Pathway Controlled Phase Separation of Minimal Building Blocks Utilizing a Dissociative Chemical Transformation

ABSTRACT

Metastable states are utilized by biology to power the construction of large and motile macromolecules and also to realize out‐of‐equilibrium phase separation. Energy transduction from orthogonal and unrelated exergonic reactions drives the contra‐thermodynamic transformation, which acts as the catalyst for the exergonic reaction. Herein, we show that thermodynamically stable building blocks can undergo phase separation when the process is coupled with an exergonic degradation of their thermodynamically activated precursor. The thermodynamically stable products alone are incapable of accessing the droplets. The chemical transformation is critical to achieve pathway‐controlled phase separation, which catalyzes the chemical transformation. The activated precursor undergoes a β‐elimination reaction to produce an aromatic substrate along with trimethylammonium cations, which provide temporal stabilization to the phase‐separated droplets. Droplet formation is not observed with precursors incapable of undergoing the β‐elimination reaction. The generated droplets can imbibe diverse guest molecules, and their transition from droplets to proto‐tissue‐like structures is observed in the presence of porphyrin. Importantly, the kinetically accessed metastable liquid droplets are shown to augment the catalytic potential of hemin, cofactor of natural peroxidase, and accelerate the hydrolase‐peroxidase cascade reaction.

Kinetically controlled phase separation of thermodynamically stable products takes place when the process is coupled with dissociative chemical transformation of thermodynamically activated precursors. The energy consumption pathway is necessary, as the thermodynamically stable products alone are incapable of phase separation.

1. Introduction

Spontaneous chemical reactions proceed energetically downhill, moving toward local or global energy minima. Non‐spontaneous reactions (characterized by a positive Gibbs free energy change, ΔG>0), by contrast, require energy to proceed and do not occur spontaneously [1, 2, 3]. Biological anabolism, which encompasses the construction of large and diverse macromolecules or phase separation as seen in biomolecular condensates occur from simpler precursors. Importantly these processes are driven by energy stored in thermodynamically activated and kinetically stable molecules, such as nucleoside triphosphates [4]. The degradation of such activated molecules is utilized for thermodynamically uphill synthesis and thus involves an energy transduction process [5, 6]. For instance, in GTP driven cytoskeleton polymerization, synthesis of high energy structures importantly catalyze the degradation of GTP at accelerated rates leading to complex behavior, such as dynamic instability, treadmilling, and so forth [7, 8, 9, 10, 11, 12, 13, 14].

In recent years, chemical transformations have been realized by exploiting an external energy source where directionality in ratchet mechanisms plays critical role [3, 15, 16, 17]. Reports have shown contra‐thermodynamic processes of deracemization by utilizing light and redox cycling [18]. Recently, high energy coupling agents based on carbodiimides have been used to demonstrate synthesis of Diels‐Alder adducts under thermodynamically unfavored conditions [5, 19]. In addition to that, light energy has also been harnessed to achieve ratcheted high energy supramolecular complex in water under non‐equilibrium conditions [2]. Herein, we report for the first time that, thermodynamically stable small building blocks undergo phase separation, only when the process is coupled with degradation of their thermodynamically activated precursor (Figure 1). This coupling with a chemical transformation is critical as the thermodynamically stable products alone are incapable of the droplet formation in the absence of the spontaneous dissociative chemical transformation. The precursor features a trimethylammonium cation in the β position of the ester carbonyl group and when subjected to a pH close to physiological milieu, it undergoes spontaneous β‐elimination to form hydrophobic aromatic alkene ester along with ammonium cations, which in combination drive the droplet formation. Notably, the β‐elimination is facilitated by the phase separation making this phenomenon akin to the dynamic self‐assembly processes seen in extant biology. The droplets accessed via kinetically controlled pathway act as catalytically active crucibles as these accelerate a hydrolase‐peroxidase cascade reaction by entrapping hemin, cofactor of natural peroxidases; the reaction is otherwise not promoted in absence of these phase separated droplets despite the presence of all constituents.

FIGURE 1.

(A) Schematic representation of pathway‐controlled phase separation of droplets via β‐elimination reaction. Grey arrow indicates that the transformation is not feasible under the experimental conditions. (B) Schematic representation of phase separated droplets formation in water via β‐elimination reaction. Bright field optical microscopy image of C3QE containing system after 60 min. Inset shows the vial containing turbid solution due to the formation of phase separated structures. (C) Different molecules used in the study.

2. Results and Discussion

We started with small molecules that featured hydrophobic aromatic part and a cationic group as these group can involve in cation‐π interactions, which have been shown to play important role in simple coacervation [20, 21, 22, 23, 24]. We started with mixing 5 mM of 3‐((4‐chloronaphthalen‐1‐yl)oxy)‐N,N,N‐trimethyl‐3‐oxopropan‐1‐aminium bromide (C3QE) in aqueous medium (pH 7.1, TRIS buffer 200 mM, Figure 1C). Interestingly, C3QE based system showed the appearance of turbidity with time. Optical microscopy of the turbid solution showed the formation of spherical droplets with a typical size of ∼1‐5 µm (Figure 1B, representative library in Figure S1, the minimum concentration of C3QE required for phase separation was found to be around 250 µM, ∼0.1% w/v from turbidity studies, Figure 2A). Further, optical microscopy also showed the process of fusion in the native state of droplets (confocal microscopy images in Figure 2, vide infra, frames taken from Movie 1, Figure S2). Importantly, droplets maintained the spherical shape even after the fusion, which proved the liquid‐like nature. Polarized optical microscopy (POM) images suggested the absence of crystalline phase, which further supported the liquid‐like nature of the droplets (Figure S3). The droplets were able to encapsulate a range of water‐soluble dyes, such as Rhodamine 6G, ThT, and Rhodamine B (confocal laser scanning microscope, CLSM images, Figure 2B,C,F). Moreover, the partitioning behavior of hydrophobic dyes, such as, Nile red, Coumarin 343 could be seen under CLSM (Figure 2D,E). Importantly, the apparent partition coefficient (K = C_droplets/C_solution) of Nile red partitioning was found to be 2498 (Figure S4). These data underpinned the presence of both hydrophobic as well as hydrophilic microenvironments within the liquid droplets. The coalescence event of the droplets could also be visualized under CLSM as the fluorescent droplets (enriched with Rhodamine B) fused with time (Figure 2F). Extensive fluorescence recovery after photobleaching experiments at different times showed the fluorescence recovery of the photobleached area of droplets, which suggested the maintenance of the liquid‐like nature of droplets throughout the time (Figure 2G–I). Importantly, almost all droplets coalesced with time and turbidity diminished within ca. 5 h (Figure S5).

FIGURE 2.

(A) Concentration dependent change in absorbance of C3QE system at 700 nm wavelength. (B‐E) CLSM image of C3QE droplets incubated with Rhodamine 6G (B), ThT (C), Nile red (D) and Coumarin 343 (E). All images in (B‐E) are recorded after ∼60 min and [C3QE] = 5 mM, [Nile red] = 20 µM, [ThT] = [Coumarin 343] = 50 µM. (F) Different frames (time increases from left to right) obtained from CLSM images of C3QE containing system incubated with Rhodamine B shows the fusion of liquid droplets; time gap of each frame is ∼3.2 s. [C3QE] = 20 mM, [Rhodamine B] = 20 µM. (G‐I) Different frames during FRAP experiment of the C3QE system after ∼30 min (G), 120 min (H), and 180 min (I). (J) Time dependent change in normalized fluorescence intensity at different times of FRAP experiment. [C3QE] = 10 mM, [Rhodamine 6G] = 1 µM. Scale bar for (G, H) and (I) is 1 µm.

Next, ¹H NMR of the C3QE containing system was performed to understand the interaction in the droplets (see Supporting Information for details). Strikingly, the peaks corresponding to C3QE started showing rapid decline with subsequent generation of C3AE (Figure 3A, orange circle). In addition to that, the generation of trimethylammonium cation (TMAC) could be seen in ¹H NMR (Figure S6). These data suggested that, C3AE was rapidly getting generated in the medium via a β‐elimination (possibly triggered by hydroxide ion present in the medium), a spontaneous dissociative chemical transformation accompanied by the formation of TMAC. Importantly, the peaks showed broadening in D₂O due to the generation of the hydrophobic microenvironment and involvement of the C3AE in the formation of phase separated droplets (Figures 1B and 3A) [25]. Notably, almost complete conversion of C3QE to C3AE could be observed in ¹H NMR within 1 h. The progress of the β‐elimination reaction was further monitored by high performance liquid chromatography (HPLC), which also showed almost complete conversion of precursor C3QE to building block C3AE within 1 h corroborating the data obtained from ¹H NMR (Figure 3B). However, retention of spherical droplets up to 2 h could be observed under optical microscopy (Figure S7). This suggested that the reaction and subsequent generation of hydrophobic C3AE along with TMAC helped in the realization of the droplets (Figures 1B and S6). Intriguingly, when synthesized C3AE (see Supporting Information for synthesis procedure) was separately introduced in the same condition (pH 7.1, TRIS buffer 200 mM), the formation of spherical droplets could not be observed. Next, the C3AE was supplemented with mechanical energy via probe sonication (probe sonicator, power 227 W, 20 min) with the expectation that the influx of energy would promote the energetically uphill droplets formation. However, optical microscopy showed the presence of crystalline solid aggregates in the solution (Figure S8). Since some amount of CN formation in the system could be seen from the time dependent HPLC chromatograms, experiments were done by mixing C3AE, TMAC (as bromide salt, see Supporting Information for synthesis procedure) with different percentage of CN (0%, 5%, 10%, and 20%, respectively with respect to TMAC) under identical conditions. Notably, no turbidity was observed. Optical microscopy revealed the formation of crystalline structures rather than liquid droplets, and no droplet‐like morphologies could be detected (Figure S9). These results strongly support that mixing of the elimination products (C3AE and TMAC) did not lead to droplet formation. These data suggested the importance of concomitantly generated TMAC and the possible role of pathway control in the otherwise inacessible process of the droplet generation. It is worth noting trimethylamine N‐oxide (TMAO, structurally similar to TMAC) is a well known osmolyte that helps in maintaining the liquid‐like nature of different hydrophobic protein condensates. This process helps to restrict aberrant protein aggregation [26, 27, 28]. When TMAO and C3AE were mixed in buffer solution, no droplets could be seen. However, when the mixture was subjected to mechanical energy via probe sonication, spherical droplets could be observed under optical microscope; although the population remained significantly lower than the C3QE system (Figure S10). Nonetheless, the results suggested that the interaction between hydrophilic TMAO and hydrophobic C3AE could indeed facilitate droplet formation only when mechanical energy was supplemented suggesting the energetically uphill phase separation process. To probe further, all‐atom molecular dynamics (MD) simulations were performed using the GROMACS software package (version 2019.6, see Supporting Information for details). Four different systems were used for MD simulations: (i) C3QE molecules dispersed in water, (ii) C3AE molecules dispersed in water, (iii) C3AE molecules dispersed in water with TMAC at a 1:1 molar ratio, and (iv) C3AE molecules dispersed in water with TMAO at a 1:1 molar ratio. MD simulation studies supported the C3AE molecules were indeed undergoing rapid aggregation followed by merger into larger aggregates in the aqueous medium (experimentally seen as crystalline solid aggregates, Figures S8, 3C, and S11). Strikingly, in the presence of TMAC, the aggregation propensity slowed down significantly (Figures 3D, E, and S12). The TMAC molecules preferentially localized near the interfacial regions of the C3AE clusters inducing net positive charges on these colloidal particles (Figure 3F, G, and Movie 2). Similar results were obtained for TMAO (Figure 3E–H). The MD simulations suggested that, the weakly repulsive electrostatic barrier from the localized cations (generated via the β‐elimination transformation) led to the stabilization of the liquid‐like droplets and thus helped in the process of kinetically accessed phase separation. In the absence of such stabilizers, C3AE would rapidly phase separate into the insoluble aggregates. In addition to that, the mean square displacement (MSD) of the C3AE molecules in the presence and absence of TMAC inside the aggregates suggested that, the C3AE molecules move much freely in presence of TMAC, representing the liquid‐like nature (Figure S13). However, in case of system containing only C3AE, the movement is restricted which further supports the crystalline aggregate nature of only C3AE observed in optical microscopy. Molecular dynamics simulations were also performed to calculate free energy (chemical potential) of individual C3AE molecules in different phases. It was observed that the chemical potential of C3AE in the aggregate in the absence of TMAC is more negative compared to that in the presence of TMAC (Figure S14). This data proved that the liquid‐like aggregates observed in the presence of TMAC are thermodynamically less stable (metastable, ΔΔG = 1.64 KJ/mol) as compared to the aggregates formed in water (in absence of TMAC). In combination, the results underpinned the importance of concomitantly produced TMAC to realize the kinetically phase separation of thermodynamically stable building blocks (sustained up to 2 h, despite the conversion being complete in ca. 1 h).

FIGURE 3.

(A) Schematic representation of β‐elimination reaction and time dependent change in ¹H NMR spectra of C3QE system in TRIS buffer in D₂O and ¹H NMR of C3AE in DMSO‐d₆ for comparison. [C3QE] = 10 mM. (B) HPLC chromatograms (extracted at 290 nm) of time dependent generation of C3AE from C3QE. [C3QE] = 1 mM, pH 7.1, TRIS buffer 200 mM. CN corresponds to 4‐chloro‐1‐naphthol which is generated due to background hydrolysis. (C) Snapshots from the MD simulation trajectory showing the aggregation process of C3AE in presence of TMAC (orange). Clustering is seen at t = 50 ns, with structures colored according to their assigned cluster labels. The same color scheme is preserved for frames at t = 0 ns and t = 1000 ns to visualize the temporal progression and stability. (D) Temporal evolution of the size of the largest cluster of C3AE in presence and absence of TMAC. (E) Total number of clusters of C3AE aggregates formed over time with only C3AE and in presence of TMAC, TMAO. (F) Density profile of TMAC molecules. The upper panel includes a reference snapshot of the system (TMAC as orange spheres and C3AE as cyan). The positions of the interfaces are marked using vertical dashed lines. Representative snapshots of liquid‐like blobs of C3AE molecules with surface adsorbed TMAC (G) and TMAO (H). (I) Temperature dependent change in absorbance in C3QE system. [C3QE] = 1 mM, pH 7.1, TRIS buffer 200 mM. (J) pH dependent change in absorbance in C3QE (1 mM) system. TRIS buffer 200 mM is used for experiments performed at pH 8, 7.1 and MES buffer 200 mM is used for experiments performed at pH 6.

We next probed the C3QE system at different environmental conditions to understand their effects on the kinetics of liquid droplets formation. First, temperature dependent turbidity was checked, which suggested that thermal energy promoted an increase of turbidity (Figure 3I). Interestingly, appearance of sigmoidal kinetic profile for the turbidity generation could be observed in UV–vis spectroscopy, which also indicated the presence of lag phase prior to the phase separation. This data suggested that a threshold concentration of the β‐elimination products (C3AE and TMAC) must be reached before the emergence of kinetically accessed metastable droplets. We anticipated that at higher temperature, the rate of β‐elimination would get accelerated. Indeed, higher temperatures promoted droplet formations. Also, a decrease in the lag phase for turbidity generation could be observed upon increasing the pH of the medium. Expectedly, at high pH, β‐elimination became more feasible, which promoted the phase separation (Figure 3J). To probe the importance of ester functional group for the generation of liquid droplets, control molecule C3QA was synthesized with amide functional group as amides are not prone to facile β‐elimination. When C3QA was incubated in the aqueous medium (5 mM in pH 7.1, TRIS buffer 200 mM), no emergence of turbidity with time could be observed. Expectedly, time dependent NMR showed negligible generation of the alkene amide product even after 6 h (Figure S15). Next, ester moiety was replaced with ether functional group, which resulted in the generation of C3E molecule (Figure 1C). Once again, due to inability to undergo β‐elimination reaction, the molecule remained soluble, and the appearance of turbidity could not be observed in the medium. For C6QE which featured a longer carbon chain (Figure 1C), no appearance of turbidity could be seen in the medium, which further supported the critical role of β‐elimination (β‐eliminated product could not be traced from HPLC). As a positive control, the naphthyl group was replaced with phenyl group (PC3QE) with the expectation that the β‐elimination will be facilitated yet the droplet formation may be impeded due to diminished hydrophobic effects (Figure 1C). Indeed, the solution PC3QE (1 mM) accessed a clear solution which remained transparent for longer time although NMR studies showed the progress of β‐elimination event with time (Figure S16). The formation of liquid droplets could only be seen at higher concentration (Figure S17). Interestingly, it could be observed from NMR that the consumption of C3QE was significantly faster than the PC3QE (5 mM each, Figure S18A). We were curious to investigate whether the metastable liquid droplets were indeed facilitating the β‐elimination reaction. For this purpose, a control experiment was performed in 50% DMSO/buffer system where no turbidity could be seen. Interestingly, it could be observed that the consumption of PC3QE was faster than C3QE (Figure S18B). These results suggested that the phase separation was possibly catalysing the β‐elimination reaction [5, 6].

To further probe the role of hydrophobic effects in the formation of phase separated droplets, we exposed C3QE containing system to different external stimuli. First, precursor C3QE (1 mM) was exposed to different concentrations of 1,6‐hexanediol (HD) and turbidity of the system after 30 min was measured. HD is commonly used to disrupt weak interactions (in proteins) and is selective to dissolve liquid condensate over solid aggregates [29, 30]. Increasing the concentration of HD resulted in decrease of the turbidity (Figure S19A). Importantly, when 5 M HD was introduced, no phase separated droplets could be seen under optical microscope (Figure S20). This data proved the importance of hydrophobic effects toward droplet formation. Next, C3QE based system was exposed to different vol% of DMF, which also acts as disruptor of hydrophobic effects [31]. Upon increasing the vol% of DMF, decrease in turbidity could be observed and no structures could be seen with 25% (v/v) DMF under optical microscope (Figures S19B, S20). Turbidity measurement of C3QE containing system with varying concentration of NaCl showed minimal change, suggesting electrostatic interactions had negligible effect in the phase separation (Figure S19C). Moreover, optical microscopy suggested that the structures remained stable in presence of 2 M NaCl (Figure S20). Thus, it could be concluded that hydrophobic collapse of building block C3AE was the primary factor for the phase separation.

At this stage we were curious to investigate the capabilities of the droplets to imbibe guests which can impart structural and functional features. In this context, porphyrins that are natural pigments play critical catalytic and transport roles in proteins (such as cytochromes and hemoglobin) and are essential across life forms present on Earth [32, 33]. When protoporphyrin IX (PPIX) was added to the system after 60 min (when a significant percentage of the droplets were generated), the accumulation of PPIX could be observed inside the liquid droplets (intrinsic fluorescence of PPIX imbibed within droplets, CLSM image, Figure 4A). Surprisingly, when PPIX was introduced from the beginning (time∼0 min), the formation of higher ordered multidroplets assemblies could be observed (Figure 4A). We speculated that the initial interactions (electrostatic and hydrophobic effects) between the cationic precursor C3QE and negatively charged PPIX coupled with the non‐equilibrium changes in the system's composition due to the dissociative reaction impacted the phase separation process leading to higher‐ordered tissue‐like assemblies [34, 35, 36, 37]. Notably, accumulation of the iron containing congener of PPIX i.e. hemin, cofactor of natural peroxidases, could be observed inside the liquid droplets (introduced at time∼0 min). However, higher‐ordered assemblies could not be seen possibly due to absence of net effective charge of hemin, which prevented the higher‐order assembly formation (Figure 4B).

FIGURE 4.

(A) Schematic representation of evolution of C3QE system in presence of PPIX and corresponding CLSM. (B) Schematic representation of evolution of C3QE system in presence of hemin and corresponding CLSM images. [C3QE] = 1 mM, [PPIX] = [Hemin] = 50 µM, pH 7.1, TRIS buffer 200 mM. (C) Schematic diagram showing 2‐methoxyphenyl acetate to tetraguaiacol conversion via hydrolase‐peroxidase cascade in presence and absence of liquid droplet microreactors. (D) UV–vis kinetics at ∼470 nm for comparison of the generation of tetraguaiacol from 2‐methoxyphenyl acetate in presence and absence of liquid droplets. (E) UV–vis spectra show the tetraguaiacol formation from 2‐methoxyphenyl acetate in the presence of droplets.

We next investigated whether the hydrophobic microenvironment of kinetically accessed droplets could regulate intrinsic catalytic properties of the water insoluble hemin. The hydrophobic microenvironment provided by the liquid droplets was anticipated to enhance the oxidation reaction by concentrating substrates with the cofactor in the close proximity [38]. Hence, the capability of the droplets to promote oxidation reaction and also a two‐step hydrolase‐peroxidase cascade reaction. For checking the oxidation, 2‐methoxyphenol was used in presence of hemin and H₂O₂. The product of oxidation of 2‐methoxyphenol can be followed spectroscopically (λ_max≈470 nm, HPLC could not be used for monitoring oxidation as oxidation product tetraguaiacol is not stable) [38]. Notably, oxidation in the presence of droplets was similar to background which was reasonable due to the polar nature of the phenol (Figure S21). However, when substrate 2‐methoxyphenyl acetate (Figure 4C) was used, rapid intensification of the peak corresponding to oxidized product was observed (Figure 4D,E, the rate was found to be 0.83 µM/min) [39, 40]. In contrast, the rates in only buffer in presence of hemin showed ∼7‐fold lower activity (v_i = 0.12 µM/min, Figures 4D). This data suggested that, the hydrolase‐peroxidase cascade was indeed promoted by the phase separated droplets. To further probe the role of metastable liquid droplets, hydrolysis of 2‐methoxyphenyl acetate was monitored in HPLC; the hydrolysis rate of 16.88 µM/min was observed (Figure S22). Next, we evaluated the ability of the liquid droplets to encapsulate 2‐methoxyphenyl acetate. Notably, more than 80% of the substrate was partitioned into the droplets; specifically, 81.86% ± 0.97% and 86.56% ± 0.09% were localized within the droplets at 1 and 5 mM C3QE, respectively (Figure S23). This result suggested that the higher local concentration of the cofactor hemin within the droplets facilitated oxidation step of the cascade reaction; as the phenol was locally generated due to higher concentration of sequestered esters. Similar enhancement of reactivity within droplet‐based compartments have been reported in recent literature [41]. Importantly, in the presence of the synthesized C3AE and TMAC in buffer, the activity remained similar to only buffer solution (Figure S24). These data suggested that the kinetically accessed metastable liquid droplets could function as reaction hubs as it augmented the catalytic potential of cofactor hemin and promoted the overall hydrolase‐peroxidase cascade activity.

3. Conclusion

Biological anabolism relies on chemically orthogonal transformations, such as the energy‐releasing degradation of nucleoside triphosphates, to kinetically control the formation of complex macromolecular assemblies. In this context, our results demonstrate the first experimental manifestation of kinetically accessed metastable phase separation of thermodynamically stable building blocks, exploiting a spontaneous chemical transformation. Specifically, a dissociative β‐elimination reaction of a thermodynamically activated low‐molecular‐weight precursor was found to drive the formation of such metastable phase‐separated states. Integration of the chemical transformation was found to be critical for the formation of kinetically accessed metastable droplets, as the building blocks alone were incapable of undergoing phase separation. The TMAC generated during the reaction helped in maintaining the liquid‐like nature of the C3AE aggregates. Changes in molecular architecture emphasized the importance of exergonic reaction for the kinetically controlled generation of these small molecular liquid droplets. The metastable droplets were also shown to facilitate the β‐elimination similar to the strategy seen in extant biological metabolism. These droplets accessed via kinetic pathway could effectively sequester a wide variety of guest molecules, which is an important feature of membraneless organelles for performing diverse biological processes. Moreover, liquid droplets augmented the oxidation potential of hemin and accelerated the hydrolase‐peroxidase cascade reaction. Pathway controlled phase separation of functionally active droplets via small molecules could help in our understanding of protocell models and chemical emergence of compartments.

Author Contributions

Sumit Pal: conceptualization, writing – original draft, methodology, investigation, formal analysis, validation, visualization, and data curation. Dibyendu Maity: software and validation. Janardan Chakraborty: methodology, formal analysis, investigation, validation, and data curation. Sangam Jha: investigation and methodology. Khyati Sarma: investigation and validation. Neela Koner: investigation and validation. Suman Chakrabarty: software, supervision, resources, and funding acquisition. Dibyendu Das: conceptualization, project administration, supervision, funding acquisition, writing – review and editing, and resources.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

The authors have cited additional references within the Supporting Information [42, 43, 44, 45, 46, 47, 48, 49, 50, 51].

Supporting File 1: anie73310‐sup‐0001‐SuppMat.docx.

Supporting File 2: anie73310‐sup‐0002‐MovieS1.mp4.

Supporting File 3: anie73310‐sup‐0003‐MovieS2.mp4.

Data Availability Statement

The data that supports the findings of this study are available in the supplementary material of this article.