Cytoskeleton-functionalized synthetic cells with life-like mechanical features and regulated membrane dynamicity

Abstract

The cytoskeleton is a crucial determinant of mammalian cell structure and function, providing mechanical resilience, supporting the cell membrane and orchestrating essential processes such as cell division and motility. Because of its fundamental role in living cells, developing a reconstituted or artificial cytoskeleton is of major interest. Here we present an approach to construct an artificial cytoskeleton that imparts mechanical support and regulates membrane dynamics. Our system involves amylose-based coacervates stabilized by a terpolymer membrane, with a cytoskeleton formed from polydiacetylene fibrils. The fibrils bundle due to interactions with the positively charged amylose derivative, forming micrometre-sized structures mimicking a cytoskeleton. Given the intricate interplay between cellular structure and function, the design and integration of this artificial cytoskeleton represent a crucial advancement, paving the way for the development of artificial cell platforms exhibiting enhanced life-like behaviour.

The cytoskeleton is vital for living organisms. Now an artificial cytoskeleton within amylose-based coacervates stabilized by a terpolymer membrane has been developed using polydiacetylene fibrils to mimic natural cytoskeletal functions. The integration of this artificial cytoskeleton marks an important step in creating artificial cell platforms with enhanced life-like behaviour.

Main

The cytoskeleton of the mammalian cell is one of its key determining components. It not only provides structural resilience by regulating the cell’s mechanical properties and by supporting the cell membrane but also has an important scaffolding function and plays an essential role in cell division and motility¹. Due to its essential role in natural cells, the cytoskeleton has gained increasing interest in the field of artificial cell research. This field of science uses a bottom-up self-assembly approach to construct compartments with cell-like features. This is motivated by the rationale that the development of artificial cells will provide us with a better insight on how biological processes are organized in living cells^2,3. The focus of the community has been mainly on functional behaviour, such as communication^4,5, compartmentalized biosynthesis and metabolism^6,7, adaptation^8,9 and motility^10,11. Recently, more attention has been paid to the mimicry of structural aspects of living cells. For example, coacervates have garnered renewed interest as materials that mimic the crowdedness of the cell’s cytoplasm, which is intrinsically connected to functional behaviour, as crowdedness will affect diffusivity and interactions between encapsulated biomolecules¹². Furthermore, artificial organelles have been developed that have been incorporated in artificial cells to achieve hierarchical control over biological processes^13,14. Important advances have been reported with regard to the development of an artificial or reconstituted cytoskeleton^15–18. Examples include the observation of membrane deformations of lipid vesicles using a reconstituted actomyosin ring¹⁹, irreversible directional cargo transport along DNA filaments²⁰, the mimicry of actomyosin-based contraction with a composite material based on a temperature-responsive polymer (poly(N-isopropylacrylamide)) and F-actin²¹ and the restructuring of coacervate droplets via the gelation of actin²². In particular, the role of the cytoskeleton in regulating artificial cellular mechanical features remains an interesting field of research^23–25.

For this purpose, we developed an approach to construct an artificial cytoskeleton that has a mechanical support function for the artificial cell. Our artificial cytoskeleton has broad applicability and was successfully integrated into various cellular mimics, but especially in complex coacervates. By carefully modulating its hydrophobicity, the cytoskeleton could either be used to support the membrane, thereby regulating its dynamics, or be positioned inside the lumen of the coacervates, thereby inducing mechanical properties to the artificial cell, similar to living cells. Furthermore, its scaffolding role was demonstrated by the spatial and organizational control of cargo molecules. As cellular structure and function are intricately connected, this artificial cytoskeleton design is an important aspect to be included to achieve artificial cell platforms with improved life-like behaviour.

Results

Design of artificial cytoskeleton

To create an artificial cell platform with cytoskeletal function, we started from our recently developed amylose-based coacervate platform. Coacervate formation is attained by the coalescence of an excess of positively charged quaternized amylose (Q-Am) with negatively charged carboxymethylated amylose (Cm-Am), which are stabilized by the addition of a terpolymer that forms a semi-permeable layer around the coacervate²⁶. With this platform, we can mimic the crowdedness of the cytoplasm and the barrier function of the cell membrane. To install a cytoskeleton, we opted to include a fibrous network of polydiacetylenes (PDAs). Our specific PDA of interest is shown in Fig. 1a. These PDAs are a well-investigated system^27–29 in which self-assembly is governed by the interplay between the bisurea stabilizing hydrogen-bonding moieties and the hydrophobic interactions of the alkyl chains. Upon assembly, the diacetylenes are well positioned for topochemical polymerization, which allows covalent stabilization of the fibrous assembly.

Fig. 1. The build-up of the artificial cytoskeleton.

a, The chemical structure of diacetylene (DA) building blocks. Azido and DBCO functionality are depicted in green. b, Fibril assembly shown for a single ribbon. A fibril consists of several ribbons. The schematic shows the self-assembly of DA building blocks and their polymerization into PDA to afford stable fibrils. Azido and DBCO functionality are depicted as a green sphere. c, A cryo-TEM micrograph of PDA fibrils. Scale bar, 50 nm. d, PDA bundling. The schematic shows bundling of PDA fibrils induced by a positively charged polyelectrolyte. Left: the PDA fibril in detail. Right: an approximation of PDA fibrils (green) bundled by Q-Am (red). e, A confocal micrograph of bundled PDA. Scale bar, 5 µm.

The PDA is an excellent candidate for a biomimetic cytoskeleton as it matches the key features of a natural cytoskeleton, being nanometre-sized semi-flexible fibrils that become viscoelastic after physical entanglement²⁷. To endow the PDAs with the ability to be taken up in the net positively charged coacervates and to be bundled into filamentous structures, carboxylate end groups were installed. Furthermore, for cargo loading and for control over the polarity of the fibrils, DAs with clickable units were included via a co-assembly approach (Fig. 1b).

The polymerization of the diacetylene moieties was followed by ultraviolet–visible spectroscopy (Supplementary Fig. 1). Before reaction, the monomers showed no absorbance in the visible spectrum (400–700 nm). During the polymerization of the diacetylene units, a covalent connection of alternating double and triple bonds was formed between the monomers, resulting in the delocalization of electrons, which led to an increase in absorption in the visible spectrum. After irradiating the sample with ultraviolet light (λ = 254 nm) for 35 min, the absorption plateaued, suggesting the completion of the polymerization. The successful formation of fibrils with the newly designed functionalities was verified by cryogenic transmission electron microscopy (cryo-TEM; Fig. 1c). In the case of a PDA consisting of only carboxylate-terminated DA, the mean contour and persistence lengths were determined to be 163 ± 54 and 161 ± 35 nm, respectively. The thickness was determined to be 5.8 ± 0.8 nm. Comparable data were obtained for the co-assembled variants, which, depending on the ratio of different terminal groups, led to fibrils with lengths ranging from 100 to 400 nm (Supplementary Fig. 2).

As mentioned above, the carboxyl-terminated DA had two main functions. On the one hand, it enabled the uptake of the PDA fibrils into the positively charged coacervates by electrostatic interactions. On the other hand, the carboxylate end group facilitated the aggregation of the PDA fibrils into micrometre-sized bundles.

We performed experiments investigating the uptake of PDA fibrils into coacervates (Supplementary Fig. 20). Here, we investigated PDA fibrils containing only azide end groups, only carboxylate end groups and the mixture, which was used for the rest of this Article (90 wt% carboxylate and 10 wt% azide end groups). Fibrils containing only azide end groups mostly aggregated outside of coacervates, whereas bundles containing only carboxylate end groups were taken up effectively. For co-assembled azide and carboxylate PDA bundles, we observed a similar behaviour as seen for the PDA bundles based only on carboxylates.

The nanometre-sized fibrils were assembled into micrometre-sized bundles to affect mechanical properties at the micrometre scale. This hierarchical aggregation was facilitated by interaction of the negatively charged carboxyl-terminated DA with the positively charged polyelectrolyte (Q-Am) (Fig. 1d). The formation of micrometre-sized entanglements was observed with confocal laser scanning microscopy (CLSM). Before the addition of Q-Am, the homogeneous distribution of fibrils below the diffraction limit (nanometre scale) prevented observation with CLSM. The addition of Q-Am to PDA immediately resulted in the formation of densely entangled micrometre-sized networks of bundles (Fig. 1e). To demonstrate that bundling as a result of the interaction of the PDAs with positively charged polymers was a robust process, we formed PDA bundles into giant unilamellar vesicles, water-in-oil droplets and membrane-stabilized coacervates by simply ensuring the presence of both the PDA fibrils and Q-Am (Supplementary Fig. 3). The aggregation was also induced by other positively charged polyelectrolytes such as poly(l-lysine) but was observed neither for negative electrolytes (CM-Am) nor for uncharged crowding agents such as dextran, Ficoll 400 or poly(ethylene glycol) (PEG) (Supplementary Figs. 4 and 5).

Spatial assembly of PDA-based cytoskeleton inside coacervates

After demonstrating the bundling capacity, we focused on the positioning of the PDA in the coacervates. To engineer a biomimetic cytoskeleton that can affect both the membrane as well as the bulk properties of a cell, we tried to find a way to control the interaction of the PDA with the terpolymer membrane. Similar to lipid membranes the terpolymer membrane has a hydrophobic domain that interacts with hydrophobic molecules. At the same time, we also wanted to introduce an easy way for post-modification after cytoskeleton formation to enable the scaffolding of functional materials. Therefore, we chose the dibenzocyclooctyne (DBCO) moiety, which can be utilized for copper-free strain-promoted azide-alkyne cycloaddition while also being sufficiently hydrophobic to enable localization on the terpolymer membrane. For the same dual reason, we introduced terminal azides on the DA as they are less bulky and less hydrophobic and would therefore lead to the formation of a cytoplasmatic cytoskeleton with conjugation ability. To combine these features with the ones offered by the carboxyl-terminated DA (uptake and micrometre-scale aggregation), 10% of either DBCO- or azide-functional DA was mixed with 90% of the negatively charged carboxyl-terminated DA. Even though charged and bulky terminal groups can have negative influences on the self-assembly, because they increase the repulsion between individual fibrils, fibrils were formed in both cases in the range mentioned before.

When mixing either hydrophobic (DBCO, PDA-M) or hydrophilic PDA (azide, PDA-L) together with the amylose derivatives, coacervates were formed in which both PDA bundles were efficiently taken up and homogeneously distributed in the artificial cells (Supplementary Fig. 6). After the addition of terpolymer and the subsequent membrane formation, a differentiation in localization of hydrophobic and hydrophilic PDA took place. Confocal microscopy clearly showed that PDA-M was associated with the membrane, forming micron-sized entanglements (Fig. 2b,d), while the PDA-L remained evenly distributed in the lumen, forming a widespread network of densely entangled micrometre-sized bundles (Fig. 2c,e). This proved to be robust behaviour, as this effect was observed for the large majority of the population as quantified by line profile fluorescence measurements of the PDA bundle distribution (Supplementary Fig. 7).

Fig. 2. Spatial assembly of the cytoskeleton in artificial cells.

a, A schematic of PDA fibril introduction into coacervates. (i) Mixing of amyloses and PDA fibrils results in membrane-less coacervates with PDA bundles. (ii) Addition of terpolymer leads to the formation of a membrane around the coacervate. (iii) Hydrophobic PDA bundles (PDA-M) are associated to the membrane. (iv) Hydrophilic PDA bundles (PDA-L) are distributed in the lumen. Q-Am is depicted in red, CM-Am in blue, PDA bundles in green and terpolymer membrane in grey. b, A confocal micrograph of PDA-M bundles encapsulated in terpolymer-stabilized coacervates. Scale bar, 10 µm. c, A confocal micrograph of PDA-L bundles encapsulated in terpolymer-stabilized coacervates. Scale bar, 10 µm. d,e, A 3D reconstruction of confocal z-stack micrographs of PDA-M (d) and PDA-L (e) bundles encapsulated in terpolymer-stabilized coacervates. Top view (left), side view of merged channels PDA and membrane (top right) and side view of membrane channel (bottom right). Cyan represents the terpolymer membrane, and magenta represents the PDA bundles. Scale bar, 5 µm.

Next, we quantitively determined the encapsulation efficiency of PDA into coacervates using fluorescence spectroscopy. PDA-L was taken up into the coacervates with an efficiency of 50%, reaching a maximum concentration of 40 mM related to the diacetylene building blocks (Supplementary Fig. 8). The concentration of the fibrils depended partly on their size but was limited mainly by the negative charge introduced through the terminal carboxylate. At higher PDA concentrations, the coacervates could not compensate the increasing negative charge anymore and remained disassembled (Supplementary Fig. 9). PDA-M was also effectively taken up into coacervates. At lower concentrations, this variant associated exclusively with the membrane until approximately 30% of the membrane surface was covered. At higher concentrations, the excess of PDA that did not associate with the membrane accumulated in the lumen.

Cytoskeletal control over enzyme positioning and activity

A key feature of the cytoskeleton, which is essential to its function, is its ability to organize molecules spatially. It can aid in regulating complex cellular functions by facilitating the interaction of proteins that are scaffolded on the cytoskeleton. To show the benefit of spatial organization and to demonstrate the scaffolding effect of the cytoskeleton, we developed a co-assembly assay based on the Nanoluc luciferase. Luciferase creates luminescence by catalysing the oxidation of its substrate furimazine. By splitting luciferase into two inactive subunits, luminescence is observed only when the original protein is reconstituted. The final protein concentration and the substrate concentration were kept constant throughout the experiments. Therefore, an increase in luminescence was a measure of the improved spatial organization and reconstitution of the protein when arranged on the PDA scaffold. The split luciferase consisted of a larger 18 kDa fragment LgBiT-c-Raf S233/S259-Histidine (LgBiT-His) and a smaller fragment called mNeonGreen-SmBiT₁₀₁-His (mNeonGreen-SmBiT-His). The smaller fragment was fused to the fluorescent protein mNeonGreen to enable its visualization by fluorescence microscopy. As mNeonGreen absorbs the luminescence generated by the luciferase, bioluminescence resonance energy transfer (BRET) occurred, and the green luminescence was used as readout to determine the efficiency of colocalization leading to protein reconstitution (Fig. 3a).

Fig. 3. Spatial assembly of proteins on the artificial cytoskeleton.

a, A scheme of the Nanoluc luciferase reconstitution assay. The enzyme is split into two parts: a large domain and a small domain (LgBiT and SmBiT, respectively). SmBiT is fused to a fluorescent protein, mNeongreen. Upon reconstitution by bringing the two domains in close proximity, luminescence capacity is restored, leading to BRET to mNeongreen. The emission of green light indicates the success of this process. b, BRET as quantified by luminescence spectroscopy. The data represent mean ± standard deviation of four technical replicates. The boxes extend from the 25th to the 75th percentiles, with a line at the median and a black square at the mean. The whiskers span 1.5× s.d. The data points are shown as white circles. c–e, A schematic (left) and confocal micrograph displaying mNeonGreen fluorescence (right) of mNeonGreen-SmBiT-His and LgBiT-His free in bulk (c), mNeonGreen-SmBiT-His and LgBiT-His conjugated to PDA fibrils by His-tag Ni²⁺–NTA interactions (d), mNeonGreen-SmBiT-His and LgBiT-His conjugated to PDA bundles by His-tag Ni²⁺–NTA interactions (e) and mNeonGreen-SmBiT-His and LgBiT-His conjugated to PDA bundles by His-tag Ni²⁺–NTA interactions encapsulated in coacervates (f). Scale bar, 25 µm. Panels a and c–e adapted from BioRender.com.

Both constructs were expressed with a His-tag. The azide-functional PDA-L fibril was modified via strain-promoted azide–alkyne cycloaddition with a DBCO-functionalized nickel nitrilotriacetic acid (Ni²⁺–NTA) moiety to enable the conjugation of both His-proteins to the fibril via His Ni–NTA interaction. Before we set out to investigate the attachment of this anchor, we first confirmed the reactivity of the azide and DBCO functionalities introduced on the fibre bundles by covalently attaching dyes, DNA or proteins (Supplementary Fig. 10).

The dissociation constant (K_D) for LgBiT and SmBiT amounts to 2.5 µM according to ref. ³⁰. For the protein reconstitution experiments (Fig. 3b), we chose a protein concentration of 33 nM, which is well below the K_D of this construct. Hereby we ensured that the subunits remained separated when in bulk (Fig. 3c). Indeed, under these conditions we observed low luminescence (Fig. 3b), suggesting that the majority of the subunits are separated and thereby inactive. When mixing Ni²⁺–NTA-modified PDA with the His-tagged protein subunits, both were spatially organized on the fibril (Fig. 3d), enabling successful protein reconstitution, resulting in a 620-fold increase in luminescence. Through the addition of the positively charged polyelectrolyte Q-Am, bundling of the individual fibrils was induced (Fig. 3e). This led to a further increase in luminescence, suggesting an even higher degree of spatial organization. When the bundled PDA were formed in the coacervates (Fig. 3f) we observed a 135-fold increase in luminescence compared with the non-scaffolded proteins, demonstrating the robustness of the scaffolding function of PDAs in the complex environment found in a coacervate. The reduced luminescence in coacervates compared with PDA bundles in bulk conditions can be explained by a lower overall enzyme concentration caused by the sample preparation, in which non-encapsulated enzymes and PDA were washed away, and by scattering effects observed with the coacervates. Overall, we demonstrate that our artificial cytoskeleton can effectively act as a scaffold for the spatial organization of proteins to ensure their functional reconstitution.

Manipulation of membrane dynamics

Besides a scaffolding function, cytoskeletons also play a key role in regulating membrane dynamics. We therefore investigated with fluorescence recovery after photobleaching (FRAP) how the PDA-M-based cytoskeleton, which formed a fibrous network around the membrane, could regulate the lateral diffusivity of the terpolymer membrane components (Fig. 4a). Three samples were prepared: one sample contained no PDA (referred to as empty) whereas the other two contained an amount of PDA equivalent to a membrane surface coverage of 5–10% (Fig. 4b and Supplementary Fig. 11) and 30–35% (Fig. 4c and Supplementary Fig. 11), respectively. From our previous work, we know that the terpolymer membrane is dynamic, similar to lipid membranes³¹. This was again confirmed for this system, as a diffusion coefficient of the terpolymer membrane of 0.068 µm² s⁻¹ and an immobile fraction of around 20% were measured for a regular 25-µm-sized coacervate with no PDA. In the case of PDA-supported membranes, we observed a 3- and 6-fold deceleration of the membrane diffusivity and a 2- and 2.5-fold increase of the immobile fraction with increasing PDA concentration relative to coacervates without any PDA. Lastly, to determine whether the attachment of the artificial cytoskeleton to the membrane is necessary to affect membrane dynamics, we conducted a control experiment using the same concentration of PDA-L as used for the highest concentration of PDA-M. PDA-L, which has no affinity for the membrane, was enclosed in the lumen of the coacervate. The results for the PDA-L sample were similar to those of the empty coacervates sample (Supplementary Fig. 19), indicating that membrane attachment of the artificial cytoskeleton is crucial for altering membrane dynamics.

Fig. 4. PDA fibrils enable control over terpolymer membrane dynamics.

a, FRAP of the terpolymer membrane for empty coacervates (grey), coacervates in which 10% of membrane surface is covered with PDA (medium coverage, light green) and coacervates in which 30% of membrane surface is covered with PDA (high coverage, dark green). The numbers refer to the amount of immobile fraction. The data represent mean ± standard deviation of three biological replicates. b, A confocal micrograph of coacervate in which 10% of the membrane is covered with PDA. c, A confocal micrograph of coacervate in which 30% of the membrane is covered with PDA. d, A FRAP measurement of coacervates with medium PDA surface coverage showing laser-mediated bleaching of a circular spot on the membrane and subsequent fluorescence recovery of the bleached spot. Scale bars, 10 µm (b–d). In the confocal micrographs, green represents the PDA and cyan represents the polymer membrane. All micrographs were taken at the top plane of the coacervate.

Replication of cell-like mechanical properties

As a final cytoskeletal feature, we wanted to explore the effect of a cytoplasmatic network on the mechanical properties of the artificial cell. To study this effect, real-time deformability cytometry (RT-DC) measurements were performed for determining the mechanical properties of individual cells. In brief, suspended cells were passed through a 20 × 20 µm microfluidic channel, in which hydrodynamic forces led to characteristic bullet-like deformations of the initially spherical cells, reaching a steady state at the rear end of the channel (Fig. 5a, red square). Based on the circularity c of each cell, the deformation D was quantified (D = 1 − c). Although the coacervates showed large heterogeneity in size, which made it difficult to directly compare the deformation between different populations because deformation and size are highly correlated, we found that the deformation of all coacervates samples scales approximately linearly with size (Supplementary Figs. 12 and 13). This allowed us to analyse the deformation per area to correct for any differences in the coacervate size distribution between the samples.

Fig. 5. Mechanical properties of cytoskeleton-functionalized coacervates.

a, The microfluidic channel used for RT-DC measurements. Adapted from BioRender.com. b, Brightfield images of coacervates filled with PDA-L (top) and regular coacervates containing no additives (bottom) in the reservoir (left) and in the narrow channel (right) of the microfluidic device; a convex hull is fitted around them, which was used for determining the deformation (red). Scale bar, 5 µm. c, The deformation per area for different coacervate-based architectures. The data represent three biological replicates for coacervates without any additives (n = 1,106), coacervates filled with dextran (n = 1,921), coacervates filled with PDA-M (n = 1,151) and coacervates filled with PDA-L (n = 1,553). The white dot represents the median, the thick grey bar in the centre represents the interquartile range (IQR) and the thin grey line represents the 1.5× IQR. d, Apparent Young’s moduli of coacervates filled with hydrophobic PDA-M (n = 962), hydrophilic PDA-L (n = 1,553) and HL60 cells (n = 1,248). The white dot represents the median, the thick grey bar in the centre represents the IQR, the thin grey line represents the 1.5× IQR and the circles represent individual data points. Reproduced from BioRender.com.

Four different coacervate architectures were investigated under equal experimental conditions (Fig. 5c). The terpolymer-stabilized coacervate without cytoskeleton showed the largest deformation. Coacervates filled with dextran, as a non-fibre forming polymer constituent, showed similar results. This was expected, as here the main contribution for the mechanical stability arises from the surface tension of the coacervate droplet and the elastic shell formed by the terpolymer membrane. Adding the polysaccharide led to higher intracellular crowding, which should lead to an increased viscosity but which had only a small effect on the coacervate’s elastic properties as measured with RT-DC. Next, cells with a hydrophobic PDA cytoskeleton (PDA-M) were investigated. In this case, the PDA associated with the membrane. Here, we saw again a minor decrease in mean deformation. These membrane-stabilized coacervates were relatively soft, and their mechanical properties are best approximated as a droplet defined by its surface tension and an elastic shell. By contrast, using a cytoplasmatic network of hydrophilic PDA (PDA-L), in which the whole lumen is filled with a widespread network of densely entangled micrometre-sized bundles, we observed a significant decrease in deformation. The artificial cells became significantly stiffer and their mechanical properties were best represented as an elastic sphere because the volume of the elastic component was homogeneously distributed over the whole volume of the cell. The Young’s modulus of the lumen sample was independent of the size, in contrast to the other samples, suggesting that it was a good fit to the elastic sphere model (Supplementary Fig. 14). Hence, the mathematical model of an elastic sphere^32,33 was used to determine the Young’s modulus of coacervates filled with PDA-L as E = 0.43 ± 0.20 kPa (n = 1,553). In case of the PDA-M filled coacervates, the elastic sphere model fit less effectively but was used to calculate an apparent Young’s modulus E_app = 0.27 ± 0.11 kPa (n = 962) to demonstrate the difference between the two systems (Supplementary Fig. 15). For comparison, we measured a commonly used human leukaemia cell line (HL60) under the same experimental conditions as used before. For the HL60 cells, we determined a Young’s modulus of E = 0.48 ± 0.12 kPa (n = 1248). The stiffness and also the distribution of the HL60 cells matched well with our synthetic architecture (Fig. 5d and Supplementary Fig. 16) showing that our PDA-L coacervates closely mimic a natural cell from a mechanical perspective.

Discussion

We have designed a self-assembled polymer-based fibrous network that closely mimics the natural cytoskeleton, is compatible with different artificial cell architectures and can be utilized to perform key cytoskeletal functions. In particular, we iinstalled biomolecule scaffolding, regulation of membrane dynamics and cell-like mechanical features into our artificial cell platform based on coacervates.

The spatial organization of biomolecules is crucial for their proper functioning within the cellular environment². Our cytoskeleton shows high affinity for different cargos, by both reversible and irreversible interactions, enabling spatio-(temporal) control over biomolecule concentration and stoichiometry. Scaffolding of proteins at high local concentrations and within 10 nm proximity was demonstrated by the reconstitution of luciferase from its two separate domains, and effective signal transduction was achieved with BRET between luciferase and the fluorescent protein mNeonGreen. We envisage that, due to the chemically robust and versatile approaches we have developed, the scaffolding process can be extended easily to a range of other biomolecules and processes that rely on proximity for efficient performance.

Controlling membrane properties is crucial for interactions both between cells and within a cell. In natural cells, membrane properties are controlled by the lipid composition but also by the cytoskeleton. The relative mobility of individual lipid molecules and their association with the cytoskeleton is key for fulfilling their biological tasks, including the regulation of cell shape, tissue integrity and signal transduction pathways. The importance of membrane fluidity as a functional feature has been outlined in literature, for example, for motility and for T cell activation via antigen-presenting cells^31,34. We demonstrated how the fluidity and the amount of immobile domains can be controlled by the degree of PDA surface coverage.

Arguably one of the least explored features in artificial cell research are the bulk mechanical properties. Traditionally used platforms, such as giant unilamellar vesicles and coacervates, have a liquid-like interior that is, from a mechanical property perspective, very distant from the properties found in living cells. Artificial cells can, from the mechanistic side, rather be perceived as relatively simple entities defined by their surface tension. In our case, our terpolymer membrane-stabilized coacervates behave as soft droplets with a thin elastic shell³⁵. Through the incorporation of our biomimetic cytoskeleton, their internal structure is adjusted to result in more life-like artificial cells, which can tolerate higher mechanical stresses, by distributing interacting forces more evenly throughout the cell. The cell-like mechanical resilience opens perspectives to modulate the communication and interaction of artificial cells with natural ones by changing their mechano-biological behaviour.

We have shown that the incorporation of a cytoskeleton in artificial cells, with control over intracellular positioning and with a scaffolding function, is a valuable asset that not only enables the regulation of mechanical features but also translates into improved function. The capability of mimicking all these features with one type of cytoskeleton demonstrates this system’s uniqueness. In contrast to other cytoskeletons used in artificial cells, however, it lacks dynamicity^20,35. We are therefore currently working towards a more dynamic cytoskeleton in which the fibril network can be reversibly assembled using environmental cues. At the moment, bundling is facilitated through electrostatic interactions. To disassemble the bundles, we need to disrupt these interactions. This can be done theoretically under acidic conditions or high salt concentrations, but it would also lead to the disassembly of coacervates. Therefore, we need to reengineer the bundling process to make it reversible under physiological conditions.

Methods

Coacervate formation

Coacervate formation, PDA for maximum membrane coverage

Q-Am and CM-Am were dissolved separately in phosphate-buffered saline (PBS) at a concentration of 1 mg ml⁻¹. Consecutively, Q-Am (66 µl, 1 mg ml⁻¹, degree of substitution (DS) 1.0), PDA (PDA-M; Supplementary Information section 5.6) (5 µl, 0,1 mg ml⁻¹) and CM-Am (33 µl, 1 mg ml⁻¹, DS 0.5) were added. After shaking for 10 min at 1,500 rpm, 3.3 µl terpolymer (50 mg ml⁻¹ in PEG350) was added and the sample was shaken for another 5–10 s.

Coacervate formation, PDA for maximum lumen coverage

Q-Am and CM-Am were dissolved separately in PBS at a concentration of 1 mg ml⁻¹. Consecutively, Q-Am (66 µl, 1 mg ml⁻¹, DS 1.0), PDA (PDA-L; Supplementary Information section 5.6) (1 µl, 10 mg ml⁻¹) and CM-Am (33 µl, 1 mg ml⁻¹, DS 0.5) were added. After shaking for 10 min at 1,500 rpm, 3.3 µl terpolymer (50 mg ml⁻¹ in PEG350) was added and the sample was shaken for another 5–10 s.

Coacervate formation, PDA-NTA

Q-Am, CM-Am and NTA-Am were dissolved separately in PBS at a concentration of 1 m. First, 1 µl NTA–PDA (10 mg ml⁻¹) was added to 7.5 µM of NiSO₄ (final concentration) to reach a final NTA–PDA concentration of approximately 1 mg ml⁻¹. The proteins were added and incubated by shaking at 500 rpm for 30 min. Consecutively, Q-Am and CM-Am were added to induce coacervation. After shaking for 10 min at 1,500 rpm, 3.3 µl terpolymer (50 mg ml⁻¹ in PEG350) was added and the sample was shaken for another 5–10 s.

Coacervate formation for FRAP measurements

Q-Am and CM-Am were dissolved separately in PBS at a concentration of 1 mg ml⁻¹. Consecutively, Q-Am (66 µl, 1 mg ml⁻¹, DS 1.0), PDA-sulfo-Cy5 (1–10 µl (depending on the final membrane coverage), 0.01 mg ml⁻¹) and CM-Am (33 µl, 1 mg ml⁻¹, DS 0.5) were added. After shaking for 10 min at 1,500 rpm, 3,3 µl terpolymer (50 mg ml⁻¹ in PEG350 containing 15% of Alexa Fluor 488 dye (AF488)-labelled poly(ethylene glycol)-poly(caprolactone-gradient-trimethylene carbonate) (PEG68-b-PCL50-g-PTMC50) was added and the sample was shaken for another 5–10 s.

Fluorometric quantification of encapsulation efficiencies

Coacervates were prepared as mentioned in ‘Coacervate formation, PDA for maximum lumen coverage’ varying the amount of PDA added. One-hundred microlitres of the coacervate suspension was diluted to 500 µl with PBS. To obtain the supernatant without coacervates, 250 µl was filtered through a 400-nm-pore-size polyvinylidene fluoride syringe filter. The corresponding fluorescence intensities were analysed using a Tecan Spark 10M plate reader equipped with a monochromator. In a 384-well black non-treated plate (Thermo Fisher Scientific), 30 µl of the coacervate samples was added to each well. Encapsulation efficiencies were calculated by the fluorescence ratio of the whole sample in relation to the supernatant.

The concentration of molecules inside coacervates was calculated in accordance with a previously published method³¹. In brief, the total volume of coacervates was calculated according to $V_{\mathrm{total}}=\frac{m_{\text{Q-Am}}+m_{\text{CM-Am}}}{\mathrm{Density}}$. Finally, the concentration determined by the plate reader was adjusted to the total volume of the coacervate phase.

Bioluminescence spectroscopy

For analysis of bioluminescence in coacervate samples, a Tecan Spark 10M plate reader equipped with a monochromator was used. Twenty microlitres of the coacervate sample was added to each well in a white 384-well non-treated plate (Thermo Fisher Scientific). Directly before the measurement, a 1:40 dilution of the Nano-Glo substrate mix was added to reach a final substrate concentration of 1:240. The plate was covered with an EASYseal transparent sealing film (Greiner Bio-One). Bioluminescence was measured at 458 ± 12.5 nm and BRET was measured at 533 ± 12.5 nm with an integration time of 100 ms. The bioluminescence experiment was conducted in bulk for reference. For the experiments, a batch of the PDA fibril was incubated with the enzymes. The same fibrils were further processed to be used for the bundle and coacervate sample. Each sample was split into three replicates and furimazine was added independently to each sample before being measured.

CLSM

CLSM measurements were performed on a Leica TCS SP8. Three-dimensional (3D) rendering of confocal micrographs was performed using the 3D rendering wizard in the Leica Application Suite X software. Images were taken with a Leica HC PL APO CS2 63× water immersion objective with a numerical aperture of 1.20. Image resolution, scanning speed, detector gain and laser intensity were optimized for each sample. If necessary, image deconvolution was performed with plug-ins included in FIJI. Scale bars were added in FIJI.

FRAP

For imaging, the sample preparation was adapted from a previously published procedure and is described in ‘Coacervate formation for FRAP measurements’³⁶. At 60 °C, 25 µl of coacervate were mixed with 25 µl of agarose solution (2%). The mixture was transferred to the imaging slide (IBIDI, 18-well glass bottom), was allowed to harden for 10 min at room temperature and was subsequently covered with 50 µl PBS.

FRAP measurements were performed on a Leica TCS SP8 using the FRAP wizard in the software. Images were taken with an HC PL APO CS2 63× water immersion objective with a numerical aperture of 1.20. Images were acquired with a resolution of 512 × 512 pixels and a scanning speed of 600 Hz. Artificial cells between 20 μm and 30 µm were selected. Ten images were acquired before the bleaching. Subsequently, the region of interest (ROI) (d = 3.5 µm located at the top of the cell) was bleached for ten iterations for a total of 7.8 s with 100% laser power. The recovery was monitored with a 3 s interval until full recovery was reached. The intensities of the bleached ROI, reference area, part of the membrane that was not bleached, and background were extracted from the images with FIJI.

Data were analysed as follows with Origin 2019 following the procedure from ref. ³⁷. Fluorescence intensity at time t, I(t) was background subtracted and corrected for unintentional photobleaching, which was calculated as

$$I\left(t\right)=\frac{\mathrm{ROI}\left(t\right)-\mathrm{Bg}}{\mathrm{Ref}\left(t\right)-\mathrm{Bg}},$$

where ROI(t) is the average intensity of the bleached area at time t, Ref(t) is the average intensity of an unbleached fluorescent area (same size as bleached area) and Bg is the average intensity of a background area. I(t) was further normalized such that the pre-bleach intensity was set to 1. This was done by dividing I(t) by the average background-subtracted pre-bleach intensity within the bleach area. The normalized intensity I(t)_normalized was plotted against the recovery time. The recovery curve was fitted with

$${I\left(t\right)}_{\mathrm{normalized}}=B-A{\mathrm{e}}^{-\tau t},$$

assuming exponential kinetics, to obtain parameter τ. In the above equation, τ is the recovery time constant, t is time and A and B are two constants.

Finally, recovery half-time (τ_1/2) and apparent diffusion coefficient (D_app) were calculated according to the following equations for a circular bleaching spot where ω is approximated as the radius of the bleaching spot:

$${\tau }_{1/2}=\frac{-\ln \left(0.5\right)}{\tau }$$

$$D_{\mathrm{app}}=\frac{0.88{\omega }^2}{2{\tau }_{1/2}}.$$

RT-DC experiments

RT-DC experiments were performed with a commercially available RT-DC device (AcCellerator, Zellmechanik). The measurement principle of RT-DC is described in detail in ref. ³⁸. In brief, a microfluidic chip containing the measurement channels with a cross-section of 20 × 20 µm was mounted on an inverted microscope (Axiovert 200M, ZEISS) and flow was introduced with syringe pumps (Cetony Nemesys). Images were captured with a complementary metal oxide semiconductor camera (Mikrotron). The syringe pumps and camera were controlled with the measurement software ShapeIn (Zellmechanik), which analyses contours in real time.

Coacervates were prepared according to the procedures mentioned in ‘Coacervate formation’. Then, coacervates were centrifugated at 5,000g for 5 min, the supernatant was removed and the pellet was resuspended in 100 µl of CellCarrier solution (Zellmechanik). CellCarrier is a measurement solution for RT-DC experiments and consists of 0.49 w/w% methyl cellulose dissolved in PBS. CellCarrier is described in detail in ref. ³³. All samples were measured at a flow rate of 0.01 µl s⁻¹.

For cell experiments, 1 ml of HL60-cell solution (2 × 10⁵ cells ml⁻¹) was centrifuged at 200g for 4 min. After the supernatant was removed, the cell pellet was resuspended in 100 µl of CellCarrier solution to a final cell concentration of 2 × 10⁶ cells ml⁻¹ and measured at a flow rate of 0.01 µl s⁻¹.

For the analysis, all samples were filtered for projected areas with 50–200 μm². Furthermore, only events with a ratio of convex-hull area to the raw area of 1.0–1.05 were considered^39,40. This step guarantees that noisy contours that could lead to overestimated deformation values were not considered in the analysis.

The deformation in RT-DC was quantified on the basis of the circularity of the convexified contour $\mathrm{Deformation}$ $=1-\mathrm{Circularity}=1-\frac{2\sqrt{\mathrm{\pi }\times \mathrm{Area}}}{\mathrm{Perimeter}}$. A deformation of 0 represents a perfect circle, and increasing values indicate increasing deviation from a circular shape. Deformation and cell size were always correlated in RT-DC experiments because bigger objects were exposed to greater stresses acting closer to the channel walls³⁵. If measured objects could be approximated as a fully elastic sphere with homogeneous Young’s modulus, deformation and size could be decoupled and the mechanical properties were described by the Young’s modulus³². The Young’s moduli for the lumen coacervates and HL60 cells for this study were calculated using the mechanical model from the study of Wittwer et al., using the Python package for RT-DC analysis dclab⁴¹.

To test for statistical significance between samples, we used an analysis of variance P value computed with linear mixed-effects models as described by Herbig et al.⁴² and implemented in the lme4 package in dclab.

Statistics and reproducibility

For all micrographs, similar results were obtained in three independent experiments.

Reporting summary

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

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41557-024-01697-5.

Author contributions

S.N. designed the experiments and analysed the experimental results. F.R. led the experiments and analyses related to RT-DC and was supported by S.N. Y.L. and H.W. performed the cryo-TEM measurements. T.v.V. performed the molecular cloning and helped S.N. with the nanoluc complexation assay. H.J. and J.v.H. conceived and supervised the project. J.G. provided feedback on the results and helped conceive the experiments related to the cell mechanics-related measurements. S.N. wrote the manuscript with the aid and input of all other authors.

Peer review

Peer review information

Nature Chemistry thanks Jianbo Liu, Oskar Staufer and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Data availability

Raw data for Figs. 3b, 4a and 5c,d are publicly available on the 4TU.ResearchData repository and can be accessed through 10.4121/0960fdee-dbeb-438c-b907-d9792265521d. Raw confocal microscopy image files are also available on request, but due to the large number of files they are not provided with the Supplementary Information. Other data are presented in the Article and the Supplementary Information.

Code availability

The Young’s moduli for the lumen coacervates and HL60 cells for this study were calculated using the mechanical model from ref. ⁹, using the Python package for RT-DC analysis dclab39.

Competing interests

J.G. is co-founder of the company Rivercyte GmbH, which offers commercial products and consumables related to deformability cytometry. The other authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41557-024-01697-5.