Complex Energy Landscapes of Self-Assembled Vesicles

The field of supramolecular chemistry has witnessed tremendous progress in bringing the system away from equilibrium for traditionally inaccessible structures and functions. Vesicular assemblies with complex energy landscapes and pathways, which are reminiscent of diverse cellular vesicles like exosomes, remain exceedingly rare. Here, relying on the activation of oligo(ethylene glycol) (OEG) interdigitation and the encoded conformational freedom in monodisperse Janus dendrimers, we reveal a rich landscape and a pathway selection of distinct vesicles. The interdigitation can be selectively switched on and off using temperature ramps, and the critical temperatures can be further determined by molecular design. Our findings suggest that synthetic vesicles, with different energy states and unexpected transition pathways, emulate dynamic cellular vesicles in nature. We anticipate that vesicles with an activated OEG corona conformation will open new routes for nanomedicine and advanced materials.


Table of Contents
hydrophobic alkyl chains, hydrophilic oligo(ethylene glycol) and penthaerythritol core was synthesized in nine steps, as described previously. 1 In a nutshell, the hydrophobic unit was 4 synthesized convergently via direct esterification of 3,4-hydroxybenzoates with 1bromododecane. The hydrophilic unit began with the tosylation of methyl-terminated triethylene glycol, followed by the reaction with methyl ester-protected gallic acid by Williamson ether synthesis. Corresponding acid was obtained by the removal of methyl ester group through hydrolysis. A benzylidene-protected strategy was used to differentially substitute the penthaerythritol core firstly with hydrophobic unit and then the hydrophilic unit.
All products of each step were purified and characterized with nuclear magnetic resonance (NMR) and matrix-assisted laser desorption/ionization-time-of-flight (MALDI-TOF) mass spectrometry.

S2 | Methods of molecular dynamics (MD) simulation
All MD simulations were conducted using the GROMACS 2021.5 package. 2 The CHARMM36 force field 3 with the TIP3P model 4 were adopted for the Janus dendrimers with complete −OMe or −OH end groups and water, respectively. The parameters of the dendrimers were assigned using CHARMM-GUI Ligand Reader & Modeler. 5 We prepared two simulation systems consisting of a fully hydrated single bilayer for each type of the dendrimer. The bilayer was composed 128 dendrimer molecules and was built using the MemGen web server. 6 The system temperature mentioned below was maintained with a Nosé-Hoover thermostat. [7][8] A Parrinello Rahman barostat with semiisotropic coupling was used to control the system pressure to be 1 atm. 9 Electrostatic interaction was calculated using the particle mesh Ewald method. 10 The cutoff length of Lenard-Jones interactions was set to 1.2 nm with a force switching distance of 1.0 nm. The water geometries were maintained constant using the SETTLE algorithm. 11 All bonds involving hydrogen atoms were constrained employing the LINCS algorithm. 12 The simulation time step was set to 2 fs. Energy minimization with the steepest decent algorithm was conducted to prevent unstable structure in the initial configurations. We performed 200 ns MD simulations to equilibrate the single bilayer systems at 353 K. Then, further 100 ns equilibration runs were performed at three different temperature of 340, 310, and 280 K, respectively. We carried out 1 µs MD simulations of the systems as described above. The last 500 ns MD trajectories were used for analyses.

S3.1 | NMR Spectroscopy
Nuclear magnetic resonance (NMR) spectra were recorded on a Bruker 400 MHz Avance III HD nanobay spectrometer with a BBFO probe. Tetramethylsilane (d = 0.0 ppm) or the residual protons of the deuterated solvent was used as the internal reference. 1 H NMR spectra were acquired using 32 or 64 scans and a relaxation delay of 5 s.

S3.2 | Mass Spectrometry
Matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry was performed on a Bruker Microflex LRF MALDI-TOF system equipped with a nitrogen laser (337 µm) and operating in reflection mode. Saturating matrix solution was prepared by The analytical sample was obtained by mixing a 1/1 (v/v) ACN/H2O or THF solution of analyte (2 mg/mL). The solution of the analyte and matrix was mixed in equal ratio and 1.0 µL solution was loaded on the MALDI plate and allowed to dry at room temperature before inserting into the vacuum chamber of the MALDI instrument. The laser steps and voltages applied were adjusted accordingly depending on the nature of analyte.
For temperature trend measurements in heating/cooling cycles, samples were equilibrated for 300 seconds at each temperature before each measurement. Based on the Stokes-Einstein equation, the size and distribution of the particles were derived from the fluctuations in scattered light intensity due to the Brownian motion of the particles by assuming a hard sphere model.

S3.4 | Nanoparticles tracking analysis (NTA)
NTA was performed with Nanosight LM10-HS instrument with a Marlin camera and a 60 mW blue laser illumination (405 nm). NTA offers a particle-by-particle methodology with high 10 resolution results of the particle size and concentration. Typically, sample solution (0.5 mg/mL) was diluted by 1000 times to ensure an optimized number of particles (10 7 -10 9 particles/mL) for analysis. The diluted solution was injected in a sample chamber. The Brownian motion of the nanoparticles was recorded in three videos of 60 seconds at 30 frames/s. The size and concentration of the sample were obtained by averaging the results as derived from the videos.
Multiple measurements were conducted for each sample by the injection of sample solution.    We performed quantitative measurement of the percent frequency of each morphology for each condition (Figure 2e, Figure 3h and Figure 4b−the bar chart). Images from different areas and different batches of samples were taken and counted to minimize the error. In average, images of each condition were randomly taken and more than 500 particles were counted.

S4 | Supplementary Figures & Discussion
As an example, the analysis of images of self-assemblies prepared by direct injection in the presence of ethanol was shown in Figure S5. The percentages for each type of vesicles were averaged and the results were presented as bar charts, illustrating the frequency distribution of each morphology. The results derived from Figure S5            A fully hydrated bilayer for each type of the dendrimer molecule (i.e. with −OMe and −OH end groups) was prepared using molecular dynamics (MD) method. Figure S15a (Table S1). The results showed that the number of water molecules hydrating the −OH end groups was more than twice the number of water molecules hydrating the −OMe end groups at all three temperatures. Based on these findings, it is evident that the OEG of −OH end-group dendrimer exhibits a higher degree of hydrophilicity compared to the −OMe endgroup dendrimer across all temperatures investigated in the study.