On the Assembly of Actin Polymerization‐Powered Motors

ABSTRACT

Nano‐ and micromotors are a class of active colloids that can self‐propel outperforming Brownian motion. Polymer synthesis or degradation are alternative ways to enzyme‐based or externally‐driven strategies to induce self‐propulsion in particles, but they are often limited due to the reaction conditions. Nature leverages biopolymerization reactions to sustain locomotion either of whole microorganisms or of organelles inside cells. With the aim of integrating natural locomotion strategies into engineered motors, we have begun to explore the propulsion mechanism of the food‐borne pathogen Listeria monocytogenes, which expresses the actin‐recruiting protein ActA on its surface to harness host cell actin polymerization for rapid intracellular movement. Here, we compare the locomotion of silica particles depending on the ActA immobilization strategy on the motor surface, using either homogeneous coatings, Janus‐type coatings, or ActA immobilization within polymer brushes. An up to 5‐fold increase in the propulsion of the motors compared to their Brownian motion is observed when Janus motors are considered. The motors orbit around or dock onto larger tracer particles depending on the environmental pH and on whether they are individuals or in clusters. Altogether, these motors illustrate how integration of concepts of the natural and synthetic world can yield unique engineered units.

Actin polymerization‐powered motors have the highest speed when the actin recruiting‐polymerizing proteins are immobilized in polymer brushes on their surface. These motors orbit around larger tracer particles at pH above the isoelectric point of ActA but dock onto the tracer particles’ surface at lower pH.

1. Introduction

Motion is fundamental for all living organisms, and this has inspired broad interest in creating autonomous motion in artificial systems. Nano/micromotors (also referred to as swimmers) have emerged as a class of active matter able to self‐navigate different environments. These (sub)micrometer particles convert various forms of input energy into kinetic energy, allowing them to move in ways that exceed random Brownian motion. Significant advances have been made, not only in motor design (e.g., from spherical particles to more complex shapes) and propulsion mechanisms, but also in expanding the diversity of environments where the motors can move (e.g., from water to living organs), as highlighted in several reviews [1, 2, 3, 4, 5] and recently in a technology roadmap [6].

Popular mechanisms of motion include the use of external magnetic fields [7, 8], light [9, 10, 11] or (bio)chemical reactions [12, 13, 14]. These approaches have shown considerable control over the motors’ directionality or achievable top speeds. As an alternative, we [15, 16, 17, 18, 19] and others [20, 21] have previously explored the use of (de)polymerization processes to induce motion. The Sen's group reported the first effort on using polymer reactions to propel particles [20], where they employed the Grubbs catalyst decorating Janus (asymmetric) particles to polymerize norbornene in bulk solution. The formation of the polymer in solution led to an imbalanced distribution of molecules around the motors that resulted in locomotion due to self‐diffusiophoresis. Since then, polymer‐powered motors have attracted great attention as recently summarized in a review by Yan and co‐workers [22]. Examples include nanomotors that can move due to depolymerization (degradation) of their environment, polymer‐mediated reactions or (de)polymerization processes on the motors’ surface. In the former case, we have for instance shown that motors equipped with the collagenase enzyme could move both in collagen‐ [23, 24] and gelatin‐based [25] environments at very high speeds. Similarly, Khoe and co‐workers have employed the enzyme lipase to specifically degrade ester bonds along polymer chains and produce locomotion [26, 27, 28]. Alternatively, nanomotors made of polymerized L‐arginine that could respond to reactive oxygen species produced after inflammation in human umbilical vein endothelial cells (HUVEC) and macrophages (Raw264.7 cells) and moved due to the gradual production of nitric oxide [29]. Alternatively, Janus motors incorporating thermolabile acyl azide moieties were employed to propel due to the thermal‐based self‐diffusiophoresis, thanks to the chemical gradient produced when the acyl azides underwent a phase rearrangement upon heating above 32°C [30]. We showed that the surface polymerization of poly(hydroxyethyl methacrylate) from Janus silica particles resulted in enhanced diffusion [15]. These motors were able to exhibit swarming (collective motion) when high numbers of motors were considered. Conversely, polymer brush detachment from the motors’ surface by cleaving a pH‐responsive linker was evaluated as a method to induce locomotion [17]. This approach showed enhanced motor mobility when the pH of the medium was lower than 4, displaying ballistic motion during long trajectories (ca. 15 µm). As a comparative example, polymer fragmentation using self‐immolative polymers was considered [16]. This strategy showed enhanced Brownian diffusion when the trigger molecule for the fragmentation was present. Another report illustrated the potential of photo‐labile azo‐based motifs that could decompose with the simultaneous release of nitrogen gas, resulting in fast motion [31]. However, all these strategies considered synthetic polymers and required harsh reaction conditions (e.g., extreme pH, organic solvents), which limit the use of these motors in biological environments. We therefore started to explore a nature‐mimicking locomotion approach. Inspired by the food‐borne bacterium Listeria monocytogenes, which polymerizes actin filaments using the inherent machinery of mammalian cells upon infection [32, 33], we have recently reported that polystyrene nanoparticles coated with an actin recruiting‐polymerizing protein (ActA) were able to form cytoskeleton‐like networks in the lumen of giant vesicles [34]. However, this first effort did neither explore nor compare different motor designs beyond the size of the carrier particle.

Therefore, we expanded on the topic by focusing on the locomotion characteristics of motors driven by actin polymerization. Specifically, we i) employed different strategies to deposit ActA onto silica particles, either as a homogeneous monolayer, in a Janus configuration, or embedded within poly(diethylaminoethylmethacrylate) polymer brushes in order to increase the density of immobilized ActA, with the aim to identify the design that yielded the motors with the most efficient locomotion, and ii) determined the capacity of these motors to interact with passive tracer particles depending on the environmental pH (Scheme 1).

SCHEME 1.

Overview of the actin‐polymerization driven motors. The actin recruiting protein ActA is deposited onto the surface of silica nanoparticles (SiO₂ NP) to yield motors with different surface decoration able to move due to polymerization of G‐actin into F‐actin, supported by endogenous proteins in cellular lysate (left). These motors interact with tracer particles and orbit or dock depending on the pH of the environment due to the isoelectric point (pI) of the ActA protein (right). Legend: PLL‐g‐PEG, poly(L‐lysine)‐grafted‐poly(ethylene glycol); PDEAEMA, poly(diethylaminoethylmethacrylate); ^0.5JS‐ActA, Janus‐shaped motor; ^0.5S‐ActA, homogenously coated motor; ^0.5S^B‐ActA, in‐polymer‐brush coated motor.

2. Results and Discussion

2.1. Actin‐Polymerization Driven Motors

First, we compared different motor designs and their ability to polymerize G‐actin.

2.1.1. Motor Design

The motors consisted of a silica (SiO₂) core decorated with actin recruiting‐assembling proteins (ActA) [35]. ActA was expressed in an Escherichia coli strain (≈30% purity) using a pET30a vector, as previously detailed [34]. SiO₂ particles are advantageous cores because they are easy to produce in high quality, can be readily chemically modified, and enable straightforward Janus particle fabrication, even when organic solvents are required. SiO₂ particles were synthesized by the Stöber method [36], aiming for an average diameter of ≈500 nm (^0.5S, Figure S1a). Three motor designs were considered to determine the ActA deposition conditions that yielded the highest velocities. Specifically, Janus (asymmetric) motors were assembled by first making a Pickering emulsion [37], which allowed for the passivation of one hemisphere of ^0.5S with poly(L‐lysine)‐grafted‐poly(ethylene glycol) (PLL‐g‐PEG) to avoid protein binding. Upon removal of the wax, the second hemisphere was decorated with the ActA proteins resulting in ^0.5JS‐ActA. In addition, ActA was directly deposited on ^0.5S to obtain ^0.5S‐ActA (Figure 1ai). The third type of motor was based on ^0.5S equipped with pre‐polymerized poly(diethylaminoethylmethacrylate) (PDEAEMA) brushes [25, 38], before ActA was immobilized to obtain ^0.5S^B‐ActA (Figure S1b). In addition, we assembled control motors by depositing BSA instead of ActA resulting in ^0.5JS‐BSA, ^0.5S‐BSA, and ^0.5S^B‐BSA.

FIGURE 1.

Motor design and G‐actin surface polymerization. a) Schematic of ActA protein deposition on silica (SiO₂) nanoparticles to result in ^0.5JS‐ActA, ^0.5S‐ActA and ^0.5S^B‐ActA (i). Representative SR‐CLSM images of the ActA_OG deposition (ii) (scale bars in the insets are 1 µm) and line scan profiles (iii). b) Representative TEM (i) and SR‐CLSM (ii) images of ^0.5JS‐ActA_OG, ^0.5JS‐BSA_FITC, ^0.5S‐ActA_OG, ^0.5S‐BSA_FITC, ^0.5S^B‐ActA_OG, and ^0.5S^B‐BSA_FITC after G‐actin_AT polymerization. The dotted circles indicate the cores of the particles. (Cyan: ActA_OG; green: BSA_FITC; magenta: G‐actin_AT).

The ActA deposition was assessed by superresolution confocal laser scanning microscopy (SR‐CLSM) using Oregon Green 488 maleimide‐modified ActA (ActA_OG, Figure 1aii). The fluorescent signal from ActA_OG originated in a ring for ^0.5S‐ActA_OG and ^0.5S^B‐ActA_OG, confirming the homogeneous coating, whilst it originated from a half‐moon shape for ^0.5JS‐ActA_OG, which was indicative of the Janus morphology. The line scan profiles supported the presence of two different morphologies, as one peak was found for ^0.5JS‐ActA_OG, while two peaks were found for ^0.5S‐ActA_OG and ^0.5S^B‐ActA_OG (Figure 1aiii). In addition, the fluorescence emission spectra of ^0.5S‐ActA_OG and ^0.5S^B‐ActA_OG were recorded and compared to the emission spectrum of bulk ActA_OG (Figure S1c). The maximum emission at λ _em = 524 nm was ≈4× and ≈2× higher for the bulk ActA_OG compared to ^0.5S‐ActA_OG and ^0.5S^B‐ActA_OG, respectively. We used these data to estimate that ≈3×10⁴ and ≈1×10⁵ ActA proteins were immobilized per ^0.5S‐ActA and ^0.5S^B‐ActA, respectively (assuming no losses during functionalization). We would like to note that this calculation could not be performed for ^0.5JS‐ActA due to insufficient accuracy in estimating its concentration.

2.1.2. Actin Surface Polymerization

Actin polymerization is a dynamic process where ATP‐bound actin monomers (G‐actin) add to the barbed end of filaments, promoted by proteins such as ActA. Spontaneous nucleation can occur but is inefficient and usually yields small filament clusters [39]. Once formed, filaments (F‐actin) grow to hundreds or thousands of nanometers. Disassembly occurs at the pointed end as actin subunits hydrolyze ATP to ADP, regulated by proteins such as cofilin and other ADF family members [40, 41].

We and others have previously reported on actin surface polymerization on micrometer sized polystyrene beads [34, 42, 43, 44, 45] and lipid vesicles [46]. Here, we aimed to determine how the three different ActA immobilization strategies affected the subsequent G‐actin polymerization (Figure 1bi). We incubated ^0.5JS‐ActA, ^0.5S‐ActA and ^0.5S^B‐ActA with human umbilical vein endothelial (HUVEC) cell lysate supplemented with G‐actin (actin monomers, 0–4 µm final concentration) and ATP (1 mm final concentration) for 20 min at room temperature followed by washing in 4‐(2‐hydroxyethyl)piperazine‐1‐ethane‐sulfonic acid (HEPES) buffer. (For details on the cell lysate preparation, please refer to our previous publication [34]). The cell lysate contained ≈4 µm actin (G+F), ≈25 µm ATP and a protein content ranging from 200 to 1000 µg mL⁻¹. ^0.5JS‐BSA, ^0.5S‐BSA and ^0.5S^B‐BSA were employed as controls in addition to using cytochalasin D as an inhibitor for actin polymerization [32].

Transmission electron microscopy (TEM) images of ^0.5JS‐ActA, ^0.5S‐ActA, ^0.5S^B‐ActA as well as the controls ^0.5JS‐BSA ^0.5S‐BSA and ^0.5S^B‐BSA were collected after incubation with cell lysate supplemented with G‐actin and ATP (Figure 1bi). The samples coated with ActA showed a grey shell after incubation, typical of organic material, F‐actin in this case. Especially, the organic layer was much more obvious and thicker (compared to the solid particle diameter) in ^0.5S^B‐ActA, likely due to the polymer shell together with the F‐actin. However, BSA‐coated samples did not show such a layer, indicating F‐actin did not form on these samples.

SR‐CLSM images of the same assemblies were also recorded after incubation in cell lysate supplemented with ATTO 647 actin (G‐actin_AT) and ATP (Figure 1bii). ATTO 647 actin is a chemically modified fluorescent G‐actin that possesses the characteristics of native G‐actin and can integrate into the F‐actin assemblies. In this case, ActA_OG and fluorescein isothiocyanate (FITC) labelled BSA (BSA_FITC) were employed. ^0.5JS‐ActA_OG, ^0.5S‐ActA_OG and ^0.5S^B‐ActA_OG showed both a fluorescent signal originating from the ActA_OG (cyan) and an overlapping fluorescent signal stemming from F‐actin_AT (magenta). The ActA on the surface of the motors was able to recruit the G‐actin from the solution and initiate the polymerization of F‐actin with the assistance of ATP and proteins such as the Arp2/3 complex, confirming the TEM observations. We would like to emphasize that ^0.5S^B‐ActA_OG showed the strongest fluorescent signal from F‐actin_AT, illustrating the ability of the polymer brushes to accommodate high amounts of ActA and thus produce more F‐actin. Further, neither ^0.5JS‐BSA_FITC, ^0.5S‐BSA_FITC nor ^0.5S^B‐BSA_FITC exhibited any fluorescent signal originating from G‐actin_AT, i.e., no F‐actin was formed. However, small amounts of F‐actin filaments or G‐actin aggregates were found in the solution or close to the surface of these three assemblies, likely due to spontaneous F‐actin formation even in the absence of recruiting proteins or the formation of G‐actin aggregates. Additionally, cytochalasin D was used to inhibit the F‐actin formation on ^0.5JS‐ActA_OG, ^0.5S‐ActA_OG, and ^0.5S^B‐ActA_OG, showing very similar results to when BSA‐coated particles were employed (Figure S2). Therefore, we used ^0.5JS‐BSA, ^0.5S‐BSA, and ^0.5S^B‐BSA as controls for further experiments instead of the inhibitor proteins.

2.2. Locomotion

The locomotion of the three designs was compared in cell lysate with the aim to identify the configuration that resulted in the fastest moving motors.

2.2.1. Locomotion of ^0.5JS‐ActA

We assessed the locomotion of ^0.5JS‐ActA in cell lysate that followed a Brownian ratchet mechanism‐like propulsion (Figure S3 and related text) [33, 47, 48, 49]. An evaluation of the best conditions for locomotion, including G‐actin concentration, viscosity of the medium and temperature was carried out first using ^0.5JS‐ActA, identifying that 10 vol% cell lysate and 4 µm G‐actin were the optimal parameters for locomotion, whereas temperature did not seem to influence the active motion of the motors (Figure S4 and the related text). Then, the mobility of ^0.5JS‐ActA was evaluated at different time points, namely after 30, 60, and 90 min of incubation, by dispersing the motors in uncoated µ‐Slides VI^0.4 microfluidic channels in an environment consisting of 10 vol% cell lysate in HEPES buffer, 0.5 µm G‐actin and 1 mm ATP at room temperature (Figure 2ai and Movie panel S1). Only time points of 30 min or longer were considered, since F‐actin formation in the presence of ActA is expected to start only after ≈20 min [42]. The trajectories of ^0.5JS‐ActA were directional and became slightly longer over time, while ^0.5JS‐BSA displayed shorter and non‐directional trajectories. The MSD plots corroborated this information, showing a parabolic trend that increased over time for ^0.5JS‐ActA, cluster whilst a linear trend was observed for ^0.5JS‐BSA (Figure S4b). The whisker plots of the effective diffusion coefficient (D _eff) for ^0.5JS‐ActA and ^0.5JS‐BSA showed a clear difference, where D _eff increased over time for ^0.5JS‐ActA (from ≈0.60 to ≈0.75 µm² s⁻¹ together with a broadening of the D _eff distribution within 60 min) but remained at a constant low level for ^0.5JS‐BSA (≈0.40–0.55 µm² s⁻¹, Figure 2aii).

FIGURE 2.

Comparison of the locomotion of ^0.5JS‐ActA, ^0.5S‐ActA and ^0.5S^B‐ActA (10 vol% lysate in HEPES buffer, 0.5 µm G‐actin, 1 mm ATP, RT). Mobility of a) ^0.5JS‐ActA (and ^0.5JS‐BSA), b) ^0.5S‐ActA (and ^0.5S‐BSA) and c) ^0.5S^B‐ActA (and ^0.5S^B‐BSA) represented as trajectory maps (i) and whisker plots (ii) (purple: ActA‐coated motors; gray: BSA‐coated motors). n = 2, at least 100 particles were analyzed per repeat, *p‐value ≤ 0.001.

2.2.2. Locomotion Comparison of ^0.5JS‐ActA to ^0.5S‐ActA and ^0.5S^B‐ActA

The lead conditions tested for ^0.5JS‐ActA were employed to assess the locomotion of ^0.5S‐ActA and ^0.5S^B‐ActA. Specifically, either ^0.5S‐ActA or ^0.5S^B‐ActA were mixed in a solution made of 10 vol% lysate in HEPES buffer, 0.5 µm G‐actin and 1 mm ATP, and transferred into uncoated µ‐Slides VI^0.4 microfluidic channels, and their motion was recorded and analyzed as outlined before. ^0.5S‐BSA and ^0.5S^B‐BSA were used as controls.

First, the trajectories of ^0.5S‐ActA became shorter over time and their mobility seemed impaired compared to ^0.5JS‐ActA (Figure 2bi and Movie panel S2). Quantitatively, the MSD plots displayed a milder slope (Figure S5a) and the whisker plots showed a D _eff ≈0.15 µm² s⁻¹ (2× lower than for ^0.5JS‐ActA) (Figure 2bii), which confirmed the slower motion. A closer look at the videos revealed that ^0.5S‐ActA began to gather during G‐actin polymerization, resulting in clusters of ^0.5S‐ActA with limited mobility. This observation was consistent with what others have reported for L. monocytogenes aggregation to form bacterial biofilms [50]. In addition, we previously observed clustering in our earlier reported polystyrene‐based ActA motors, which had a higher D _eff of ≈0.25 µm² s⁻¹ under similar conditions [34]. This increased diffusivity relative to ^0.5S‐ActA is consistent with polystyrene's ≈2× lower density compared to silica. However, the clustering required the polymerization of G‐actin in all cases because no ^0.5S‐ActA clustering was observed when no ATP was present. It should be noted that ^0.5S‐ActA clustering was independent of temperature, G‐actin concentration, the viscosity of the medium, or particle size, if G‐actin and ATP were present (Figure S6a). In addition, ^0.5S‐BSA did not show any clustering, and their Brownian motion did not change, which was quantified in the whisker plots, displaying D _eff ≈0.30 µm² s⁻¹.

Second, ^0.5S^B‐ActA was analyzed to determine the impact of the higher amount of ActA present in the polymer brushes on the locomotion. Surprisingly, ^0.5S^B‐ActA showed directed trajectories that became longer over time, in contrast to what was observed for ^0.5S‐ActA or the controls and despite the absence of an inherent asymmetry (i.e., a Janus shape, Figure 2ci and Movie panel S3). A similar observation was reported by Cameron et al. [42] and us [34], attributed to rapid polymerization occurring on specific regions of the motor surface, which preferentially promotes F‐actin elongation. Additionally, ^0.5S^B‐ActA did not show any kind of clustering, likely due to the steric repulsion stemming from the polymer brushes (Figure S5b). The MSD plots (Figure S6b) and the whisker plots (Figure 2bii) of ^0.5S^B‐ActA confirmed locomotion over time, displaying a parabolic trend typical of directed motion, and a D _eff that increased over time from ≈0.7 up to ≈0.9 µm² s⁻¹, while D _eff of ^0.5S^B‐BSA remained constant at ≈0.1–0.2 µm² s⁻¹.

Altogether, ^0.5S^B‐ActA moved the fastest from all the tested motors, establishing protein‐in‐a‐polymer brush coated motors as lead candidates for enhanced mobility, which we have also previously observed for collagenase‐driven motors [25].

2.3. pH‐Dependent Clustering of ^0.5S‐ActA

L. monocytogenes biofilm formation has been proposed to depend on pH, but experimental confirmation of pH effects on biofilm per se is confounded by the fact that low pH also negatively impacts bacterial viability [50, 51, 52]. With the aim of gaining insight into the clustering behavior of ^0.5S‐ActA, we polymerize G‐actin in different pH environments. We suspended ^0.5S‐ActA in a solution containing the necessary components for G‐actin polymerization as outlined above and adjusted the pH to either 7.4 (physiological pH and above ActA's pI), 5.2 (in the range of ActA's pI) or 1 (below ActA's pI). We should note that at a pH like the ActA's pI, we expected an overall neutral charge of the ActA amino acid residues, resulting in aggregation similar to what L. monocytogenes exhibit. In general, G‐actin polymerization is expected to be optimal at neutral or slightly basic pH due to the overall stability of all proteins participating in the process, and the ATP‐G‐actin binding is favored. At slightly acidic pH (pH 5.2), G‐actin polymerization still occurs, but slower, and the F‐actin filaments might be less stable. At very acidic pH (pH 1), denaturation of G‐actin and other auxiliary proteins occur, resulting in no F‐actin formation. First, we visualized the F‐actin formed in the different pH environments in the presence of ^0.5S‐ActA (Figure 3a). Fluorescence intensity line scans on the SR‐CLSM images were used to semi‐quantify the amount of the produced F‐actin, keeping all the imaging settings constant. Surprisingly, the signal of F‐actin_AT polymerized from G‐actin_AT by ^0.5S‐ActA at pH 5.2 was ≈3× higher than when the F‐ actin_AT was produced by ^0.5S‐ActA at pH 7.4. This observation was attributed to the conversion of G‐actin into F‐actin at a pH close to the pI (pH 5.2), as observed by others [53], together with the fact that ^0.5S‐ActA tended to cluster much more at this pH, which could result in an apparent higher fluorescent signal. However, we would like to note that it was not possible to determine whether the fluorescent signal originated from F‐actin or G‐actin aggregates on the surface of the motors, especially for pH 5.2. In contrast, ^0.5S‐ActA at pH 1 gave a F‐actin_AT signal that was ≈10‐fold lower compared to the signal for F‐actin_AT associated with ^0.5S‐ActA at pH 7.4, indicating very limited to no polymerization of G‐actin, reflecting the fact of G‐actin and other proteins being unstable at this low pH.

FIGURE 3.

pH‐dependent clustering of ^0.5S‐ActA. a) Representative SR‐CLSM images (top panel) and fluorescence intensity line scans (bottom panel) of F‐actin polymerization from ^0.5S‐ActA at pH 7.4, 5.2 and 1.0 (10 vol% lysate in HEPES buffer, 0.5 µm G‐actin, 1 mm ATP, RT). The inset is a zoom‐in of the dotted rectangle. Magenta: G‐actin_AT. b) Representative bright field images (i) and pixel distribution curves (ii) of ^0.5S‐ActA reversible clustering at different pH. Insets: Schematics of ^0.5S‐ActA in a cluster at pH 7.4 (yellow background) and pH 5.2 (orange background) as well as individuals at pH 1 (pink background). c) Representative bright field microscopy images and the corresponding fractal analysis of ^0.5S‐ActA at different pH after 30 min. (n = 2).

Clustering of ^0.5S‐ActA was observed at pH 7.4 and 5.2, while ^0.5S‐ActA was completely dispersed at pH 1 as expected due to the absence of G‐actin polymerization (Figure 3bi). Specifically, at pH 5.2, ^0.5S‐ActA had very limited mobility and seemed to form 3D branched structures, suggesting an overproduction of F‐actin that bridged between individual ^0.5S‐ActA. In order to quantitatively evaluate this clustering, a pixel distribution analysis was performed for ^0.5S‐ActA in the different pH environments (Figure 3bii). The pixel analysis collected all white pixels of a binary image and performed a normal distribution profile based on the number of pixels gathered and the brightness (contrast) of the image. Please note that the center of the curves only indicated the brightness of the picture and had no relevance for the comparative analysis. The variance (or width) of the curves was used to interpret these data. Broader curves suggested a larger heterogeneity of the image due to the averaging of low‐intensity pixels (i.e., single particles) and high‐intensity pixels (i.e., groups of particles), whereas narrower distributions meant the opposite, namely a homogeneous distribution of particles across the image. For instance, the pixel analysis of ^0.5S‐ActA at pH 7.4 showed a broad distribution (Figure 3bii, yellow bars) that became narrower at pH 1 (Figure 3bii, pink bars), meaning that the clusters of motors formed at pH 7.4 but dispersed at pH 1. This observation was in agreement with the bright field images in Figure 3bi, and consistent within all the samples at different pH conditions.

Finally, a fractal analysis was performed on the bright field images to quantitatively determine the dimensionality of the clusters (Figure 3c). Specifically, ^0.5S‐ActA was considered after 30 min G‐actin polymerization at pH 7.4 and 5.2, and pH 1 was used as a control (since no G‐actin polymerization occurred). The fractal analysis considers two main parameters, namely lacunarity, i.e., the heterogeneity of the agglomerates, and the fractal dimension, i.e., 1D, 2D, or 3D. Higher values of lacunarity and fractal dimension point towards large heterogeneous structures with 3D arrangements. The fractal analysis revealed that ^0.5S‐ActA at pH 5.2 met these criteria, followed by ^0.5S‐ActA at pH 7.4. On the other end, ^0.5S‐ActA at pH 1 showed the lowest lacunarity values, indicating more homogeneous samples. In other words, ^0.5S‐ActA at pH 5.2 displayed a fractal‐like arrangement of the particles upon clustering, closely followed by ^0.5S‐ActA at pH 7.4, whilst ^0.5S‐ActA at pH 1 did not show any kind of fractal arrangement. Such pH‐dependent differences in cluster architecture are highly relevant for bottom‐up artificial cell design, as they demonstrate that simple environmental cues can be used to program cytoskeletal organization and potentially drive the emergence of higher‐order actin structures inside artificial cells.

2.4. Environmental pH‐Dependent Locomotion

Our next aim was to assess the mobility of ^0.5S‐ActA in the different pH environments. However, the high number of motors, especially when clustering, made our conventional tracking protocols insufficient. Therefore, we decided to introduce 4 µm polystyrene particles (⁴P) as tracers to observe the mobility of the densely packed ^0.5S‐ActA. ⁴P themselves had very low random motion per se, which became even more limited over time (Figure S7a). Our hypothesis was that if ^0.5S‐ActA were able to generate convection due to enhanced mobility, this would be reflected in an enhanced motion of ⁴P. We tracked the mobility of ⁴P exposed to ^0.5S‐ActA at pH 7.4, 5.2, and 1 at times 20 min (noted as t = 0 h) and 4 h later, and we employed ^0.5S‐BSA as controls. The trajectory maps of ⁴P showed very short and slow trajectories at times 0 and 4 h (Figure S7bi,ii). However, the MSD analyses revealed an increase in diffusivity when ^0.5S‐ActA were employed compared to ^0.5S‐BSA, which was even more evident after 4 h (Figure S7biii). Whisker plots were used to quantify these data (Figure 4a). No pH dependent differences were found when ^0.5S‐BSA were employed. In contrast, ⁴P exhibited higher D _eff values when exposed to ^0.5S‐ActA at pH 7.4 and at pH 1, while D _eff of ⁴P remained very low when exposed to ^0.5S‐ActA at pH 5.2 after 4 h. The latter observation was expected since ^0.5S‐ActA formed large and entangled clusters at this pH. In contrast, we attributed the increase of D _eff for ⁴P exposed to ^0.5S‐ActA at pH 1 after 4 h to a strong electrostatic repulsion due to the protonation of the amino acid residues, combined with protein denaturation at this low pH, which increased the Brownian motion of ^0.5S‐ActA. These two aspects together likely increased the Brownian motion of ⁴P. However, we found it quite surprising how high the D _eff for ⁴P exposed to ^0.5S‐ActA at pH 7.4 was. A careful mobility analysis of ^0.5S‐ActA at the interface with ⁴P was carried out with the aim of understanding the interaction in more detail. It has been reported that active particles had unique behavior when exposed to boundaries, such as walls, steps, or larger passive particles, e.g., latex beads and lipid vesicles [54]. For instance, others have shown that active particles orbited around larger tracers particles based on their motion dynamics and size, which governed the residence time (i.e., the time the active particle interacts with the surface of the tracer particle) and the scattering (i.e., the probability of the active particle to leave the orbit) [55, 56]. Similar examples have also been reported considering the interaction of active colloids with soft matter tracers, such as giant vesicles [57, 58] or droplets [58, 59]. In our case, the trajectories of ^0.5S‐ActA were analyzed in an area ca. 20 × 15 µm² around ⁴P for 20 s (Figure 4b and Movie panel S4). The results showed on the one hand that few ^0.5S‐ActA interacted with ⁴P at pH 7.4, probably due to the formation of clusters that limited the mobility of ^0.5S‐ActA to encounter ⁴P. Those ^0.5S‐ActA that were able to interact with ⁴P remained orbiting around it throughout the duration of the experiment. We hypothesized that this was because G‐actin could not polymerize from the hemisphere of ^0.5S‐ActA that was in close contact with ⁴P. Simultaneously, G‐actin polymerized from the free hemisphere (i.e., the one where ^0.5S‐ActA did not interact with ⁴P), making the motor propel continuously. This continuous polymerization created enough energy to overcome the adhesion forces between ^0.5S‐ActA and ⁴P, making ^0.5S‐ActA orbit around ⁴P (schematics in Figure 4b). On the other hand, ^0.5S‐ActA at pH 5.2 showed limited to no interaction with ⁴P, which was expected due to their fast cluster formation at this pH, which left few individual ^0.5S‐ActA free to interact with ⁴P. Finally, ^0.5S‐ActA at pH 1 demonstrated the highest number of interactions with ⁴P, as well as their capacity to escape the orbital motion, likely due to the absence of F‐actin formation that restrained the motion to the surface of ⁴P. Conversely, ^0.5S‐BSA docked at the surface of ⁴P independent of the pH (Figure S7c and Movie panel S4). We believed that this behavior was related to the fact that ^0.5S‐BSA, which did not show self‐propulsion, displayed a strong adhesion force towards ⁴P via the non‐specific binding of BSA, which immobilized them on the ⁴P's surface. Altogether, these different kinds of interactions were responsible for the observed D _eff values of ⁴P.

FIGURE 4.

Motion and interaction of ^0.5S‐ActA with ⁴P tracer particles. a) MSD plots of ⁴P tracer particles in the presence of ^0.5S‐ActA at different pH at times 0 and 4 h. For the whisker plots, ≤ 10 ⁴P tracers were analyzed. *p‐value ≤ 0.001; ns: not significant. b) Representative ^0.5S‐ActA trajectories’ overlays in the proximity of ⁴P (indicated by the dotted, white circles) at different pH (top panel) and corresponding schematics (bottom panel). The heat map represents the time ^0.5S‐ActA moved. (n = 2).

3. Conclusion

We have evaluated the deposition of actin‐recruiting proteins (ActA) on the surface of silica particles with the aim of assembling motors that exhibited enhanced locomotion due to actin polymerization. Grafting‐from of poly(diethylaminoethylmethacrylate) polymer brushes on silica particles increased the ActA deposition and resulted in a ≈5× enhanced mobility of these motors compared to their Brownian motion. We have also demonstrated that ^0.5S‐ActA formed clusters in a pH‐dependent manner. Further, the motors interacted with passive tracer particles typically by orbiting around them when they were individuals, or by trapping the tracers within the F‐actin network when clustering.

The next steps for these motors will evaluate whether reversible clustering behavior dependent on pH is feasible in order to create dynamic assemblies inside confined spaces. On a different note, the grafting‐from of polymers in Janus (asymmetric) particles could be useful to attain high‐speed motors by combining the high‐load of proteins within the polymer brushes with a break in symmetry, which would promote directionality.

Overall, this bacteria‐inspired class of motors that employ biopolymerization reactions to trigger locomotion is a versatile and powerful alternative to existing motors owing to the great variety of biological monomers that could be used.

4. Experimental Section

4.1. Materials

Sodium chloride (NaCl, 99%), poly(L‐lysine) hydrobromide (PLL, MW 30–70 kDa), poly(sodium 4‐styrenesulfonate) (PSS, MW 70 kDa), 4‐(2‐hydroxyethyl)piperazine‐1‐ethane‐sulfonic acid (HEPES), hexadecyltrimethylammonium bromide (CTAB), tetraethylorthosilicate (TEOS), (3‐aminopropyl)triethoxysilane (APTES), ammonium hydroxide (NH₄OH, 29 wt%), rhodamine B (λ _ex/em = 546/568 nm), bovine serum albumin (BSA), fluorescein isothiocyanate labelled bovine serum albumin (BSA_FITC, λ_ex/em = 495/515 nm) were purchased from Sigma‐Aldrich. Methoxypolyethylene glycol carboxylic acid (mPEG‐COOH, MW 2 kDa) was purchased from Abbexa Ltd. Actin assembly‐inducing protein (ActA) was expressed from Escherichia coli by GenScript. G‐actin and ATTO 647 actin were purchased from Hypermol. Human Large Vessel Endothelial Cell Basal Medium, Large Vessel Endothelial Supplement (LVES), Pen‐strep (100 µg mL⁻¹ streptomycin, and 100 U mL⁻¹ penicillin), Pierce Protease Inhibitor Mini Tablets, Pierce BCA Protein Assay Kit, Bolt sample reducing agent, Bolt LDS sample buffer, 4–12% Bis‐Tris polyacrylamide gel, Nitrocellulose/Filter Paper Sandwich, 0.2 µm, 8.3 × 7.3 cm, Bolt Transfer Buffer (20×), 1‐Step Ultra TMB‐Blotting Solution, and Oregon Green 488 maleimide (λ_ex/em = 496/524 nm) were available from Thermo Fisher. Paraffin wax was commercially available from Merck KGaA (Germany). µ‐Slides VI^0.4 uncoated were purchased from ibidi GmbH. Cell lysis buffer was purchased from Cell Signaling Technology. Prestained Protein Ladder—Broad molecular weight (10–245 kDa), Anti‐Arp2 antibody (ab128934), anti‐Actin antibody (ab124964), Anti‐Cofilin antibody (ab42824), and ATP Assay Kit (Colorimetric/Fluorometric) were purchased from Abcam. Ultrapure water (18.2 MΩ cm resistivity) was provided by a Synergy & Synergy UV water purification system (Merk Millipore).

Poly(L‐lysine)‐grafted‐poly(ethylene glycol) (PLL‐g‐PEG, grafting ratio ≈17%, i.e., 1 out of 6 PLL chains contained 1 PEG chain) was synthesized according to a previously published protocol [60].

HEPES buffer consisted of 10 mm HEPES and 150 mm NaCl dissolved in ultrapure water and adjusted to pH 7.4. Phosphate buffer consisted of 2.5 mm Na₂HPO₄ and 50 mm NaCl dissolved in ultrapure water and adjusted at pH 5.2 or pH 1.

ActA protein (stock concentration 0.67‐0.85 mg mL⁻¹, 30% purity) was dissolved in 50 mm Tris‐HCl buffer, containing 500 mM NaCl and 10% glycerol adjusted to pH 8.

G‐actin and ATTO 647 actin (stock concentration 24 µM) were reconstituted according to the manufacturer protocol, resulting in a solution of 2 mm Tris‐HCl buffer at pH 8.2, containing 0.4 mm ATP, 0.5 mm DTT, 0.1 mm CaCl₂, 1 mm NaN₃ and 0.3% disaccharides.

Oregon Green 488 labelled ActA (ActA_OG) was prepared as described in a previous publication [34].

4.2. Silica (SiO₂) Particle Synthesis

SiO₂ nanoparticles of ≈500 nm in diameter were made by a modified Stöber method [36]. Briefly, 2.72 mL of ultrapure water and 15.8 mL of absolute ethanol were mixed in a 250‐mL Erlenmeyer flask. Next, 5 mL of NH₄OH (28 wt%) and 1.4 mL TEOS were added, and the solution was stirred (800 rpm) for 2 h at RT, resulting in ^0.5S. The colloids were collected and washed by centrifugation in ethanol (6500 rpm, 20 min, 3×). The final suspension was stored in ethanol at 4 °C, and the particle concentration was determined after drying a portion of the sample and calculated to be ≈15 mg mL⁻¹.

Transmission electron microscopy (TEM) images were taken in a Tecnai G2 Spirit working at an acceleration voltage of 120 kV. The samples were first cast on a Formvar‐carbon grid (300 mesh, EMS) and dried in air.

4.3. Homogeneous Motor (^0.5S‐ActA)

50 µL of ^0.5S were transferred to HEPES buffer and 100 µL ActA (0.67‐0.85 mg mL⁻¹) were added and let to incubate for 20 min at room temperature. Next, the colloids were washed by centrifugation in HEPES buffer (6500 rpm, 3 min, 3×), resulting in ^0.5S‐ActA. Alternatively, 30 vol% of the protein solution contained ActA_OG to obtain ^0.5S‐ActA_OG. BSA (or BSA_FITC) was used instead of ActA to obtain the control motors ^0.5S‐BSA or ^0.5S‐BSA_FITC. All the samples were stored at 4 °C and used within 24 h.

4.4. Preparation of Janus Particles

Janus (asymmetric) particles were made by a Pickering emulsion process [37, 61]. In short, 200 µL of ^0.5S stock dispersion (15 mg mL⁻¹) were added to 0.1 mm CTAB and stirred (250 rpm) in a water bath at 80 °C for 5–10 min. Next, 3 pellets (≈120 mg) of paraffin wax were added and allowed to dissolve. Subsequently, the mixture was stirred at 2000 rpm for 10 min and let to cool down at 4 °C. The wax droplets containing ^0.5S were washed with ethanol and dried under vacuum, resulting in wax‐^0.5JS.

4.5. Janus Motor (^0.5JS‐ActA)

≈1 mg of wax‐^0.5JS was exposed to PLL‐g‐PEG (2 mg mL⁻¹ in HEPES buffer) for 20 min and subsequently washed with ultrapure water and dried under pressure. Next, the wax was removed by centrifugation and resuspension in chloroform (2–3×), ethanol (2×) and HEPES buffer (2×) (6500 rpm, 3 min). Finally, the exposed hemisphere was coated with ActA (with or without 30 vol% ActA_OG) (0.67–0.85 mg mL⁻¹) or BSA (with or without 30 vol% BSA_FITC) (1 mg mL⁻¹) for 20 min at RT and washed by centrifugation in HEPES buffer (6500 rpm, 3 min, 3×), yielding ^0.5JS‐ActA or ^0.5JS‐BSA. All the samples were stored at 4 °C and used within 24 h.

4.6. Grafting‐from of DEAEMA from the SiO₂ Particles

Polymer brushes were grown from the surface of the as‐synthesized SiO₂ particles as outlined in a previous publication [25, 38]. First, 1 mL of ^0.5S (50 mg mL⁻¹) was redispersed in 20 mL ethanol and transferred to a flask. APTES were added, keeping a 1:20 APTES to ethanol volume ratio, and heated at 56 °C for 2 h. The APTES‐modified particles were washed (6500 rpm, 5 min, 3×), redispersed in toluene, and then transferred to a clean flask. Trimethylamine was added (200 µL), followed by the careful addition of 150 µL α‐bromoisobutyryl bromide under stirring, to react overnight at room temperature. The functionalized ^0.5S were washed (6500 rpm, 5 min, 5×) and redispersed in ethanol. Next, the DEAEMA polymerization was carried out. To do so, 400 µL CuSO₄ (0.1 mg mL⁻¹), 200 µL sodium ascorbate (0.5 mg µL⁻¹), 16.7 µL Me₆TREN and 109.2 µL DEAEMA (0.7 m) were mixed with 200 µL of the surfaced‐modified ^0.5S (10 mg) and mixed for 6 h at room temperature. The samples were then dialyzed (3500 MWCO) against water for 3 d, washed in phosphate buffer at pH 1, and sonicated for 30 min to dissolve the polymer aggregates. Finally, the particles were washed in HEPES buffer at pH 5.2, resulting in ^0.5S^B.

4.7. Homogeneous Motor with Polymer Brushes (^0.5S^B‐ActA)

50 µL ^0.5S^B were transferred to HEPES buffer and 100 µL ActA (0.67‐0.85 mg mL⁻¹) were added and left to incubate for 20 min at room temperature. Next, the colloids were washed by centrifugation in HEPES buffer (6500 rpm, 3 min, 3×), resulting in ^0.5S^B‐ActA. Alternately, 30 vol% of the ActA solution contained ActA_OG, which resulted in ^0.5S^B‐ActA_OG. BSA (or BSA_FITC) was used to assemble the controls referred to as ^0.5S^B‐BSA or ^0.5S^B‐BSA_FITC. All the samples were stored at 4 °C and used within 24 h.

4.8. Motor Characterization

The protein deposition was characterized by super‐resolution confocal laser scanning microscopy (SR‐CLSM) using a Zeiss LSM800 Airyscan microscope (Carl Zeiss, Germany). At least five images were taken from different areas for each sample. The following settings were used: 63×/1.40 oil DIC M27 objective, 1168 × 1168 ppi resolution, imaging speed of 5 µm s⁻¹, 8‐line averaging and Airyscan SR mode (2D, auto). An excitation wavelength of λ_ex = 488 nm (2% laser power) was used to visualize the ActA_OG or BSA_FITC. Line scans were taken with the “Profile” tool from the software ZEISS ZEN Blue lite v3.9.

Fluorescence intensity spectra were recorded in a multiplate reader (EnSight, PerkinElmer), using fully coated ActA_OG and compared to the free protein in a solution.

4.9. Actin Polymerization

10 µL ^0.5JS‐ActA, ^0.5S‐ActA or ^0.5S^B‐ActA (and the corresponding controls ^0.5JS‐BSA, ^0.5S‐BSA or ^0.5S^B‐BSA) were redispersed in 100 µL cell lysate (50 vol% HEPES buffer), mixed with 5 µL G‐actin (24 µm, 30 vol% ATTO 647 actin) and incubated for 20–30 min at room temperature. Next, the particles were washed by centrifugation in HEPES buffer (6500 rpm, 3 min, 2×) and stored in HEPES buffer at 4°C.

TEM images were collected as previously described. SR‐CLSM images were taken in a Zeiss LSM800 Airyscan microscope (Carl Zeiss, Germany). At least five images were taken from different areas for each sample. The following settings were used: 63×/1.40 oil DIC M27 objective, 1168 × 1168 ppi resolution, imaging speed of 5 µm s⁻¹, 8‐line averaging and Airyscan SR mode (2D, auto). Excitation wavelengths of λ _ex = 488 nm (2% laser power) and λ _ex = 640 nm (3% laser power) were used to visualize the ActA_OG or BSA_FITC, and the ATTO 647 actin, respectively.

4.10. Optical Tracking

Uncoated µ‐Slides VI^0.4 microfluidic channels (Ibidi GmbH) were used to assess the locomotion of the motors. Specifically, 1 µL of ^0.5JS‐ActA, ^0.5S‐ActA, or ^0.5S^B‐ActA was mixed with 87 µL of cell lysate (10 vol.% lysate in HEPES buffer, unless specified otherwise), 10 µL ATP (10 mm) and 2 µL G‐actin solution (stock concentrations ranging from 240 to 2.4 µm) were added. The mixture was vortexed for some seconds and transferred into the microfluidic channel. BSA‐coated particles or ^0.5JS‐ActA in the presence of cytochalasin D (2 µL, 1 mg mL⁻¹) were used as controls. Movies were recorded after 5 min stabilization in an Olympus IX inverted microscope using a 60×/1.6× oil objective at a frame rate of 16.67 fps for 20 s. At least 2 movies were recorded per sample, and the experiments were repeated twice. For visualization purposes, the movies were accelerated 2×.

Experiments considering different medium viscosities were performed using ^0.5JS‐ActA and G‐actin (24 µm stock concentration) together with undiluted cell lysate or lysate diluted in HEPES buffer (50 vol% and 10 vol%).

300 frames were extracted per movie, and the trajectories were software tracked using the plugin TrackMate from Fiji [62]. The mean squared displacement (MSD) plots were obtained using a protocol published elsewhere with the software Matlab [63]. The MSD plots are related to the squared displacement (L ²) of a particle in a short time interval (Δt), and provide information about the type of motion (e.g., Brownian, enhanced diffusion, nano propulsion) the motors may display, upon fitting of the following (simplified) equation:

$$〈L^2〉=4D_{\mathrm{eff}}\mathrm{\Delta }t+v^2\mathrm{\Delta }{t}^2$$ (1)

where D _eff stands for the effective diffusion coefficient and v for the average velocity of the ensemble. For linear plots, the quadratic term becomes zero and they refer to particles governed by Brownian (stochastic) motion. In contrast, parabolic curves account for ballistic (directed) motion where the first part of the equation becomes negligible [64, 65]. Note that the error bars of the MSD plots become larger at longer Δt as fewer time points are considered, and the error accumulates.

The D _eff was calculated from the slope of the MSD plots and presented as whisker plots (box plots). In a whisker plot, the lower and upper limits refer to the first quartile (Q1, 25% of the data) and the third quartile (Q3, 75% of the data), respectively, whereas the line crossing the box represents the median (Q2, second quartile, 50% of the data) and the triangle related to the average. Outliers are represented as crosses and refer to the data points that exceed the lower or upper limits, according to the rule $\mathrm{outlier}\_\mathrm{low}=Q1-1.5(Q3-Q1)$ or $\mathrm{outlier}\_\mathrm{up}=Q3+1.5(Q3-Q1)$.

4.11. Clustering Experiments

The motors were mixed with 87 µL of cell lysate (10 vol% cell lysate in HEPES buffer), 10 µL ATP (10 mm) and 2 µL G‐actin solution (stock concentrations 24 µm) and allowed to react for 30 min before collection by centrifugation (6000 rpm, 2 min). Next, the motors were redispersed in a solution containing the same final concentrations of G‐actin and ATP, with the pH adjusted to either 7.4, 5.2, or 1.0. The pixel distributions were performed using the “Histogram” tool from Fiji.

4.12. Cell Work

The human umbilical vein endothelial cells (HUVEC) used in this study were purchased from Thermofisher Scientific. The cells were cultured in 75 cm² (TC‐treated) culture flasks in human large vessel endothelial cell basal medium at 37 °C and 5% CO₂. The medium was supplemented with large vessel endothelial supplement (LVES) and 1% Pen‐strep (100 µg mL⁻¹ streptomycin, and 100 U mL⁻¹ penicillin).

4.13. Cell Lysate Extraction

The HUVECs cultured in a 75 cm² culture flask were washed twice with PBS followed by 10 min incubation on ice with 1.5 mL of cell lysis buffer (Cell Signaling Technology, diluted 1:10 in MQ with 1 tablet of Pierce Protease Inhibitor Mini Tablets). Next, a cell scraper was used to detach the cells from the flask, and the cell suspension was transferred to a 1.5 mL Eppendorf tube. The cell suspension was centrifuged (10 000 rpm, 10 min, 4 °C) after which the supernatant containing the cell lysate was transferred to a clean Eppendorf tube. The cell lysate was stored in aliquots of 100 µL at ‐20 °C until use. We previously characterized the cell lysate in detail [34].

4.14. Artificial Intelligence Generated Content

The authors used the AI‐based language model ChatGPT developed by OpenAI for language optimization and assistance with phrasing in the preparation of this manuscript.

Funding

Carlsberg Foundation Distinguished Associate Professor Fellowship (B.S. CF16‐0233) and the Lundbeck Foundation.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File 1: marc70312‐sup‐0001‐SuppMat.docx.

Supporting File 2: marc70312‐sup‐0002‐MovieS1.mp4.

Supporting File 3: marc70312‐sup‐0003‐MovieS2.avi.

Supporting File 4: marc70312‐sup‐0004‐MovieS3.avi.

Supporting File 5: marc70312‐sup‐0005‐MovieS4.avi.

Supporting File 6: marc70312‐sup‐0006‐MovieS5.avi.

Supporting File 7: marc70312‐sup‐0007‐MovieS6.avi.

Supporting File 8: marc70312‐sup‐0008‐MovieS7.avi.

Supporting File 9: marc70312‐sup‐0009‐MovieS8.avi.

Supporting File 10: marc70312‐sup‐0010‐MovieS9.avi.

Data Availability Statement

The raw data for this manuscript is available upon request to the authors within reasonable time.