Homopolymer self-assembly of poly(propylene sulfone) hydrogels via dynamic noncovalent sulfone–sulfone bonding

Natural biomolecules such as peptides and DNA can dynamically self-organize into diverse hierarchical structures. Mimicry of this homopolymer self-assembly using synthetic systems has remained limited but would be advantageous for the design of adaptive bio/nanomaterials. Here, we report both experiments and simulations on the dynamic network self-assembly and subsequent collapse of the synthetic homopolymer poly(propylene sulfone). The assembly is directed by dynamic noncovalent sulfone–sulfone bonds that are susceptible to solvent polarity. The hydration history, specified by the stepwise increase in water ratio within lower polarity water-miscible solvents like dimethylsulfoxide, controls the homopolymer assembly into crystalline frameworks or uniform nanostructured hydrogels of spherical, vesicular, or cylindrical morphologies. These electrostatic hydrogels have a high affinity for a wide range of organic solutes, achieving >95% encapsulation efficiency for hydrophilic small molecules and biologics. This system validates sulfone–sulfone bonding for dynamic self-assembly, presenting a robust platform for controllable gelation, nanofabrication, and molecular encapsulation.

precipitating the homogeneous solution in THF. Apparent differences were discovered for PPSU20 and the random copolymers of sulfoxides and sulfones. First, PPSU20 powders are demonstrated by WAXD to be crystalline while the sulfoxides/sulfone mixtures are mostly amorphous. Second, the sulfoxides/sulfone mixtures are readily dissolved by water while PPSU20 solids are completely insoluble in water.
All-atom explicit solvent molecular dynamics simulations. Classical all-atom molecular dynamics simulations were performed using the CHARMM 36m force field 1 . The recommended CHARMM TIP3P water model 2 was applied with the structures constrained using the SETTLE algorithm 3 . The simulations were performed using the package GROMACS (version 2016. 3) 4 . In all the simulations, the degree of polymerization (DP) of 20 was employed for the polymer chains, the same as that PPSU20 in the experiments. The polymer chains were created in the extended form. Six extended PPSU20 chains were randomly dissolved in each of the three water boxes initially. Whereas in the DMSO systems, the initially extended PPSU20 chains were equilibrated in a vacuum, forming coiled configurations, and were then dissolved in the DMSO boxes. The initial configurations are provided in Supplementary Figure 3. See Supplementary Table 2 for the components of the systems. For both solvent conditions, three parallel simulations were performed.
The system potential energy was first minimized using the steepest descent algorithm, followed by the equilibration of 1 ps in the NVT ensemble (constant number of particles, volume and temperature). Subsequently the NPT ensemble (constant number of particles, pressure and temperature) was applied. An equilibration of 10 ps using the time step of 1 fs was followed by another equilibration of 0.1 ns using a time step of 2 fs in the DMSO system, or 2.5 fs in the aqueous system. Subsequently long equilibration simulations were performed. The periodic boundary conditions were applied in all three dimensions. The neighbor searching was performed up to a cut-off distance of 1.2 nm by means of the Verlet particle-based approach and was updated every 20 time steps. The potential-switch method was applied for the short-range Lennard-Jones (LJ) 12-6 interactions from 1 nm to 1.2 nm. The short-range electrostatic interactions were calculated up to 1.2 nm, and the long-range electrostatic interactions were calculated by means of the Particle Mesh Ewald algorithm 5 . A time step of 2 fs (2.5 fs) was employed by constraining all the covalent bonds using the LINCS algorithm 6 in the DMSO system (water system). The temperatures of the PPSU20 solute and the solvent molecules were separately coupled using the Nosé-Hover algorithm (reference temperature 298 K, characteristic time 1 ps). The isotropic Parrinello-Rahman barostat was utilized with the reference pressure of 1 bar, the characteristic Additionally, control simulations were performed. In the control simulations, initially extended polymer chains were employed for both DMSO and water solvent systems. The annealing simulations were performed to speed up the convergence of the equilibrations 7 . In the annealing simulations, the temperatures of polymer and solvent (DMSO or water) were separately coupled.
The temperatures started at 298 K initially, which increased to 353 K within 1 ns. They stayed at 353 K for 9 ns, then dropped to 298 K within 1 ns. Finally, the temperatures stayed at 298 K for another 9 ns. Therefore, each annealing cycle lasted 20 ns. 5 annealing cycles were performed for the DMSO systems (100 ns in total), and 8 cycles (160 ns in total) for the water systems.
Subsequently, the production simulations were performed for 40 ns each. Agreements were found in regards with the polymer chain distribution (molecularly dissolved in DMSO and aggregated in water), the polymer structures (end-to-end distance and persistence length) and the radial distribution function of the polymer sulfur atoms.
Calculation of the dipolar energy between neighbor charge-neutral units from atomistic molecular dynamics simulations. In all the atomistic simulations, the PPSU repeat units and the solvent (DMSO or water) molecules are charge-neutral. The monopole interactions between them could be reasonably expected negligible. By following a previous work 8 , we calculated the dipolar interactions between the neighbor units. Each PPSU repeat unit is defined as one charge-neutral unit, as well as one DMSO molecule and one H2O molecule. The dipolar interaction energy between the charge-neutral units is calculated as below: 1) The sulfur atoms of PPSU repeat units, the sulfur atoms of DMSO and the oxygen atoms of  Table 3). The first minima were chosen to define the upper distance of the neighbors, which was 6.  field were originally presented, and around 20% larger than the experimental value 10 . The dipole moment was calculated to be 6.534 D for PPSU repeat units.
3) The dipolar interaction energy between units i and j could be thus obtained by Humidity induced-aggregation of PPSU20 in DMSO. DMSO solutions of PPSU20 were exposed to humidity in air and the phase transition was tracked. Sol-to-gel phase transition was observed for a highly concentrated solution (200 mg mL -1 ) overnight and a low concentration solution (25 mg mL -1 ) became cloudy in 3 days. As the cloudy solution was allowed to age further (110 days), fluffy precipitates (can be observed in 7 days) were obtained by centrifugation. The fluffy precipitates were demonstrated by WAXD to be mostly amorphous (Supplementary Figure 9).
After dispersing in water, these fluffy precipitates were recollected by centrifugation. In samples treated this way, WAXD showed an increased crystallinity in Supplementary Figure 9.
CryoTEM imaging. Samples were prepared by applying 3 µL of sample (5 mg mL -1 ) on pretreated holey or lacey carbon 400 mesh TEM copper grids (Electron Microscopy Sciences).
Following a 3 s blot, samples were plunge-frozen (Gatan Cryoplunge 3 freezer). Images of samples entrapped in vitreous ice were acquired using a field emission transmission electron microscope (JEOL 3200FS) operating at 300 keV with magnification ranging from 2,000 × to 12,000 × nominal magnification. Digital Micrograph software (Gatan) was used to align the individual frames of each micrograph to compensate for stage and beam-induced drift. Any further image processing conducted on the aligned frames was completed in ImageJ.

SAXS measurements. SAXS measurements were performed at the DuPont-Northwestern-Dow
Collaborative Access Team (DND-CAT) beamline at Argonne National Laboratory's Advanced Photon Source (Argonne, IL, USA) with 10 keV (wavelength λ = 1.24 Å) collimated X-rays. All the samples (5 mg mL -1 ) were analyzed in the q-range (0.001-0.5 Å -1 ), with a sample-to-detector distance of approximately 7.5 m and an exposure time of 1 s. The diffraction patterns of silver behenate were utilized to calibrate the q-range. The momentum transfer vector q is defined as q = 4π sinθ λ -1 , where θ is the scattering angle. Data reduction, consisting of the removal of solvent/buffer scattering from the acquired sample scattering, was completed using PRIMUS 2.8.2 software while model fitting was completed using SasView 4.0.1 software package. The core-shell cylinder, vesicle and core-shell sphere models were utilized to analyze the data.
Transmission electron microscopy of negatively stained PPSU20 nanostructures. 1.5% uranyl formate (UF) was prepared in ultrapure water. The pH was adjusted to 4.5 by addition of 10 N KOH. 4 L of nanostructures were applied to glow discharged (25 W, 10 s) formvar carbon film copper grids (400 mesh, Electron Microscopy Sciences, Inc.). Samples were gently washed twice via passage through ultrapure water, and were negatively stained by passage through two 30 L volumes of 1.5% UF. Excess stained was removed by blotting each sample with Whatman filter paper. After this procedure, ~0.5 L stain remains on the grid with an activity of 2.55 × 10 -5 Ci grid -1 . Images were acquired at 30,000 × on a JOEL 1400 Transmission Electron Microscope operating at 120 kV.
Energy dispersive X-ray spectroscopy. 1 µL of PPSU20 nanobundles (5 mg mL -1 in water) were mixed thoroughly with 1 µL of 2% methyl cellulose and 3 µL of water. 5 µL of the mixture was deposited on formvar carbon film copper grids (200 mesh, Electron Microscopy Sciences, Inc.) that were glow discharged to create a hydrophilic surface. After resting for 15 s, the grid was incubated in 1.5% UF for 15 s. Excess stained was wicked away with filter paper. Scanning transmission electron microscopy images were taken with high angle annular dark field (HAADF) mode where bright contrast is indicative of a species with a higher atomic number. In this case, uranium from UF provides the contrast, and the absence of that contrast shows the structure of the Three parallel simulations were performed for both solvents (6 chains, 12.5 mg mL -1 ). Hydrogen atoms on PPSU20 and solvent molecules are omitted for clarity. The six PPSU20 chains are colored differently. The blue solid line denotes the simulation box. The Coul-SR and LJ-SR interactions were calculated up to a cutoff distance of 1.2 nm. The Ree and Rg were calculated using the sulfur atoms on PPSU. The dipolar interactions were calculated between PPSU monomers which are covalently connected (⃗⃗⃗⃗ − +1 ⃗⃗⃗⃗⃗⃗⃗⃗⃗), or neighbors of ⃗⃗⃗⃗ − +2 ⃗⃗⃗⃗⃗⃗⃗⃗⃗, or ⃗⃗⃗⃗ − +3 ⃗⃗⃗⃗⃗⃗⃗⃗⃗. All the calculations supported that the simulations were roughly converged after around 100 ns. Figure 4. Distribution of PPSU20 clusters in the DMSO simulations. PPSU20 chains are viewed as clusters if the distance of any inter-chain sulfur atoms is less than 0.67 nm (the first minimum in the radial distribution function in Supplementary Figure 8). A maximum probability occurs at the cluster size of 1 supports that the PPSU chains are dispersed in DMSO solvent 11 . The GROMACS program gmx clustersize was employed for the calculations.

Supplementary Figure 5. Atomistic simulation snapshots of PPSU20 in DMSO and water. (a)
The initially coiled PPSU20 chains turned into extended conformation in DMSO. (b) Under the application of solvent replacement from DMSO to water, network structure of PPSU20 formed due to inter-chain associations. Two parallel simulations were performed for both systems (22 chains, 25 mg mL -1 ).  Table 3). The RDFs between the sulfur atoms on PPSU20 in the aqueous solutions supporting a crystalline structure for PPSU20. Figure 9. WAXD patterns for PPSU20 precipitates collected from DMSO or water. The fluffy precipitates were obtained by exposing 200 μL of DMSO solution of PPSU20 (25 mg mL -1 ) to humidity in air for 110 days. After centrifugation, part of the precipitates was applied for WAXD directly, and the other precipitates was dispersed in water and recollected for WAXD. Figure 10. Quick hydration leading to large scale spatial redistribution of PPSU20 chains on network. (a) Exposure and aging of a highly concentrated DMSO solution of PPSU20 in air leads to a colorless gel. Thoroughly mixing of the gel with water results in a cloudy solution. (b) Cryo-TEM images of the cloudy solution showing PPSU20 nanostructures in a water-DMSO system (1/9, V/V) and in 100% water. Non-uniform aggregates including ribbons formed in 90% water, which reorganized into bundles and vesicular nanogels after dialysis to remove residual DMSO. Figure 11. CryoTEM of PPSU20 nanogels in water. The nanogels were prepared by stepwise hydration of DMSO solutions of PPSU20 (25 mg mL -1 , 200 μL) with 400 μL of water (see Fig. 2a . Statistically significant differences in nanostructure uptake in the absence (PBS pre-treated "Control") or presence of inhibitor (CytD or CPZ) was determined using Sidak's multiple comparisons test (*** p<0.0005, **** p<0.0001).