Unraveling an Alternative Mechanism in Polymer Self-Assemblies: An Order–Order Transition with Unusual Molecular Interactions between Hydrophilic and Hydrophobic Polymer Blocks

Polymer self-assembly leading to cooling-induced hydrogel formation is relatively rare for synthetic polymers and typically relies on H-bonding between repeat units. Here, we describe a non-H-bonding mechanism for a cooling-induced reversible order–order (sphere-to-worm) transition and related thermogelation of solutions of polymer self-assemblies. A multitude of complementary analytical tools allowed us to reveal that a significant fraction of the hydrophobic and hydrophilic repeat units of the underlying block copolymer is in close proximity in the gel state. This unusual interaction between hydrophilic and hydrophobic blocks reduces the mobility of the hydrophilic block significantly by condensing the hydrophilic block onto the hydrophobic micelle core, thereby affecting the micelle packing parameter. This triggers the order–order transition from well-defined spherical micelles to long worm-like micelles, which ultimately results in the inverse thermogelation. Molecular dynamics modeling indicates that this unexpected condensation of the hydrophilic corona onto the hydrophobic core is due to particular interactions between amide groups in the hydrophilic repeat units and phenyl rings in the hydrophobic ones. Consequently, changes in the structure of the hydrophilic blocks affecting the strength of the interaction could be used to control macromolecular self-assembly, thus allowing for the tuning of gel characteristics such as strength, persistence, and gelation kinetics. We believe that this mechanism might be a relevant interaction pattern for other polymeric materials as well as their interaction in and with biological environments. For example, controlling the gel characteristics could be considered important for applications in drug delivery or biofabrication.


Polymer synthesis
Briefly, the initiator MeOTf (1 eq.) was added to a dried and argon flushed flask and dissolved in the 21 mL of solvent (PhCN). The first monomer EtOx or MeOzi (35 eq.) was added to the reaction mixture and heated to 100°C for several hours. After complete consumption (verified via 1 H NMR), the monomer for the second block PheOzi (15 eq.) was added at room temperature. The reaction mixture was heated to 120°C over night. After complete monomer consumption was confirmed, the 3 rd block EtOx or MeOzi (35 eq.) was added and stirred for 2 several hours at 100°C. Termination was carried out by the addition of 3 eq. of 1-Boc-piperazine (PipBoc) or Ethylisonipecotate (EIP) at 50°C and kept on stirring for 6 hours. The solvent was removed at reduced pressure. The raw product was dissolved in deionized (DI) water and dialyzed (MWCO 1 kDa, cellulose acetate) against DI water for 3 days. The polymer solution was lyophilized, and the polymers were obtained as a white powder.

Differential Scanning Calorimetry (DSC)
All measurements were performed using aluminum crucibles on a calibrated DS 204 F1 Phoenix system from NETZSCH (Selb, Germany) equipped with a CC200 F1 controller unit from -50°C to 200°C with three heating and two cooling phases and a cooling rate of 10°C/min. The third heating cycle was used to analyze the glass transition temperature of dried polymer powders.

Rheology
All experiments were performed using an Anton Paar (Ostfildern, Germany) Physica MCR 301 system utilizing a plate-plate geometry (25 mm diameter) equipped with a solvent trap and Peltier element for temperature adjustment. All aqueous 15 wt.% samples were dissolved at room temperature stirring constantly and incubated at 5°C for 48 h. In addition, pictures were taken to visualize the gels.
A temperature-sweep was performed in oscillation mode from 5-50°C (heating rate: 0.05°C/s) using a fixed amplitude of 0.1% and angular frequency of 10 rad/s. The long-time gelation experiment at 5°C was performed at an amplitude of 0.1% and an angular frequency of 10 rad/s for several hours.

Transmission electron microscopy
For transmission electron microscopy (TEM) experiments, the polymers were dissolved in DI water to a final concentration of 20 g L −1 and stored at room temperature. 400 mesh copper-rhodium grids (maxtaform) with a homemade carbon layer were glow discharged in air for 1.  Viscosities determination. Viscosities were calculated at 5, 12, 22 and 37°C by the eq. S1 log = log + log (S1) where -fluorescence lifetime of BPC12/DASPMI in the solution of a given viscosity , and and are constants. The linear part of the log τf versus log η dependencies is taken as calibration plot and the viscosities are calculated using the equations of the linear fits. Calibration plots and equations for DASPMI are shown in ( Figure S3.1, right) and those of BPC12 rotor are taken from the literature (Table   S3.1). 7 Four series of nine water/glycerol mixtures of different viscosities at concentrations of glycerol between 80 wt.% and 100 wt.% were prepared as a calibration set. Viscosities (η) of each binary mixture were measured at 5, 12, 22 and at 37°C using a LOVIS 2000M rolling ball microviscosimeter from Anton Paar (Graz, Austria) with a LOVIS 1.8 capillary and a steel ball of 1.5 mm diameter. Prior to viscosity measurements, the density of the sample at the specific temperature was recorded using a DMA 4100M density meter from Anton Paar (Graz, Austria). Each water/glycerol mixture was used to dissolve 5 µM of DASPMI and the corresponding fluorescence intensity decays were recorded in quartz cuvettes using the TCSPC system (PicoQuant, GmBH) described above (
Spectra are shown as an average spectrum of 3 spectra at different locations on the same sample, which were recorded with an integration time of 5 s and 10 accumulations. The resulting data were processed with cosmic ray removal and background subtraction. The bulk water signals at 3100 cm -1 -3700 cm -1 were fitted using a Gaussian deconvolution method as described elsewhere. 8-9

Nuclear magnetic resonance (NMR) experiments in solution
All experiments in solution were performed at a Bruker Avance III HD 600 spectrometer (Karlsruhe, Germany) operating at 600.4 MHz equipped with a BBFO 5 mm probe using a BCU-02 temperature control unit. 1  Multiple-quantum (MQ) NMR using a pulse sequence based on the experiment of Baum and Pines was applied to probe the overall magnitude of residual dipolar couplings in chains subject to restricted conformational dynamics. 10

Solid-state nuclear magnetic resonance
Solid-state NMR (ssNMR) measurements were performed using a 4 mm double-channel Bruker probe at 9.4 T using between 3 and 7 kHz magic angle spinning (MAS). The hydrogel sample was cooled to 273 K prior to measurement. For the 13 C CP MAS experiment, a 2 ms ramp (50 to 100 %) on the 1 H channel was used during the cross-polarization (CP) contact time for all samples. 13

Molecular modeling
Three systems, each containing eight chains of a single polymer type, were modeled: Me-pMeOx35-b- pMeOzi35-PipBoc (Me = methyl group, EIP = ethyl isonipecotate, PipBoc = 1-Boc-piperazine). The hydrophobic pPheOzi blocks faced each other to form a single inner strand along the Z axis and the hydrophilic A-blocks were bent outwards. Four individual molecules made up the strand and were subsequently duplicated and moved next to the original polymers along the Z axis, ultimately resulting in two layers of polymers in each simulation box. The stretched-out, hydrophilic A-blocks were subjected to an energy minimization and a short, 50 ps long simulation with the Noisé-Poincaré-Andersen method [11][12] (applying the Amber14:EHT force field [13][14] with the R-field implicit solvation model 15 ) to yield a more compact starting conformation, while keeping the inner strand in a straight orientation ( Figure S8.1). All modeling was performed with MOE (Molecular Operating Environment 2019.01). 16 This setup was inspired by previous modeling studies regarding worm-like micelles of small molecules, in which the generation of a continuous micelle was also achieved via initial placement of hydrophobic parts in the inner and hydrophilic moieties in the outer regions of the threadlike structure, which was aligned along one axis of the simulation box. [17][18][19] RESP partial charges 20 of single monomers used as building blocks were derived from calculations with Gaussian 09 Rev. C.01 21 (Hartree-Fock level of theory, 6-31G* basis set); parameters based on the Amber14ffSB 13 and GAFF2 22 force fields were assigned via antechamber and parmchk2 of AmberTools18. [23][24] During charge derivation monomers were capped with residues of the same type and terminal groups with an A-block monomer. The calculated parameters were used to generate polymers with an initial straight conformation using tleap. [23][24] Starting structures were solvated with TIP3P water 25 in a simulation box with a minimum border-topolymer distance of 20 Å in the X and Y directions. Water molecules found inside the inner hydrophobic strand after this initial placement were removed if the distance to the pPheOzi blocks was less than 10 Å. Periodic boundary conditions with minimum image convention were applied during simulations, which allowed for infinitely sized worm-like micelles along the Z axis and ensured a sufficient distance between polymers of neighboring boxes along the X and Y dimensions. Simulations were performed using NAMD 2.13 26 with 2 fs time steps. An initial energy minimization of 10,000 steps was conducted before slowly heating the system from 100K to 278K over the course of 500 ps. Harmonic constraints were initially applied on all polymers and gradually reduced over an additional 1.6 ns, allowing rapid reordering of solvent molecules around the polymers. Langevin dynamics and the Nosé-Hoover Langevin piston method (1 atm) were used for temperature and pressure control in an NPT ensemble.
After another 2 ns of additional equilibration, the production run was performed for 600 ns. Semiisotropic coupling allowed for fluctuations along the Z axis, independent from the X and Y axes. The particle mesh Ewald method 27 with a cut-off of 1.2 nm was applied and coordinates were saved every 10 ps. Subsequent analyses were performed using CPPTRAJ. 28 Average densities for polymer groups around pPheOzi monomers were retrieved as follows: All pPheOzi residues were iteratively aligned onto the same monomer. Next, binned occupancy histograms of different moieties around the center of the aligned residue were calculated for the last 100 ns using the grid command in CPPTRAJ. This was performed on a 1.6 x 1.6 x 1.6 nm³ grid with a 1 Å resolution. After this procedure, the obtained values around each pPheOzi monomer were added up at each grid element and divided by the number of analyzed frames (10000) and monomers (104). Thus, densities represent the average amount of atoms of interest found at each grid element per frame around a single monomer. The first and last pPheOzi monomer of each pPheOzi block (16 out of 120) were excluded from this calculation, as these are always situated near neighboring A-blocks. Additionally, several distances between these pPheOzi monomers and the other polymer residues were analyzed, as well as the angle ω between the plane of nearby amide (N-(C=O)-C) groups and the phenyl ring plane for every 10 ps of the last 100 ns.

Complementary Material
Chapter S1 -SAXS and WAXS measurements The intensity I as a function of Q from the SAXS measurements was plotted for different temperatures for P1. In the hydrogel state (5°C, blue), a pronounced structure peak ( Figure S2 29 Above Tgel, the second species at high Q-values (worm-like micelles) disappears, confirming that the order-order transition found at low concentration also occurs at higher concentration. In addition, the structure peak is shifted towards higher Q-values indicating a lower particle/particle distance in the sol state due to the formation of small spherical micelles (vertical lines). This is a logical consequence of the disintegration of relatively few worm-micelles into much more numerous spherical micelles. The extrapolation of the absolute intensity I to Q0 (horizontal lines) was used as a measure for relative mean particle size. In the gel state (5°C), a higher I(Q0) value compared to the liquid state was observed, indicating once more larger particles in gel state (worm-like micelles) in comparison to the sol state (spherical micelles). The qualitative analysis of the SAXS scattering profiles is summarized in more detail in Table S1 below. Additionally, temperature-dependent wide angle X-ray scattering of P1 ( Figure S2.1 B) can provide insights into intra-und interpolymer interactions as previously described for biopolymers 30 and thermogelling peptide. 31 In the gel state (5°C, blue), a rather defined peak centered around 4.2 Å was observed, which could be interpreted to hint towards π-π interactions of the phenyl moieties. 32 With increasing temperature, the peak position was maintained, but a noticeably broadening between 4.2 and 7 Å indicates reduced order and increased degrees of freedom.     Fluorescence intensity together with the shift of fluorescence spectrum maximum provides information about the phase transition of the polymer as well as the polarity changes of the probe microenvironment. For example, changes in the fluorescence intensity and wavelength of DASPMI have been attributed to sol-to-gel transition as well polarity changes of the microenvironment. 6,33 Polarity-sensitive properties of BPC12 and its derivatives have also been noticed, although with a lower degree of responsiveness. 34 The same probes have been used to determine microviscosities of hydrophylic (DASPMI) and hydrophobic (BPC12) compartments of the self-assembled systems. 6 With molecular rotors such as these, the viscosity sensitivity can be seen as changes in the fluorescence lifetime, which is affected by the rotation ability of the structural segments with respect to each other.
This in turn is strongly dependent on the immediate molecular environment. However, it has to be kept in mind that it is not always clear what this immediate molecular environment is exactly.   Table S3.1) at four temperatures (5, 12, 22 and 37 °C). In contrast to bulk viscosity (macroviscosity) of P1, which is higher in the gel state, higher microviscosities were obtained in the sol state and the hydrophilic probe gave almost two times higher values than the hydrophobic one. The microviscosities decreased upon gelation and were similar for both probes below 25 °C ( Figure S3.3). In case of P2, a similar but less pronounced trend was observed for DASPMI i.e. a decrease in microviscosity upon gelation, while for BPC12 the microviscosity remained almost the same throughout the whole temperature range. This observation may indicate that upon gelation of this copolymer, the condensation takes place in the hydrophilic corona, but the hydrophobic core is less affected during spheres-to-worms transition. Moreover, it seems that slightly lower temperatures are needed for gelation of P2 compared to P1 (Figure S3  This increase suggests that the probes can be partially expelled from polymeric self-assemblies and become more exposed to polar aqueous solvent. For P2, the change in polarity of the microenvironment for DASPMI is less pronounced while BPC12 does not show significant spectral shift. The bathochromic shift of fluorescence spectrum maximum for DASPMI still indicates that the environment of the probe becomes more polar upon micelles-to-worms transition. An even less pronounced temperature dependence was observed for both BPC12 and DASPMI when studying P3 thermogelation ( Figure S3.4), where both dyes exhibited negligible hypsochromic shifts. It is noteworthy that the initial polarity of the DASPMI microenvironment in P3 micelles is significantly lower than that for the other two studied block-copolymers. The conclusion is drawn given the fluorescence maximum of DASPMI in P3 micelles is centred at ca. 570 nm, which is 20 nm shorter than those of P1 and P2 sols.
Ultimately, the time-resolved and steady-state fluorescence measurements clearly show that the microenvironment of both molecular rotors is more polar and less viscous in the gel state, suggesting that the gelation of P1 causes a probe migration out of the condensed polymeric assembly closer to the polymer-water interface. In the case of P2 this effect can be seen only for DASPMI, which is presumably co-localized in the hydrophilic shell, indicating the crucial role of EtOx moieties upon thermogelation. It seems that rather weak intermicellar interactions of P3 do not lead to any changes in the microenvironment of either dye in terms of polarity or microviscosity.

Chapter S4 -Raman measurements
In the gel state of P1, a sharp and moderately intense signal is observed at 731 cm -1 , which is much weaker and barely resolved in the sol state. For P2, also a moderate signal at 731 cm -1 is observed that is slightly weaker in the sol state, while a signal at 739 cm -1 barely distinguishable in the gel state becomes markedly stronger in the sol state. For P3, we observed a clear peak at 804 cm -1 in the gel state, which is absent in the sol state. In this spectral region, we expect C-C stretching modes, which are abundant in our polymers. At this point, we cannot hypothesize on the assignment of this signal without further understanding of the system. All three polymers show strong signals around 1000 cm -1 , attributed to the phenyl ring and the C-H in plane bending mode. Interestingly, for P1 and P2, the main signal shows a hypsochromic shift, while in P3, we observe a bathochromic shift. At 1464 cm -1 , a small but clearly distinguishable peak is exclusively present in the gel state of all polymers. Unfortunately, both aromatic ring vibrations as well as CH3 and CH2 deformation vibrations ubiquitous in the polymer backbone and hydrophilic sidechains appear in this region, making an unambiguous assignment challenging.
In addition, for P1 and P2, a clear difference between sol and gel state is also observed in the OH region of 3100 cm -1 to 3600 cm -1 , originating from water molecules. The different types of bonding modes in water molecules can be categorized using Gaussian deconvolution to divide the OH region into areas with different binding strength. 8,35 For P1, this suggests that water is less mobile in the hydrogel compared to the sol ( Figure S4.1B) as indicated by the increased contribution of the peak at 3250 cm -1 ( Figure S4.1B, red line, 1). Qualitatively, the situation for P2 appears very comparable. For P3, the water evaporated too quickly at 40°C and we could not obtain suitable spectra despite various adjustments in measurement parameters.

Chapter S5 -NMR experiments in solution
For a more quantitative assessment of the changing 1 H NMR signal intensities during the temperature induced phase transition the fraction p was calculated with the integrals I(T) and I(T0) at the respective temperatures T and T0 using the following equation (S1). 1 H NMR spectra were measured from 5°C to 40°C with the highest signal intensity at 40°C, which is therefore defined as T0 giving a p-ratio of 0.
Decreasing signal intensities at lower temperatures could be quantified by 0≤p≤1.       Occupancy density analyses around aligned pPheOzi residues (white sticks), showing hotspots for different polymer structures as meshes from two different perspectives. In 1) and 2) the violet densities represent pPheOzi side chains and the gray densities pPheOzi backbone atoms (isovalues: 0.08). Structures in 3) and 4) depict densities (isovalues: 0.03) for A-block backbone atoms (green) and side chain atoms (orange) from two different perspectives. 5) illustrates an exemplary MD snapshot in which residues at the surface of the micelle overlap with occupancy hotspots. Herein, A-blocks are shown with blue, turquoise and green carbon atoms, respectively, pPheOzi residues with magenta carbon atoms, and the aligned monomer of interest is highlighted in yellow. Densities are shown analogously as in the illustrations 1) to 4).

Figure S8.3
Histograms (bin size: 0.1 Å) for all distances up to 6 Å between pPheOzi moieties and the nearest polymer atoms (excluding hydrogen atoms). Dark blue describes P1, light blue P2 and green P3. Plots show the total amount of occurrences for all snapshots of all 104 pPheOzi residues which were used for density calculation. Barplots next to each histogram further compare the total amount of occurrences below 4 Å for each polymer type.