Mechanoresponsive Metal-Organic Cage-Crosslinked Polymer Hydrogels

Graphical Abstract

The utilization of homobifunctional poly(ethylene glycol) chains terminally substituted with bipyridines as ligands for Pd₆L₄ cages enabled crosslinking of the metal-organic cages and subsequent network formation, leading to supramolecular polymer hydrogels. The hydrogels allow for non-covalent guest uptake and release using ultrasound from aqueous solution.

Abstract

We report the formation of metal-organic cage-crosslinked polymer hydrogels. To enable crosslinking of the cages and subsequent network formation, we used homodifunctionalized poly(ethylene glycol) (PEG) chains terminally substituted with bipyridines as ligands for the Pd₆L₄ corners. The encapsulation of guest molecules into supramolecular self-assembled metal-organic cage-crosslinked hydrogels, as well as ultrasound-induced disassembly of the cages with release of their cargo, is presented in addition to their characterization by nuclear magnetic resonance (NMR) techniques, rheology, and comprehensive small-angle X-ray scattering (SAXS) experiments. The constrained geometries simulating external force (CoGEF) method and barriers using a force-modified potential energy surface (FMPES) suggest that the cage-opening mechanism starts with the dissociation of one pyridine ligand at around 0.5 nN. We show the efficient sonochemical activation of the hydrogels HG_3–6, increasing the non-covalent guest-loading of completely unmodified drugs available for release by a factor of ten in comparison to non-crosslinked, star-shaped assemblies in solution.

Introduction

Hydrogels are soft polymer networks that trap and are swollen by large amounts of water. They can be formed by covalent or physical crosslinking, chain entanglement, or by supramolecular self-assembly of more complex molecules.¹ Metal-organic cage-based gels allow for a modular design, enabling access to soft materials by self-organization, offering promising solutions to current scientific challenges by adding an additional layer of complexity.^{1, 2} They form a three-dimensional network with distinct pore sizes and shapes. Their structures rely on non-covalent interactions and can thus be dynamic, changing properties in response to a variety of external stimuli to exert functions, thus providing smart materials for a wide range of applications.³ This can be accomplished primarily by first constructing the cage/polyhedral or by polymerization of a macroligand.⁴ They can also be formed by orthogonal metal coordination and host-guest interplay⁵ and can be used to assemble cages into nanowire-based soft films.⁶ Several linked metal-organic polyhedral (MOP) gels that combine processability with permanent porosity and respond to stimuli have been reported by the group of Furukawa,⁷ with spatiotemporal control of the gel formation using photoacids demonstrated by several groups.^{7d, 8} The group of Nitschke was one of the first to show selective encapsulation of guests within the two phases of hydrogels containing cages and was also able to fabricate them into responsive microparticles.^3e Employing Fe-based cages grafted with oligoethylene glycol imidazolium chains instead, conformational change upon heating allowed for the construction of a heat engine traversing multiple phase boundaries,^3b in addition to creating permanently porous ionic liquids based on coordination cages.^3c The groups of Nitschke and Marchesan embedded pristine cages in tripeptide gels, forming hybrid cage-based nanostructured materials.⁹ Johnson and co-workers also contributed switchable materials based on the interconversion of Pd₃L₆-type cages and Pd₂₄L₄₈-type spheres,^10b assemblies based on M₁₂L₂₄ spheres,^{10d, 11} multi-component assemblies,^10c as well as complex Cu-based gels that undergo reversible, photoredox-induced disassembly and enable catalytic reactions within the networks.^10a We recently reported a star-shaped polymer with an M₆L₄-type Pd-cage in its centre, enabling mechanochemical release of non-covalently bound guests from aqueous solution.¹²

The on-demand, mechanochemically-induced release or activation of compounds, such as catalysts, drugs, or monomers for self-healing, is highly desirable because it allows for the activation of pristine small molecules from macromolecular frameworks.^{12, 13, 14, 15, 16} While this can also be achieved using supramolecular mechanoresponsive assemblies, the structural integrity of the carrier and leaching can be problematic in these systems.^13f M₆L₄-type Pd-cages are generally flexible, stable, and excellent hosts, as established by Fujita and others,¹⁷ rendering them an appealing choice for hydro- and organogel applications.

Here, we report the formation of metal-organic cage-crosslinked polymer hydrogels. In addition to their characterization, including comprehensive small-angle X-ray scattering (SAXS), the encapsulation of pharmaceutically active compounds into supramolecular self-assembled metal-organic cage-crosslinked hydrogels, as well as ultrasound-induced disassembly of the cage with release of its cargo, is presented. Experimental results are additionally supported by the force-modified potential energy surface (FMPES) and constrained geometries simulating external force (CoGEF) method.¹⁸ We utilise homodifunctionalized poly(ethylene glycol) (PEG) chains terminally substituted with bipyridines as ligands for the Pd₆L₄ corners to enable crosslinking of the cages and subsequent network formation (see Figure 1).¹⁹ We anticipate these cages to be susceptible to the force-induced scission of the weak Pd−N bond (Figure 1), which is analysed by computations using the Constrained Geometries simulating External Force (CoGEF) method¹⁸ and by the force-modified potential energy surface (FMPES)²⁰ barrier heights.

Figure 1

a) Formation of metal-organic cage-crosslinked polymer hydrogels by terminal functionalization of PEG with bipyridine ligands, with hydrogels formed from the corresponding palladium nitrate precursors in combination with 2,4,6-tri(4-pyridyl)-1,3,5-triazine (TPT) (Figure S1); b) ¹H NMR (600 MHz, D₂O) of HG₃ without a guest (top), HG₃ ⋅ (ibuprofen)₂ encapsulating two ibuprofen moieties (middle) and SC ⋅ (ibuprofen)₂ as reference (bottom),^12a see also Figure S2 for an additional detailed representation of ¹H NMRs of guest encapsulation experiments.

Results and Discussion

We synthesised a series of metal-organic cage-based gels varying in their linker lengths, ranging from an average molar mass (M_n) of 1 to 6 kDa (Figure 1). For the small hydrogel HG₁, commercially available PEG (M_n=1 kDa) was functionalized with 2 equiv. of 4-bromomethyl-4′-methyl-2,2′-bipyridine in a nucleophilic substitution reaction yielding linker 4 (see Supporting Information, Figure S1). Followed by a two-step reaction, the corresponding palladium complex 6 bearing nitrate counterions was obtained in over 89 % yield. A comparison between the characteristic aromatic bipyridine signals in the ¹H NMR and those of the PEG-backbone gave a 12 : 92 ratio, which corresponded well to the anticipated average molar mass of the used PEG for 6 (Figure S39). ¹³C NMR, heteronuclear 2D measurements, and mass spectrometry further confirmed the successful formation of 6. The isostructural linkers for HG₃ and HG₆ were synthesised analogously using PEG with M_n=3 or 6 kDa. When 2,4,6-tri(4-pyridyl)-1,3,5-triazine (TPT) (4 equiv.) was added to an aqueous solution of 6 (3 equiv.) and heated to 50 °C for 1 h, a pale-yellow, clear gel (HG₁) was obtained. Annealing for 12 h ensured complete conversion of the starting material, which was confirmed by ¹H NMR (Figure 1b top and S57). To our surprise, 5 wt.% of the building blocks in H₂O resulted only in a viscous solution, but network formation occurred using 10 wt.% and 20 wt.% samples, forming stable gels. For HG₃, gel formation was observed after 1 h for 10 wt.%, 20 wt.%, and after 2 weeks, even for the 5 wt.% sample. In the case of HG₆, all samples formed a hydrogel. The dynamic behaviour of the hydrogels was investigated by swelling experiments. The hydrogels of all three linker lengths were freeze-dried to yield aerogel-like structures, and a 20 wt.% sample of HG3 was investigated by scanning electron microscopy (SEM, Figure 2a and 2b). The addition of H₂O to regain a 10 wt.% hydrogel resulted in gel formation in all cases after 1 d. To these hydrogels, the same amount of H₂O was subsequently added once per day, leading to further swelling until the gel structure ultimately collapsed and formed a solution that was not stable to inversion anymore (Figure S13). Under these conditions, HG₁ was capable of absorbing 18× its own mass in H₂O before further addition led to dissolution. The swelling ratio could be increased to 27× for HG₃ and 45× for HG₆. These results confirm that the maximum mesh size of HG₆, which is obtainable by uncoiling the PEG linkers, is significantly greater than that of HG₁. All three hydrogels were further characterised in a comparable state using SAXS measurements.²¹

Figure 2

a) SEM image of a 20 wt.% HG₃, freeze-dried prior to the measurement (scale bar 100 μm); b) SEM image of a 20 wt.% HG₃, freeze-dried prior to the measurement (scale bar 10 μm); c) SAXS profiles of the freeze-dried hydrogels in blue: HG₁ (top), HG₃ (middle) and HG₆ (bottom) with corresponding fits comprised of a power law and broad peaks (solid black lines, profiles are shifted by fixed multipliers for better visibility; d) SAXS profiles of HG₃ after freeze drying before sonication (light blue) and after sonication (red). The scattering profiles are shifted by fixed multipliers for better visibility.

For this, freeze-dried samples offer the advantage of minimising scattering contributions from dynamic network fluctuations and thus enabling focus on the internal network structure. Figure 2c shows the respective SAXS profiles recorded from freeze-dried hydrogels HG₁–HG₆. At first glance, the profiles appear very similar. In the low q region, the scattering intensity increases linearly in the presented double logarithmic representation, with the scattering following a simple power law with slopes close to q⁻⁴ (Porod law). This scaling is indicative of scattering from sharp interfaces. The fact that the scattering intensities do not reach values at low q points to global structures being larger than the covered q range. We can conclude that the aerogel-like samples feature sharp interfaces between the matrix and the included air. In the mid-to-high q region, all scattering profiles show several distinct peaks with relatively large peak widths. These peaks can be well described by simple, broad peak contributions. The solid black lines in Figure 2c correspond to fits based on the sum of the power law contribution and several broad peaks. The broad peak contributions are indicative of well-defined internal length scales of the samples, providing information on the structure factor of the gels. We attribute the first structure peak at lowest q to the average mesh size of the gels, with real space values of 4.03 nm (HG₁), 9.79 nm (HG₃), and 10.54 nm (HG₆). All peak positions, correlation lengths, and real space values are listed in Tables S1 and S2. The increase in mesh size from HG₁ to HG₆ is in good agreement with the results from our swelling experiments. The peaks at highest q appear at nearly the same position for all three samples (q≈6 nm⁻¹). This q value corresponds to a real space distance of approximately 1 nm. This agrees well with the dimensions of the crosslinking metal complexes but could also originate from local inhomogeneities in the PEG chain. The distinction between both is difficult due to structural differences in the gel compared with its components.

A more detailed discussion of the SAXS profiles and additional SAXS profiles, for example, hydrogels with a defined water content, can be found in the Supporting Information (Figure S17). To further explore the mechanical property space of the hydrogels, we carried out shear rheology using a plate rheometer. Initially, we measured storage (G′) and loss (G′′) moduli during polymerization of HG₃ and HG₆ at 60 °C (Figure 3a). We found that gelation occurred in seconds, and the typical crossover point could not be measured on this timescale. Complete polymerization was, however, already achieved after ca. 60 min, as indicated by the asymptotic development of G′′. Notably, HG₃ was about one order of magnitude stiffer than HG₆, confirming the mesh size trends obtained from the swelling and SAXS experiments. Cooling the hydrogels to 25 °C (Figure 3b) and heating them to 85 °C (Figure 3c) then proved that, on the one hand, the mechanical properties, as expressed through the complex viscosity η*, were widely temperature invariant. On the other hand, these measurements revealed a thermal stability that is unusually high for hydrogels. The minor stiffening effect upon heating was possibly caused by the lower critical solution temperature (LCST) behaviour of PEG in the nitrate salt solution which was received after the in situ gel formation.²² Frequency-dependent measurements were then carried out to better understand the potentially dynamic character of the supramolecular hydrogel (Figure 3d). Both HG₃ and HG₆ showed typical hydrogel behaviour with a frequency-invariant G′. Dynamic bond rearrangement is usually indicated by a crossover point and was not observed in the coverable frequency regime.

Figure 3

Shear rheology of 10 wt.% HG₃ and HG₆. a) G′ and G′′ at strain amplitude γ₀=1 %, frequency ω/2π=1 Hz, and temperature T=60 °C upon polymerization time sweep. b) η* at γ₀=1 % and ω/2π=1 Hz upon cooling to T=25 °C; c) η* at γ₀=1 % and ω/2π=1 Hz upon temperature sweep; d) G′ and G′′ at γ₀=1 % and T=25 °C upon frequency sweep.

Guest-binding studies were carried out for all three hydrogels. HG₃ can indeed be loaded with ibuprofen, progesterone, or drospirenone, respectively, by adding an excess (6 equiv. per cage) of the guest to the reaction mixture, facilitating complete conversion to the host-guest complex. ¹H NMR confirmed the successful encapsulation of the guests by observing the distinctive upfield shift of the signals in the area of −0.5 to −1.3 ppm, caused by the shielding of the employed TPT panels (Figure 1b and S2b and S83). The encapsulation of ibuprofen was repeated for HG₁ and HG₆, confirmed by ¹H NMR (Figures S76 and S93), and is identical to HG₃.

In addition, we synthesised a Pd₆L₄-type cage SC as a reference bearing no polymer chains, allowing us to obtain ¹H, ¹³C, ¹H DOSY NMRs, and heteronuclear 2D measurements of the cage and its incorporated guests (Figures S71 and S75). As seen in Figure S2, guests encapsulated in the hydrogels show considerable broadening of the resonances, with similar chemical shifts to SC. Subsequent sonication experiments were conducted to determine the mechanochemical conversion in relation to polymer chain length. We hypothesised that due to the crosslinked nature of the polymer network, an increase in the relative cage concentration with a simultaneous decrease in average polymer segment length in the hydrogels (Figure S10b–d) would still warrant efficient mechanochemical activation comparable to supramolecular cages SP in solution.¹² It is known that mechanochemical reactivity scales with chain length.²³ Hence, we envisioned that the nearly infinite molar mass of the hydrogels would significantly increase their susceptibility to shear forces, analogously to microgels.²⁴ Therefore, sonication experiments were performed with an immersion probe sonicator (20 kHz) in H₂O, monitored by ¹H NMR (Figure S10). After 3 h of effective sonication time, no considerable amount of activation was observed for the shortest hydrogel HG₁ (Figure S10b), showing no change in ¹H NMR. Conversely, HG₃ already showed a significant amount of activation, with fragmentation products appearing during sonication, which are highlighted in blue (Figure S10c). For HG₃, the integral at δ=7.85 ppm increased in magnitude to 8.26, which corresponds to roughly 25 % of activation. Further increasing the polymer length of the hydrogel (HG₆) resulted in an increased activation of around 37 % (Figure S10d). This activated fraction of HG₆ (6 kDa) is equivalent to mechanoresponsive SP (20 kDa) (Figure S10e). As the hydrogel HG₃ indicated mechanochemical activation, it was used for guest release studies. To enable a quantitative analysis of the release, we added an external reference (maleic acid, 20 mM) to the NMR sample (10 mg in 0.7 mL D₂O) before and after sonication (Figure 4). As seen in Figure 4a, we normalised the cage to its characteristic aromatic bipyridine signals before sonication and received an integral for the maleic acid of 3.43 at δ=6.4 ppm. Transferring this to the ¹H NMR of the sonicated HG₆ provides detailed information about the fragmentation. A significant decrease in cage signals (δ=7.6, 8.5, 9.0 and 9.5 ppm) can be observed, while fragmentation product signals increase (δ=7.8 and 8.6 ppm) (Figure 4b). An activation of approximately 40 % can be attributed to HG₆, which is in accordance with the results discussed earlier. HG₆ ⋅ (ibuprofen)₂ was sonicated for 3 h and analysed by ¹H NMR with maleic acid as an external reference (δ=6.4 ppm) ultimately resulting in a verifiable release of its cargo (Figure 4d), judging from disappearing signals corresponding to the loaded guest precipitating from the aqueous solution and the occurrence of fragments of the cage. Since the uptake of the non-covalently bound guest within the cavity of the host is largely facilitated by the hydrophobicity of the guest, the guest precipitates from the solution upon release.^{12, 17} The decrease of the signals at δ=1.3, 0.6, 0.3, and −0.5 ppm confirms the release of encapsulated ibuprofen from the hydrogel and corresponds to an estimated release of ca. 66 % after 3 h (further prolonging the reaction time leads to increased release). The impact of sonication on sample HG₃ was also studied by SAXS. Figure 2d additionally compares the SAXS profiles prior to and after sonication. For the sonicated sample, broad peaks stay in the same position but lose significantly in intensity. The high q peak vanishes completely. This is in excellent agreement with the NMR analysis of the mechanochemical activation.

Figure 4

¹H NMR (600 MHz, D₂O) with maleic acid as a reference. All samples had a concentration of 1.0 mg mL⁻¹ during the sonication experiments. a) HG₆ prior to sonication, concentration of the ¹H NMR sample 14 mg mL⁻¹; b) HG₆ after 3 h of sonication, concentration of the ¹H NMR sample 14 mg mL⁻¹; c) HG₆ ⋅ (ibuprofen)₂ prior to sonication, concentration of the ¹H NMR sample 16 mg mL⁻¹; d) HG₆ ⋅ (ibuprofen)₂ after 3 h of sonication, concentration of the ¹H NMR sample 16 mg mL⁻¹; blue bands indicate novel appearing signals originating from cage fragmentation products in solution.

To gain additional insight into the mechanochemical activation of the Pd−N bonds within the crosslinking cages, the system was investigated with CoGEF¹⁸ calculations using density functional theory at the B3LYP+D3/6-31G* level (Figure 5).

Figure 5

a) Free activation energies as a function of force for the two investigated pathways (red: trans, blue: cis) for Pd−N dissociation; b) potential energy (black) and Pd−N distance (red) are shown along the CoGEF path (cis-pulling), please see Supporting Information for the case of trans-pulling; c) snapshots along the CoGEF path (cis-pulling) from relaxed [Pd(4,4′-dimethyl-2,2′-bipyridine)(4-methylpyridine)₂]²⁺ complex. After strongly distorting the dissociating pyridine-ligand, an H₂O molecule engages in the Pd²⁺ coordination sphere.

The CoGEF method has been used to study complex reactions such as the peeling off of metallocene mechanophores and even similar Pd−N bonded systems.^{12b, 23} Therefore, it can be assumed that it describes the heterolytic bond breaking in this system well. We used a [Pd(4,4′-dimethyl-2,2′-bipyridine)(4-methylpyridine)₂]²⁺ model complex to simulate the mechanochemical dissociation of one vertex of the Pd₆L₄ cage. The complex was surrounded by eight H₂O molecules, as these are required to fill the coordination sphere of the Pd²⁺ ion after the nitrogen ligands dissociate. The nitrate counterions are not included, as the ions do not form contact ion pairs but rather diffuse freely in the solution. External force was applied in cis- and trans direction (Figure 5a). Both CoGEF paths of cis- and trans-pulling revealed that the pyridine ligands dissociate more readily than the bipyridine ligand. The dissociation of the pyridine ligands is immediately followed by a re-coordination of a surrounding water molecule by the Pd²⁺ ion to recover the square planar coordination. The highest potential energies (E_max) along both CoGEF paths are quite similar: 27.8 kcal mol⁻¹ (cis) and 26.5 kcal mol⁻¹ (trans). The corresponding values of F_max are 1.87 nN (cis) and 2.57 nN (trans), implying an easier dissociation of the pyridine ligand in the case of cis-pulling. However, the CoGEF method yields rupture forces, which tend to lie above the actual forces in action.^18c Therefore, more complex methods are applied to evaluate the barrier heights of these two competing reaction mechanisms.

Free energy barrier heights based on the force-modified potential energy surface (FMPES) approach were used to better determine the strength of the pulling force in action.^{18d, 18e} Since these computations focus on the barrier of the first Pd−N bond dissociation and not on the re-coordination process of the Pd²⁺ ion, a closed-shell description of the studied system is deemed reliable. The barrier for the pyridine dissociation along the cis-pathway is smaller throughout the range where transition state structures can be obtained. On the 0.5 nN-PES, the barrier heights are 12.6 kcal mol⁻¹ and 16.0 kcal mol⁻¹ for the cis- and trans-pulling cases, respectively. At a force of 0.8 nN, the free energy barrier is lowered to 8.2 kcal mol⁻¹ (cis) and 12.6 kcal mol⁻¹ (trans). In this force regime, the barrier heights of the cis-pulling case are small enough that a reaction can take place, and we can therefore assume that, indeed, a subtle pulling force of between 0.5 and 1.0 nN is sufficient to dissociate the pyridine-ligand in cis-direction. Pulling in trans direction, the Pd−N bond is directly stretched, which requires a comparatively high pulling force. In cis-pulling, the N−Pd−N angle is strongly distorted, resulting in weaker Pd−N bonding due to smaller orbital overlap. As a result, the mechanical force is more strongly coupled to the reaction and leads to an easier breaking of the Pd−N bond. This can open the octahedral cage, while the pyridine-ligand in trans direction would stay coordinated to the Pd²⁺ ion.

While a reported star-shaped metal-organic cage-based polymer was capable of fully releasing its cargo load,¹² it lacked one main aspect, which is the cargo-load ratio in relation to the polymer backbone. The used polymeric system, SC, had an Mn of ca. 62 kDa, whereas the guests only had molar masses of around 206 g mol⁻¹ for ibuprofen and 366 g mol⁻¹ for drospirenone. This implies that only around 0.7 wt.% of our host-guest system was attributed to the guest. The hydrogels HG_1–6 reported herein are capable of overcoming this issue. Due to the crosslinked systems, the cargo-load-ratio can be increased to 1.9 % for the longest hydrogel (HG₆) and to approximately 6.2 % for our shortest hydrogel (HG₁).

Conclusion

Here, we have demonstrated that supramolecular crosslinkers provide access to mechanoresponsive hydrogels. In addition to light and temperature, this broadens the capability for spatiotemporal control of soft matter by integration of metal-organic cages. Whereas guest uptake in differentially addressable spaces within metal-organic cage-crosslinked polymeric hydrogels was one of the earliest contributions published, we could show the efficient sonochemical activation, increasing non-covalent guest loading of completely unmodified drugs available for release by a factor of ten. The transcription of the Fujita-type M₆L₄ cage into soft matter applications will enable the development of novel, responsive, and more complex nanomaterials. Because the release of small molecules from their latent macromolecular carriers via polymer mechanochemistry usually requires the use of cargo molecules that are specifically functionalized, we anticipate that our combination of universal supramolecular encapsulation and force as an external stimulus will contribute to further supramolecular release systems and dynamic materials.

Acknowledgments

We thank the Strategic Research Fund of Heinrich Heine University (F-2018/1460-4), the Manchot Foundation for a fellowship to R.K., and the North Rhine-Westphalian Academy of Sciences, Humanities and the Arts (B.M.S.). R.G. and L.P.N. are thankful for a Freigeist-Fellowship of the Volkswagen Foundation (92888). R.G. is grateful for support by the German Research Foundation DFG (492017525, 503981124). Computational infrastructure and support were provided by the Centre for Information and Media Technology at Heinrich Heine University Düsseldorf. The authors thank the Center for Structural Studies (CSS) that is funded by the Deutsche Forschungsgemeinschaft (DFG Grant numbers 417919780 and INST 208/761-1 FUGG) for access to the SAXS instrument. Open Access funding enabled and organized by Projekt DEAL.

Conflict of interest

The authors declare no conflict of interest.

Open Research

Data Availability Statement

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