Fast-Relaxing Hydrogels with Reversibly Tunable Mechanics for Dynamic Cancer Cell Culture

Abstract

The mechanics of the tumor microenvironment (TME) significantly impact disease progression and the efficacy of anti-cancer therapeutics. While it is recognized that advanced in vitro cancer models will benefit cancer research, none of the current engineered extracellular matrices (ECM) adequately recapitulate the highly dynamic TME. Through integrating reversible boronate-ester bonding and dithiolane ring-opening polymerization, we fabricated synthetic polymer hydrogels with tumor-mimetic fast relaxation and reversibly tunable elastic moduli. Importantly, the crosslinking and dynamic stiffening of matrix mechanics were achieved in the absence of a photoinitiator, often the source of cytotoxicity. Central to this strategy was Poly(PEGA-co-LAA-co-AAPBA) (PELA), a highly defined polymer synthesized by reversible addition-fragmentation chain transfer (RAFT) polymerization. PELA contains dithiolane for initiator-free gel crosslinking, stiffening, and softening, as well as boronic acid for complexation with diol-containing polymers to give rise to tunable viscoelasticity. PELA hydrogels were highly cytocompatible for dynamic culture of patient-derived pancreatic cancer cells. It was found that the fast-relaxing matrix induced mesenchymal phenotype of cancer cells, and dynamic matrix stiffening restricted tumor spheroid growth. Moreover, this new dynamic viscoelastic hydrogel system permitted sequential stiffening and softening to mimic the physical changes of TME.

Graphical Abstract

1. Introduction

Cellular processes are regulated by the extracellular matrix (ECM), a highly dynamic microenvironment composed of proteins, glycosaminoglycans (GAGs), and other matrix-associated growth factors. During tumor progression, activated stromal cells secrete excess amounts of ECM molecules, including collagen, fibronectin, and hyaluronic acid (HA),^{1, 2} leading to stromal stiffening and alteration in cancer cell morphology, proliferation, and migration.^{3, 4} It is now established that the ECM undergoes varying degree of stiffening during tissue development, liver fibrosis, myocardial infarction,^{5, 6} and tumor progression.⁷ Higher matrix stiffness strongly correlates with progression and severity of many diseases, including cancer.⁸ For instance, it was reported that the elastic moduli of pancreatic ductal adenocarcinoma (PDAC) tissues were much higher (G’ ~ 5.46 ± 3.18 kPa) than that of pancreatitis (G’ ~ 2.15 ± 0.41 kPa) and normal tissues (G’ ~ 1.06 ± 0.25 kPa).⁹ In addition to elastic moduli, the role of matrix stress-relaxation was increasingly recognized as an important physical cue altering cell functions and morphogenesis.^9–13 Interestingly, it was reported that even though the PDAC and pancreatitis tissues were stiffer than the healthy pancreatic tissue, the relaxation time scales of the diseased tissues (τ_1/2 ~ 66.1 ± 20.8 s and 27.6 ± 14.0 s for PDAC and pancreatitis tissue, respectively) were both faster than that of the healthy tissue (τ_1/2 ~ 92.7 ± 46.4 s).⁹ Nonetheless, these relaxation time scales (τ_1/2 ~30 to 100 s) were considerably faster than that of most chemically crosslinked and elastic hydrogels (τ_1/2 ~ ∞). Hence, there exists a need to develop fast-relaxing hydrogels capable of being dynamically stiffened to mimic the evolving nature of TME.

Various polymeric matrices have been developed for biomedical applications, including three-dimensional (3D) culture of healthy and diseased cells. These matrices are designed by using a diverse range of functional macromers, including natural matrices such as alginate,^{12, 14} collagen,¹⁵ gelatin,^{16, 17} chitosan,^{18, 19} dextran,²⁰ and hyaluronic acid (HA),²¹ as well as synthetic polymers like poly(ethylene glycol) (PEG) and highly defined RAFT (reversible addition-fragmentation chain transfer polymerization) polymers.^{22, 23} While hydrogels fabricated from synthetic polymers do not possess inherent bioactive moieties for immediate cell adhesion and protease-mediated degradation, they do provide various benefits as three-dimensional (3D) cell culture platforms owing to their high tunability and reproducibility, as well as abundant availability of the raw materials. Recent decades have witnessed remarkable advances in synthetic polymer science and engineering, paving the way for the development of ‘smart hydrogels’ with tailorable and stimuli-responsive physicochemical properties.^{22, 24-26} For instance, polymer networks can be designed to undergo tunable hydrolysis, photolysis, and proteolysis, leading to user-controlled dynamic changes in matrix mechanics that subsequently affect cell fate process.^{25, 27–30} Nevertheless, the majority of these synthetic dynamic hydrogels are elastic, making them suboptimal for mimicking the fast-relaxing nature of the human tissues.

Owing to its reversible nature, disulfide bond exchange is a common strategy in bioconjugation chemistry and has been leveraged in the modification of biomaterials. In addition to participating in the thiol-disulfide exchange, disulfide bonds can be cleaved by light-generated radicals.^{8, 31} Disulfide bonds formed from the oxidation of thiols (SH) typically require a harsh oxidation process incompatible with encapsulated cells.⁸ To address this issue while preserving the dynamic nature of disulfide linkages, polymers containing 1,2-dithiolanes are increasingly been used in hydrogel crosslinking. Lipoic acid (LA), a disulfide compound with a five-membered ring, is a particularly attractive small molecule that undergoes a photo-induced crosslinking/decrosslinking process under mild reaction conditions.³² This versatile and reversible chemistry supports an initiator-free ring-opening polymerization upon irradiation of light at 365 nm. On the other hand, depolymerization occurs under the same condition but in the presence of photoinitiators, such as lithium phenyl(2,4,6-trimethylbenzoyl) phosphinate or LAP. The dynamic mechanism makes such materials highly useful for constructing responsive and dynamic networks, including supramolecular adhesives,³³ protein-disulfide conjugates for the delivery of protein therapeutics,³⁴ and advanced self-assembled networks via thiol-disulfide interconversion.³⁵ Another interesting property of dithiolane chemistry is that the addition of mono-functional SH-bearing molecules to the dithiolane-crosslinked hydrogels could induce thiol-disulfide exchange that gives rise to faster stress-relaxation of the otherwise elastic networks. Anseth and co-workers demonstrated that the stress-relaxation of 20 wt% of 4-arm PEG-LA hydrogels crosslinked initiator-free (25 mW/cm2, 365 nm, 10 min) could be accelerated by passive disulfide metathesis and thiol-disulfide exchange through adding 4-arm PEG-SH (equimolar of LA and SH).³⁶ Unfortunately, with this approach the fastest relaxation time scale was on the order of 1,000 sec, a relaxation time scale 10 times longer than that in many biological tissues, including PDAC stroma.^{11, 22}

Reversible bonding chemistry is typically required to create hydrogels with fast relaxation. To this end, we have previously reported the development of gelatin-based and synthetic RAFT polymers containing norbornene and boronic acid moieties.^{22, 37} The installation of norbornene groups on the polymers permitted thiol-norbornene photo-click reaction (with multi-functional thiol-bearing crosslinkers) for fabricating elastic hydrogels with highly tunable stiffness. On the other hand, the incorporation of boronic acid groups rendered the hydrogels viscoelastic owing to their reversible cyclic boronate ester bond formation (with polymers containing 1,2- or 1,3-diols). The fast reversible equilibrium of covalent B-O bonds in the form of 5-, 6- or 7-membered rings in aqueous media allows the reversible molecular assemblies, which typically are unstable in the presence of glucose. However, the impact of glucose on B-O dissociation is less pronounced within the covalent thiol-norbornene network. While hydrogels with fast relaxation were created using this strategy, these viscoelastic hydrogels cannot readily be stiffened or softened to mimic the dynamic of tumor stroma.

In this work, we aimed to create a fast-relaxing hydrogel system with reversibly tunable mechanical property, which has not been reported in the literature. To achieve this, we integrated initiator-free dithiolane crosslinking and boronate-diol bonding, both of which are reversible chemistry. To permit initiator-free crosslinking of fast-relaxing hydrogels, we synthesized a new set of functional RAFT polymers, including boronic acid containing poly(poly(ethylene glycol)methylether acrylate-co-lipoic acid acrylamide-co-3-(acrylamido)phenylboronic acid) or PELA, poly(poly(ethylene glycol) methyl ether acrylate-co-lipoic acid acrylamide) or PEL (control polymer for PELA), and dopamine-modified poly(hydroxyethyl acrylate) or PHD. We demonstrated initiator-free photo-crosslinking of PELA and PEL hydrogels with tunable elasticity, as well as PELA/PHD hydrogels with fast relaxation and reversibly tunable mechanics. In addition to characterizing the new dynamic fast-relaxing PELA-based hydrogels, we conducted pilot cell studies to evaluate the cytocompatibility and the effect of matrix viscoelasticity and dynamic stiffening/softening on pancreatic cancer cell behaviors.

2. Materials and Method

2.1. Materials

DL-∞-lipoic acid (>99.0%, TCI), 2-(2-carboxyethylsulfanylthiocarbonylsulfanyl)propionic acid (CPA, Sigma-Aldrich), 3-(acrylamido)phenylboronic acid (98%, Sigma-Aldrich), 3,4-dihydroxyphenylacetic acid (DOPAC, Ambeed), anhydrous dichloromethane (DCM, 99.8%, Acros), 4-(dimethylamino)pyridine (DMAP, Sigma-Aldrich), N,N’-dicyclohexylcarbodiimide (Sigma-Aldrich), 1-(3-dimethylaminopropyl)-3-ethylcarbodiimide hydrochloride (EDC.HCl, Acros Organics), hydrochloric acid (HCl, 6N, J.T.Baker), sodium bicarbonate (NaHCO₃, ACS reagent, ≥99.7%, Sigma-Aldrich), four-arm PEG (4ARM-NH₂HCl-20K, Jenkem Technology), poly(vinyl alcohol) (PVA, average molecular weight 146 −186 kDa, 87–89% hydrolyzed, Sigma-Aldrich), lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP, Sigma-Aldrich), glutathione (Fisher Scientific), hexafluorophosphate azabenzotriazole tetramethyl uronium (HATU, CHEM-IM EX INT’L INC.), 4-methylmorpholine (Fisher Bioreagents), acryloyl chloride (≥99.7%, Sigma-Aldrich), diethyl ether (ACS, VWR International), dimethylformamide (DMF, ACS grade, ≥99.8%, Sigma-Aldrich), dimethylformamide (DMF, anhydrous, amine-free, 99.9%, Alfa Aesar), methanol (ACS grade, ≥99.8%, Sigma-Aldrich), sodium hydroxide (NaOH, Fisher Scientific), anhydrous sodium sulfate (Na₂SO₄, ACS grade, AMRESCO) were used as received. The liquid monomers: N-hydroxyethyl acrylamide (HEAA, 97%, Sigma-Aldrich), and poly(ethylene glycol) methyl ether acrylate (PEGA, average M_n 480 g/mol, Sigma-Aldrich) were passed through the column filled with inhibitor removers before use. The thermal initiator: 2,2’-azobisisobutyronitrile (AIBN) was recrystallized from methanol.

2.2. Synthesis and characterization of monomers and macromers

2.2.1. Lipoic acid acrylamide (LAA)

The monomer synthesis was performed via EDC/DMAP esterification reaction as follows. Lipoic acid (5 g, 24.23 mmol) and HEAA (2.79 g, 24.23 mmol) dissolved in anhydrous DCM (60 mL) was stirred in ice-cold condition, followed by dropwise addition of DMAP (1.48 g, 12.11 mmol) and EDC.HCl (6.96 g, 36.31 mmol) to the solution in 15 mL DCM. Then the solution was reacted by stirring at room temperature overnight. The crude was then purified by extraction with 1N HCl (100 mL ×3), saturated NaHCO₃ (100 mL × 3), and brine (100 mL × 1). Then the collected organic layer was dried over anhydrous Na₂SO₄. After vacuum filtration, the solvent was removed under reduced pressure to yield the product as a yellow viscous oil (yield ~90%).

2.2.2. 3-(Acrylamido)phenylboronic acid (AAPBA)

3-(Acrylamido)phenylboronic acid was prepared as described in our previous report.²² In detail, 3-aminophenyl boronic acid (1 g, 7.30 mmol) was dissolved in 2M NaOH solution (13.3 mL). After the dropwise addition of acryloyl chloride (0.87 g, 9.61 mmol) in an ice-cold condition, the solution was reacted for 30 min. Then the reaction solution was brought to room temperature and continued stirring for another 1.5 hr. Subsequently, the pH of the solution was adjusted to ~1 by dropwise addition of 6N HCl. Then the precipitates observed during the addition of HCl were collected by vacuum filtration and washing with cold water. The obtained solid powder was dissolved in 10 mL of water preheated at 60 °C. Upon the complete dissolution, the solution vial was kept at 4 °C overnight to allow the formation of crystals. The pure monomer in crystal form was finally collected by vacuum filtration and washing with cold water. After lyophilization, the structure of the monomer (yield ~20%) was determined by ¹H-NMR.

2.2.3. Statistical copolymers PELA and PEL

The syntheses of linear chain statistical copolymers were performed via RAFT polymerization. For the synthesis of PELA, PEGA (1.1 g, 2.28 mmol), LAA (160 mg, 0.53 mmol), AAPBA (201.40 mg, 1.05 mmol), CPA (8.94 mg, 35.15 μmol), and AIBN (577.23 μg, 3.51 μmol) were dissolved in DMF (amine-free, 3.87 mL) at 1 M of monomer concentration. The ratio between the components (PEGA: LAA: AAPBA: RAFT agent: AIBN) was kept at 65: 15: 30: 1: 0.1. The mass of components added for the PEL synthesis was as follows: PEGA (1.30 g, 2.70 mmol), LAA (160 mg, 0.53 mmol), RAFT agent (8.94 mg, 35.13 μmol) and AIBN (0.58 mg, 3.51 μmol) with the ratio between them at 77: 15: 1: 0.1. The solution was then degassed under nitrogen for 45 min in ice-cold condition. After 9 hr of polymerization at 65 °C, the reaction was terminated by introducing air to the solution and placing the vial in an ice bath. The resulting polymer was then purified via dialysis against methanol for 2 days and water for 1 more day, followed by freeze-drying to yield yellow polymer product. The monomer conversion and the molecular weight of purified polymers were then determined via ¹H-NMR and SEC systems.

2.2.4. Poly(ethylene glycol)-tetra-lipoic acid (PEG-4LA)

Four-arm PEG was modified with lipoic acid via HATU coupling reaction. Briefly, in a glass vial, lipoic acid (165 mg, 800 μmol), and HATU (304 mg, 800 μmol) were first dissolved in 4 mL anhydrous DMF (amine-free). After the dropwise addition of 4-methylmorpholine (161.84 mg, 1.60 mmol), the reaction was activated under nitrogen for 1 hr. 4-arm PEG-amine (1 g, 200 μmol of NH₂) dissolved in 6 mL DMF was then added to the activated lipoic acid solution, followed by reacting at room temperature for one day. The pure PEG-4LA was obtained by precipitation-centrifugation in diethyl ether once (3000 rpm, 5 min), dialysis against DMF (ACS grade) for one day (200 mL × 3), and dialysis against water for two more days (water was changed frequently). After freeze-drying, the pure functional polymer was anlayzed via ¹H-NMR.

2.2.5. Poly(N-hydroxyethyl acrylamide)-DOPA

Poly(N-hydroxyethyl acrylamide) (PHEAA) with theoretical molecular weight 23000 g/mol was first synthesized via RAFT polymerization. The side-chain functionalization was then conducted via DCC/DMAP/pyridine esterification reaction as follows. The solution of 3,4-dihydroxyphenylacetic acid (DOPAC) (1 g, 6.46 mmol) and DCC (1.24 g, 6.46 mmol) in 6.3 mL of DMF (amine-free) was stirred under nitrogen atmosphere and ice-cold condition for 1 hr. Subsequently after the activation step, the solid precipitates in the reaction were removed by vacuum filtration. After mixing the filtrate with PHEAA (300 mg, 13.04 μmol), DMAP (158 mg, 1.29 mmol), and pyridine (1 g, 12.91 mmol), the solution was left reacting overnight under nitrogen (final solvent volume was 9.3 mL). The functionalized polymer was purified by a few cycles of precipitation-centrifugation in diethyl ether until the clear supernatant was observed. Then the residual impurities were removed via dialysis against methanol for 18 hrs. The pure polymer sample was analyzed by ¹H-NMR.

2.2.6. Characterization methods

¹H-NMR spectra were recorded by using Bruker Avance III 400 MHz NMR. Samples were dissolved and analysed in deuterated chloroform (CDCl₃), dimethyl sulfoxide (DMSO) or deuterium oxide (D₂O). NMR spectra were processed by using either Bruker Topspin 3.6.2 software or MestReNova. The molecular weight and molecular weight distribution of co-polymers were determined by the Waters Breeze gel permeation chromatography system. The instrument consists of a 1525 binary HPC pump, a 1500 column heater, a manual injector, an inline degasser AF, a 2414 refractive index detector, and a 2487 dual wavelength absorbance UV/Vis detector. The system is equipped with three Styragel high resolution column (HR2, HR4, and HR5, 7.8×300 mm). Tetrahydrofuran (inhibitor free) with a flow rate of 1mL/min was used as the continuous phase at 35 °C. The system was calibrated based on the commercially available linear polystyrene standards with molecular weights ranging from 570 to 2×10⁶ g/mol.

2.3. Hydrogel fabrication and characterization

Stock solutions of PELA (or PEL), PEG-4LA, PVA (or PHD) were prepared in PBS. Hydrogel rheometry measurements were performed by using a sandblasted 8 mm probe on an Anton Paar rheometer (MCR102 with True Strain attachment). In situ rheometry was performed using a quartz upper plate connected with a Omnicure S2000 light source with a wavelength set at 365 nm and specific intensity (e.g., 5, 10, 15, and 20 mW/cm²). 10 μL of gel precursor solution was loaded onto the in situ curing platform with 0.1 mm gap, and the modulus changes were observed at 1 % strain and 1 Hz frequency at the desired temperature. Hydrogel static moduli were measured with strain-sweep rheometry with the following parameters: 1 mm gap, frequency of 1 Hz, normal force of 0.25 N, and strain of 0.1 to 5 %, at room temperature. Hydrogels were fabricated with desired formulations and crosslinked between two glass slides treated with hydrophobic solution (e.g., Rainaway) and sandwiched with 1 mm thick spacers. The glass-slide assembly was gently placed on the mini heat block (Benchmark) set at 37 °C and irradiated with 365 nm light (20 mW/cm²) for 3 to 10 min (as noted in the figure captions). The gels (diameter: 8 mm) were then removed and incubated in PBS for 30 min before the stiffness measurement. The hydrogels were incubated in PBS at 37 °C and the moduli (G’, and G”) were recorded at different time intervals. The stress relaxation analyses of pre-fabricated hydrogels were conducted using the following parameters: a 1 mm gap, a frequency of 1 Hz, a normal force of 0.25 N, and a strain of 10 %, at room temperature. The gap size was set based on the gel thickness whereas the frequency was selected within the linear viscoelastic region.

2.4. Reversible tuning of hydrogel stiffness

The dynamic tuning of hydrogel mechanical properties (i.e., reversible stiffening and softening) of the pre-fabricated hydrogels were investigated as follows. For stiffening tests, the hydrogels were incubated in 6 wt% PEG-4LA dissolved in PBS or cell culture medium (DMEM/high glucose) for 1 hr. Then the gels were stiffened under UV irradiation for 10 min at 37 °C (λ = 365 nm, 20 mW/cm²). For softening tests, the hydrogels were incubated in the photo-initiator solution containing 10 mM LAP and 5 mM GSH (dissolved in PBS or cell culture medium DMEM/high glucose) for 30 min. Then the hydrogels were softened under UV irradiation for 50 sec (λ = 365 nm, 20 mW/cm²) at room temperature.

2.5. Cell encapsulation

A patient-derived pancreatic cancer cell line (Pa03C, a generous gift from Dr, Melissa Fishel, Indiana University School of Medicine) was cultured in high glucose DMEM supplemented with 10 % FBS and 1 % Penicillin Streptomycin as antibiotic. Pa03C line was derived from stage IV liver metastasized pancreatic ductal adenocarcinoma.^{38, 39} When the cells were above 70 % confluent in cell culture plate, they were washed with PBS twice, and then detached from the well by treating with 0.05 % trypsin solution. After collecting the cells by centrifugation at 1000 rpm for 3 min, the cells were counted by a haemocytometer and used directly for encapsulation or assembly into spheroids in Aggrewell^™ 400 microwell plate. All cell-laden hydrogels were fabricated between two glass slides separated by 0.8 mm Teflon spacers. The glass slides (1.0 – 1.22 mm thickness) were sterilized by first treating with water repellent solution (e.g., rain-X), followed by immersing the slide assembly in 70% ethanol for 15 min, and germicidal UV irradiation in cell culture hood for 30 min. 15 – 20 μL of pregel solution containing cells was injected between the two glass slides, followed by initiator-free light-triggered gelation.

2.5.1. Encapsulation of dispersed single cells

Dispersed cells were mixed with the precursor solutions at a final density of 5 ×10⁶ cells/mL prior to gel crosslinking. Two groups of initiator-free gel formulations are designated as elastic (EL) and viscoelastic (VE) gels. All gels contained 5 wt% of PEL (for EL gels) or PELA (for VE gels) and 3 wt% PHD. Extra PEG-4LA was supplemented to EL gels (i.e., 8 wt% and 7 wt% for EL and VE gels, respectively) to offset the slightly higher stiffness of the VE gels. The cell-containing precursor solutions were irradiated with 365 nm (20 mW/cm²) for 5 min at 37 °C. Cell viability and morphology were monitored over 7 days.

2.5.2. Encapsulation of pre-formed spheroids

The cancer cells were assembled into spheroids in aggrewell^™400 microwell culture plate. The number of cells in each spheroid was ~50 cells. After 3 days, spheroids were collected and suspended in EL and VE gel precursor solutions, at approximately 450 spheroids per hydrogel. Spheroids encapsulation was performed as described above.

2.5.3. Evaluation of cell viability and morphology

Cell viability was monitored by live/dead staining. Briefly, the cell-laden hydrogels were gently washed with phosphate-buffered saline (PBS) twice, followed by incubation in commercially available live/dead reagent: calcein AM (for staining live cells) and ethidium homodimer-1 (for staining dead cells), at 37 °C. After 30 min of incubation, the cell-laden hydrogels were imaged by a benchtop confocal microscope (Oxford BC43). In addition to live/dead staining, cell spheroids morphology was also monitored with F-actin staining (by Alexa Fluor 488 Phallacidin) and nuclei counter-staining (by 4’,6-diamidino-2-phenylindole or DAPI). Briefly, cell-laden hydrogels were gently washed in PBS three times, followed by fixation with 4 % (w/v) paraformaldehyde for 1 hr at room temperature. After PBS washes (×3), the gels were incubated in the solution containing Alexa Fluor 488 Phalloidin and DAPI overnight at 4 °C. The structure of spheroids was then observed by a confocal microscope (Oxford BC43). The employed confocal parameters are as follows: Z-stack slices = 35 – 40, step size = 5 μm, scan size = 175 – 200 μm, and magnification = 20×. The images were captured by a Fusion software and processed by ImageJ software.

2.6. Dynamic tuning of mechanics of cell-laden viscoelastic hydrogels

Cell-laden hydrogels (EL and VE) were subjected to light-based dynamic tuning of stiffness as described in section 2.4. To confirm the biocompatibility of dynamic stiffening/softening process, the cell-laden hydrogels were stained using live/dead kit and imaged as described above.

2.7. mRNA expression

mRNA expression was determined using qRT-PCR. The spheroids cultured in viscoelastic gels (dithiolane crosslinked hydrogels containing boronate ester bonds) for 7 days were collected by degrading the network upon UV exposure at 365 nm (20 mW/cm²) for 50 sec at room temperature. The gel degradation was conducted straight after incubating the spheroids-embedded hydrogels in an initiator solution: 10 mM LAP and 5 mM GSH dissolved in DMEM/HG, for 30 min. Afterward, the total RNA was extracted with a NucleoSpin RNA II kit (Clontech) according to the manufacturer’s protocol. The concentration of eluted highly pure RNA was then quantified by a DeNovix (DS-11 FX) Spectrophotometer/Fluorometer. After total RNA was reverse transcribed into complementary DNA (cDNA), mRNA expression was analysed using SYBR Premix Ex TaqII kit (Clontech) with primers: ZEB1, Vimentin, EPCAM, PIEZO1, CDH1, and CDH2. The primer sequences used in this study are listed in Table S2. The PCR reactions were then conducted using a 7500 Real-Time PCR system (Applied Biosystems). Each measurement was performed in three replicates to calculate the means and standard deviations (n = 3).

2.8. Statistical analysis

Each experiment was repeated three times. The statistical analyses were carried out by using GraphPad Prism 10 software and the differences between groups were determined by one-way or two-way analysis of variance (ANOVA). Data are represented as mean ± standard deviation (SD). The number of asterisks (*): one, two, three and four indicates p value of < 0.05, 0.01, 0.001 and 0.0001, respectively.

3. Results and Discussion

3.1. Synthesis of monomers and RAFT polymers

Lipoic acid containing polymers have been used in prior studies to fabricate dynamically tunable hydrogels.^{32, 40} However, none of the prior work integrated boronate-diol bonding into lipoic acid containing hydrogels for adjusting matrix stress-relaxation. To enable RAFT polymerization of lipoic acid containing polymers, lipoic acid acrylamide (LAA) was first synthesized by EDC/DMAP mediated coupling of carboxylic acid moiety in DL-∞-lipoic acid (LA) with commercially available N-hydroxyethyl acrylamide (HEAA) monomer. To include boronic acid on the RAFT polymers, 3-(acrylamido)phenylboronic acid (AAPBA) was synthesized by reacting 3-aminophenyl boronic acid with acryloyl chloride according to the previously developed methodology.^{22, 41} In the polymer chain consisting of AAPBA, phenylboronic acid moieties will enable the formation of viscoelastic hydrogel network through the boronate-diol bonding with 1,2- or 1,3-diols containing polymers.^{22, 41} The ¹H-NMR shown in Figure S1A and S1B confirmed the successful syntheses of the two monomers, with yields of ~90% and ~20% for LAA and AAPBA, respectively.

LAA, AAPBA, and commercially available PEGA monomers were mixed at desired weight ratio and polymerized via RAFT polymerization. The linear polymers were referred to as PEL and PELA for polymer chains without or with AAPBA. Of note, the monomer units of PEGA, LAA and AAPBA were kept at 76: 15: 0 for PEL and 65: 15: 30 for PELA. The concentration of LAA monomer was kept low to prevent premature gelation or intermolecular disulfide exchange. As confirmed by ¹H-NMR, the monomer units in resulting polymer chains were labelled as PEGA₅₁-co-LAA₁₂ for PEL and PEGA₅₂-co-LAA₁₃-co-AAPBA₂₄ for PELA (Table S1). The theoretical molecular weight determined by ¹H-NMR revealed that both polymer chains were generated with M_w approximately 30 kDa (Table S1). Besides, SEC traces in Figure S2 revealed the monomodal distribution with a dispersity (Đ) of <1.5 for both synthesized polymers, confirming that the polymerization occurred in a controlled manner. To be noted is that the five-membered rings of lipoic acid in bulk can undergo homolytic cleavage above its melting temperature (>61 °C), which would result in the subsequent branching of the polymer structure.³³ The obtained SEC results show that such phenomenon did not occur in our system, indicating the successful preparation of 1,2-dithiolane containing copolymers with a desired structure.

3.2. Initiator-free photocrosslinking of PELA hydrogels

PELA is a linear polymer containing both dithiolane and boronic acid, the former permits initiator-free photocrosslinking whereas the later contributes to reversible boronate-diol bonding. The initiator-free photocrosslinking of single component PELA hydrogels under mild aqueous conditions (Figure 1A) was first validated by dynamic time-sweep rheological experiments. PELA was dissolved in PBS at 20 wt%, which is typical for gelation of linear cyclic dithiolane containing polymers. As the disulfide-containing five-membered rings serve as chromophores with absorption maximum at ~365 nm,^{32, 42} we first explored the suitable gelation condition at such wavelength at room temperature. Through in situ photo-rheometry, we verified light intensity dependent gelation of single-component PELA hydrogels (Figure 1C), with the G’/G” crossover time ranging from ~410 sec to ~130 sec for light intensity of ~5 mW/cm² to ~20 mW/cm². A prior report by Maes et al. demonstrated similar initiator-free photocrosslinking of linear polymethacrylate-based dithiolane-containing polymer hydrogels.⁴⁰ However, cytotoxic volatile solvent (e.g., chloroform) was used for the gelation, precluding the use of the polymethacrylate-based copolymers for cell-laden hydrogels. In contrast, our single-component PELA linear polymer crosslinked into hydrogels without the use of photoinitiator and under ambient conditions. The plateau storage moduli of 20 wt% PELA hydrogels reached gelation in 2–7 minutes of light irradiation in aqueous solution and at ambient temperature, suggesting the potential utility of this mild initiator-free hydrogel crosslinking system for biological applications. We further optimized the initiator-free photopolymerization process by attempting the gelation at the light intensity of 20 mW/cm² and at 37 °C, a cell-friendly temperature that may also enhance the ring-opening efficiency of dithiolanes.^{33, 43} Indeed, raising temperature from 25 °C to 37 °C accelerated the G’/G” crossover time by about 50 sec (Figure 1D). Thus, 365 nm light irradiation, 20 mW/cm² intensity, and 37 °C were chosen as the ideal parameters for further experiments.

Figure 1. Initiator-free ring-opening polymerization of dithiolane and boronic acid containing hydrogels.

(A) Schematic of initiator-free ring-opening polymerization of dithiolane-/boronate-containing PELA into a single component hydrogel. (B) Representative photographs of a transparent PELA hydrogel (λ = 365 nm, Light intensity = 20 mW/cm², 37 °C). (C) Effect of light (λ = 365 nm) intensity on in situ gelation and evolution of storage (G’) and loss (G”) moduli of single component PELA hydrogels (20 wt%, at room temperature). (D) Effect of temperature on in situ gelation of single component PELA hydrogels (20 wt%). (E) Effect of PEG-4LA on in situ gelation dithiolane-containing hydrogels. [PELA] = 5 wt% for all conditions, [PEG-4LA] = 0, 1, 3 and 5 wt%, λ =365 nm, 20 mW/cm², 37 °C. (F) Effect of boronic acid on in situ gelation between of dithiolane-containing hydrogels. [PEG-4LA] = 1 wt% in all conditions. λ =365 nm, 20 mW/cm², 37 °C. In all experiments, light was turned on at 50 sec (dashed lines). All in situ rheometry studies were performed under γ = 1 %, f = 1 Hz and F_N = 0 N.

While the gelation of single-component PELA hydrogels was achieved in 2–7 minutes, a relatively high PELA concentration (20 wt%) was needed. Furthermore, after 10 min of continuous light irradiation the final storage moduli of the hydrogels were still under 3 kPa. While this phenomenon was typical for gelation using linear polymers, there is a need to improve efficiency of PELA hydrogel crosslinking beyond optimization of gelation parameters. Hence, we sought to develop a PELA-based hydrogel system that enables faster gelation rate at low precursor concentration under ambient conditions. To achieve this, we synthesized and incorporated PEG-4LA (Figure S3A) into the PELA crosslinking system. In one example, we lowered PELA concentration to 5 wt% and adjusted PEG-4LA from 0 wt% to 5 wt%. As shown in Figure 1E, no gelation occurred for pure PELA alone at a low concentration of 5 wt% (i.e., 0 wt% PEG-4LA). On the other hand, the addition of mere 1 wt% PEG-4LA into 5 wt% PELA resulted in a crossover time of ~100 sec. Further increasing PEG-4LA to 5 wt% substantially reduced crossover time to under 1 minute and plateau storage moduli of over 9 kPa.

As 5 wt% PELA alone did not permit initiator-free photo-gelation, we asked if PELA at this low concentration participated in hydrogel crosslinking. This was an important question as un-crosslinked linear PELA would leach out of the hydrogels, resulting in the loss of boronic acid groups needed for adjusting hydrogel relaxation. To this end, we demonstrated the crosslinking of 1 wt % PEG-4LA alone, which was slow (G’/G” crossover at ~240 sec. Figure 1F). However, upon the addition of 5 wt% PELA into the solution, the gelation time decreased to ~180 sec, affirming that both polymers involved in the hydrogel crosslinking. Furthermore, we examined the possible impact of bulky phenylboronic acid groups through using a control RAFT polymer without the boronic acid groups (i.e., PEL). In situ rheometry results show that no apparent difference was found in between using PELA or PEL, suggesting that boronic acid groups did not hinder initiator-free crosslinking of PELA-based hydrogels. Based on these results, the remaining studies were performed with PELA (or PEL) at 5 wt% and PEG-4LA supplemented at 1 to 5 wt%.

3.3. Testing viscoelasticity of PELA-based dynamic hydrogels

Biological tissues are known to exhibit various degrees of stress-relaxation, which regulates cell fate processes. We asked if the phenylboronic acid (PBA) pendant groups on the purely synthetic PELA could be leveraged for forming reversible boronate-diol bonds that give rise to tissue-like stress-relaxation. To answer this question, we first employed PVA (146 −186 kDa), a linear polymer with abundant 1,3-diols for reversible complexation with the PBA groups on PELA. The PELA-based (5 wt%) hydrogels supplemented with PEG-4LA were fabricated by 365 nm light irradiation at 20 mW/cm² for 10 min at 37 °C. As expected, increasing PEG-4LA from 1 wt% to 5 wt% significantly increased the storage modulus of PELA-based hydrogel, with G’ increased from less than ~100 Pa to ~ 8 kPa (Figure 2A). In addition, the incorporation of 1 wt% PVA to the network resulted in noticeably higher stiffness owing to the presence of additional reversible boronate-diol bonds. It was also noted that without PVA the two-component PELA/PEG-4LA hydrogels became softer in 10 days (Figure S4A), presumably due to gradual reduction of disulfide bonds overtime. This was less noticeable in hydrogels containing PVA (Figure S4B). In addition, the inclusion of PVA resulted in gels with increase opacity (Figure S5).

Figure 2. Mechanical properties of viscoelastic hydrogels.

(A) Storage modulus of hydrogels with increasing PEG-4LA content (1 to 5 wt%) in the absence or the presence of 1 wt% PVA. In this study, PELA concentration was maintained at 5 wt% and the hydrogels were prepared upon UV exposure at 365 nm (20 mW/cm²) for 10 min at 37 °C. B) Stress-relaxation profiles of fabricated hydrogels (G’ ~1 kPa) with increasing PVA content from 0 wt% to 3 wt%. The gel formulations: 5 wt% PELA + 2 wt% PEG-4LA, 5 wt% PELA + 1 wt% PEG-4LA + 1 wt% PVA, 5 wt% PELA + 1 wt% PEG-4LA + 2 wt% PVA and 5 wt% PELA + 1 wt% PEG-4LA + 3 wt% PVA. (C) The effect of PHD on the stiffness of fabricated hydrogel using the formulation as 5 wt% PELA + 4 wt% PEG-4LA. (D) The stress-relaxation profiles of viscoelasic hydrogels with increasing PHD content from 0 wt% to 3 wt%. Parameters for stain-sweep and stress-relaxation analyses via rheology: f = 1 Hz, T = 25 °C and F_N = 0.25 N (γ = 0.1 – 5 % for strain-sweep and γ = 10 % for stress-relaxation). Each experiment was performed in three replicates.

Next, the stress-relaxation profiles of PELA/PEG-4LA hydrogels containing different amounts of PVA (0 wt% to 3 wt%) were tested. The formulations of the hydrogels were tuned to reach initial G’ (i.e., G’₀) of ~1 kPa. As shown in Figure 2B, soft pure PELA/PEG-4LA hydrogels (i.e., 0 wt% PVA) stress-relaxed after prolonged straining, with a stress-relaxation halftime (τ_1/2) of ~1,800 sec. While this result was in agreement with that reported in the literature,³⁶ the relaxation time scale was too slow to mimic the highly viscoelastic pancreatic cancer tissue (τ_1/2<100 sec).⁹ When PVA was included in the hydrogel formulations, however, the characteristic relaxation times were significantly shortened τ_1/2 from ~516 sec for 1 wt% PVA to ~116 sec for 3 wt% PVA, respectively (Figure 2B). It is worth noting that while prior report of pure 4-arm PEG-LA hydrogels (at 20 wt%) also demonstrated tunable relaxation, additional high amount of PEG-thiol (PEG-SH) was needed (at stoichiometric ratio of LA to SH) only to reduce the relaxation half-time to ~800 sec,³⁶ still far slower than the sub-100 sec needed for mimicking pancreatic tissues. In contrast, we were able to reduce the relaxation half-time to ~100 sec with addition of 3 wt% PVA.

We next asked if the new PELA/PEG-4LA/PVA hydrogel system could be dynamically stiffened and softened using reversible light induced dithiolane ring opening polymerization. Further, we were interested in understanding the effect of this reversible stiffening/softening process on the relaxation of the hydrogels. Taking advantage of the photo-induced rearrangement of disulfides in the PELA/PEG-4LA hydrogels, we demonstrated post-gelation stiffening and softening of PELA/PEG-4LA hydrogels both in the absence or presence of PVA. Specifically, additional 10 min of UV irradiation at 37 °C resulted in ~6.5-fold and ~4-fold increase in G’ for soft hydrogels (G’₀ ~1 kPa) in the absence and presence of 1 wt% PVA, respectively (Figure S6A, post-stiffened group). Moreover, through LAP and GSH facilitated thiol-disulfide exchange, all stiffened hydrogels were subsequently softened significantly after 50 sec of light irradiation at ambient temperature (final G’ below 1 kPa, Figure S6A, post-softened group). It is worth noted that all dynamic stiffening/softening experiments were performed in DMEM with high glucose concentration (4500 mg/L), mimicking the maximum glucose content found in pancreatic cancer patients (>2000 mg/L).⁴⁴ While the introduction of PVA significantly accelerated relaxation of the PELA-based hydrogels, the stiffening process unfortunately increased the relaxation time scale (i.e., less stress-relaxation) (Figure S6B).

We postulated that the loss of fast stress-relaxation characteristics in the stiffened PELA/PEG-4LA/PVA hydrogels was a result of lower affinity of 1,3-diols (on PVA) with BA groups (on PELA). Thus, we hypothesized that replacing cis-1,3 diols with cis-1,2 diols would strengthen the reversible condensation reaction of the cyclic BA-diol ester bonds. To achieve this, we utilized a previously reported 1,2-diol bearing modified RAFT polymer poly(N-hydroxyethyl acrylamide)-dopamine (PHEAA-DOPA or PHD) (Figure S3B).²² As shown in Figure 2C, adding PHD (1 wt% to 3 wt%) slightly increased G’ from ~3.2 kPa to ~3.9 kPa. In contrast to soft PELA hydrogels that exhibited some relaxation even in the absence of diol-containing polymers (Figure 2B), no relaxation was observed for stiffer PELA/PEG-4LA hydrogels with 0% PHD (Figure 2D). This suggests a strong correlation of stiffness and the relaxation induced by pure disulfide bond exchange. As PELA/PEG-4LA hydrogels at ~3.9 kPa exhibited no noticeable stress-relaxation, these hydrogels were considered elastic (EL). In contrast, increasing PHD content to 1 wt%, 2 wt% and 3 wt% drastically shortened τ_1/2 to ~475 sec, ~156 sec and ~63 sec, respectively (Figure 2D). PHD-containing PELA/PEG-4LA hydrogels outperformed those with PVA in terms of the speed of relaxation even at moderately high stiffness. More importantly, the relaxation halftime approached that of native pancreatic tumor tissues (τ_1/2 ~ 70 to 100 sec) as stated earlier.

3.4. Reversible stiffening/softening of fast-relaxing hydrogel network

We sought to explore light-induced dynamic tuning of hydrogel mechanics to mimic tumor stiffening and stroma softening. Matrix stiffening is of great interest to cancer research as a stiffened tumor matrix negatively correlates to poor therapeutic treatment outcome, potentially due to abnormal cell growth and dysregulated signaling.^{7, 45} Furthermore, It was previously reported that the stiffer and more stress-relaxing matrix-induced cell proliferation and chondrogenic ECM deposition.⁴⁶ With these examples in mind, we developed a viscoelastic cell culture system with tunable mechanics which was monitored by in situ rheological measurements. We first prepared soft gels with G’ around 1 kPa from 5 wt% PELA and 1 wt% PEG-4LA (10 min of 365 nm light exposure as described above). Hydrogels as fabricated were incubated in additional PEG-4LA (6 wt%) dissolved in PBS for 1 hr to allow the penetration of polymer precursors in the hydrogel network, followed by additional light exposure to induce dynamic stiffening (Figure 3A). The in situ rheometry results in Figure 3B demonstrated a gradual and substantial increase in G’ of the hydrogels (from ~1 kPa up to ~6 kPa). This suggests that disulfide bond arrangement through photo-activation and the additional dithiolane rings contributed to the increase of matrix stiffness. To show that the increase in G’ occurred upon the photo-stiffening mechanism, we conducted two control experiments using the same hydrogel system. The hydrogels were incubated with additional PEG-4LA precursors but without UV irradiation or with additional light irradiation but without PEG-4LA. In both cases, there was no increase in G’ overtime, suggesting that both light and additional dithiolane were needed for the dynamic stiffening. It is worth noting that the stiffness change (i.e., from 1 kPa to 6 kPa) was within the range of observed stiffness difference in the pancreatic healthy and tumor tissues.⁹

Figure 3. Dynamic tuning of PELA hydrogel viscoelasticity.

(A) Schematic of on-demand stiffening. (B) In-situ rheology profiles presenting the network stiffening from ~1 kPa towards ~6 kPa upon continuous UV irradiation at 365 nm (20 mW/cm²) at 37 °C. The hydrogels were incubated in 6 wt% PEG-4LA prior to stiffening. (C) In-situ rheology profile of stepwise network stiffening upon intermittent light exposure (λ= 365nm, 20 mW/cm²) at 37 °C. (D) Schematic illustration on the photo-degradation mechanism of dithiolane crosslinked hydrogels. (E) In-situ profile presenting the photo-softening behaviour of fabricated hydrogels in the presence of LAP and GSH. Prior to the degradation test, the hydrogels were incubated in the GSH solution with and without 20 mM LAP. (F) The tunability of matrix stiffness upon variation of UV exposure time. The hydrogels were incubated in 6 wt% PEG-4LA for 1 hr prior to the stiffening and incubated in photo-initiator solution (10 mM LAP + 5 mM GSH) prior to the softening. (G) Stress relaxation profiles of viscoelastic hydrogels throughout the dynamic mechanical variation. Parameters for in-situ rheology analyses: γ = 1 %, f = 1 Hz, T = 37 °C/25 °C and F_N = 0 N. Parameters for stain-sweep and stress-relaxation analyses via rheology: f = 1 Hz, T = 25 °C and F_N = 0.25 N (γ = 0.1 – 5 % for strain-sweep and γ = 10 % for stress relaxation). Each experiment was performed in three replicates. (H) Viability of encapsulated Pa03C cells within EL and VE hydrogels during dynamic photo-tuning process. Cells were imaged at day 0, day 4 (after stiffening) and day 6 (after softening). The hydrogels were fabricated by 365 nm light irradiation (20 mW/cm²) at 37 °C for 5 min. Stiffening was performed by incubating the cell-laden gels in 6 wt% PEG-4LA solution for 1 hr, followed by additional light irradiation for 5 min. Softening was conducted by incubating the hydrogels in photo-initiator and mono-thiol solution (i.e., 10 mM LAP + 5 mM GSH) for 30 min followed by additional light irradiation (365 nm, 20 mW/cm² at room temperature) for ~50 sec.

One of the unique features of light-mediated crosslinking is their excellent temporal control over polymerization kinetics. After incubating the pre-formed PELA/PEG-4LA hydrogels in 6 wt% PEG-4LA solution, we attempted intermittent on-off irradiations and observed light-dependent stiffening (Figure 3C), a result similar to that reported in the literature.^{32, 47, 48} Furthermore, the same hydrogels could also be dynamically softened, granted that additional photoinitiators and free thiols were provided. Mechanistically, light irradiation induces photolysis of additional photoinitiators (e.g., LAP), which abstract hydrogens from free thiols (e.g., GSH) and produced new thiyl radicals.⁴⁹ The later participate in additional thiol-disulfide exchange and reduce crosslinking density by terminating the disulfide exchange (Figure 3D). To demonstrate on-demand softening of PELA/PEG-4LA hydrogels, we fabricated stiff hydrogels (~7 kPa) using 5 wt% PELA and 5 wt% PEG-4LA. The hydrogels were equilibrated in PBS overnight, followed by swelling in LAP (20 mM) and GSH (5 mM) for 30 min. On demand softening was initiated by 365 nm light exposure (20 mW/cm²) at room temperature. As shown in Figure 3E, hydrogel moduli decreased rapidly soon after light was turned on and reached ~30% of the initial stiffness. Control experiment using GSH alone (i.e., no secondary light initiator) did not result in any degradation.

Encouraged by the on-demand stiffening or softening of different hydrogels, we next investigated the photo-induced dynamic stiffening/softening of the same viscoelastic hydrogels (5 wt% PELA, 1 wt% PEG-4LA, 3 wt% PHD). Photo-stiffening and softening were conducted with the same protocol as mentioned above apart from that a slightly lower LAP concentration (10 mM) was used, considering future applications of the system for cell study. With this hydrogel formulation, 5 min of light irradiation resulted in hydrogels with initial G’ of around 700 Pa (Figure 3F). The hydrogels were stiffened to approximately ~2.4 kPa upon additional 5 min of light exposure. A second 5 min of light irradiation further increased G’ to ~3.7 kPa. This group of hydrogels were subsequently photo-softened to ~2 kPa and then to ~0.5 kPa after two-stage (50 sec and 70 sec) of light exposure (Figure 3F). Finally, we tested the stress-relaxation profiles of the same hydrogels underwent the sequential stiffening and softening experiments and found that all hydrogels (initial, stiffened, softened) maintained fast relaxation (τ_1/2 ~ 90 to 240 sec) (Figure 3G). It was worth noting that while the reversible stiffening and softening experiments reduced the degree of relaxation, all hydrogels were still considered fast-relaxing and will be used for subsequent cell studies.

Next, we tested the cytocompatibility of the dynamic dithiolane-based hydrogels using Pa03C, a patient-derived pancreatic cancer cell line. Pa03C cells were encapsulated in hydrogels crosslinked by PEL/PEG-4LA/PHD (denoted as EL) or PELA/PEG-4LA/PHD (denoted as VE). The only difference between these two gel formulations was the use of boronic acid-free PEL or boronic acid-containing PELA. As the stress-relaxation was granted by boronate-diol bonding, PEL-based hydrogels did not display noticeable stress-relaxation and were denoted as EL gels. On the other hand, PELA-containing hydrogels were labelled as VE gels owing to the fast-relaxing behavior. Regardless, all hydrogels were crosslinked by 5 min of 365 nm light (20 mW/cm²) exposure at 37 °C and exhibited initial G’ of ~2.5 kPa (data not shown). Even in the absence of any cell adhesive motif, the encapsulated cells exhibited high viability (>90% live cells) in both EL and VE gels, indicating the cytocompatibility of this purely synthetic hydrogel platform (Figure 3H). The cell-embedded hydrogels were cultured for 4 days and then incubated in 6 wt% PEG-4LA solution for 1 hr and stiffened under the same process as initial gelation condition. After 2 more days, the gels were incubated in the initiator solution (10 mM LAP + 5 mM GSH) for 30 min, followed by softening upon 365 nm light irradiation for 50 sec at room temperature. No substantial cell death throughout the sequential dynamic stiffening/softening process was observed in both EL and VE hydrogels and the cells proliferated from single cells to multicellular spheroids (Figure 3H). Interestingly, while the sequential stiffening/softening process did not adversely affect cell viability, the spheroids in the stiffened-then-softened VE gels appeared larger than those in the EL counterpart. Thus, further investigation was warranted.

3.5. Effect of matrix relaxation on cancer cell fate

Encouraged by the high cytocompatibility of the PELA-based hydrogels, we further explored the effect of matrix relaxation on pancreatic cancer cell fate. To this end, we encapsulated dispersed single Pa03C cells in hydrogels crosslinked initiator-free using PELA (or PEL for elastic control) PEG-4LA, and PHD. While there was no notable difference between the elastic control PEL hydrogels (i.e., EL) and viscoelastic PELA hydrogels (i.e., VE) in the first 4 days of culture, cells in the VE gels grew into significantly larger spheroids than in EL gels by day 7 (Figure 4A, 4B). It was likely that the rearrangement of boronic ester bonds in the VE gels permitted higher degree of polymer chain flexibility, which allowed the cells to grow into larger spheroids. As both the EL and VE hydrogels were crosslinked in the absence of any biological motif (e.g., integrin ligand RGD), the morphological differences seen between these two groups can be largely attributed to the physical cues (e.g., intracellular traction forces). However, additional studies in the future are needed to elucidate the underlying mechanisms.

Figure 4. The impact of matrix viscoelasticity on Pa03C cellular behavior in 2D and 3D models.

(A) Representative live/dead images of cells in EL and VE gels. (B) Quantitative analysis of cells size obtained from live/dead confocal images. (C) Confocal images presenting the structure of 3D spheroids cultured in EL and VE hydrogels. (D) mRNA expression of selected EMT genes. EL hydrogels: 5 wt% PEL, 8 wt% PEG-4LA, 3 wt% PHD; VE hydrogels: 5 wt% PELA, 7 wt% PEG-4LA, 3 wt% PHD.

In a separate experiment, we used Aggrewell to pre-aggregate Pa03C cells into small spheroids prior to encapsulation. The encapsulated spheroids were stained for F-actin (by Alexa Fluor 488 Phalloidin), with cell nuclei counterstained by DAPI (Figure 4C). In the EL gels, the spheroids remained rounded over 5 days. In contrast, spheroids in the VE gels adapted more irregular morphology. We reason that the elastic PEL-based hydrogels were more restrictive for cell morphogenesis, whereas the introduction of boronate groups on the PELA polymer enabled stress dissipation and network rearrangement, a prerequisite of cellular fate processes, such as migration and spreading. The irregularly shaped spheroids in the VE gels warranted further investigation. After culturing the spheroids in EL and VE gels for 5 days, the cells were collected via light-mediated disulfide bond scission (Figure 3E). As shown in the Figure S7, the spheroids recovered from photo-mediated gel degradation did not negatively affect spheroid integrity and viability. The recovered spheroids were prepped for RNA extraction and qRT-PCR analysis of a panel of six genes related to epithelial-mesenchymal transition (EMT) and mechanosensing. As shown in Figure 4D, the expression of mesenchymal markers vimentin, ZEB1 (i.e, Zinc finger E-box binding homeobox 1), and CDH2 (i.e., N-cadherin) were significantly higher in VE gels than in EL gels, suggesting that VE gels induced more EMT phenotype in the cancer spheroids.⁵⁰

EMT is a cellular process preceding invasion and metastasis. Many genes are implicated in the EMT process, including vimentin,⁵¹ and N-cadherin (CDH2),⁵² and ZEB1.⁵³ Increase in vimentin expression in epithelial cells changes cell shape and motility, whereas upregulation of CDH2 signifies the loss of cell-cell adhesion.⁵⁴ On the other hand, upregulation of ZEB1 correlates with tumor-promoting pathways TGF-β, β-catenin, as well as miRNA expression associated with multidrug resistance.⁵⁵ ZEB1 upregulation also weakens cell-cell adhesion by repressing the epithelial marker E-cadherin and enhances the mesenchymal markers N-cadherin.^{55, 56} Our imaging and gene expression results of this pilot study suggest that, without the complication of biological motifs, fast matrix relaxation directly upregulated promoted mesenchymal phenotype of the cancer cells. While the expressions of Vimentin, ZEB1, and N-cadherin were significantly upregulated at the mRNA levels, we did not observe significant differences in the immunostaining results (data not shown), potentially due to the lack of any biological motifs that may play a role in the protein expression. Nonetheless, to the best of our knowledge, fast relaxing disulfide crosslinked hydrogels have not been used to study viscoelasticity induced cellular response. Using the purely synthetic PELA-based hydrogels and without the influence of any biological motifs, we successfully demonstrated the impact of matrix viscoelasticity on cancer cell fate. We believe the PELA/PHD-based hydrogels were the first purely synthetic and fast-relaxing (τ_1/2 <100 sec, Figure 2D, 3G) hydrogels capable of supporting cancer cell fate processes. With these promising results, other biologically active ligands may be added in the fast-relaxing hydrogels to permit a reductionist approach in studying the impact of specific biological cues on cancer cell behaviors. This can be achieved via physical entrapment or conjugating desired proteins/proteoglycans with lipoic acid or lipoic acid-PEG-amine prior to co-polymerization into the PELA-based hydrogels.

3.6. Reversible stiffening/softening of fast-relaxing hydrogels to probe cancer cell response

A hydrogel that can be dynamically stiffened can be used to recapitulate progressive matrix crosslinking, whereas one that can be dynamically softened is highly useful to predict the effect of stroma-targeting on cancer cell behavior. While engineered hydrogels with reversible stiffening and softening capability have been reported, to the best of our knowledge, none of the reported examples exhibit fast relaxation time scale akin to pancreatic cancer. Using our fast-relaxing PELA-based hydrogels (i.e., PELA/PEG-4LA/PHD, G’ ~3 kPa), we demonstrated the effects of reversible matrix stiffening/softening on the growth of pancreatic cancer cells Pa03C. Figure 5A illustrates the experimental designs and timelines. It is worth noting that all hydrogel groups displayed fast-relaxation akin to that of the pancreatic tumor tissues (Figure 3G). After the encapsulated spheroids were cultured for 2 days, the hydrogels were either stiffened to G’ ~5 kPa (i.e., the Stiffen group) or softened to G’ <1 kPa (i.e., the Soften group) using the protocols presented above. In the reversible Stiffen-Soften group, hydrogels were stiffened at day 2, followed by softening at day 5. Between day-2 and day-5, Pa03C spheroids in the control group became less rounded and irregular in cell shape (Figure 5B). In contrast, spheroids in the Stiffen group remained largely rounded even after culturing through day 8 in the highly viscoelastic PELA-based hydrogels (Figure 5C). It was likely that the stiffened matrices were too rigid for the cells to deform, especially in the absence of cell-adhesive ligands in the purely synthetic PELA-based hydrogels. We also performed hydrogel softening and observed progressive growth and irregular shapes of the Pa03C spheroids (Figure 5D).

Figure 5. Confocal images visualizing the mechanosensing behaviour of Pa03C cell spheroids in viscoelastic hydrogels.

(A) Representative illustration on the photo-induced mechanical variation of viscoelastic hydrogels embedding cancer spheriods. Live/dead confocal images monitoring behaviour of spheroids’ morphology on altering the matrix mechanics (B: control gels, G’ ~3 kPa), (C: stiffened gel), (D: Softened gel) and (E: Stiffened-Softened gel). Live cells were stained with calcein AM (green), and dead ells were stained with Ethidium Homodimer III (red). The hydrogels were fabricated by irradiation (365 nm, 20 mW/cm²) at 37 °C for 5 min. The stiffening was performed by incubation in 6 wt% PEG-4LA for 1 hr followed by irradiation at 365 nm (20 mW/cm², 37 °C) for 5 min, and the softening was performed by incubating the hydrogels in photo-initiator solution (10 mM LAP + 5 mM GSH) for 30 min followed by irradiation at 365 nm (20 mW/cm², r.t.) for ~50 sec.

We further explored whether the history of matrix mechanics would have an impact on cancer cell growth. To achieve that, the stiffened spheroids-laded PELA hydrogels (G’ from 2 to 5 kPa) were further softened on day 5 to decrease the modulus similar to the hydrogels softened to < 1 kPa (Figure 5C, 5E). Interestingly, after incubating the hydrogels for another 3 days, the cell spheroids that underwent the reversible stiffening-softening process became larger in size but maintained the circular morphology (Figure 5E), similar to that observed before softening (Figure 5C). This qualitative observation suggests that cancer cell growth depends on not just matrix stiffness but also the history of the stiffness. Specifically, cells that never experienced a stiff microenvironment might grow significantly due to fast matrix relaxation (Figure 5D). Once experienced a stiffened matrix, cancer cells may be restrained from excessive outgrowth even the stiffened matrix was later softened (Figure 5E). It is worth noting that additional quantitative evaluation of cell fate is needed to draw firm conclusion about the potential ‘mechanical memory’ of pancreatic cancer cells. Future work will focus on building the purely synthetic, fast-relaxing, and reversible tunable hydrogels with bioactive motifs to investigate the specific matrix cues on cancer cell fate processes, including migration, invasion, and drug sensitivity.^57–60

4. Conclusion

In this study, we developed a new cytocompatible fast-relaxing hydrogel based on RAFT polymers (PELA and PHD) and granted the hydrogels with initiator-free crosslinking, fast-relaxation, and reversible tuning of matrix mechanics. These were achieved via combining dithiolane ring-opening chemistry and reversible boronate-diol bonding. The hydrogels were highly modular, reversibly tunable, and with high cytocompatibility. With this new hydrogel system, we performed pilot cell studies to evaluate the effect of matrix relaxation and stiffness on PDAC cell growth and potential EMT. Similar results were obtained using cancer cell spheroids, indicating that the cells in fast-relaxing gels have higher tendency to expand than those in the elastic gels. Finally, we designed reversible stiffening/softening experiments to probe the potential mechanical memory of the encapsulated spheroids in the absence of any cell adhesive ligand. Future work includes studying the mechanisms underlying the observed relaxing-induced cell fate changes, as well as designing PELA-based hydrogels containing biological moieties to facilitate a reductionist approach in studying cell-matrix interactions.

Scheme 1.

Schematic illustration for the synthesis of lipoic acid acrylamide (LAA) monomer.

Scheme 2.

Schematic illustration for the synthesis of AAPBA monomer.

Scheme 3.

Schematic illustration for the synthesis of statistical copolymers: poly(PEGA-co-LAA-co-AAPBA) PELA, and poly(PEGA-co-LAA) PEL.

Scheme 4.

Schematic illustration for the synthesis of PEG-4LA.

Scheme 5.

Schematic illustration for the side-chain modification of PHEAA with catechol functionalities.

Highlights.

• New dynamic fast-relaxing hydrogels were developed by combining initiator-free dithiolane ring-opening chemistry and boronate-diol bonding.

• The hydrogels were highly modular, with reversibly tunable stiffness, and high cytocompatibility for in vitro cancer cell research.

• With this new hydrogel system, we discovered that pancreatic cancer cells exhibited higher degrees of growth in fast relaxing matrices.

• We also showed that reversible tuning of fast-relaxing hydrogel stiffness led to potential mechanical memory in the cancer cells.