Tuning Hydrogels by Mixing Dynamic Cross‐Linkers: Enabling Cell‐Instructive Hydrogels and Advanced Bioinks

Abstract

Rational design of hydrogels that balance processability and extracellular matrix (ECM) biomimicry remains a challenge for tissue engineering and biofabrication. Hydrogels suitable for biofabrication techniques, yet tuneable to match the mechanical (static and dynamic) properties of native tissues remain elusive. Dynamic covalent hydrogels possessing shear‐thinning/self‐healing (processability) and time‐dependent cross‐links (mechanical properties) provide a potential solution, yet can be difficult to rationally control. Here, the straightforward modular mixing of dynamic cross‐links with different timescales (hydrazone and oxime) is explored using rheology, self‐healing tests, extrusion printing, and culture of primary human dermal fibroblasts. Maintaining a constant polymer content and cross‐linker concentration, the stiffness and stress relaxation can be tuned across two orders of magnitude. All formulations demonstrate a similar flow profile after network rupture, allowing the separation of initial mechanical properties from flow behavior during printing. Furthermore, the self‐healing nature of hydrogels with high hydrazone content enables recyclability of printed structures. Last, a distinct threshold for cell spreading and morphology is observed within this hydrogel series, even in multi‐material constructs. Simple cross‐linker mixing enables fine control and is of general interest for bioink development, targeting viscoelastic properties of specific cellular niches, and as an accessible and flexible platform for designing dynamic networks.

Dynamic hydrogels are an important tool for the development of advanced bioinks. Individually, dynamic cross‐linkers introduce time‐dependent properties including shear thinning, self‐healing, and stress relaxation. By mixing oxime and hydrazone cross‐links in an oxidized alginate hydrogel, hybrid networks can combine desirable properties of both, leading to tunable mechanical properties, improved processability and printability, and the ability to control fibroblast morphology.

1. Introduction

Progress in tissue engineering and regenerative medicine (TERM) has created a demand for the creation of complex 3D scaffolds that mimic native tissue and tissue development. The native extracellular matrix (ECM) microenvironment can be considered as a soft, bioactive hydrogel with carefully controlled mechanical (physical) and signaling (biochemical) cues. Early work has established the importance of static mechanical properties such as substrate stiffness (elasticity)^{[ 1} , ^{2 ]} on cellular behavior, while more recently, the importance of material dynamics (viscoelasticity) and spatial organization have been demonstrated.^{[ 3} , ⁴ , ^{5 ]} When creating synthetic ECM environments, different mechanical regimes have been shown to direct the behavior of diverse cell types including fibroblasts,^{[ 3 ]} human adipose‐derived stem cells (hASCs),^{[ 6 ]} human endothelial colony‐forming cells (hECFCs),^{[ 7 ]} human mesenchymal stem cells (HMSCs),^{[ 5 ]} and chondrocytes.^{[ 8 ]} Control over the mechanical properties (both stiffness and stress relaxation) of biomaterials is necessary to induce desired cell behavior and facilitate tissue formation.

In addition, the creation of advanced tissue and organ models also requires processable materials, which can be fabricated to achieve control over cellular arrangement in 3D. Bioprinting is an attractive approach, though there is a known tradeoff between the printability of a material and its suitability (permissivity, mechanical biomimicry) for cell culture. These conflicting mechanical goals force compromises in processability and biomimicry to be made, limiting the biofabrication window.^{[ 9 ]} Currently, the challenge remains to balance and merge processability with tissue formation post‐fabrication, especially in synthetically modified materials.

To overcome the dichotomy between cell culture and biofabrication, dynamic hydrogels have emerged as a promising approach.^{[ 10 ]} Built from reversible interactions, these hydrogels can have tunable elastic and viscoelastic properties while maintaining processability via shear‐thinning and self‐healing. Dynamic hydrogels can be broadly classified as either dynamic covalent,^{[ 11 ]} whereby a covalent bond is formed and broken in equilibrium, or supramolecular,^{[ 12 ]} in which physical attractions (electrostatics, H‐bonding, host‐guest, hydrophobic, metal‐ligand coordination) lead to reversible (self‐)assembly.^{[ 13} , ^{14 ]}

Dynamic covalent chemistry (DCvC) has recently been highlighted as a versatile tool for the development of dynamic hydrogels. Various classes of DCvC have been explored including imines,^{[ 15} , ^{16 ]} boronic acid esters,^{[ 17} , ^{18 ]} disulfides,^{[ 19} , ²⁰ , ^{21 ]} and diels–alder.^{[ 11} , ²² , ^{23 ]} Via control over the molecular equilibria of the underlying DCvC, stiffness, stress relaxation, and self‐healing of the hydrogels can be controlled in a rational manner. For example, the control over mechanical properties has been attributed to thermodynamic control of the reversible reaction.^{[ 24 ]} Changing a single atom of the cross‐linker can induce large changes in equilibrium, rate constants, and the resulting hydrogel properties.^{[ 16} , ²⁵ , ^{26 ]} More precisely, the equilibrium constant (K _eq) is proportional to the number of active cross‐links, and the stiffness of a hydrogel (Figure 1A). Similarly, the forward rate constant (k ₁) is proportional to rates of gel formation and self‐healing kinetics, while the reverse rate constant (k ₋₁) is proportional to stress relaxation behavior.

Figure 1.

Dynamic covalent cross‐links based on secondary aldimine formation between oxidized alginate and dihydrazides/bishydroxylamines. A) DCvC leverages the relationship between molecular constants such as the equilibrium constant, and the resulting mechanical properties. *Reference equilibrium values taken from ref. [27]. B) General schematic for the formation of a secondary aldimines from and aldehyde and an amine‐terminated functional group; variation in the R leads to different equilibria and rate constants for the reaction. In this work, we use oxidized alginate combined with a dihydrazide and/or bishydroxylamine to form hydrazone and oxime cross‐links.

A prominent class of DCvC cross‐links is Schiff bases such as hydrazone and oxime. These reactions have historically been used as bioconjugation tools,^{[ 27 ]} and more recently, McKinnon et al. and Wang et al. have used hydrazone DCvC to prepare dynamic hydrogels for viscoelastic 3D cell culture, and injectable stem cell delivery, respectively.^{[ 28} , ^{29 ]} Tunable DCvC hydrogels have also been prepared using boronic acid esters by Yesilyurt et al.,^{[ 30 ]} and a detailed review of molecular tuning available using boronic acid ester systems was recently presented by Marco–Dufort and Tibbit.^{[ 31 ]}

Due to the modular and autonomous cross‐linking enabled by DCvC, one can tune the properties of a hydrogel by simply mixing cross‐linkers with different kinetic and thermodynamic properties. To date, only a few examples of this mixing of dynamic cross‐linkers have been explored. Richardson et al. employed the mixing of an alkyl and a benzyl aldehyde to tune hydrogel viscoelasticity independent of stiffness, and showed that the mixed systems allowed greater ECM production in chondrocytes.^{[ 32 ]} Similarly, Yesilurt et al. explored mixed PEG networks cross‐linked using two distinct boronic acid esters with different diol complexation rates, and also observed fine control over the stress relaxation behavior independent of stiffness.^{[ 30 ]} Notably, the measured stress relaxation was faster for the phenylboronic acid with a higher pKa, reinforcing the molecular control over macroscopic mechanics accessible using DCvC systems.

Here, we sought to investigate a similar cross‐linker mixing strategy, building upon our previous work with oxidized alginate cross‐linked by Schiff base formation.^{[ 16 ]} By mixing the most dynamic (hydrazone) and least dynamic (oxime) cross‐links with oxidized alginate (Figure 1B), we hypothesized that we could finely tune the mechanical properties of our dynamic hydrogels without changing the cross‐linker concentration. We also sought to see if this tuning could facilitate more printable, stable hydrogels, with fine control over cell‐matrix interactions. Using a combination of rheology and self‐healing tests, we investigated the effect of mixed cross‐linkers on the hydrogel properties. In addition, we investigated the differences in shear‐thinning, injectability, and (bio)printability of these hydrogels as a function of the cross‐linker composition. During these investigations, we also observed impressive recycling and reprocessabilty of our reversibly cross‐linked hydrogels. Last, we investigated the behavior of human dermal fibroblasts (HDFs) as a function of hydrogel composition, and explored their use as bioinks as well as their ability to control cell morphology in multi‐material constructs.

2. Results and Discussion

2.1. Designing Schiff‐Base Dynamic Covalent Hydrogels

Oxidized alginate is inexpensive, synthetically facile to prepare, biocompatible, and biodegradable, making it an excellent material for diverse applications in tissue engineering. The presence of aldehydes along the backbone also facilitates the bioconjugation of bioactive molecules/peptides such as RGD. We chose a low polymer content (2 wt.%), as higher polymer contents are known to increase network density, inhibit cell migration, and limit development.^{[ 33 ]} For the design of a series of mechanically tunable hydrogels that leverage the mixing of cross‐linkers possessing different molecular constants, we sought to keep all other variables constant while choosing a regime relevant for bioprinting applications. In the current study, we have mixed the cross‐linkers (adipic acid dihydrazide and O,O′‐1,3‐propanediylbishydroxylamine) at a constant total concentration with oxidized alginate (2 wt.%) in order to investigate the influence of their ratio on the mechanical properties of hydrogels. All our hydrogels were formed using 2 wt.% oxidized alginate (Ox‐Alg, M _n = 114 000, Ð = 2.0, theoretical degree of oxidation = 11.5%) with 1.0 mol equiv of cross‐linker functions with respect to the aldehyde on the Ox‐Alg backbone (10.1  mM at 2 wt.%).

We first wanted to explore the pristine materials and their properties with respect to the different molecular cross‐linkers. First, we observed that the hydrazone cross‐linked hydrogel was softer when directly compared to the oxime cross‐linked hydrogel (Figure 2 ). Turning towards rheology, we observed that the shear storage modulus (G’) of the hydrazone hydrogel was approximately an order of magnitude higher than the oxime hydrogel (Figure 3A), in line with an expectedly larger K _eq for oxime. We then measured the gelation kinetics of both hydrazone and oxime hydrogel formation by rheometry to determine which cross‐linker has a faster forward rate constant. We found that gel formation via oxime cross‐links was faster, with an observed gel point (defined here as the crossover of G’ and G’’ indicating a shift to an elastically dominated regime) of 6 min compared to 20 min for hydrazone gelation (Figure S1, Supporting Information). To compare the relative magnitudes of the reverse rate constant, the stress relaxation behavior of both hydrazone and oxime hydrogels was measured (Figure 3C), and is discussed in more detail later. As expected, the hydrazone hydrogel exhibited a faster stress relaxation, indicative of a larger k ₋₁. The larger k ₁ but smaller k ₋₁ measured for oxime hydrogels are self‐consistent with the larger K _eq, showing that our molecular control over macroscopic mechanical properties follows the expected relationships.

Figure 2.

Translating equilibria constants to macroscopic observations: Dependence of hydrogel stiffness on K _eq. A) Schematic hydrogel networks formed with 2% (w/w) oxidized alginate (from left to right) using 1 mol equiv hydrazone cross‐links, 0.6 mol equiv hydrazone plus 0.4 mol equiv oxime cross‐links, and 1 mol equiv oxime cross‐links. All mol equiv are w.r.t. the aldehyde concentration present in a 2% (w/w) solution of oxidized alginate. B) Photos of the same series of oxidized alginate hydrogels illustrating the difference in stiffness as the average equilibrium constant of the cross‐links increases. Hydrazone possesses a lower K _eq, leading to a lower proportion of bound states at any given moment, and thus a lower stiffness. Consequently, the 100% hydrazone hydrogel stretched and deformed under its own weight, wrapping around the spatula. By replacing 40% of the hydrazone cross‐links with oxime cross‐links that have a larger equilibrium constant, we observe a stiffening of the hydrogel, while the 100% oxime hydrogel is markedly stiffer.

Figure 3.

Mechanical properties of dynamic hydrogels can be tuned by mixing oxime and hydrazone cross‐links. A) The shear storage modulus (G’) obtained for each formulation. By successively replacing 0.2 mol equiv hydrazone with oxime, we can tune the final hydrogel modulus from 0.22 to 2.78 kPa. B) The normalized and smoothed (see Figure S6, Supporting Information, for unprocessed data) strain sweep until rupture of each formulation shows a decrease in strain‐stiffening onset as the proportion of oxime increases. Notably, all formulations display strain‐stiffening behavior, a common property of native tissue. C) Normalized stress relaxation of the hydrazone and oxime hydrogels, measured by rheometry at 20% strain. Hydrazone cross‐linked hydrogels exhibit a fast stress relaxation similar to that of native tissues,^{[ 4 ]} compared to the much slower oxime hydrogels. D) The swelling ratio of a 150 µL hydrogel of each formulation over 24 h in 10 mL PBS. As the cross‐link density (and proportion of oxime) increases, the swelling capacity of the hydrogel decreases. Notably, the replacement of only 0.2 mol equiv hydrazone with oxime significantly increases the stability of the hydrogel compared to pure hydrazone hydrogel. *The latter began breaking up into gelatinous chunks when handled after 24 h. The error reported in panels A and C are the mean ± SD with n = 3–5 and n = 2 respectively.

Interestingly, if we consider the pK _a of our cross‐linkers, we observe a lower pK _a presenting faster stress relaxation behavior; adipic dihydrazide has a reported pK _a of ca. 2.5,^{[ 34} , ^{35 ]} while O‐ethylhydroxylamine (a representative molecule for O,O′‐1,3‐propanediylbishydroxylamine) has a reported pK _a of 4.65.^{[ 36 ]} This is the opposite trend in stress relaxation to that reported by Yesilyurt et al. for their phenylboric acid esters.^{[ 30 ]} Given that Schiff base formation proceeds via a different mechanism to that of boronic acids and diols, different molecular properties can guide reactivity. Indeed, the standard rate‐limiting step in hydrazone/oxime formation at neutral pH is the breakdown of the tetrahedral intermediate, and local electronic effects have been shown to have a large impact on react rates.^{[ 25} , ^{26 ]} In the current work, we explore only aliphatic hydrazone and oxime cross‐links, however, it merits mentioning that a wide range of possible reactive groups have already been studied in the context of bioconjugation.^{[ 27 ]} These previous studies provide many different chemical variants that have dramatically different reactivities, highlighting the potential to extend the principles of molecular engineering explored here to a broader range of molecular constants, and consequently, macroscopic mechanics.

By varying the ratio of hydrazone and oxime cross‐linkers, we aimed to highlight the potential of mixing dynamic cross‐linkers possessing different molecular constants to tune mechanical properties entirely independently of polymer content, molecular weight, or branching. We primarily explored a series that also maintains a constant total cross‐linker concentration to further highlight how very small changes in molecular design have significant ramifications for macroscopic mechanical properties.

2.2. Mixing Dynamic Cross‐Linkers Produces Hydrogels with Intermediate Stiffness

By mixing hydrazone and oxime cross‐links, we aimed to develop a highly tunable and processable hydrogel system. To verify the basis of our study, we prepared a 100% hydrazone hydrogel, a 100% oxime hydrogel, and a mixed hydrogel with 60% hydrazone and 40% oxime, and qualitatively analyzed their mechanical properties by hanging the resulting hydrogels at the edge of a spatula (Figure 2A,B). As expected, the oxime hydrogel resulted in a stiffer hydrogel able to support its own weight under gravity. In contrast, the hydrazone hydrogel stretched and deformed under its own weight, wrapping around the spatula, while the mixed hydrogel displayed behavior in between the pure systems, bending a lot more while retaining its circular shape. The relative proportion of hydrazone and oxime cross‐linkers clearly had a noticeable impact on hydrogel mechanics.

2.3. Effect of Cross‐Linker Ratio with a Fixed Total Cross‐Linker Concentration on Hydrogel Mechanics

To understand the effect mixing our cross‐linkers has on the mechanical properties of the hydrogel, we chose to analyze the stiffness, stress relaxation, and strain response of these materials, as well as their swelling behavior. From progress in mechanobiology, we know that both the stiffness and stress relaxation properties of a hydrogel can have a significant impact on cellular behavior,^{[ 37} , ^{38 ]} and consequently, that these properties are highly relevant for applications in TERM. The role of strain‐stiffening in cellular mechanosensing is less clearly understood, but is becoming an important parameter for investigation.^{[ 39} , ⁴⁰ , ⁴¹ , ^{42 ]} Finally, the swelling behavior is important for the stability of hydrogels under cell culture conditions. By analyzing this combination of mechanical properties, we aimed to provide the main relevant information for various cell culture applications.

By varying the proportion of hydrazone and oxime cross‐links comprising the 1.0 mol equiv, we prepared a series of six different formulations: HyOx(χ _Hy:χ _Ox) where χ _Hy and χ _Ox denote the mole fraction of hydrazone and oxime cross‐linkers, respectively. We characterized the rheological and swelling behavior of these six formulations in Figure 3.

Looking first at the stiffness of this series, HyOx(1.0:0.0) possessed the lowest shear storage modulus (G’) of 220 ± 30 Pa, compared to HyOx(0.0:1.0) with a shear storage modulus of 2780 ± 90 Pa (Figure 3A). Comparing the mean stiffness of sequential formulations, each increment replacing 0.2 mol equiv hydrazone with oxime yielded a mean stiffness increase (Δ[mean G’]) of 510 ± 210 Pa with the smallest values being the first and last increments, while a linear fit gives a slope of 550 ± 20 Pa (Figure S2, Supporting Information). This linear trend allows us to target a specific hydrogel stiffness simply by changing the ratio of hydrazone and oxime cross‐links. A similarly linear trend is also observed in the case of using only a single cross‐linker and varying the total concentration (Figure S3, Supporting Information). Shifting the window of mechanical accessible properties is further possible through modification of traditional parameters, such as the molecular weight, degree of oxidation, or polymer content, to give a few examples, but is beyond the scope of this study.

Next, we investigated the strain response of these hydrogels. A comparison of the normalized shear storage modulus reveals that all formulations display strain‐stiffening behavior reminiscent of native ECM (Figure 3B). Strain‐stiffening is ubiquitous in natural systems, but rarely present in synthetic hydrogels and is non‐trivial to engineer into a material. From the softer HyOx(1.0:0.0) to the stiffer HyOx(0.0:1.0) we observed a decrease in the strain‐stiffening onset, with strain‐independent behavior ending at 40% strain and 10% strain, respectively. Across the series, we observe a general trend of increasing strain‐stiffening magnitude (fold‐increase in G’ before rupture) with increasing hydrazone content until HyOx(1.0:0.0), which decreases compared to HyOx(0.8:0.2). This trend shows that the addition of multiple cross‐linkers is able to increase a hydrogel's failure strain, and that there exists an optimal ratio to maximize this effect. The exact mechanism of this strain‐stiffening is unknown, yet is under further investigation.

Looking next at the stress relaxation behavior of our formulations (Figure 3C), HyOx(1.0:0.0) showed the fastest stress relaxation, with a t_½ of 300 s. Introducing 0.2 mol equiv oxime into the network increased the t_½ to 2600 s for HyOx(0.8:0.2), an ≈8.5‐fold increase. HyOx(0.6:0.4) then increased to 16 000 s, a ≈sixfold increase. However, as more oxime is introduced to the network, these slower relaxing cross‐links quickly dominate the hydrogel behavior, with very little difference seen between HyOx(0.4:0.6) (t_½ = 25 500 s), HyOx(0.2:0.8), and HyOx(0.0:1.0) (t_½ = 33 000 s).

Stress relaxation curves were fitted using a stretched exponential model, which confirmed that very little difference in characteristic stress relaxation times was observed beyond HyOx(0.4:0.6) (Figure S4 and Equations S1 and S2, Supporting Information). These findings are consistent with the frequency sweeps of pure hydrazone and pure oxime hydrogels, where hydrazone showed a crossover point (τ ∝ 1/ω; larger crossover frequency corresponds to a faster relaxation time) at 2×10⁻⁴ rad s⁻¹ while the crossover point for oxime was lower than the experimentally accessible range (Figure S5, Supporting Information). Of note, the stress relaxation profile of HyOx(1.0:0.0) is on the same order of magnitude as native soft tissues such as the brain and the liver.^{[ 4 ]}

We also studied maintaing a constant ratio of oxime to hydrazone and varying the total cross‐linker concentration. This series exhibited an interesting inflection point in the stress relaxation data; increasing hydrazone concentration led to faster short‐term relaxation, while increasing oxime concentration led to slower long‐term relaxation (Figure S7, Supporting Information). These results imply that we could use the different degrees of dynamicity present in our cross‐linkers to selectively influence specific regions of the stress relaxation behavior, though a more thorough investigation of this observation is beyond the scope of the current work. A control series measuring the stress relaxation of hydrogels with 0.1–0.4 mol equiv oxime was also measured and confirmed that even low concentrations of oxime yielded similarly long relaxation times, with slightly slower relaxation as oxime concentration increase (Figure S8, Supporting Information).

In contrast to the rapid dynamics of HyOx(1.0:0.0), the slower dynamics of HyOx(0.0:1.0) are able to increase hydrogel stability; this is reflected in the swelling behavior of these hydrogels (Figure 3D). In a large excess of solvent (PBS, pH 7.4, RT), the pure hydrazone hydrogel swells rapidly, lacking less dynamic cross‐links to limit network expansion. Consequently, HyOx(1.0:0.0) becomes challenging to handle within 24 h. However, even a small proportion of oxime cross‐links limit the rapid network expansion, and we observe a consistent decrease in the swelling ratio from 2.05 for the softer HyOx(0.8:0.2) to 1.02 for the stiffer HyOx(0.0:1.0). The swelling behavior of HyOx(0.8:0.2) is a clear example of how mixing in a small amount of a second cross‐link can have an important effect on one behavior, swelling in this case, while having a relatively small impact on another, such as stiffness. All formulations except HyOx(1.0:0.0) were still stable after 7 days in a large excess of PBS (Figure S9, Supporting Information).

2.4. Injectability and Self‐Healing Behavior of Mixed Dynamic Hydrogels

Next, we wanted to investigate how the mixing of dynamic cross‐linkers would affect the injectability and self‐healing behavior of our formulations. We began by manually injecting all formulations through a 1.27 cm (½ inch), 0.33 mm internal diameter (I.D.), cylindrical nozzle onto a glass slide. Pure hydrazone and pure oxime hydrogels are compared in Figure 4A, which shows that the hydrazone hydrogel allowed clean fibers to be extruded while the oxime hydrogel was challenging to extrude manually in an even manner. The full series can be seen in Figure S10, Supporting Information, and shows that HyOx(1.0:0.0) and HyOx(0.8:0.2) produced smooth fibers while all formulations with at least 0.4 mol equiv oxime produced irregular and broken fibers.

Figure 4.

All mixed cross‐linker hydrogels are injectable and flow similarly under shear but only hydrogels with high hydrazone content self‐heal and form fibers. A) Hydrazone and oxime hydrogels being injected by hand through a 1.27 cm (½ inch) 23 gauge needle (I.D. = 0.33 mm). The hydrazone fiber has greater uniformity than the oxime chunks that are extruded. The complete series is shown in Figure S10, Supporting Information. B) Flow curves of the mixed cross‐linker hydrogels series. The initial viscosity follows the initial moduli of these hydrogels, but all formulations have a similar viscosity at shear rates above 1.0 s⁻¹. C) A self‐healing experiment revealed that the oxime hydrogels are unable to self‐heal after 24 h in contrast to the hydrazone hydrogels that can. This trend appears gradually, with HyOx(0.8:0.2) showing good self‐healing and HyOx(0.6:0.4) showing only slight self‐healing. Interestingly, the hydrazone hydrogel is capable of self‐healing to the oxime hydrogel, suggesting that the hydrazone diffusion across the interface is the determining factor.

We moved from manual extrusion to printing hydrogels with a BioX printer using pressure‐based extrusion. The required initial extrusion pressure increased as the formulations increased in stiffness, with values above 200 kPa. (Figure S11, Supporting Information). We noticed that the pressure had to be reduced after flow had started in order to achieve a consistent flow. Furthermore, the required initial extrusion pressure for subsequent prints of HyOx(1.0:0.0) from the same syringe was inconsistent and could vary by up to 100 kPa (Figure S11, Supporting Information). The observed differences in required pressure for initial extrusion are consistent with a Bingham pseudoplastic material, possessing a yield stress that must be surpassed to induce flow. The variable pressure required to maintain flow is also indicative of a shear‐thinning response (viscosity decreases as the imposed shear rate increases) once the yield stress is overcome (Figure S12A, Supporting Information). These complex behaviors made consistent printing of pristine hydrogels challenging.

To better understand the variable pressures required to repeatedly extrude the series from a syringe, we measured the flow curves of each formulation on the rheometer (Figure 4B). Interestingly, while the initial viscosity mimics the trend in the initial shear storage modulus, the flow behavior after yielding was similar across all formulations. The molecular hydrogel formulation in all cases is nearly identical: 2 wt.% polymer content with 1 mol equiv small molecule cross‐linker. The only difference is the molecular equilibria and kinetics of the small molecule cross‐linkers. Once the dynamic cross‐links are ruptured, the resulting non‐cross‐linked material is essentially identical across all formulations. We aggregated the flow data post‐rupture and fitted them to a power‐law model (Equation S3, Supporting Information), obtaining a consistency index (K) of 149 ± 6 Pa sⁿ and a flow index (n) of 0.18 ± 0.07 (Figure S12B,C, Supporting Information). The flow index is also known as the shear‐thinning parameter and represents the extent to which an increase in shear rate decreases the viscosity (lower values of n correspond to a higher degree of shear‐thinning) while the consistency index is the proportionality constant between the exponentiated shear rate and the apparent viscosity for a given value of n. Our obtained flow index is low and indicated that our hydrogels have a high degree of shear‐thinning post‐rupture. Unsurprisingly, the obtained consistency index is much lower than those found by fitting the pre‐ruptured regime (Figure S12B, Supporting Information).

We then performed the analysis described by Hernandez et al. to quantify the injectability of our hydrogels in the pristine versus post‐ruptured states.^{[ 43 ]} Following the provided method, all pristine formulations with the exception of HyOx(1.0:0.0) are calculated to be non‐injectable using a 0.41 mm I.D. 1.27 cm cylindrical tip, which is in line with our experimental observations of the required printing pressure for pristine formulations (Figures S12 and S13 and Tables S1 and S2, Supporting Information). In contrast, the aggregate ruptured state is found to be well within the injectability regime, a result that is supported by our lower printing pressures for pre‐ruptured hydrogels. Notably, these results demonstrate that despite large differences in the static mechanical properties of our mixed hydrogels (Figure 3), they all possess similar flow behavior once ruptured. The dynamic nature of our hydrazone cross‐links then enables a corresponding proportion of the ruptured network to self‐heal, recovering its static mechanical properties. This approach allows us to decouple final mechanical properties from the hydrogel processability, which is a major advantage for application as a bioinks.

To investigate the self‐healing behavior of our formulations, we prepared pairs of hydrogel discs, one transparent, and one containing a small amount of a hydrazone‐functionalized dye. These discs were cut in half, matched in pairs so that only one half contained the dye, returned to their molds for 24 h, and finally pulled on with tweezers to assess self‐healing (Figure 4C). We observed (Figure 4D) that hydrogels with high hydrazone content self‐heal over 24 h while those with a high proportion of oxime do not. HyOx(1.0:0.0) and HyOx(0.8:0.2) self‐healed completely while HyOx(0.6:0.4) and HyOx(0.4:0.6) showed a moderate degree of self‐healing in this timeframe. Interestingly, the blue dye we used for visualization is hydrazone functionalized and the rapid propagation of the blue color from one‐half to the other shows how easily hydrazone functions can diffuse across the interface. If the self‐healing behavior on these materials is driven by the kinetics of bond exchanges, then it follows that a hydrazone hydrogel half may be able to self‐heal to any other formulation due to its ability to freely diffuse across the interface. To test this hypothesis, we performed the self‐healing experiment using halves of different compositions (Figure 4E). HyOx(1.0:0.0) was able to completely self‐heal to HyOx(0.0:1.0), supporting our hypothesis. Similarly, HyOx(0.8:0.2) and HyOx(0.2:0.8) self‐healed, while HyOx(0.6:0.4) and HyOx(0.4:0.6) did not.

We have summarized the self‐healing behavior of our formulations in Table S3, Supporting Information. Following the same principle of kinetic control over cross‐link exchange, Lou et al. used a biocompatible catalyst to temporarily increase the exchange dynamic to allow injection of cross‐linked hydrogels.^{[ 44 ]} Interestingly, the ability to create both “adherent” and “non‐adherent” formulations that will or will not self‐healing to one another is an interesting feature of our system that could find use in the design of mixed biomaterial constructs, though a more robust exploration of these possibilities is beyond the scope of the current work.

2.5. Biomaterial Ink Printing and Recyclability of HyOx(0.8:0.2)

Following the quantification of the extrusion characteristics of our hydrogels, we realized that the ruptured state of our hydrogels possessed desirable flow characteristics for extrusion bioprinting. Given the identical flow behavior we observed of the ruptured networks, we hypothesized that all formulations should have similar printability in a post‐ruptured state. To investigate this, we elected to use HyOx(1.0:0.0), HyOx(0.8:0.2), and HyOx(0.6:0.4), as they have the most biomimetic mechanical properties, and all possess some degree of self‐healing. We began by printing 2‐layers of a square grid scaffold to assess the printing fidelity and ensure that fibers were cohesive (Figure 5A,B). In all cases, the printed fiber could be picked up using tweezers. We were able to print this lattice structure up to 16‐layers while retaining the porosity of the scaffold (Figure 5C). Measuring the average fiber diameter, as well as the overall dimension of the printed scaffolds, we see that the fibers are thicker than the internal diameter of the needle by ≈50% (≈600 µm vs 410 µm), but that they are consistent across all three tested formulations (Figure 5D). The large standard deviation is due to the aggregation of filaments at the edges resulting in thicker fibers towards the extremities of the scaffold. In contrast, the overall scaffold dimensions are in very good agreement with the 15 mm x 15 mm model (Figure 5E).

Figure 5.

Pre‐ruptured hydrogels are printable and show similar fiber qualities independent of the formulation. Using three consecutive formulations (HyOx(1.0:0.0), HyOx(0.8:0.2), and HyOx(O.6:0.4)) as a way to test the variation in printability of mixed ruptured networks, A) we see that the fiber and grid quality of the printed 2‐layer construct is similar across these formulations, in agreement with our flow behavior of ruptured networks. B) Render of the 15 mm x 15 mm 2‐layer gcode file. C) We are able to print 16 layers of the grid pattern and retain the porosity of the printed scaffold. D,E) The measured dimensions of both the fibers and scaffolds. The fibers are thicker than the 1.27 cm (½ inch), 0.41 mm I.D. cylindrical nozzle used for printing, while the length and width of the scaffold remain very close to the desired 15 mm across all formulations. The horizontal lines indicate the nozzles internal diameter and expected dimensions respectively. Scaffolds are surprisingly robust and can easily be handled with a spatula (Movie S1, Supporting Information). F) Illustrates the flexibility of these materials; we are able to print a 16‐layer scaffold, then mold it into an octopus able to support its own weight, before printing a new 16‐layer scaffold from the same material. Scale bars in panels (A) and (C) are 5 mm. The error reported in panel (D) is the mean ± SD for n = 22–32 fiber measurements distributed across the entire construct, while the error reported in panel (E) is the mean ± SD for n = 5 (for each length and width).

The printed scaffolds are surprisingly robust, could be comfortably manipulated with a spatula, and even cut in half with a razor blade, showing the preserved internal structure (Movie S1 and Figure S14, Supporting Information). Long‐term storage and stability over time is often cited as a concern with dynamic materials, so we also stored some HyOx(0.8:0.2) in the fridge for extended periods. A 2‐layer print was stable after a week even when soaked in PBS (500 µL), while a 16‐layer HyOx(0.8:0.2) scaffold showed only slight sagging (creep) after 4 months of storage, highlighting the robust nature of these materials post‐printing (Figure S15, Supporting Information).

Considering the self‐healing nature of these hydrogels, we wanted to explore the ability to re‐use or recycle these materials using different fabrication processes. To test this clearly, we first printed a 16‐layer lattice scaffold, and then compressed it in an octopus cookie mold for 2 h under a 300 g weight. Once removed from the mold, the octopus was able to sit upright under its own weight with ease. Finally, we centrifuged the octopus in a syringe before printing a new 16‐layer scaffold from the same material (Figure 5F). From our experience, the hydrogel was surprisingly malleable when handled, but required enough force to deform that we would describe it as robust once printed. Of note, while our self‐healing test was performed after 24 h, in this experiment, a mere 2 h under mild pressure was sufficient for HyOx(0.8:0.2) to self‐heal.^{[ 45 ]} The entire cycle from bulk hydrogel to the second 16‐layer scaffold was performed in less than 4 h during which time the material was presented in 3 different robust solid forms with the only requirement to transition from one processing method to the next being the application of force. The ease with which HyOx(0.8:0.2) was recycled demonstrates an unexpected but promising aspect of our system that will enable its application with more diverse fabrication techniques such as injection molding, which is particularly relevant for filling moderate to large defects in damaged tissues.

2.6. Human Dermal Fibroblast Spreading and Morphology in 2D as a Function of Cross‐Linker Ratio

Armed with a highly tunable and processable series of materials, we wanted to investigate how HDFs would respond to the gradient of mechanical properties present in our hydrogels. Fibroblasts have been shown to be sensitive to both viscous and elastic components of hydrogel mechanics, as well as stress relaxation dynamics.^{[ 3} , ⁴⁶ , ^{47 ]} Here, we wondered whether our materials would allow us to control fibroblast morphology and/or spreading, and in particular, whether any observed changes in fibroblast response would be sudden or gradual.

HDFs were cultured for 72 h on top of our series of mixed hydrogels, with images taken after 24 and 72 h to quantify the cell area and aspect ratio. All formulations with at least 0.4 mol equiv showed the highest aspect ratios and similarly large cell areas. In comparison, HyOx(1.0:0.0) showed a rounded fibroblast morphology after 24 h and had fully dissolved by 72 h, which is unsurprising given its swelling behavior (Figure 3D and Figure 6A–C). Immediately prior to imaging, the wells were gently agitated to ensure that rounded cells were attached to the hydrogel and unable to be dislodged. HyOx(0.8:0.2) showed a round morphology and small cell area at both 24 and 72 h, while HyOx(0.6:0.4) exhibited a sharp increase in both cell area and aspect ratio. Initially, this appeared to be a very sudden transition, so we added an intermediate condition, HyOx(0.6:0.3), to try and capture intermediate behavior between these two regimes. This approach was ultimately successful, as we see a clear trend of increasing cell area and aspect ratio from HyOx(0.8:0.2) to HyOx(0.6:0.3) and then HyOx(0.6:0.4). Continuing to increase the proportion of oxime in the hydrogel formulation yielded no significant change in either cell area of aspect ratio. These observations follow the measured trend for stress relaxation, with large increases from HyOx(1.0:0.0) to HyOx(0.6:0.4), but negligible change as the proportion of oxime was further increased.

Figure 6.

HDFs show increased spreading and cell area on mixed hydrogels with slower stress relaxation but remain rounded on fast relaxing hydrogels. A) Images of HDFs seeded on top of our series of dynamic hydrogels after 24 and 72 h. HyOx(1.0:0.0) was dissolved by 72 h. The green color represents actin filaments (phalloidin) while the blue color represents the nuclei (DAPI). We see that as the stiffness (and oxime content) of the hydrogel increases, the cell morphology transitions from rounded to spread, with little increase in cell spreading seen past 0.4 mol equiv oxime. B) Quantification of the cell area and C) aspect ratio at day 3 using ImageJ confirms this observation. Sample sizes are N = 2 with n = 57–277. A D'Agostino normality was performed to assess the distribution of the data, followed by Kruskal–Wallis test and Dunn's post‐hoc multiple comparisons test (****P ≤ 0.0001); statistically significant comparisons between high oxime formulations have been omitted for visual clarity. Data points from two separate experiments are shown as grey or black circles while the plotted bars represent the mean ± SD. D) An image of the interface of a self‐healed HyOx(0.8:0.2) and HyOx(0.6:0.4) zonal hydrogel shows that the control over fibroblast morphology is retained after reassembling different hydrogel formulations. Complete statistical comparison of aspect ratio, area, and their differences can be found in Tables S4–S7, Supporting Information. Scale bars in (A) and (D) = 100 µm.

These results support the idea that the stress relaxation characteristic of our dynamic hydrogel is primarily responsible for the observed HDF phenotypes (as opposed to hydrogel stiffness). While fibroblasts are known to be sensitive to substrate stiffness, our entire hydrogel series varies in shear storage modulus by only 2.56 kPa which is very small compared to studies on the effect of substrate stiffness.^{[ 6} , ^{47 ]} Furthermore, previous work has shown that fibroblasts spreading is predominantly influenced by ligand density and stress relaxation for very soft (< 2 kPa) hydrogels.^{[ 3} , ^{48 ]}

We also analyzed the difference in observed cell area and aspect ratio between day 1 and day 3 (Figure S16, Supporting Information) to follow the development of these phenotypes. We see that there is a maxima present for both cell areas an aspect ratio centered around the intermediate formulations; The difference in cell aspect ratio is maximal for HyOx(0.6:0.4) whereas the difference for cell area is maximal across HyOx(0.6:0.4) to HyOx(0.2:0.8). While we did not set out to investigate the kinetics of morphological development, it is worthwhile to note that our hydrogels are also able to influence the rate at which cell area and aspect ratio change as a function of mechanical properties.

Our results demonstrate that HDFs indeed present a distinct, gradual shift in morphology as the dynamicity of their local mechanical environment changes. Considering these results in the context of our printing study, we now have two mechanically distinct hydrogels, HyOx(0.8:0.2) and HyOx(0.6:0.4), that are able to induce significant changes in cell area and aspect ratio while retaining similar processability and self‐healing capacity. This opens the possibility of creating zonal constructs where different regions of a multi‐material hydrogel could induce different fibroblast morphologies. To demonstrate the feasibility of these zonal constructs, we prepared HyOx(0.8:0.2) and HyOx(0.6:0.4) hydrogel discs, cut them in half, self‐healed them together again, and cultured HDFs on them for 72 h. We can clearly see that the fibroblasts on the HyOx(0.8:0.2) half of the construct present the expected rounded morphology, while fibroblasts on the HyOx(0.6:0.4) half of the construct exhibit the spread morphology they previously showed on complete HyOx(0.6:0.4) hydrogels (Figure 6D). We also controlled for the effect of binding ligand concentration by testing RGD concentrations of 200, 500, and 1000 µm and found no noticeable effect of ligand density on morphology. (Figure S17, Supporting Information).

2.7. Human Dermal Fibroblast Viability and Morphology after 3D Encapsulation and Extrusion

To validate the applicability of our materials as injectable cell carriers and bioinks, we aimed to investigate whether encapsulated cells could be extruded while maintaining high viability and retaining similar morphological control in 3D. With this in mind, we loaded HDFs into each formulation, extruded them through a 0.58 mm I.D. 1.27 cm cylindrical nozzle, and cultured them for 24 h before staining and imaging (Figure 7A). Despite imaging the entire well to ensure even cell distribution and enable robust quantification of cell populations (Figure 7B and Figure S18, Supporting Information), we were unable to locate dead cells encapsulated within the hydrogels; some wells had zero dead cells in the imaged planes while others had less than 10 (>99% viability). We then chose to also image to the bottom of the well plate to obtain any dead cells which may have sedimented or been shed during extrusion (Figure S19, Supporting Information). This approach provided further insight into the global cell viability; we found >85% viable cells outside of the gel across all formulations (Figure S20, Supporting Information). This suggests that the shear‐thinning behavior of the hydrogels is able to protect the fibroblasts from shear stress during the extrusion process and that these materials can be used as bioinks.

Figure 7.

HDFs show similar morphology in 3D to that seen in 2D. A) Schematic of the 3D cell culture experiment. Briefly, cell‐laden hydrogels were extruded through a 1.27 cm, 0.58 mm I.D. printing needle and cultured for 24 h before staining and imaging. B) Representative image of an entire well highlighting the homogenous distribution of cells throughout the hydrogel. C) Zoomed in regions of the HyOx(0.8:0.2) and HyOx(0.6:0.4) hydrogels, showing a rounded or spread morphology respectively; the same morphology is observed in 2D culture as shown in Figure 6D. The full series of whole well and higher magnification insets is shown in Figure S17, Supporting Information. Cells seemed to aggregate in HyOx(0.8:0.2), whereas single cells were predominantly found in HyOx(0.6:0.4), shown in Figure S21, Supporting Information. Green is calcein‐AM staining while blue is Hoechst (nuclei). Scale bars in panels (B) and (C) are 1000 and 500 µm respectively.

Our 3D imaging of the encapsulated HDFs also revealed that the difference in fibroblast morphology observed in 2D culture is present in the 3D culture (Figure 7C). HyOx(0.8:0.2) showed mainly rounded fibroblasts, while HyOx(0.6:0.4) induced noticeably more elongated and spread cell populations, mimicking the 2D results seen in Figure 6D. At first glance, cells in HyOx(0.8:0.2) appear to be slightly larger than those seen in HyOx(0.6:0.4), but this is due to the formation of small aggregates of 3–4 cells, while cells in HyOx(0.6:0.4) were found mainly as single cells or in groups of two (Figure S21, Supporting Information). The morphological control that we are able to exert over HDFs with our mixed dynamic hydrogels extends to 3D cell culture and further highlights their strong potential as cell‐instructive matrices.

3. Conclusion

We have found that simple mixing of dynamic covalent cross‐linkers allowed fine control over the stiffness and stress relaxation behavior of the resulting hydrogels. Notably, we were able to vary the shear storage modulus on the order of 10²–10³ Pa and the characteristic relaxation times from 10²–10⁵ s by varying only the relative proportion of hydrazone and oxime cross‐links while maintaining a constant total cross‐linker concentration and polymer content. Furthermore, these formulations are shown to possess shear‐thinning and self‐healing properties, allowing them to be injected, molded, printed, and recycled. HDFs showed dramatic differences in cell area and aspect ratio on the hydrogels with different cross‐linker ratios. Our data using a dynamic hydrogel platform based on oxidized alginate positions these materials as promising candidates to bridge the gap between highly processable and cell‐instructive biomaterial inks and bioinks independent of traditional design parameters. Possessing highly tunable mechanical properties, good printability, biomimetic stiffness regimes, and strain‐stiffening behavior, as well as high cell viability post‐extrusion and cytocompatability, these materials will be of interest to mechanobiologists and labs looking to expand into bioprinting for TERM.

4. Experimental Section

Materials

Adipic dihydrazide (Sigma, ≥98%), O,O′‐1,3‐propanediylbishydroxylamine dihydrochloride (Sigma, 98%), activated charcoal (Sigma, Norit), sodium (meta)periodate (Sigma, ≥98%, NaIO₄), ethylene glycol (Sigma, ≥99.5%), Dulbecco's Phosphate Buffered Saline (Sigma, without MgCl₂ or CaCl₂), and dialysis membranes (VWR, Spectra/por 6, diameter = 28.6 mm, MWCO = 3.5 kDa) were purchased and used as received. Sodium alginate powder (FMC Manugel GMB, Lot No. G9402001, ≈380 kDa) was purified as described below before subsequent oxidation. Filter papers (Whatman) with pore sizes of 11, 1.2, 0.45, and 0.2 µm were purchased from VWR (11 µm) and Sigma (1.2, 0.45, and 0.2 µm). For the phosphate‐buffered D₂O, deuterium oxide (Sigma, 99.9 atom % D, D₂O), potassium dideuterium phosphate (Sigma, 98 atom % D, KD₂PO₄), potassium phosphate tribasic (Sigma, ≥98%, K₃PO₄), and 3‐(Trimethylsilyl)‐1‐propanesulfonic acid‐d₆ sodium salt (Sigma, 98 atom % D, DSS‐d₆) were purchased and used without further treatment. CF 647 hydrazide (sigma) was diluted in 500 µL PBS after arrival, yielded a stock concentration of 0.67 mm, and was used to color hydrogels for ease of visualization.

For cell culture, Dulbecco's modified eagle medium (DMEM, Gibco) containing 4.5 g L⁻¹ glucose, GlutaMax, and pyruvate, was supplemented with 10% (v/v) Fetal Bovine Serum (FBS, Sigma, Batch BCBX5318), and 100 U mL⁻¹ Penicillin/Streptomycin (P/S, 10 000 U mL⁻¹, 100X, Gibco). Trypsin‐EDTA (0.05%, Gibco, containing phenol red) was used as received, while formaldehyde (Sigma, 37 wt.% in H₂O & containing 10–15% methanol as a stabilizer) was diluted to 4 wt.% in PBS prior to use as a fixation agent. For cellular staining, Hoechst 33 342 trihydrochloride, trihydrate (Invitrogen), ethidium homodimer‐1 (Invitrogen), calcein AM (Invitrogen), Alexa Fluor 488 Phalloidin (Invitrogen), and 4′,6‐diamidino‐2‐phenylindole dihydrochloride (DAPI, Sigma, ≥95.0% (HPLC)) were diluted as necessary in PBS to obtain the required working solutions (specified in the relevant methods sections). Cellular adhesion was facilitated using and RGD binding peptide ((AOAC)‐GGGRGDS, Chinapeptides, 98.05%, molecular weight = 677.65 g mol⁻¹). Adult HDFs were purchased from ScienCell (Catalog #2320, San Diego, USA) and cultured according to the manufacturer's recommendations.

Purification of Sodium Alginate

Sodium alginate (15 g) was dissolved overnight at 4 °C in MilliQ ultrapure water (1.5 L) with mechanical mixing giving a concentration of 1% (w/w). The following morning, 7.5 g activated charcoal (0.5% (w/w)) was added and the solution was stirred for a further 24 h at 4 °C. The charcoal loaded solution was then passed multiple times (2–4) through an 11 µm filter under vacuum until no more charcoal was captured by the membrane. This process was repeated using 1.2 µm, 0.45 µm, and finally 0.2 µm filters. Filtration took 2–3 full days and solutions were stored at 4 °C overnight. The purified alginate solution was then frozen and lyophilized, yielding a white fibrous powder. Typical yields were 50–70% following volume loss during the filtration step due to the large number of filters required.

Oxidation of Sodium Alginate

In a 100 mL beaker, purified alginate (1.00 g, 5.05 mmol of uronic acid units) was dissolved with stirring in 95 mL distilled water (dH₂O) over 4–6 h to yield a homogenous, slightly viscous solution. Meanwhile, a solution of NaIO₄ (121 mg, 0.565 mmol, 0.112 equiv w.r.t. moles of uronate units) was prepared by dissolution in 4.0 mL dH₂O. The beaker containing the oxidized alginate solution was wrapped in aluminum foil to protect it from light and the NaIO₄ solution was added in one portion—giving a final alginate concentration of 1% (w/w)—and the reaction was left to proceed at RT (Scheme S1, Supporting Information). The oxidation of each uronate unit generates two aldehyde moieties.^{[ 49 ]} Here, the equiv of NaIO₄ gives a theoretic degree of oxidation (mol % oxidized uronic acid units) of 11.5%. After 17 h, ethylene glycol (31.66 µL, 0.568 mmol, 1.0 equiv w.r.t. NaIO₄) was added and left for ≈1 h to quench the oxidation. The quenched reaction mixture was divided into two portions of 50 mL, which were then dialyzed against 5 L NaCl solution of decreasing concentration. The dialysis solution was changed each morning and evening a total of 5 times over 3 days (100; 50; 25; 12.5; and 0 mm). The initial 50 mL was observed to swell to ≈3 times its initial volume over this period. Finally, the two dialyzes were mixed before being frozen and lyophilized to yield a fibrous white solid. Typical yields were 73–74% (730–740 mg).

NMR of Oxidized Alginate

Phosphate buffered D₂O (pH 7.4, 45–50 mm PO₄ ³⁻) was prepared using D₂O, K₃PO₄, and KD₂PO₄. DSS‐d₆ was used as an internal standard. Oxidized alginate was then dissolved in phosphate‐buffered D₂O at a concentration of 1 or 2% (w/v) and 700 µL of this solution was transferred to a clean disposable glass NMR tube. NMR spectra were recorded on a Bruker Avance III HD 700‐MHz spectrometer equipped with a cryogenically cooled three‐channel TCI probe. Spectra of two batches of oxidized alginate can be found in Figure S22, Supporting Information.

GPC of Oxidized Alginate

Oxidized alginate samples were dissolved in aqueous 0.1 mol L⁻¹ sodium nitrate (NaNO₃) running buffer at a concentration of 1 mg mL⁻¹, passed through a 0.2 µm filter, and loaded into a Shimadzu Prominence‐i LC‐2030C 3D liquid chromatograph equipped with Shodex SB‐803 HQ & SB‐804 HQ columns (Showa Denko America). Columns were calibrated using PEG standards from 0.575–700 kDa (PEG calibration kit, Agilent Technologies). Analysis of the M _n, M _w, and dispersity was performed using the accompanying LabSolutions software (Shimadzu). Plots of retention time and molecular weight distribution can be found in Figure S23, Supporting Information.

Degree of Oxidation

To confirm that at least a 10% degree of oxidation was achieved, hydrogels were formed using either only hydrazone or only oxime from 0.1–1.0 mol equiv with respect to the expected concentration of aldehyde moieties to test for a linear relationship between cross‐linker concentration and the resulting shear storage modulus. (Figure S3, Supporting Information).

Stock Solutions and Equivalents

Stock solutions of oxidized alginate at 4% (w/v), adipic dihydrazide (0.101 mol L⁻¹), and O,O′‐1,3‐propanediylbishydroxylamine dihydrochloride (0.101 mol L⁻¹) were prepared in advance in PBS (without MgCl₂ CaCl₂), and stored for up to 1 week in the fridge (example preparation values can be found in Table S8, Supporting Information). The number of equivalents was calculated by functional group, assuming 11.5% of the uronic acid units of the alginate chain had been oxidized, each generating 2 aldehydes, and accounting for the bifunctional nature of the cross‐linker.^{[ 49 ]} Thus, the addition of 1 equiv total cross‐linker indicated that there should be an equimolar amount of hydrazide/hydroxylamine and aldehyde.

Pre‐Formed Hydrogel Discs

Hydrogels were formed in silicon molds 12 mm in diameter with a height of 1350 µm and covered (top and bottom) with a 15 mm glass coverslip. Briefly, the oxidized alginate solution was diluted where necessary with PBS while the cross‐linker solutions were mixed in the appropriate proportions. The cross‐linking solution was then added quickly to the oxidized alginate solution and the resulting solution vortexed for 10 s to ensure a homogenous mixture. Finally, 150 µL was pipetted into the mold and sealed with the second 15 mm glass coverslip. 2 µL CF 647 hydrazide (Stock concentration 0.67 mm) was added just prior to sealing to color the hydrogel blue in the case of hydrogels for self‐healing tests. The molds were then left in the fridge overnight to allow hydrogel formation. Hydrogels were prepared with a final alginate content of 2% (w/w) and varying equivalents of one or both cross‐linkers. Example volumes for hydrogel formation can be found in Table S9, Supporting Information.

In Situ Hydrogel Formation for Rheology

Hydrogels formed in situ were formed using the same stock solutions. Briefly, the oxidized alginate solution was diluted where necessary with PBS while the cross‐linker solutions were mixed in the appropriate proportions. The cross‐linking solution was then added quickly to the oxidized alginate solution and the resulting solution vortexed for 10 s to ensure a homogenous mixture. An 84 µL aliquot was then immediately loaded into the rheometer for in situ evaluation of hydrogel formation. See “Gelation kinetics” for further details.

Rheological Characterization

Rheological measurements were performed using a DHR‐2 from TA instruments equipped with a Peltier heating element and solvent trap. For the preformed hydrogels, an 8 mm parallel plate geometry was used, compared to the 20 mm cone‐plate with an angle of 2.002° used for in situ gelation and all subsequent measurements. All measurements were performed at 20 °C.

Gelation Kinetics

Using the stock solutions described for pre‐formed hydrogels, gelling solutions were prepared in the same manner as above, then pipetted (84 µL) onto the rheometer and quickly loaded under the 20 mm cone. An oscillatory time sweep was then measured at 1% strain and 10 rad s⁻¹ for 2.5 h to following the cross‐linking process to completion.

Stress Relaxation Measurements

Immediately following the 2.5 h gelation, a stress relaxation measurement was performed on the same sample. The strain was set at 20% and maintained for 15.5 h. The solvent trap prevented the evaporation of the sample during this long measurement.

Frequency Sweep Measurements

Oscillatory frequency sweeps were also performed following in situ gelation—in place of stress relaxation—from 100 to 1.6×10⁻⁴ rad s⁻¹ and at 1% strain with 5 points per decade. This measurement took 30.7 h and the solvent trap was used to prevent the evaporation of the sample during this time.

Measurements of Shear Moduli

In the case of in situ gelation, the shear storage modulus was taken to be the plateau value after 2.5 h. For the pre‐formed gels, samples were loaded into the 8 mm parallel plate and then trimmed using a razor blade. The gap size was set to 1000 µm and oscillatory time sweeps at 2% strain and 10 rad s⁻¹ were performed for 10 min—long enough to allow sample equilibration after loading and short enough to avoid drying. Average values were taken to be the shear storage modulus.

Strain Sweep Measurements

Oscillatory strain sweeps were also performed following stress relaxation measurements from 1–1000% strain at 10 rad s⁻¹.

Flow Curves

Flow curves were also performed following the 2.5 h in situ gelation increasing the shear rate from 0.0001–1000 s⁻¹. Steady‐state sensing was enabled in the TRIOS software, with an equilibration time of 120 s, a sample period of 30 s, and 5% tolerance within 3 points. Scaled time average was also enabled with a scaling factor of 1 to scale the equilibrium time and averaging time inversely to the applied shear rate.

Swelling of Hydrazone and Oxime Cross‐Linked Hydrogels

Hydrogels of the desired composition were first formed in duplicate according to the “Pre‐formed hydrogel discs” protocol described earlier. The hydrogels were left to cross‐link overnight in their molds. Each hydrogel was excised from its mold after running a spatula around the edge—to prevent it sticking to the mold—weighed, and transferred to a small, labeled petri dish. Then each petri dish was filled with 10 mL PBS (pH = 7.4, without Mg²⁺ or Ca²⁺) and the hydrogels were gently moved around in the PBS with a spatula to ensure that none were stuck to the bottom and protecting one surface from the surrounding liquid. Hydrogels were maintained at RT. At t = 0.5, 1, 2, 4, 6, 12, & 24 h, each sample was carefully transferred to a plastic weighing boat and tilted at an angle using a spatula to drain excess water by gravity. This draining process was repeated a second time using a dry weighing boat before the swollen mass was measured using a final dry weighing boat.

The swelling ratio was calculated as:

$$S=\frac{m_{\mathrm{t}}}{m_{\mathrm{i}}}$$ (1)

Where m _t is the mass at time, t, and m _i is the initial mass.

Injection of Hydrogels

Preformed hydrogels were prepared as described in “in situ hydrogel formation”, in a 3 mL syringe (9.3 mm internal diameter). Once formed these hydrogels were manually injected through a 0.33 mm I.D. 1.27 cm cylindrical nozzle onto a glass slide to assess the injectability of each formulation and the injected fiber.

Self‐Healing of Preformed Hydrogels

Hydrogel discs were prepared as described in “Pre‐formed hydrogel discs”. Pairs of discs, with only one containing the blue hydrazone dye, were cut in half with a razor blade. The opposite colored halves (so one blue, one clear) were reassembled, returned to the 12 mm silicon mold, and recovered with a glass coverslip to prevent drying out. After 24 h, the reassembled gels were removed from the mold and the original halves were tugged in opposite directions with a tweezer to assess whether the disparate halves had healed together.

Printing of Lattice Structures

Lattice structures where evaluated using 2 mL of HyOx(1.0:0.0), HyOx(0.8:0.2) and HyOx(0.6:0.4). Each hydrogel formulation was prepared as described in “in situ hydrogel formation”, in a 3 mL syringe (9.3 mm internal diameter) and colored blue using 15 µL CF 647 hydrazone dye (Stock concentration 0.67 mm). Printing was tested using a CellINK BioX printer connected to a 7 bar external pressure source. The hydrogels were pre‐extruded through the print tip to create the ruptured network for extrusion printing. After extruding the 2 mL of hydrogel, all material was collected and centrifuged at 2000 rcf for 10 min to compress it and remove large air bubbles. Printing was done using a 1.27 cm (½ inch), 0.41 mm internal diameter cylindrical tip. The structure used to test printing was a 16‐layer square lattice with an initial trailing fiber (first two layers rendered in Figure S24, Supporting Information) to allow real‐time pressure adjustment at the beginning of the print and to allow the flow to stabilize. Printing was stopped after the desired number of layers was achieved each run. A layer‐height equal to the I.D. was used during printing. A summary of the printing pressures and head speeds can be found in Tables S1 and S2, Supporting Information.

Stability of Printed Structures

To assess the stability of printed lattice structures, a 2‐layer print of HyOx(0.8:0.2) was covered in 500 µL of PBS in a small petri dish, sealed with Parafilm, and left in the fridge for a week. To check for longer term creep of a printed structure, a 16‐layer print of HyOx(0.8:0.2) was stored in a petri dish, sealed with Parafilm, and left for 4 months before the inspection.

Measurement of Fiber and Scaffold Dimensions

Images of scaffolds were all taken with an element of known size in frame. This element was used to calibrate the pixel:mm conversion in ImageJ. A selection of 20 fiber diameters taken from both the edge and center of 4 quadrants were measured using this method and averaged to obtain final average fiber diameters. Similarly, a selection of 10 lengths and widths (5 each) were measured using the same method to obtain the average scaffold dimensions.

Recycling and Molding

To evaluate the recyclability of the ruptured network used for printing, an extruded 16‐layer scaffold was compressed for 2 h in a silicon mold of a stylized octopus under a 300 g weight (box). The molded hydrogel was then removed, and shifted to an upright position to illustrate the structural integrity following compression. It was then loosely broken with a spatula and centrifuged once again at 2000 rcf for 10 min. Finally, another 16‐layer scaffold was printed using the same material to complete the cycle.

2D Fibroblast Morphology and Spreading Study

In this study, HDFs were used from passage P8‐P12. HDFs were cultured in DMEM (4.5 g L⁻¹ glucose, L‐glutamine, pyruvate, 10% FBS v/v, 100 U mL⁻¹ P/S) in a humidified incubator (21% O₂, 5% CO₂, 37 °C). RGD binding peptide ((AOAC)‐GGGRGDS) was added to 4 wt.% stock oxidized alginate solution with a final concentration of 400 µm, vortexed for 60 s and allowed to react for a minimum of 10 min prior to hydrogel formation. Hydrogels (100 µL, 2 wt.%, 200 µm RGD) were cast in 96‐well glass‐bottomed plates, sealed to prevent evaporation, and allowed to cross‐link overnight at 4 °C. For the study of the effect of RGD concentration, the stock concentration was varied accordingly to obtain final concentrations of 200, 500, or 1000 µm RGD. Prior to cell seeding, plates were placed in the incubator for at least 30 min to equilibrate thermally. Cells were trypsinized using standard procedures and cell density was adjusted to seed 640 cells in 150 µL of medium in each well (2000 cell cm⁻²). At 24 and 72 h, samples were washed once with 100 µL PBS and then fixed with 4% paraformaldehyde in PBS for 15 min at RT and washed once more with PBS. Samples were permeabilized using 0.1% Triton‐X100 (v/v) in PBS for 10 min and then stained with 0.5 µm Alexa Fluor 488 Phalloidin and 0.2 µg mL⁻¹ DAPI for 1 h at RT. Samples were then washed twice in 100 µl PBS for 5 min before imaging.

Imaging and Quantification of Cell Area and Aspect Ratio

Fluorescent image acquisition for 2D cell spreading was performed on an automated inverted Nikon Ti‐E microscope, equipped with a Lumencor Spectra X light source, Photometrics Prime 95B sCMOS camera, and an MCL NANO Z500‐N TI z‐stage. The cell area and aspect ratio were calculated using ImageJ by adjusting the image threshold and manually selecting individual cells. Only single cells or small clusters where each individual cell could be identified and separated using the pencil tool were measured. A minimum of three pictures per sample were quantified from two separate experiments. The number of data points obtained ranged from 57 to 277. Data presentation and statistical analysis were performed using GraphPad Prism Software 9.1 (GraphPad Software, Inc. La Jolla, CA, USA). A D'Agostino normality was performed to assess the distribution of the data. Kruskal–Wallis with Dunn's post‐hoc analyses were performed to determine the statistical significance of data at day 3. The differences in cell area and cell aspect ratio between day 1 and day 3 were calculated by subtracting the means and calculating the propagation error of the standard error. One‐way ANOVA with Tukey's post‐hoc multiple comparisons test was performed to determine statistical significance. Differences were considered statistically significant at p ≤ 0.05. Complete statistical comparison of aspect ratio, area, and their differences can be found in Tables S4–S7, Supporting Information.

Fibroblast Spreading on a Zonal Construct

HyOx(0.8:0.2) and HyOx(0.6:0.4) hydrogel discs were prepared as described in “Pre‐formed hydrogel discs” with a few differences: A 8 mm x 7 mm silicon mold was used and the volume of each hydrogel was 200 µL. After 24 h in the fridge to fully cross‐link, each hydrogel was cut in half and differently formulated halves were reassembled inside the silicon mold and left for a further 24 h to self‐heal. The self‐healed zonal hydrogels were then used to culture fibroblasts in 2D as described in “2D fibroblast morphology and spreading study”.

3D Fibroblast Viability Post Extrusion Study

The hydrogels were prepared as described in “Injection of hydrogels”, with small modifications. The stock oxime solution (prepared using hydrochloride salt) was adjusted to pH 7.4 to prevent pH shock upon the initial addition of cells. Subsequent tests of the effect of this neutralization revealed that no detrimental effect was observed in the absence of neutralization (Figure S19, Supporting Information). Additionally, 1.08 mm oxime functionalized cell adhesion peptide ((AOAC)‐GGGRGDS) was added to the 4 wt.% Ox‐Alg stock solution, leading to a final RGD concentration in hydrogels for cell culture of 0.54 mm (0.053 equiv w.r.t. theoretically available aldehydes). Hydrogels were then mechanically broken by extrusion through a 1.27 cm (½ inch) 0.58 mm internal diameter cylindrical nozzle to improve printability and facilitate the addition and mixing of a cell suspension. HDFs were cultured in a humidified incubator (21% O₂, 5% CO₂, 37 °C) with supplemented DMEM (4.5 g L⁻¹ glucose, L‐glutamine, pyruvate, 10% FBS (v/v), 100 U mL⁻¹ P/S). Cells were trypsinized following standard procedures, resuspended, and seeded in the hydrogel at a density of 3.70×10⁵ cells per hydrogel (5.50×10⁵ cells mL⁻¹). Gentle mechanical agitation via rotation of a spatula within the syringe barrel was used to mix the cells into the hydrogel. The cell‐laden hydrogel was manually extruded into a 48‐well plate slowly with ≈200 µL hydrogel per well. Extruded fibroblasts were then cultured for 24 h before staining with ethidium homodimer, calcein AM, and Hoechst 33342 for fluorescent imaging. The working concentrations of ethidium homodimer (EthD‐1), calcein AM and Hoechst 33342 in PBS were 6 µm, 1 µm, and 1 µg mL⁻¹ respectively. Hydrogels were first washed once with 200 µL PBS before being left to incubate for 30 min with 300 µL staining solution. Finally, samples were washed twice more with 200 µL PBS, with 30 min incubations between washing steps before being imaged.

Imaging and Quantification of Fibroblast Viability Post Extrusion

Images of encapsulated cells post extrusion were acquired on an automated inverted Nikon Ti‐E microscope, equipped with a Lumencor Spectra light source, an Andor Zyla 5.5 sCMOS camera, and an MCL NANO Z200‐N TI z‐stage. Large stitched images comprised of 64 individual images were taken using a 10x objective, and covering a large proportion of the entire well (48‐well plate), were used to quantify the number of live and dead cells. To avoid any bias against dead cells that may have sedimented on the bottom of the well, quantification was performed at the focal plane of the bottom of each well, where a maximum number of dead cells were found. Subsequently, 5 circular regions of interest (ROI) were defined, avoiding large clusters and air bubbles, which can lead to errors in the counting program (Figure S18, Supporting Information). The total number of stained cell nuclei (Hoechst) and the total number of dead cells (EthD‐1) were quantified using the General Analysis dialogue in NIS Elements software. First, each image was denoised and background noise was removed. Then the threshold and circularity of the “bright spot detection” tool were adjusted to optimize the number of identified nuclei and dead cells. Once defined, the same threshold and circularity parameters were used for all quantified images. The cell viability within each ROI was calculated using:

$$\%\mathrm{Viability}=100-\left(\frac{\#\mathrm{dead}}{\#\mathrm{total}}\right)\ast 100$$ (2)

The average ± standard deviation for each well was then calculated from the 5 ROIs.

Data Analysis and Statistics

Statistical analyses were performed using either Origin 2018 SR1 or GraphPad Prism 9.1. The exact statistical test was specified in figure legends and the relevant methods sections. Unless otherwise stated, error bars represent the mean ± SD (standard deviation).

Conflict of Interest

M.B. and L.M. are co‐inventors on a patent related to bioinks based on dynamic covalent alginates (PCT/EP2019/080507).

Supporting information

Supporting Information

Supporting Information

Morgan F. L. C., Fernández‐Pérez J., Moroni L., Baker M. B., Tuning Hydrogels by Mixing Dynamic Cross‐Linkers: Enabling Cell‐Instructive Hydrogels and Advanced Bioinks. Adv. Healthcare Mater. 2022, 11, 2101576. 10.1002/adhm.202101576

Data Availability Statement

The data that support the findings of this study are openly available in DataverseNL at http://doi.org/10.34894/HBKOLJ.