5.0 wt% dextran, 5.0 wt% PEG, 1.0 wt% AzSAP) (b), with rev-AzSAP (5.0 wt% dextran, 5.0 wt% PEG, 1.0 wt% rev-AzSAP) (c), with AzSAP-K (5.0 wt% dextran, 5.0 wt% PEG, 1.0 wt% AzSAP-K) (d), and with AzSAP-D (5.0 wt% dextran, 5.0 wt% PEG, 1.0 wt% AzSAP-D) (e). The photographs were taken at 0-15 min (a, d, e) or 0–24 h (b, c) after vortexing. Phase separation interfaces are highlighted by green dashed lines in (a, d, and e). f, i, j Time-course phase-contrast microscopic observation of Dex-droplets during 0–10 min (f), Dex-droplets/AzSAP-K during 0–10 min (i), and Dex-droplets/AzSAP-D during 0-10 min (j) after vortexing. Scale bars: 100 μm. g, h, Differential interference contrast microscopic observation of Dex-droplets/AzSAP during 0–24 h (g), Dex-droplets/rev-AzSAP during 0–24 h (h) after vortexing. Scale bars: 20 μm. k Viscosity of Dex-droplets without additive (blue), Dex-droplets/AzSAP (brown), and Dex-droplets/rev-AzSAP (orange). Sample size for each group was n = 3, and error bars represent mean ± SD. Statistical significance was assessed by ANOVA (p < 0.01), one-sided Student’s t-test (*p < 0.01), and Welch’s t-test (df <2, p < 0.01, Cohen’s d < 15.5, 95% CI: 16.38 − 25.29). l Frequency-dependent G’ (red) and G” (blue) profiles of Dex-droplets/AzSAP and Dex-droplets/rev-AzSAP under a strain of 0.5% at 20 °C. Source data are provided as a Source Data file.

Fluidic systems are advantageous in terms of compatibility with microfluidic devices. We therefore investigated the fluidity of Dex-droplet systems containing the fiber-forming peptides. A vortexed mixture of dextran and PEG in water showed a viscosity of 10.5 ± 0.2 mPa•s (1 mPa•s = 1 cP) at 20 °C (5.0 wt% dextran, 5.0 wt% PEG, 1.0 wt%). Interestingly, the viscosity of an aqueous dispersion of dextran and PEG containing AzSAP (1.0 wt%) was 11.1 ± 1.8 mPa•s, retaining high fluidity (Fig. 3k). However, the viscosity of Dex-droplet systems containing rev-AzSAP (1.0 wt%) was approximately 3-fold higher (31.9 ± 0.1 vs 11.1 ± 1.8 mPa•s, p < 0.01, Fig. 3k). This result suggests that the viscosity of the Dex-droplet system containing the fiber-forming peptides is influenced by the structure of the hydrophobic moiety of the peptides, and rev-AzSAP fibers may possess a longer persistence length than AzSAP fibers, leading to the assembly of a more robust network, thereby increasing macroscopic viscosity⁴⁶. Frequency-dependent rheological analyses (shear strain = 0.5%) of Dex-droplet systems containing AzSAP and rev-AzSAP showed a common profile with a slope of 0.4 at a frequency of 0.10–1.0%, indicating that both systems were essentially in a fluidic solution state (Fig. 3l)⁴⁷.

Spatiotemporal control of droplet fusion using light irradiation

Thanks to the photo-responsive functionality of the azobenzene unit of AzSAP, size tuning of the Dex-droplets^48,49 and optical control of infection events can be demonstrated (Fig. 4a). Upon UV light irradiation (λ = 350 nm) of an aqueous solution of AzSAP fibers (1.0 wt%) for 3 min, absorption spectroscopy showed a decrease in the absorption band intensity at 330 nm and an increase in the band intensity at 430 nm due to trans-to-cis isomerization of the azobenzene group (Fig. 4b)⁵⁰. In CD spectroscopy, signals assigned to the β-sheet structure and azobenzene stacking decreased upon UV irradiation (Fig. 4e). In conjunction with these spectroscopic changes, TEM observation revealed a transition of the self-assembly structure of AzSAP from nanofibers (Fig. 4c) to nanoparticles (Fig. 4d) upon UV irradiation. This structural transition of AzSAP fibers was reversible, as AzSAP nanofibers re-formed following irradiation with 450-nm light to prompt the backward cis-to-trans isomerization after exposure to UV light (Supplementary Fig. 21). We also confirmed that the localization of AzSAP fibers in the system composed of a mixture of dextran and PEG changed only minimally after photo-isomerization (Supplementary Fig. 22). Interestingly, the depolymerization of AzSAP nanofibers destabilized the Dex-droplets, thereby facilitating fusion. Exposure of the Dex-droplet/AzSAP system (5.0 wt% dextran, 5.0 wt% PEG, 1.0 wt% AzSAP) to 350 nm light for 15 min resulted in a rapid change in the turbid mixture through macroscopic phase separation (Fig. 4f, g), indicating that droplet fusion proceeded. Indeed, phase-contrast microscopy observation revealed an increase in the size of Dex-droplets in the UV-irradiated area during droplet fusion (Fig. 4h–j). It is reasonable that droplet fusion commenced immediately after UV irradiation (Fig. 4i), since the concentration threshold of AzSAP for the stabilization of Dex-droplets ( > 0.9 wt%, Supplementary Fig. 23) was close to the experimental conditions. In addition, the destabilized Dex-droplets were again stabilized by reconstruction of AzSAP fibers upon irradiation with 450-nm visible light (Supplementary Fig. 24), and this photo-reversible process was repeatable for at least 20 cycles (Supplementary Fig. 25). When we varied the duration of irradiation with UV light before the visible light-triggered suppression of droplet fusion (Fig. 4k), phase-contrast microscopy observation clearly showed that droplet size increased with increasing duration of UV-light irradiation (Fig. 4l–o). The diameter of Dex-droplets was 22.2 ± 3.8 µm (before light irradiation), 26.9 ± 3.9 µm (after 2 min UV irradiation), 36.0 ± 11.5 µm (after 4 min UV irradiation), and 58.0 ± 25.1 µm (after 5 min UV irradiation) (Fig. 4p), demonstrating the potential for irradiation time-dependent control of Dex-droplet size by employing the photo-responsiveness of AzSAP fibers. Furthermore, phage infection was suppressed when the Dex-droplet/AzSAP system containing M13 phages was mixed with E. coli cells. Subsequent light irradiation to trigger fusion of Dex-droplets/AzSAP rapidly facilitated M13 phage infection (Supplementary Fig. 26), while Dex-droplets/rev-AzSAP did not suppress infection even before the light irradiation (Supplementary Fig. 27). These results demonstrate that the optical manipulation of droplets to trigger/suppress fusion is applicable not only to non-contact size control of the droplets but also to temporal control of bioreactions.

Fig. 4. Optical control of Dex-droplets.

a Schematic illustration of reversible nanofiber-to-particle transition of AzSAP by photoisomerization of the azobenzene group upon UV (350 nm) and visible light (vis; 450 nm) irradiation. b Change in UV-vis absorption spectra of an aqueous solution of AzSAP (1.0 wt%) during irradiation with 350-nm light for 190 s at 25 °C. c, d Transmission electron micrographs of AzSAP (1.0 wt%) before (c) and after irradiation with 350 nm light for 15 min (d). Scale bars: 200 nm. Stain: uranyl acetate. e CD spectra of AzSAP (1.0 wt%) before (blue) and after irradiation with 350 nm light for 60 (light blue), 180 (orange), and 300 s (red). f, g Photographs of a mixture of dextran, PEG, and AzSAP (5.0 wt% dextran, 5.0 wt% PEG, 1.0 wt% AzSAP) before (f) and after (g) irradiation with 350-nm light for 15 min. The phase separation interface is indicated by a green dashed line. h–j Phase-contrast micrographs of a mixture of dextran, PEG, and AzSAP (5.0 wt% dextran, 5.0 wt% PEG, 1.0 wt% AzSAP) before (h) and after irradiation with 365-nm light for 2 s followed by incubation for 20 s (i) and 15 min (j). Scale bars: 100 μm. k Schematic illustration of the process to control the size of Dex-droplets by UV light irradiation and prompt Dex-droplet fusion through nanofiber-to-particle transition of AzSAP and subsequent visible-light irradiation to arrest the fused Dex-droplets through particle-to-nanofiber transition. UV light irradiation times of 2, 4, and 5 min correspond to the experimental results shown in (l–o). l–o Phase-contrast micrographs of Dex-droplets (5.0 wt% dextran, 5.0 wt% PEG) containing AzSAP (1.0 wt%) before light irradiation (l) and after irradiation with UV light (350 nm) for 2 (m), 4 (n), and 5 min (o), followed by irradiation with visible light (450 nm) for 1 min for each. Scale bars: 100 μm. p Bee swarm diagram of the diameter of Dex-droplets (sample size = 100) as calculated from the micrographs in (l–o). Source data are provided as a Source Data file.

Bioreactor application of Dex-droplets for microfluidic analysis of virus infection

Taking advantage of its higher fluidity and sufficient stability, we utilized a Dex-droplet system containing AzSAP (Dex-droplets/AzSAP) as a bioreactor for the analysis of M13 phage infection events using a microfluidic device. M13 phages infect E. coli cells in order to replicate, and this process is employed in a wide range of applications, including peptide affinity selection and large-scale DNA production^34–41. To determine the localization of M13 phages and E. coli cells in Dex-droplets, we first mixed SYTO 9-stained E. coli cells and fluorescein isothiocyanate (FITC)-labeled M13 phages (FITC-M13) with Dex-droplets containing Cy5-labeled dextran (Cy5-dextran) in lysogeny broth (LB) medium (5.0 wt% dextran, 4.0 × 10^–4 wt% Cy5-dextran, 5.0 wt% PEG). CLSM observation revealed green fluorescence originating from SYTO 9 and FITC in Dex-droplets, indicating localization of both E. coli cells and M13 phages within the droplets (Supplementary Fig. 28). Importantly, the localization preference of the phages and E. coli cells in Dex-droplets was also examined using CLSM observation of the Dex-droplet/AzSAP system (1.0 wt% AzSAP, 1.5 wt% lactoferrin for suppressing adsorption of M13 phages and E. coli cells to AzSAP, Fig. 5a–f). We also confirmed that 89 ± 12% of E. coli cells in the Dex-droplet/AzSAP system remained viable as evaluated by live/dead assays (Supplementary Fig. 29).

Fig. 5. Flow cytometry analysis of M13 phages using Dex-droplet systems.

a–f Confocal laser scanning microscopy (CLSM) micrographs of Dex-droplets/AzSAP (5.0 wt% dextran, 4 × 10^–4 wt% Cy5-dextran, 5.0 wt% PEG, 1.0 wt% AzSAP, magenta) with lactoferrin (1.5 wt%) in lysogeny broth (LB) medium with SYTO 9-stained E. coli cells (200 μM, green) (a-c) and fluorescein isothiocyanate (FITC)-labeled M13 bacteriophages (1.6 × 10¹⁰ virions ml^–1, green) (d–f) with the merged images (c, f). Scale bars: 10 μm. g, h Schematic illustration (g) and fluorescence micrograph (h, λ_ex = 650 nm) of Dex-droplets/AzSAP flowing through the flow cytometry device. Scale bar: 100 μm. i, j Plots of forward scatter (FSC) and side scatter (SSC) in flow cytometry analyses of Dex-droplets/AzSAP (i) and corresponding fluorescence intensity profile (j) in the region of the magenta dashed rectangle in (i). k Schematic illustrations of amplification of M13 bacteriophages in Dex-droplets and Dex-droplets/AzSAP. l Change in number of M13 bacteriophages from an initial concentration of 4 × 10⁵ (triangles) or 4 × 10⁶ (circles) virions ml^–1 with incubation for 24 h with IPTG (1 mM) in the presence of Dex-droplets without (blue triangles and circles) and with AzSAP (red triangles and circles). m Schematic illustration of quantification of M13 bacteriophage titer in Dex-droplet/AzSAP system containing 9H-(1,3-dichloro-9,9-dimethylacridin-2-one-7-yl) β-D-galactopyranoside (DDAO galactoside) dye. n–q CLSM images of Dex-droplets/AzSAP (5.0 wt% dextran, 2 × 10^–2 wt% FITC-labeled dextran; green, 5.0 wt% PEG, 1.0 wt% AzSAP) with IPTG (1 mM), DDAO galactosidase (magenta), E. coli cells and M13 bacteriophages at a concentration of 2 × 10² (n, o) or 2 × 10³ virions ml^–1 (p, q). White and magenta dashed circles indicate small fluorescent and fluorescent Dex-droplets, respectively. Scale bars: 10 μm. r Flow cytometry profiles of Dex-droplets/AzSAP with DDAO galactosidase (10 μM) encapsulating E. coli cells and M13 bacteriophages at initial concentrations of 0, 2 × 10², 6 × 10², and 2 × 10³ virions ml^–1 after a 24-h incubation. Higher dilutions of M13 phage samples were used to prevent overlap of M13 phages within Dex-droplets. Source data are provided as a Source Data file.

To evaluate applications to microfluidic analyses, Dex-droplets/AzSAP containing Cy5-labeled dextran (5.0 wt% dextran, 4 × 10^–4 wt% Cy5-dextran, 5.0 wt% PEG, 1.0 wt% AzSAP, 1.5 wt% lactoferrin) in LB medium was injected into a ⌀400-μm microfluidic channel⁵¹. Due to macroscopic fluidity, the Dex-droplets/AzSAP flowed into the channel with negligible leakage to the external solution, as visualized using fluorescence microscopy (λ_ex = 650 nm, Fig. 5g, h), suggesting the Dex-droplet/AzSAP system is applicable to microfluidic analyses. On the other hand, a water-in-oil droplet⁵² caused disordered flow in the flow cytometry device (Supplementary Fig. 30). Analysis of Dex-droplets/AzSAP containing Cy5-dextran using flow cytometry (core size = 40 μm) revealed spots corresponding to the fluorescence of Cy5-labeled dextran in the forward scatter (FSC) profile of 48 K–200 K and the side scatter (SSC) profile of 10 K–65 K (Fig. 5i, j). Notably, flow cytometry analysis of aqueous samples containing only dextran, PEG, or AzSAP showed few spots in this region (Supplementary Fig. 31), indicating that the profile displayed in Fig. 5i was derived from Dex-droplets/AzSAP; thus, each droplet can be analyzed by flow cytometry.

Next, we investigated the restraining effects of the Dex-droplet/AzSAP system as a bioreactor on secondary infection in the amplification of M13 phages. M13 phages (4 × 10⁵ or 4 × 10⁶ virions ml^–1) and E. coli cells were cultured in LB medium, and the incubation time-dependent M13 phage concentration was determined using real-time polymerase chain reaction (PCR) analysis (Fig. 5k). In Dex-droplets without AzSAP (5.0 wt% dextran, 5.0 wt% PEG), the number of M13 phages increased exponentially over time and reached saturation after incubation for 24 h (Fig. 5l), indicating that repetitive amplification of M13 phages proceeded via multiple unrestrained infection events during droplet fusion. By contrast, the increase in the number of M13 phages over time proceeded more slowly in the Dex-droplet/AzSAP system than in Dex-droplets alone (Fig. 5l, Supplementary Figs. 32, 33). We also found that Dex-droplets prepared by continuous vortexing without AzSAP showed a lower suppression effect on infection than the Dex-droplets/AzSAP (Supplementary Fig. 34). Importantly, in the Dex-droplet/AzSAP system, the number of M13 phages at a given time point after addition was essentially proportional to the initial number. These results indicate that secondary infection by M13 phages was minimized in the Dex-droplet/AzSAP system; that is, the Dex-droplets effectively functioned as a bioreactor, providing a confined reaction space for M13 phage infection of E. coli cells. Notably, the proportional relationship between the amplified and initial number of M13 phages enables evaluation of the initial number by quantifying droplets containing infecting M13 phages. This approach would be useful for highly sensitive quantitative detection of viruses using flow cytometry.

To demonstrate highly sensitive quantitative detection of M13 phages, Dex-droplets/AzSAP were mixed with 9H-(1,3-dichloro-9,9-dimethylacridin-2-one-7-yl) β-D-galactopyranoside (DDAO galactoside), which emits red fluorescence after degradation, generating DDAO dye through reaction with β-galactosidase produced in M13 phage–infected E. coli cells⁵³. After incubation of M13 phages at a concentration of 2 × 10²–2 × 10³ virions ml^–1 with E. coli cells, CLSM observation revealed an increase in the number of Dex-droplets exhibiting DDAO dye fluorescence correlated with the initial concentration of M13 phages (Fig. 5m–q, Supplementary Fig. 35). Furthermore, flow cytometry analysis showed that the number of droplets exhibiting DDAO dye fluorescence increased depending on the initial M13 phage concentration (Fig. 5r). In sharp contrast to those results, incubation of E. coli cells and M13 phages in LB medium without Dex-droplets/AzSAP resulted in a saturation of the number of fluorescent droplets, regardless of the initial phage concentration (Supplementary Fig. 36). It should be noted that the concentration that could be detected using flow cytometry was >10 times higher than the detection sensitivity of conventional plaque assays performed on a two-dimensional plate (Supplementary Fig. 37). In addition, not only M13 phages but also lambda phages^54,55 can be used for the detection with comparable sensitivity (Supplementary Fig. 38), highlighting the compatibility and potential application of the 3D Dex-droplet/AzSAP system with flow cytometry analysis for highly sensitive detection of viruses.

Discussion

Here, we report a macroscopically fluid ATPS-based material stabilized by a network of supramolecular peptide nanofibers. AzSAP is composed of an iterative hydrophilic and hydrophobic amphiphilic sequence with an azobenzene unit. AzSAP affords a fluidic solution in water at 1.0 wt%, in which supramolecular nanofibers with high rectilinearity are constructed through β-sheet formation. Cryo-EM structural analyses revealed that AzSAP constructs two types of hetero clusters (3-mers and 4-mers) with azobenzene-based hydrophobic cores, which dimerize into a C₂-symmetric butterfly-shaped cross-section of 14 molecules (Fig. 2). The cross-section of AzSAP features a central cavity and loose packing, whereas rev-AzSAP, which has the reverse amino acid sequence, exhibits a denser packing in which the core is filled by the R1, D3, and R5 residues. This loose packing of AzSAP nanofibers enhances macroscopic fluidity and thermal stability compared with rev-AzSAP nanofibers, which is advantageous for microfluidic analysis and bioreactor applications. Such molecular-level understanding of the chemical and physical properties of AzSAP self-assembly, demonstrated by the comparison with rev-AzSAP, is highly valuable for optimizing molecules and materials with more sophisticated functionalities. The dextran-rich droplets formed by mixing with PEG in aqueous medium are inherently unstable and show prompt fusion within 1 h, which critically limits their application as microreactors. Addition of AzSAP nanofibers to a mixture of dextran and PEG afforded Dex-droplets/AzSAP, in which droplet fusion was effectively suppressed (Fig. 3). Suppression of droplet fusion was also observed following the addition of rev-AzSAP, which affords a self-standing hydrogel at 1.0 wt% in water. These functional differences directly mirror the nanofiber architectures revealed by cryo-EM analyses: AzSAP fibers, with their less dense packing, form a dynamic inter-droplet network that halts fusion yet preserves mobility, whereas the more compact rev-AzSAP fibers yield stronger, more viscous networks with reduced flow. Unlike Dex-droplets stabilized by liposome¹⁰ and triblock polymer¹², Dex-droplets/AzSAP are stable even in saline conditions and allow molecular transportation. Also, as compared to reported oil- or protein-based droplets^56–58, Dex-droplets and PEG minimize interactions with biomacromolecules, indicating that the Dex-droplet/AzSAP system is desirable for developing a variety of bioreactors.

A notable advantage of the Dex-droplet/AzSAP system is its photo-controllability, which allows for reversible triggering and suppression of droplet fusion for size control and photo-triggered initiation of M13 phage infection (Fig. 4). At the molecular level, this photo-responsiveness originates from the ordered π–π stacking of azobenzene units within the nanofiber cross-section, which enables structural rearrangements and polarity changes upon trans–cis isomerization that carry into macroscopic droplet behaviors. As observed in conventional Dex-droplets, M13 phages and E. coli cells localized in the droplets in the Dex-droplet/AzSAP system. Importantly, M13 phages in Dex-droplets infected E. coli cells and replicated, but secondary infection outside the droplets was effectively suppressed (Fig. 5). Nevertheless, the macroscopic fluidity, high stability of the microdroplets, and the restraining effect on secondary virus infection make it possible to use flow cytometry to quantitatively detect M13 phages in droplets of the Dex-droplet/AzSAP system. As a result, we achieved highly sensitive detection of viral titer, a complementary quantitative indicator to viral counts measurable by q-PCR⁵⁹ and microfluidic chip detection^27,60, with sensitivity >10 times higher than that of conventional plaque assays that use two-dimensional plates (Supplementary Table 1). By optimizing fluorescent probes for virus-infected cells⁶¹, the sensitivity and range of applicable target viruses for the Dex-droplet/AzSAP system could be further improved. In addition, the Dex-droplet/AzSAP system can be applicable for analyzing a variety of targets since dextran and PEG are known as biocompatible polymers, and the surface of AzSAP can be coated with blocking agents such as lactoferrin to effectively suppress undesired nonspecific interactions with proteins of interest. Detection costs could be improved in the future by optimizing the peptide structure to consist only of natural amino acids.

Thus, the present study established a direct structure–function relationship: the nanometer-scale packing geometry (visualized by cryo-EM) dictates micrometer-scale droplet stability, fluidity, and responsiveness. This correlation opens the way for the rational design of artificial droplet systems that function as high-performance, tunable microreactors for applications such as light-triggered initiation of bioreactions and ultra-sensitive virus detection.

Methods

Materials

Acetic anhydride was purchased from Kanto Chemicals (Tokyo, Japan). Acetonitrile, N,N’-dimethylformamide (DMF), Et₂O, N-methyl-2-pyrrolidone (NMP), piperidine, and trifluoroacetic acid (TFA) were purchased from Kishida Chemical (Tokyo, Japan). N,N-Diisopropylethylamine (DIEA) and tryptone were purchased from Nacalai Tesque (Kyoto, Japan). TNBS test kits, triisopropylsilane (TIS), and tetracycline hydrocholoride were purchased from Tokyo Chemical Industry (Tokyo, Japan). Fmoc-Ala-OH ∙ H₂O, Fmoc-Arg(Pbf)-OH, Fmoc-Met-OH, Fmoc-Asp(OtBu)-OH, Fmoc-Phe(4-N=NPh)-OH, Fmoc-NH-SAL resin, 2-(1H-benzotriazole-1-yl)−1,1,3,3-tetramethyluronium hexafluorophosphate (HBTU), and 1-hydrozybenzotriazole monohydrate (HOBt∙H₂O) were purchased from Watanabe Chemical Industries (Hiroshima, Japan). Ultrapure water (filtered through a 0.22-μm membrane filter, >18.2 MΩ cm) was purified by Purelab DV35 of ELGA (Buckinghamshire, UK). M13KE bacteriophage (M13 phage) and Escherichia coli ER2738 for amplification of M13 phage were purchased from New England Biolabs (Ipswich, MA, USA). Lambda gt11 phage (lambda phage) was purchased from the American Type Culture Collection (Manassas, VA, USA). E. coli ME9039 for amplification of lambda phage was purchased from National Bioresource Project (Shizuoka, Japan). Dextran (M_w = 450,000–650,000) and PEG (M_w = 35,000) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Agar powder, 5-bromo-4-chloro-3-indolyl β-D-galactopyranoside (X-gal), agarose S, isopropyl β-D-thiogalactoside (IPTG), tetracycline hydrochloride, yeast extract, bovine serum albumin, phosphate-buffered saline, lactoferrin, and gelatin were purchased from FUJIFILM (Osaka, Japan). Cover glasses (⌀13 mm) were purchased from Matsunami (Osaka, Japan). Glass substrates were purchased from Matsunami Glass Ind., Ltd. (Osaka, Japan) and sonicated in NaOH (0.1 M) solution for 60 min before use. Primers for real-time PCR (forward: CCTTTAGTGGTACCTTTCTATTCTCAC, reverse: CATGTACCGTAACACTGAGTTTCGTC) were purchased from FASMAC (Kanagawa, Japan). SPiDER-βGal was purchased from DOJINDO (Kumamoto, Japan). FITC-labeled dextran (M_w = 2000) was purchased from Cosmo Bio (Tokyo, Japan). Cyanine 5-N-hydroxysuccinimide (Cy5-NHS) was purchased from Funakoshi (Tokyo, Japan). NucleoSpin^® Virus for extraction of DNA was purchased from Takara Bio (Shiga, Japan). LIVE/DEAD BacLight™ bacterial viability kits and DDAO galactoside were purchased from Thermo Fisher Scientific (Waltham, MA, USA).

Instruments

CD spectra were recorded on a J-1100 CD spectrometer equipped with a PTC-514 Peltier temperature controller (JASCO, Tokyo, Japan). UV-vis absorption spectra and diffuse reflectance spectra were recorded on a V770 spectrometer (JASCO). Rheological analyses were conducted using a Kinexus rotational rheometer (Malvern Panalytical, Worcestershire, UK). Nuclear magnetic resonance (NMR) spectra were recorded on an ECX-500 spectrometer (JEOL, Tokyo, Japan), where the chemical shifts were determined with respect to a solvent signal as an internal standard [¹H NMR, 2.50 ppm for (CD₃)₂SO; ¹³C NMR, 39.52 ppm for (CD₃)₂SO], and NMR spectra were analyzed by Delta software. Optical rotation measurements were performed on a DIP-1000 (JASCO). TEM observations were conducted using a JEM-1400 instrument (JEOL) equipped with a Gatan UltraScan 4k×4k CCD camera using a JEOL carbon-reinforced microgrid (Cu 200 mesh) and uranyl acetate or gadolinium acetate for staining under an accelerating voltage of 200 kV. TEM images were photographed using the JEOL minimum dose system (search/focus/record modes). Cryo-TM data sets were collected using a Falcon-4i camera with a Selectris X 10-eV slit mounted on a 300-kV Titan Krios G4 (Thermo Fisher Scientific) at Okayama University (Okayama, Japan). Cryo-TM samples were prepared using a Vitrobot Mark IV instrument (Thermo Fisher Scientific). FTIR spectra were recorded on a FT/IR-6100 from JASCO (Tokyo, Japan). Peptides were synthesized using a HiPep Laboratories model shaking apparatus and polypropylene LibraTubes with a filter. Sonication was conducted using an As One model ultrasonic bath sonicator AS12GTU (35 kHz, 60 W). Lyophilization was conducted using a type FDU-1200 lyophilizer (EYELA, Tokyo, Japan). Centrifugation was conducted using a micro-refrigerated centrifuge 3700 (Kubota, Tokyo, Japan). Hydrogels were incubated using a Peltier-cooled incubator CN-40A (Mitsubishi Electric Engineering, Tokyo, Japan). Phase-contrast microscopy was performed on an Olympus model IX73 instrument equipped with a C10600-10B CCD camera (Hamamatsu Photonics, Shizuoka, Japan). Differential interference contrast microscopy was performed using an AX10 microscope (Carl Zeiss, Oberkochen, Germany) equipped with Axiocam 712 color camera (Carl Zeiss). CLSM was performed on a STELLARIS 5 instrument (Leica, Wetzlar, Germany). Flow cytometry was performed using a BD Accuri^TM C6 flow cytometer (BD Bioscience, Franklin Lakes, NJ, USA) or another flow cytometry device⁵¹. Viscosity measurements were performed using a microVISC^TM (RheoSense, San Ramon, CA, USA).

Synthesis of peptides

Ac-RAzDARMR-NH₂ (AzSAP), Ac-RMDARAzR-NH₂ (rev-AzSAP), DAzDARMD-NH₂ (AzSAP-D), Ac-KAzDARMK-NH₂ (AzSAP-K), Ac-RIDARMRADAIR-NH₂ (JigSAP)⁴², and Ac-RADARADARADARADA-NH₂ (RADA16)⁴⁵, and corresponding peptides conjugated with fluorescent Cy5 [Cy5-GRAzDARMR-NH₂ (Cy5-AzSAP), Cy5-GRMDARAzR-NH₂ (Cy5-rev-AzSAP), Cy5-GDAzRADMD-NH₂ (Cy5-AzSAP-D), Cy5-GKAzRADMK-NH₂ (Cy5-AzSAP-K), Cy5-GRIDARMRADAIR-NH₂ (Cy5-JigSAP), and Cy5-GRADARADARADARADA-NH₂ (Cy5-RADA16)] were prepared using 9-fluorenylmethyloxycarbonyl solid-phase peptide synthesis^42,43. Typically, a condensation-reagent cocktail consisting of HBTU (3.1 g, 8.0 mmol) and HOBt∙H₂O (1.3 g, 8.2 mmol) in DMF (16 ml), a mixture of DIEA (2.8 ml) and NMP (14.3 ml), and a cleavage cocktail consisting of TIS (62.5 μl), TFA (2.4 ml), and water (62.5 μl) were prepared just prior to peptide synthesis. Fmoc-NH-SAL resin (0.1 mmol) in a polypropylene tube was soaked in DMF (2 ml) for 3 h at 25 °C. After removal of DMF, piperidine in DMF (20%, 2 ml) was added and mixed by vortexing for 1 min. After the reaction solution was removed, piperidine in DMF (20%, 2 ml) was added, and the reaction tube was shaken for 10 min at 25 °C. After removal of the reaction solution, the resin was washed with DMF (2 ml, 5 times), CH₂Cl₂ (2 ml, 3 times), and DMF (2 ml, 3 times). Next, Fmoc-protected amino acids (0.3 mmol) dissolved in condensation-reagents cocktail (700 μl) and a mixture of DIEA and NMP (700 μl) were added to the resin. After shaking for 15 min at 25 °C, the reaction solution was removed, and the resin was washed with DMF (2 ml, 5 times), CH₂Cl₂ (2 ml, 3 times), and DMF (2 ml, 3 times). The Fmoc deprotection reactions with piperidine and coupling reactions of Fmoc-protected amino acids were repeated following the designed sequence. After the final Fmoc deprotection reaction and washing step, acetic anhydride in CH₂Cl₂ (25%, 2 ml) was added to the resin, and the reaction tube was shaken for 5 min at 25 °C. After removal of the reaction solution, the resin was washed with DMF (2 ml, 5 times), CH₂Cl₂ (2 ml, 3 times), and DMF (2 ml, 3 times). Cleavage cocktail (2.5 ml) was then added to the resin, and the reaction tube was left to stand for 90 min at 25 °C, with gentle shaking every 30 min. The solution was collected into a polypropylene centrifuge tube by filtration. The reaction tube was rinsed with TFA (500 μl, 3 times), which was also collected by filtration. Et₂O (40 ml) was then added to the centrifuge tube and mixed by vortexing for 1 min, followed by centrifugation at 4 °C (3500 × g, 5 min) and removal of the supernatant. This process was repeated 3 times, and the resulting peptide was dried under vacuum for 2 h at 25 °C, dispersed in water, and lyophilized. Peptide samples were purified by HPLC using a system consisting of a PU-4180 RHPLC pump, a UV-4075 UV/Vis detector, and LC-Net II/APC (JASCO) equipped with a Triart C18 column (⌀4.6 × 250 mm, YMC, Kyoto, Japan) or Actus Triart C18 column (⌀20 × 250 mm, YMC). Purified peptides were characterized using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) on an Autoflex Speed spectrometer (Bruker, Bremen, Germany) in positive reflector mode with 2,5-dihydroxybenzoic acid as the matrix.

Preparation of peptide self-assemblies

Typically, lyophilized powder of AzSAP (10 mg) was dissolved in 0.5 ml of an aqueous dispersion of PEG (10 wt%). After gentle shaking and sonication for 20 min at 25 °C, the sample was left to stand overnight at 20 °C. The sample was then mixed with 0.5 ml of an aqueous dispersion of dextran (10 wt%), and the resulting sample was vortexed for 10 s. Dex-droplets prepared using AzSAP and rev-AzSAP and the corresponding Cy5-labeled assemblies of AzSAP (10 mg AzSAP, 0.03 mg Cy5-AzSAP), rev-AzSAP (10 mg rev-AzSAP, 0.03 mg Cy5-rev-AzSAP), JigSAP (10 mg JigSAP, 0.03 mg Cy5-JigSAP), and RADA16 (10 mg RADA, 0.03 mg Cy5-RADA) were prepared according to the same method. AzSAP and rev-AzSAP fibers in aqueous dispersions were prepared using water instead of the mixture of PEG (5 wt%) and dextran (5 wt%). To prepare Dex-droplets encapsulating M13 phages and E. coli cells, a dispersion consisting of 10 wt% dextran, 40% LB medium, M13 phages, and 4% E. coli cells prepared by incubation in LB medium at 37 °C overnight was used instead of the aqueous dispersion of dextran (10 wt%).

Cryo-EM grid preparation and data collection

AzSAP fiber samples were diluted with water, and 3 µl of the diluted solution was applied to a glow-discharged holey carbon grid (Quantifoil R1.2/1.3, Cu, 300 mesh). After blotting for 3 s with a blotting force of –10 at 4 °C under 100% humidity using a Vitrobot Mark IV (Thermo Fisher Scientific), the grids were rapidly plunge-frozen in liquid ethane and stored in liquid nitrogen. Cryo-EM data were acquired on a Titan Krios G4 transmission electron microscope operated at 300 kV and equipped with a Selectris X energy filter (10-eV slit) and Falcon-4i direct electron detector (Thermo Fisher Scientific) at Okayama University (Okayama, Japan). A total of 10,020 movies were collected automatically using EPU software at a nominal magnification of 165,000×, corresponding to a calibrated pixel size of 0.727 Å, with an exposure time of 4.6 s and a total electron dose of 50 e/Ų.

Cryo-EM data processing

Collected movies were gain-corrected, aligned, dose-weighted, and summed using RELION’s implementation of MotionCor2^62,63. Contrast transfer function (CTF) parameters were estimated using  CTFFIND4⁶⁴. All subsequent image-processing steps were performed using the helical reconstruction procedure in RELION-5.0^65–67. Filaments were initially picked manually, followed by reference-free 2D classification. The resulting segments were used to train Topaz, and filaments were then picked automatically using a modified version of Topaz^68,69. From the AzSAP datasets, 7,687,512 and 3,699,956 filament segments were extracted, respectively.

Subsequent reference-free 2D classification and dendrogram-based filament sorting identified 106,867 and 550,492 homogeneous filament segments, respectively, for further processing. An initial 3D reference model was reconstructed from 2D class averages using relion_helix_inimodel2d⁷⁰ with an estimated C₂ symmetry and cross-over distance of 72.5 nm. 3D auto-refinement was performed without imposing high symmetry, revealing pseudo-21 helical symmetry from map features. Iterative CTF refinement and Bayesian polishing^64,71 were then applied, followed by 3D auto-refinement with pseudo-21 helical symmetry. High-resolution helical segments were further filtered to 92,358 using CryoSieve⁷².

The final 3D reconstructions were post-processed using standard RELION procedures, and the overall resolution was determined from Fourier shell correlations at 0.143⁷³. Map handedness was validated by examining density for main-chain carbonyl oxygens and confirmed using Ramachandran plots⁶⁹. Detailed cryo-EM data processing workflows are shown in Supplementary Fig. 5.

Atomic modeling

The initial atomic model was manually built in Coot⁷⁴, with restraint files for modified residues generated using aceDRG⁷⁵. The asymmetric unit model was refined under helical symmetry constraints using Servalcat⁷⁶, and the sharpened, weighted FoFc map from Servalcat was used for iterative model building in Coot and for visualization. The final model was validated using MolProbity^77,78. For structural visualization, the model was expanded to a multi-filament assembly using helical symmetry, and ChimeraX⁷⁹ was employed to generate structural representations, including electrostatic potential and hydrophobicity surface maps. Detailed cryo-EM data collection, reconstruction, and refinement statistics are summarized in Supplementary Table 2.

FTIR analysis of peptide self-assemblies

FTIR spectra of self-assemblies of AzSAP, rev-AzSAP, AzSAP-K, and AzSAP-D were measured in the hydrated state using BioCell CaF2 windows, in which the path length was fixed (BioTools, West Palm Beach, FL, USA). Band decomposition analysis of the spectra was carried out using six Gaussian bands⁴².

Handling of phages

M13 phages were amplified using E. coli cells (ER2738) in LB medium containing tetracycline (45 µM) and purified by two cycles of precipitation and re-dispersion using a solution containing PEG (14 wt%) and NaCl (290 mM)^80,81. The concentration of M13 phages was determined by plaque assay on X-gal plates⁸². For FITC labeling, M13 phages (1.5 × 10¹³ virions ml^–1) were incubated with FITC isomer (FITC-I, 250 µg ml^–1) in Tris-HCl buffer (0.5 mM Tris-HCl, pH 8.0) for 3 h at 25 °C; unreacted FITC-I was removed by dialysis. Lambda phages were amplified using E. coli cells (ME) in LB medium. The concentration of lambda phages was calculated based on the assumption that its plaque-forming unit (PFU) in lysate is estimated to be 1 × 10⁸ virions ml^–183, and particle/PFU ratio is estimated to be 3⁸⁴.

Flow cytometry analysis

Typically, flow cytometry analysis was performed on a BD Accuri^TM C6 flow cytometer (BD Bioscience) with gating of FSC-H on 15,000 and SSC-H on 1000 at a flow rate of 100 μl min^–1 using a flow channel with a core size of 40 μm; alternatively, a flow cytometry device with a channel width of 400 μm was used⁵¹. To prepare the device, the flow pattern was machined into a bulk sheet of polyacetal resin using computer numerical control machining followed by pouring of silicone resin. After incubation for 8 h at 60 °C for curing, the flow pattern was transferred to the resin (Supplementary Fig. 39). Fluorescent microscopy of the flow cytometry device was performed on a THUNDER Imager Model Organism (Leica Microsystems, Wetzlar, Germany) through an excitation filter (λ = 620 ± 60 nm) and fluorescence filter (λ = 700 ± 75 nm).

DNA extraction and qPCR

In the typical procedure for quantifying M13 phages by qPCR, a mixture of Dex-droplets/AzSAP (5.0 wt% dextran, 5.0 wt% PEG, 1.0 wt% AzSAP) containing lactoferrin (1.5 wt%), M13 phages (4 × 10⁵ or 4 × 10⁶ virions ml^–1), and E. coli cells (2% of confluent sample) in 20% LB medium was incubated for 0 to 24 h. The samples were diluted 10-fold with sterile water and purified using NucleoSpin^® Virus (Takara Bio). Fluorescence quantification of the purified products was performed on a CFX Opus Real-time PCR system (Bio-Rad, Hercules, CA, USA). The total reaction volume was 20 μl and consisted of 10 μl of THUNDERBIRD^® Next SYBR™ qPCR Mix, 300 nM forward primer (AGAATAATAGGTTCCGAAATAGGCAGG), 300 nM reverse primer (GTACTGGTAATAAGTTTTAACGGGGTC), 2 μl of purified product, and sterile water. The PCR procedure was set as follows: first 95 °C pre-denaturation for 30 s; second 95 °C denaturation for 5 s, 60 °C annealing for 30 s, 40 cycles. The concentration of M13 phages was estimated from the qPCR results. qPCR analyses of samples of Dex-droplets without AzSAP (5.0 wt% dextran, 5.0 wt% PEG) were performed using the same method.

Photo-control of self-assemblies of AzSAP

In the typical procedure for inducing depolymerization of AzSAP, an aqueous dispersion of a mixture of dextran and PEG (5.0 wt% dextran, 5.0 wt% PEG) in the presence of AzSAP (1.0 wt%) was irradiated with 365-nm light to induce trans-to-cis photoisomerization of the Az group using an Olympus IX-73 microscope equipped with a U-HGLGPS (130 W) and U-FUNA filter (λ = 365 ± 5 nm) (Olympus, Tokyo, Japan) under phase-contrast microscopic observation. Similarly, samples were irradiated with 420-nm light to induce cis-to-trans photoisomerization under phase-contrast microscopy using a U-FBVW filter (λ = 420 ± 20 nm). Light irradiation at 350 nm and 450 nm to induce trans-to-cis and cis-to-trans photoisomerization for macroscopic observations was performed using a MAX-303 300-W Xe-lamp (Asahi Spectra, Tokyo, Japan) through filters (λ = 350 ± 5 nm and 450 ± 5 nm, respectively).

Statistics and reproducibility

Experiments, including preparation of peptide fibers, stabilization of ATPS droplets, rheological measurements, and investigations of light-responsive behaviors of the peptide fibers, were performed at least three times to ensure reproducibility. To evaluate the size of Dex-droplets from microscopic images, the diameters of 100 Dex-droplets were measured.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

N.U. and S.K. performed most of the experiments. At.Y. and H.H. characterized the peptide self-assemblies. K.M. performed flow cytometry analyses. H.K., H.U., N.I., M.T., S.A., M.O., Y.-H.L., and S.-Y.P. performed cryo-EM analyses. S.A. and M.T. performed ESI-MS/MS analyses. Ta.S. and G.W. performed MD simulation. Ay.Y., G.H., and J.F. performed droplet stability analyses. T.T. and D.Y. fabricated the flow-device capillaries. To.S. and H.N. analyzed the data. N.U. and T.M. designed the experiments. T.M. conceived the research concept, and N.U., S.-Y.P., and T.M. co-wrote the manuscript.

Peer review

Peer review information

Nature Communications thanks anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Funding

A portion of this work was supported by the “Advanced Research Infrastructure for Materials and Nanotechnology in Japan (ARIM)” program of the Ministry of Education, Culture, Sports, Science and Technology (MEXT), grant number JPMXP12 24UT0073. This work was also supported by the following funding sources: Grant-in-Aid for Transformative Research Areas (B) JP21H05096 (to TM); Grant-in-Aid for Scientific Research (B) JP 25K01891 (to TM); Grant-in-Aid for Scientific Research (C) JP23K04927 (to NU); National Research and Development Agency, Intensive Support for Young Promising Researchers 23W1M041 (to NU); Japan Science Technology Agency (JST) Core Research for Evolutional Science and Technology (CREST) program, JPMJCR19S4 (to TM), Fusion Oriented Research for Disruptive Science and Technology (FOREST) JPMJFR2122 (to TM) and JPMJFR244H (to NU), Adopting Sustainable Partnerships for Innovative Research Ecosystem (ASPIRE) JPMJAP2526 (to TM), and ACT-X JPMJAX23DL (AtY); JSPS Program for Forming Japan’s Peak Research Universities (J-PEAKS, to TM); the Japan Association for Chemical Innovation (to NU); the Moritani Foundation (to NU); the Tanaka Foundation (to NU); the Kose Cosmetology Foundation (to NU); the Asahi Glass Foundation (to NU and TM); the Canon Foundation (to TM); Mukai Science and Technology Foundation (to TM); the Lotte Foundation (to NU); “Innovation inspired by Nature” Research Support Program, SEKISUI CHEMICAL CO., LTD. (to TM); Japan Keirin Autorace foundation (to TM); Fujimori Science and Technology Foundation (to TM); Takahashi Industrial and Economic Research Foundation (to TM); National Research Foundation of Korea (NRF) grants RS-2021-NR057690, RS-2022-NR069719, and RS-2025-25444385 (to YHL); Korea Institute of Marine Science & Technology Promotion (KIMST) grants RS-2026-25541458 and RS-2026-25508103 (to YHL); Korea Basic Science Institute (KBSI) funds C623300, C623100, C612120, and C623500 (to YHL).

Data availability

All data are available in the manuscript or the supplementary materials. Cryo-EM maps and corresponding atomic coordinates of AzSAP and rev-AzSAP were deposited in the Electron Microscopy Data Bank (EMDB), the Protein Data Bank (PDB), and Zenodo database [10.5281/zenodo.20757368] under accession codes EMD-65607 and PDB 9W3T, and EMD-62572 and PDB 9KU1, respectively. The MD simulation data generated in this study have been deposited in the Zenodo database [10.5281/zenodo.20665281]. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-75297-x.