Reversible control of biomaterial properties for dynamically tuning cell behavior

Abstract

In the past decade, significant advances in chemistry and manufacturing have enabled the development of increasingly complex and controllable biomaterials. A key innovation is the design of dynamic biomaterials that allow for user-specified, reversible, temporal control over material properties. In this review, we provide an overview of recent advancements in reversible biomaterials, including control of stiffness, chemistry, ligand presentation, and topography. These systems have wide-ranging applications within biomedical engineering, including in vitro disease models and tissue-engineered scaffolds to guide multistep biological processes.

1 |. INTRODUCTION

Advances in materials science, chemistry, and manufacturing have enabled the development of complex biomaterials that can begin to mimic the dynamic nature of the extracellular matrix (ECM). Traditional research has focused on the use of static and homogenous biomaterials, where material properties are constant in both space and time.^[1,2] Research using these static materials has elucidated the relationship between material signaling cues (e.g., topography, stiffness, and bioactive ligands) and cell behavior.^[3–6] For example, several researchers have shown that mechanical properties play an important role in cell adhesion, spreading, and differentiation.^[7–9] Recent developments in manufacturing have created materials with improved spatial control and hierarchal organization via photopatterning, lithography, electrospinning, and three-dimensional bioprinting.^[10–12] Furthermore, significant advances have been made to enable the creation of stimuli-responsive and “smart” biomaterials that exhibit temporal control over material properties. Conventional research has focused on designing materials that exhibit controlled degradation^[13,14] or biomolecule release^[15,16] with time. Here, we review more recent achievements in the development of dynamic biomaterials that allow for user-specified, reversible, temporal control over material properties.

The biophysical and biochemical signals within the in vivo cellular microenvironment are finely regulated and change over time. In particular, the dynamic nature of these cues is known to influence development, wound healing, regeneration, homeostasis, and disease progression.^[17–19] For example, during the natural healing process, a complex molecular signaling cascade occurs whereby multiple soluble factors are expressed at specific times to sequentially encourage inflammation, cell migration, differentiation, and new tissue formation.^[20,21] Therefore, the development of new biomaterials capable of replicating the unique dynamic cues of the ECM has the potential to transform our ability to create in vitro disease models, as well as design tissue-engineered scaffolds to control multi-step biological processes (e.g., migration, differentiation) for functional tissue regeneration.[22–26]

There are a number of broad reviews on stimuli-responsive biomaterials and their use within biomedical engineering.^[27–29] Additionally, a number of excellent and comprehensive reviews have covered recent advancements in spatially controlling biomaterial properties.^[30,31] In this review, we primarily focus on recent advancements in biomaterials with user-specified and reversible temporal control over material properties for in vitro cell culture models. We limited our review to material systems that exhibit completely reversible temporal control under standard cell culture conditions, via user-controlled triggers. First, we review advancements in reversible temporal control of physical cues within biomaterials, including both stiffness and topography. Modulating stiffness in particular typically involves changing crosslink density via the number of crosslinks or the distance between crosslinks (Figure 1). Next, we review recent developments in reversible temporal control of (bio)chemical cues within biomaterials, including peptides, full-length proteins, and other biomolecules. In this case, temporal control of chemical cues involves techniques that reversibly tether biomolecules to the scaffold to turn the biochemical signal “on” and “off” (Figure 1). Reversible control of these signals, either physical or chemical in nature, enables temporal control over cell behavior between the “nonactivated” and the “activated” state, where the precise type of activation is dependent on the changing signal (Figure 1). We review the use of both covalent and noncovalent (physical) mechanisms for changing these biomaterial properties. Table 1 provides an overview of common covalent and noncovalent chemistries used to enable reversible control over these material properties. Finally, we conclude by providing a perspective on potential future directions and opportunities for dynamic biomaterials, with an emphasis on the new developments in chemistry.

FIGURE 1.

(a) A schematic demonstrating reversible changes in physical properties and biochemical signals via covalent versus noncovalent mechanisms. (b) Reversible switching between the soft or signal “off” material state to a stiffer or signal “on” material state results in reversible changes in cell behavior from a nonactivated state to an activated state

TABLE 1.

Overview of commonly used reactions to reversibly control physical properties or biomolecule presentation in biomaterial scaffolds

2 |. REVERSIBILITY OF PHYSICAL PROPERTIES

Given the demonstrated importance of physical properties— such as stiffness^[7] and the nanoscale/microscale topology of the ECM—to control cell behavior, some of the earliest examples of dynamic biomaterials probed temporal control over these aspects of a hydrogel. The majority of works to date in this field allowed for unidirectional switching, for example decreasing the stiffness of a poly(ethylene glycol) (PEG) hydrogel through ultraviolet (UV)-mediated cleavage of covalent crosslinks,^[32] or increasing the stiffness of an alginate hydrogel via calcium release from gold nanorod-containing liposomes using IR light.^[33] Recently, however, there has been an increase in research into reversibly changing physical properties, either through a single mechanism (e.g., DNA hybridization), or by using two different chemistries for cleaving and reforming bonds. Below we discuss a number of approaches to achieve this and divide the discussion between covalent approaches (in which bonds are broken or formed), and noncovalent methods (which rely on reversible supramolecular interactions). Covalent approaches are advantageous due to the stability of covalent bonds, resulting in higher and more stable mechanical properties compared to noncovalent approaches. This stability, however, also makes it challenging to engineer complete reversibility into the biomaterial. On the other hand, noncovalent bonds are naturally dynamic and are more easily incorporated into biomaterial systems for reversible control over material properties. The dynamic and weak nature of noncovalent bonds can also be a limiting factor, resulting in lower mechanical properties and long-term instability (e.g., mechanical creep). The advantages and disadvantages of specific covalent and noncovalent approaches are discussed in more detail in the following sections.

3 |. COVALENT BONDS

One way to reversibly change mechanical properties is via isomerism of a covalent bond, for example using light. Rosales et al. developed a PEG-based hydrogel system that could reversibly stiffen and soften with light via photoisomerization of an azobenzene-containing crosslinker (Figure 2a).^[34] They demonstrated shear storage modulus (G′) reversibility by cycling between UV and visible light to isomerize an azobenzene double bond from cis to trans (Figure 2b). This process either lengthened (trans) or shortened (cis) the bond length, changing the stiffness of the hydrogel by tuning the distance between polymer chains. However, the difference between the ends of the two azobenzene isomers is small, so this system resulted in a modest change in G′ from 5,700 Pa (cis) to 5,580 Pa (trans). Another limitation to using azobenzene units is their thermal relaxation as the cis conformation reverts to a more stable trans conformation, resulting in softening over time. At room temperature, the isomerization took almost a week for full relaxation, while at 37°C it took approximately 18 hr, which is a limitation if the stiff form of the hydrogel must be maintained for an extended cell culture time. Although cells remained alive under the illumination conditions, no cell morphology evaluation was performed while cycling between the two conformations.

FIGURE 2.

Reversible changes in mechanical properties via covalent bonds. (a) A schematic showing how ultraviolet (UV) or visible light drive conformational changes in azobenzene units to alter hydrogel stiffness. (b) Cyclic modulus changes by alternating between UV and visible light at a high azobenzene-containing crosslinker concentration. (c) Hyaluronic acid (HA) polymers modified with the RGDS peptide, o-nitrobenzyl (ON; blue), and methacrylate (red) groups. (d) Cells on HA modified hydrogels stained for f-actin (red) and nuclei (blue) show a decrease in cell area with a decrease in stiffness while the cell area increases to that of the initial value upon an increase in stiffness. (a,b) Reproduced with permission from ref. [34]. Copyright 2015, American Chemical Society. (c,d) Reproduced with permission ref. [36]. Copyright 2017, Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Accardo and Kalow developed a system that relied on boronic ester crosslinking, where the equilibrium between a boronic acid and diol to form a boronic ester is altered by controlling the configuration of an adjacent azobenzene group using light.^[35] The authors synthesized diol- and azobenzene-terminated four-armed PEG polymers and tested for reversible gelation. The diol polymer (P1) was first tested with an o-boronic acid-azobenzene polymer (P2). The materials reached a plateau at 220 Pa for G′ after 5 hr at 365 nm. Multiple cycles at 470 and 365 nm exposure resulted in reversible solution-gelation transitions, respectively. At 365 nm, there was a slow increase (2 hr) in G′ to reach 90 Pa before a rapid decrease (2 min) in G′ to ~7 Pa when exposed to 470 nm. To optimize their system, they altered P2 by incorporating two fluorine atoms creating an o-difluoroazobenzene-boronic acid polymer (P2-F2). This alteration resulted in a faster gelation time (45 min) when exposed to 525 nm with a larger G′ (~1,900 Pa) while still remaining reversible. However, one caveat of the P1/P2-F2 hydrogels is that they fully dissolved after 6 hr at 25°C when testing swelling capabilities.

Rosales et al. later utilized a different platform to examine larger changes in reversible stiffness.^[36] They modified hyaluronic acid (HA) with three moieties: o-nitrobenzyl (ONB) acrylates for photodegradation, methacrylates for photoaddition, and an arginine–glycine–aspartic acid (RGD) peptide sequence for cell adhesion (Figure 2c). When the hydrogel was crosslinked via the ONB moieties, the material had an elastic modulus of 14.8 kPa. The hydrogels were then exposed to UV light in order to degrade the ONB groups, resulting in a decrease in modulus to 3.5 kPa. To restiffen the hydrogel, a photoinitiator was added and the material exposed to visible light, increasing modulus to 27 kPa. Compared to the earlier azobenzene model, the stiffness could be maintained until a user-defined point; however, this process only allows for a single cycle. Finally, the authors tested whether they could deliver dynamic mechanical cues to adhered human mesenchymal stem cells (hMSCs) in vitro. They seeded hMSCs onto hydrogels with the initial moderate stiffness (14.8 kPa) for 1 day. Then, they softened the system and let cells culture for 2 days before restiffening the hydrogel and letting them culture for another 2 days. Since transcriptional co-activator yes-associated protein/transcriptional coactivator with PDZbinding motif (YAP/TAZ) is a key player in controlling cell lineage commitment of MSCs, the researchers stained for YAP/TAZ as well as f-actin at Days 1, 3, and 5 (Figure 2d). Both cell area and the YAP/TAZ nucleus/cytoplasm (nuc/cyt) ratio decreased when softening the hydrogel, and increased when re-stiffening the hydrogel, confirming a reversible change in cellular mechanosensing.

4 |. NONCOVALENT BONDS

Due to their inherent dynamic nature and role in biochemical processes,^[37] noncovalent bonds have a number of advantages in creating dynamic biomaterials compared with covalent linkages. Various groups have utilized ionic interactions, hydrogen bonding, or external stimuli to create platforms with mechanical reversibility.^[38,39] For example, several researchers have studied the effects of calcium concentration to influence mechanical reversibility using the calmodulin (CaM) protein. CaM undergoes a conformational change from “closed” to “open” when four calcium ions are bound, exposing hydrophobic sites which can bind to different calmodulin binding domains (CBDs) of other proteins.^[40] Topp et al. developed a genetic toolbox of natural and engineered protein modules to form reversible networks when in the presence of calcium.^[41] Engineered proteins provide precise control over the composition of the material as well as creating a larger diversity of CBDs, resulting in variations in network topology, junction strength, and calcium sensitivity.^[42] To form calcium-sensitive junctions, CaM was used with two types of enzymes (human endothelial nitric oxide synthase [eNOS, monomer]) or petunia glutamate decarboxylase [PGD, dimer]). A tetrameric leucine zipper domain and a hydrophilic protein sequence were used for additional branching or to link components together. Three main triblock proteins were studied: CaM-(n)-Zip, PGD-(n)-PGD, and PGD-(n)-eNOS where n = hydrophilic protein (Figure 3a). The authors first probed an equimolar mixture of PGD-(n)-PGD and CaM-(8)-Zip and tested the viscosity with 0 and 30 mM of calcium, as well as after calcium removal with ethylenediaminetetraacetic acid (EDTA). They observed a 5,000-fold increase in viscosity when calcium was added, and EDTA addition reduced the viscosity back to the initial level (Figure 3b). Finally, the authors investigated stoichiometric mixtures of PGD-(n)-eNOS and CaM-(8)-Zip and tested them in the presence and absence of calcium with various crosslinker lengths. They found that the shorter crosslinker had a minimal increase in viscosity, but the longer crosslinker had a 1,000-fold increase in viscosity and could be fully reversed via EDTA addition over multiple cycles (Figure 3c).

FIGURE 3.

Reversible changes in mechanical properties via ionic interactions with the calmodulin protein. (a) Diagram illustrating the different modules used to create various triblock protein architectures. (b) Network formation in equimolar mixtures of petunia glutamate decarboxylase (PGD)-(n)-PGD and calmodulin (CaM)-(n)-Zip, showing viscosity changes in the presence and absence of Ca²⁺. (c) Viscosity changes in the presence and absence of Ca²⁺ for the network formation of PGD-(n)-endothelial nitric oxide synthase (eNOS) and CaM-(8) Zip. (d) Schematic of the CaM-hydrogel network in the unfolded (left) and folded (right) state. (e) Volume change of CaM hydrogels over multiple cycles. The dotted line denotes washes with ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid (EGTA) followed by incubation in Ca²⁺ for recovery of initial volume. (f) Light micrograph of a CaM-based/poly(ethylene glycol) (PEG) only hybrid hydrogel in Ca²⁺ buffer (left) and after TFP ligand binding (right). (a–c) Reproduced with permission from ref. [41]. Copyright 2006, American Chemical Society. (d–f) Reproduced with permission ref. [44]. Copyright 2007, Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Murphy et al. also used an engineered version of CaM to change protein conformation and to alter hydrogel porosity and swelling.^[43] They employed a four-armed PEG modified with CaM, and various concentrations of CaM were tested to determine the change in swelling volume with an increase in CaM concentration. To test reversibility, they used the antipsychotic drug trifluoperazine (TFP) to collapse the CaM conformation upon ligand binding, and the ethylene glycol-bis(β-aminoethyl ether)N,N,N′,N′-tetraacetic acid (EGTA) chelator to restore CaM to its extended conformation.

In a follow-up work, Sui et al. used the same engineered CaM protein with a different polymer structure to focus on spatial control of mechanical reversibility.^[44] The authors functionalized PEG diacrylate (PEGDA) with the modified CaM protein to form PEG–CaM–PEG with photoreactive acrylate end groups (Figure 3d). They controlled the amount of CaM included by mixing PEG–CaM–PEG with PEGDA and crosslinking it under UV radiation to tune the amount of CaM per hydrogel. The same TFP and EGTA plus a calcium buffer system was used for reversibility, giving a 65% decrease (TFP addition) and an increase (EGTA plus calcium buffer addition) in hydrogel volume over multiple cycles (Figure 3e). Due to the addition of a photochemical response, they were able to create heterogeneous materials with dynamic properties confined to specific locations. They demonstrated this by creating a hydrogel that was part CaM-based and part PEG-based. They showed that the CaM-based hydrogels decreased in volume when exposed to TFP, but the PEG-based portion remained constant (Figure 3f).

A year later, Sui et al. used this same process to understand how varying the molecular weight of the crosslinks and the crosslinking density changed hydrogel volume and optical transparency.^[45] They tested six different PEGDA molecular weights and four different polymer mass fractions, and probed changes in hydrogel volume after exposure to TFP (collapsed CaM conformation) or EGTA plus calcium buffer (extended CaM conformation). The authors also measured changes in hydrogel volume with varying PEGDA mass fraction and determined that, for low molecular weights, higher changes in volume were observed as the polymer mass fraction was decreased. Importantly, these properties could be changed over five full cycles, demonstrating that CaM-based systems are capable of reversibly controlling porosity, swelling, stiffness, and optical transparency. One concern with these systems is that calcium ions play an extensive role in many aspects of the cell cycle, and can effect changes in localization, association, and function. Repeated fluctuations in calcium concentrations have also been shown to cause cell damage, apoptosis, or necrosis.^[46,47] Another limitation is implementing this type of system in vivo, where it is difficult to adjust the calcium concentration and limit it to the wounded site.

It is also possible to tune the properties of a protein via the reduction/oxidation (redox) of disulfide bonds. Kong et al. engineered a protein-based hydrogel using an artificial elastomeric protein with a mutually exclusive protein (MEP) incorporated as a protein-folding switch.^[48] MEPs are specially designed with domain insertion proteins where there is a guest domain and a host domain. The size difference between these two domains allows for only one domain to fold at a time with the host domain being longer than the guest domain. The authors inserted cysteine residues forming a redox-responsive MEP switch. In the oxidized state, a disulfide bond is formed forcing the host domain to be folded and the guest domain unfolded; when the disulfide bond is reduced, this conformation is inverted. The authors performed stress–strain curves for the dynamic protein-hydrogel system, as well as for a control hydrogel_, in oxidized (phosphate buffered saline), reduced (10 mM dithiothreitol), and reoxidized (20 mM hydrogen peroxide) conditions. As expected, the control hydrogels displayed static mechanical properties; however, the protein-hydrogel system exhibited a decrease in elastic modulus (10 kPa) in the reduced state and then returned to the initial modulus (40 kPa) when reoxidized. Similar to calcium, redox reactions are vital in maintaining cellular homeostasis via generation and consumption of reactive oxygen and nitrogen species. An imbalance in homeostasis can lead to oxidative stress and cause irreversible oxidative damage to lipids, proteins, and DNA and has been known to play a critical role in a variety of pathologic conditions.^[49,50] This limits redoxbased systems to in vitro models and can further limit the type of in vitro cell work that is studied.

Liu et al. recently developed a method to engineer fusion protein-based crosslinkers for protein-polymer hydrogel formation.^[51] A dual-chemoenzymatic modification strategy was used to create two protein-based hydrogel crosslinkers: the first based on CaM, and the other on the photosensitive light, oxygen, and voltage-sensing domain 2 (LOV2) protein. The proteins were end functionalized with azides to react via strain-promoted azide–alkyne cycloaddition (SPAAC) between a four-armed PEG tetrabicyclononyne and a linear PEG diazide. When working with the CaM-based hydrogels, the authors assessed changes of mass swelling ratios in the presence and absence of calcium. In the presence of calcium, there was a 12.4% decrease in mass swelling and a 30% increase in G′ at 75 mol% CaM. The initial hydrogel mass and stiffness was restored when exposed to EGTA. The authors noted limitations such as relying on slow molecular diffusion to alter network mechanics, as well as not being able to spatially control protein expression within the network. To address these issues, they engineered the photoswitchable LOV2-Jα protein relying on the large conformational change when in the presence of blue light, where LOV2 initiates a photochemical reaction with a chromophore that results in the displacement of the Jα domain. To induce conformational changes, LOV2-modified hydrogels were exposed to blue light (470 nm), resulting in a 15% decrease in G′ for 75 mol% LOV2 samples. In the dark, the hydrogels fully recovered to the initial stiffness with a half-life of 35 s. The authors were able to cycle the hydrogels over 40 times in 5 hr with negligible material fatigue. Finally, the effects of cyclic loading on fibroblast differentiation into myofibroblasts were probed. Cycled light resulted in an increase in smooth muscle α-actin and periostin gene expression compared to constant light or dark, indicating activated myofibroblast differentiation.

Due to its highly programmable nature, other groups have investigated reversible DNA hybridization to alter mechanical properties. Lin et al. incorporated single-stranded DNA (ssDNA) oligomers into a polyacrylamide hydrogel network to create a platform with controllable bulk mechanical properties.^[52] This system employed three DNA strands for initial hydrogel formation: two linked to acrylamide (SA1 and SA2) and the third as a bifunctional linker strand (L3) complementary to SA1 and SA2. The central component of L3 was left single stranded; adding a complementary strand (F1) to form a duplex increased stiffness. The F1 strand contained a toehold region to allow for its removal with a fully complementary displacement strand (CF1), giving a return to the initial stiffness (Figure 6a). The elastic modulus was initially 217 Pa, the stiffened state was 640 Pa (after double strand addition), and the softened state 197 Pa after displacement strand addition. This system was later used to study changes in cell morphology in vitro. Jiang et al. started with a 50% DNA crosslinked hydrogel with a mechanical stiffness of 5.85 kPa.^[39] Additional crosslinker was incorporated to yield final crosslinking densities of 80 and 100%, increasing stiffness to 12.67 and 22.88 kPa, respectively. L929 fibroblasts were cultured on these surfaces and the authors observed a decrease in cell projection area and an increase in aspect ratio with increasing stiffness.

FIGURE 6.

Physical property and biochemical signaling changes using DNA. (a) Diagram depicting how addition and removal of DNA strands can change mechanical properties. (b) Illustration of crosslinked peptide amphiphile fibers via DNA hybridization. Confocal images show this system being used to control astrocyte reactivity on individual fibers (left, naïve), bundled fibers (center, reactive), and switching from bundles back to individual fibers (right, switch to naïve) using a displacement strand. Cells were stained for GFAP (green) and cell nuclei (blue). (c) Schematic of a toehold-mediated strand displacement system to turn RGDS peptide presentation “ON” or “OFF.” Confocal images show cycling of cell spreading in the “ON” state and cell retraction in the “OFF” state (actin: green, vinculin: red, nuclei: blue). (d) Schematic of the toehold-mediated strand displacement system used to independently control the presentation of two peptides (IKVAV and FGF-2). Immunofluorescent images show how dynamic and orthogonal presentation of IKVAV and FGF-2 modify GFAP and nestin expression in neurospheres. (a) Reproduced with permission ref. [52]. Copyright 2005, Materials Research Society. (b) Reproduced with permission ref. [53]. Copyright 2018, American Association for the Advancement of Science. (c,d) Reproduced with permission ref. [83]. Copyright 2017, Springer Nature

A very recent study conducted by Freeman et al. created hydrogels using DNA-modified self-assembling peptide amphiphiles (PAs) that reversibly form three-dimensional superstructures (Figure 6b).^[53] The authors modified PA monomers with DNA handles and co-assembled them with unmodified monomers to create nanofibers bearing pendant DNA handles. Mixing two fibers with complementary handles resulted in crosslinked hydrogels and bundled fibers with increased G′, driven by the DNA hybridization. The authors used this system to probe the effect of stiffness and nanostructure bundling on astrocyte (nonneuronal cells of the central nervous system) differentiation. Astrocytes were first seeded on individual fibers, a complementary ssDNA strand was added to form bundled fibers, and then a displacement strand of ssDNA was added to revert the bundled fibers back to individual fibers. This resulted in an increase in vimentin and glial fibrillary acidic protein (upregulated in the production of astrocytes), and phosphohistone 3 (cell proliferation marker) when the fibers were bundled; these values decreased when the bundle was unraveled. The hierarchical control over fiber structure and stiffness is highly reminiscent of natural ECM proteins like collagen, where this work demonstrated the first reversible control over such hierarchical structure in a biomaterial platform.

Rosales et al. adapted the azobenzene platform mentioned previously to a HA hydrogel with azobenzene-containing guest–host crosslinks, where the azobenzene molecule was the guest and cyclodextrin was the host.^[54] The isomeric state of the azobenzene group controls binding affinity to cyclodextrin, where a decrease in binding affinity indicates fewer guest–host linkages (softer) and an increase in binding affinity has more linkages (stiffer), respectively. The authors cycled from UV to visible light changing G′ by 36% from 1,000 to 640 Pa, which is a larger change in the storage modulus than the previous platform mentioned in Section 3.

Zou and Webber also used guest–host interactions to create hydrogels that have the ability to switch between physical and chemical crosslinks within the hydrogel network.^[55] They synthesized a four-armed PEG with trans-Brooker’s merocyanine (BM) as the end-terminated groups to serve as the guest while the host was cucurbit^[8] uril (CB^[8]). CB^[8] is known for its simultaneous inclusion of two guests, forming ternary complexes, and can reversibly catalyze a [2 + 2] photodimerization of BM. The authors used this platform to explore light-mediated reversibility of the photodimer. When PEG-BM and CB^[8] were suspended in water, they formed a hydrogel that could be molded and manipulated while remaining self-supporting (State I). After 4 hr of irradiation at 365 nm (dimerization), the hydrogel formed a disk with elastic, solid-like characteristics (State II). The hydrogel was more malleable than State II, but maintained its disk shape when exposed to 254 nm (reverse dimerization) irradiation for 12 hr (State III). When reexposed to 365 nm (dimerization), the hydrogel restored State II properties (State IV). Oscillatory rheology revealed that hydrogels with supramolecular crosslinks (State I) relax rapidly while photodimerized chemical crosslinked hydrogels (States II and IV) did not relax over the measured timeframe. State III presented approximately two orders of magnitude slower relaxation than State I, indicating that reversal of the photodimerized BM is only partially complete. Another limitation is the irradiation time requirements are long compared to other light-mediated approaches; nevertheless, this external trigger allows control over properties like swelling or self-healing by altering the function of both physical and chemical crosslinks.

Other methods have used external stimuli to alter material mechanical properties, including stresses, the application of a magnetic field, and pH.^[56–68] In particular, a number of reversible, pH-based biomaterial systems exist; however, it is a challenge to develop systems that demonstrate reversibility within a biocompatible pH range. Depending on the cell type, high and/or low pH may result in toxicity or alter cell behavior, where it is well known that pH influences cellular proliferation, protein, and DNA synthesis as well as a multitude of other functions.^[69] As a result, Yoshikawa et al.’s development of a triblock copolymer system with reversible stiffness over a relatively small, and physiologically compatible, pH range (pH 7–8) is of the utmost importance. The ABA triblock copolymer was designed where A is the pH sensitive component poly(2-[diisopropylamino]ethyl methacrylate) (PDPA) and B is the biocompatible zwitterionic component poly(2-[methacryloyloxy]ethyl phosphorylcholine).^[56] Changing the pH alters the protonation state of the nitrogen groups on PDPA, making the environment less or more hydrophobic, respectively. This cytocompatible, reversible design yielded changes in elastic modulus ranging from 1.4 to 40 kPa with changes in pH from 7 to 8, respectively. To show reversibility in vitro, the authors cultured mouse myoblast C2C12 cells onto stiff ABA hydrogel surfaces (initial pH 8) followed by softening (change to pH 7) and restiffening (change to pH 8). They reported initial cell spreading on stiff surfaces, whereas cells retracted into a spherical shape within 10 min of exchanging with pH 7 media to induce softening. Similarly, the cells regained a spread morphology within 10 min following reexchange of the media to pH 8 to induce stiffening.

Abdeen et al. used a magnetic field to alter hydrogel stiffness by encapsulating carbonyl iron particles within polyacrylamide hydrogels.^[25] Cycling the magnetic field between 0 and 0.75 T resulted in a G′ range between 0.1– 0.14 kPa and 60–90 kPa, respectively. MSCs were cultured onto static samples with (stiff, +B) or without (soft, −B) a magnetic field and compared to dynamic samples (+B−B) cultured for 4 hr on a substrate with the magnetic field on (stiff) before the field was switched off (soft). Cell area was probed as a functional measure of stiffness, with the soft substrates having an average cell area of 750 μm² while the stiff substrates gave an average area of 1,300 μm². The stiff-to-soft substrates reverted to an average cell area of 900 μm². Finally, the authors studied how changes in stiffness impact cell attachment and spreading, as well as early and late state osteogenic differentiation via the Runx2 transcription factor. An upregulation of Runx2 was observed when MSCs were initially seeded on stiff surfaces or when the substrates were stiffened at early time points, but not when the substrates were stiffened at intermediate time points. From this result, the researchers concluded that initial stiffness is important for promoting differentiation.

Cell-topography interactions play an important role in many biological processes and research over the past decade has utilized topographical changes to vary cell morphology and direct stem cell fate.^[57–65] Similarly to the previously described magnetoactive system, Kiang et al. used a magnetic field to manipulate surface topography by embedding nickel microwires within the top surface of a polyacrylamide hydrogel.^[66] The size of the microwires was optimized to 20 μm in length and 5 μm in diameter to provide maximal torque to alter the hydrogel surface. When a 0.31 T magnetic field was applied, the surface roughness increased compared to the control (no wires) and was maintained for 24 hr. Vascular smooth muscle cells were cultured onto rough (with magnetic field) and smooth (no magnetic field) substrate surfaces, where the cell area was smaller on the rougher surfaces compared to the smoother control surfaces. After 1 hr in culture, the spindle factor and cell area decreased for the microwire samples compared to the control (no wires). However, after 24 hr, the cell area and spindle factor were the same for the microwire and no microwire samples implying that the cells remodel themselves to compensate for the topography. This effect was also observed when the magnetic field was cycled between “on” and “off” for microwire samples.

Others have used stresses to manipulate topographical changes via stretching.^[67,68] Guvendiren and Burdick used poly(dimethylsiloxane) sheets and applied uniaxial or biaxial stretching prior to exposing them to UV ozone to create strain-responsive wrinkling patterns.^[67] Prestretched substrate surfaces (flat) followed by releasing to form wrinkled substrates showed changes in cell morphology where cells were spread on pre-stretched samples and then aligned over topographical patterns in a wrinkled state (Figure 4a–c). At culture times longer than 2 days, limited cell response was observed regardless of topography due to the formation of cell sheets. To mitigate this issue, the researchers blocked hMSC proliferation by treating with mitomycin C to inhibit DNA synthesis and were able to see changes in nuclear alignment up to 8 days in culture.

FIGURE 4.

Topographical changes lead to altered cell morphologies. (a) Schematic of substrate conditions through in situ dynamic pattern switching. (b,c) Corresponding bright field (b) and fluorescent images (c) (f-actin: green, nuclei: blue) of poly(dimethylsiloxane) (PDMS) substrates from prestretched to release in y-axis to stretched in y-axis and released in x-axis to fully released (left to right). (a–c) Reproduced with permission ref. [67]. Copyright 2013, Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

5 |. REVERSIBILITY OF BIOMOLECULE PRESENTATION

Aside from the physical properties discussed above, another important determinant of cell bioactivity in hydrogels is the presentation of bioactive molecules, such as growth factors. In natural regenerative processes (e.g., after injury) the ECM dynamically presents a range of various biomolecules at different time scales.^[3–6] Thus, there has been a surge of interest in recent years in dynamically controlling the presentation of bioactive signals with spatiotemporal precision. Reversibility is especially important, as is orthogonality, since cells depend on a large number of different signals for proper behavior. In this section, we describe the reversible presentation of biomolecules and, again, divide our discussion between covalent and noncovalent approaches. We note that there are only a handful of examples for covalent reversible chemistries, highlighting the utility of supramolecular chemistry for this approach.

6 |. COVALENT BONDS

Azobenzene units have also been used to reversibly tune biomolecule presentation in addition to stiffness.^[70,71] Auernheimer et al. developed photoswitchable controls using various modified RGD peptides with an azobenzene unit inserted.^[71] The RGD peptides were immobilized onto poly(methyl methacrylate) surfaces and exposed to UV or visible light to test isomerization between the E and Z forms. After 5 min of UV exposure, there was a maximum change in the E-to-Z conformation while thermal relaxation back to the initial E state took 3 days. The authors also performed a cell adhesion test resulting in an increase in cell adhesion when in the Z conformation and a decrease when in the E conformation.

Redox reactions have also been used to control peptide bioactivity. Lamb and Yousaf designed a redox-switchable surface that cycles between a cyclic and linear RGD structure (with cells binding more tightly to the former) using oxyamine-terminated ligands on hydroquinone-terminated self-assembled monolayers.^[72] Cells were cultured onto the surface and showed a 25% increase in cell diameter and a 35% slower rate of migration for the cyclic vs. the linear RGD. The authors then created an elongated RGD peptide with additional “blocking” residues, in order to create an adhesive or nonadhesive surface when in the linear or cyclic conformation, respectively. They hypothesized that the linear RGD structure would provide easy access for cell attachment, whereas cyclizing the peptide allowed the elongated linker to block access to the RGD group and render it less accessible. At 0.1% RGD surface coverage, cells adhered to the linear conformation and few cells adhered to the cyclized conformation; however, cells were able to bind to the cyclized conformation at higher RGD surface densities.

Laser photolithography has been widely used to spatially control biomolecule presentation.^[73–76] Furthermore, the Anseth group has made significant contributions to this field through the use of photoclick chemistries to tune biomolecule presentation in both space and time. Initially, DeForest and Anseth tuned cell function in real time by spatiotemporally controlling a synthetic microenvironment using orthogonal light-based chemistries.^[77] They used a four arm PEG tetracyclooctyne with a di-functionalized bis(azide) polypeptide to form a hydrogel via SPAAC coupling. The polypeptide was functionalized with two groups: a lysine(allyloxycarbonyl) amino acid, which contains a vinyl functionality to form a thiol–ene reaction for chemical patterning, and a photodegradable ONB moiety for photocleavage of the crosslinks (Figure 5a). The authors demonstrated orthogonal control by photocoupling a fluorescently labeled peptide in a three-dimensional pattern via multiphoton visible light illumination, and then selectively degraded the network using multiphoton UV light. Fluorescently labeled hMSCs were encapsulated and perpendicular lines of RGD and the synergistic proline–histidine–serine–arginine–asparagine peptide were photopatterned to have four distinct areas for cell adhesion. Subsequently, the patterned areas were photodegraded to remove RGD, and cells that originally adhered to these areas detached (Figure 5b). Gradients of RGD presentation could also be created using this method.^[78]

FIGURE 5.

Spatiotemporal control of biomolecules via covalent bonds. (a) Chemical structure of a macromolecular precursor modified with a photoreactive vinyl group and a photodegradable o-nitrobenzyl (ONB) group used to control (b) RGD (green) and proline–histidine– serine–arginine–asparagine (PHSRN; red) addition and removal. (c) Allyl sulfide mechanism used to simultaneously add and remove (d) ovalbumin (red), transferrin (green), and poly(ethylene glycol) (PEG)-SH (black) via multiphoton photolithography. (e–g) Mechanisms for (e) NPPOC-caged alkoxyamines to oxime linkages for photoaddition and (f) photocleavage of ONB moieties to (g) simultaneously control multiple full-length proteins (mCherry: red, mCerulean: blue, enhanced green fluorescent protein [EGFP]: green). (a,b) Reproduced with permission ref. [77]. Copyright 2011, Springer Nature. (c,d) Reproduced with permission ref. [80]. Copyright 2018, American Chemical Society (ACS), reader should contact ACS for further permissions (https://pubs.acs.org/doi/abs/10.1021/acscentsci.8b00325). (e–g) Reproduced with permission ref. [82]. Copyright 2019, Springer Nature

To improve temporal control, the Anseth group altered their platform by moving to an allyl sulfide crosslinker.^[79] This functional group undergoes addition–fragmentation chain transfer, simultaneously providing controlled addition and removal of biomolecules as well as regenerating the reactive functionality (Figure 5c). Multiphoton photolithography at 720 nm was used to spatially control RGD tethering or removal. The authors performed a reversible exchange of biochemical ligands using a two-photon system by uniformly patterning with one fluorophore followed by removal of a section by replacing it with another fluorophore, effectively enabling multiple cycles of ligand presentation. Finally, they cultured hMSCs onto a fluorescently labeled RGD surface to show cell attachment, then exchanged fluorescently labeled RGD with nonfluorescently labeled RGD, resulting in removal of the fluorescent tag while hMSCs remained attached to the surface.

In a subsequent work, this system was extended to control the presentation of full-length proteins, namely addition and removal of ovalbumin and transferrin.^[80] Ovalbumin was uniformly patterned, and then a 200 × 200 μm² was removed and replaced with nonfluorescent thiolated PEG. Next, a 133 × 133 μm² of transferrin replaced the thiolated PEG before a 44 × 44 μm² of transferrin was removed (Figure 5d). To see how cells responded to spatiotemporal protein cues, the surface was modified with transforming growth factor beta-1 (TGFβ−1). Mouse embryonic fibroblasts retrovirally transfected with a GFP-Smad3 fusion (to serve as an indicator of TGFβ−1 activity) were cultured onto the surface and the authors observed a 1.2-fold increase in nuclear GFP signal on patterned areas compared to the nonpatterned areas. When the TGFβ−1 signal was removed, there was a twofold decrease in GFP intensity while cell viability remained high.

DeForest and Tirrell utilized a photoaddition and photodegradation combination to control protein expression.^[81] Though still using a four-armed PEG with azide-functionalized synthetic peptides, they replaced the allyl sulfide with a photodeprotected oxime ligation to immobilize the proteins within a three-dimensional hydrogel matrix, while an ONB ester photoscission reaction facilitated protein removal (Figure 5e,f). The glycoprotein vitronectin (VTN) was photopatterned to spatiotemporally control hMSC differentiation to osteoblasts, and after 1 day in culture an increase in alkaline phosphatase (ALP) and osteocalcein (OC) production was observed. When VTN was removed on Day 4, OC and ALP activity returned to predifferentiated levels by Day 10. Also, the authors were able to control two proteins by simultaneously photoreleasing the first patterned protein then ligating the second protein using the heterobifunctional 2-(2-nitrophenyl)propyloxycarbonyl (NPPOC)photocaged alkoxyamine/azide group.

Recently, the DeForest group published work using this platform to reversibly pattern cell-laden hydrogels with sitespecifically modified proteins.^[82] Shadish et al. created 30 protein–peptide conjugates (six expression vectors: fluorescent [enhanced green fluorescent protein [EGFP], mCherry and mCerulean], enzymatic (β-lactamase [bla]), and growth factor (epidermal growth factor [EGF) and fibroblast growth factor [FGF]); and five sortaggable probes). By using the three fluorescent proteins, they were able to create a trifunctional protein pattern with spatial and temporal control (Figure 5g). Next, they tested localized control of enzymatic activity by photopatterning bla onto the hydrogel and using a thioacetate cefalotin and phenazine methosulfate solution to confirm local control over enzymatically driven precipitation. Finally, they tested control of growth factor presentation. To evaluate EGF activity, HeLa cells expressing EKAREV, a Förster resonance energy transfer reporter for mitogen-activated protein kinase (MAPK) signaling, were encapsulated within the hydrogel and then EGF was uniformly patterned onto the surface. They observed high MAPK activation when EGF is present and no upregulation of MAPK in the absence of EGF.

7 |. NONCOVALENT BONDS

As discussed in the noncovalent bonds section for physical properties, DNA is a highly programmable linker for crosslinking hydrogels. Its properties make it equally useful for tethering of peptides, proteins, and other biomolecules. Freeman et al. created a molecular system that uses peptide–DNA conjugates to add or remove two different bioactive signals on alginate-coated surfaces.^[83] The authors immobilized ssDNA handles on the surface before adding a complementary DNA strand modified with RGDS or RGES to create a bioactive or inactive surface, respectively (Figure 6c, schematic). Any unbound conjugate was rinsed away before seeding 3T3 mouse fibroblasts on the substrates. After 24 hr in culture, there were four times as many cells and a sixfold increase in cell area relative to unmodified alginate, RGES, ssDNA with no peptide, or double stranded DNA with no peptide. To confirm reversible control, a displacement strand removed the bioactive ligand. However, the tether strand remained on the surface, so the peptide could be added back in a second cycle. This type of system allowed for switching between an “ON” or “OFF” state, over multiple cycles, and the authors demonstrated concomitant changes in cell number and cell area (Figure 6c). The authors next probed orthogonal control of two different ligands by using DNA tethers with unique sequences. One complementary peptide–DNA conjugate had the signal IKVAV to mediate neuronal migration and differentiation, whereas the other displayed an FGF-2 mimetic peptide to promote proliferation. Over multiple cycles, the activity of neural stem cells was probed using GFAP, Nestin, and Tuj-1 staining, with either one or both peptides present (Figure 6d). Both migration and proliferation occurred when both peptides were present, but when IKVAV was removed the migrating cells retracted back into a neurosphere. In principle, this approach could be extended to three or more signals, given the vast number of orthogonal DNA sequences.

Boekhoven et al. used a different supramolecular interaction to control RGD peptide bioactivity.^[84] They utilized guest–host chemistry, modifying alginate surfaces with β-cyclodextrin, and using two different guest molecules for reversibility (naphthyl-RGDS, and adamantane-RGES); because adamantane is a better guest for cyclodextrin than naphthalene, the RGDS signal could be replaced by the RGES peptide, effectively switching bioactivity off. The authors tested the dynamics of their system by first binding naphthyl-RGDS to create a bioactive surface and culturing 3T3 mouse fibroblasts on the surface. Upon exposure to the adamantane-RGES, the RGDS signal was removed, and cell area decreased to match that of the control sample. Although this system was unable to control more than one ligand at a time and is only reversible for one cycle, extending it to multiple orthogonal hosts with different guests could add an element of orthogonality.

8 |. CONCLUSIONS

Taken together, the works described in this review demonstrate the great promise of reversible biomaterials for tuning the extracellular environment with excellent temporal control. The approaches taken vary—in chemistry, dynamic trigger, and the properties changed—providing a broad toolkit that can be adapted to the needs of the biological system being probed. However, compared to static biomaterials or irreversible dynamic materials, the field is still in its infancy. We foresee future research proceeding along several independent, but interrelated, tracks. First, most of the works described investigate one, specialized system rather than focusing on general design rules. As a result, the works described investigated one particular system (e.g., one hydrogel type, for a single, specific biological application). Creating a reversible dynamic platform that can be applied to a wide range of hydrogels and biological questions would greatly expand the adoption of this technology. Second, many of the approaches described require complex synthesis to create (photo)chemically responsive units. Expanding to a modular system that can be readily combined with existing biomaterial platforms would allow access to the nonexpert. In addition to this modularity, reducing the significant cost of these custom-tailored approaches would go a long way to expanding the access of these advanced materials. Third, it is still challenging to orthogonally control multiple signals in an orthogonal fashion, especially with simultaneous and independent control of physical properties. Finally, most of the systems described used hydrogels based on engineered materials (e.g., PEG, modified alginate, or HA). Imbuing natural materials that comprise the ECM—such as collagen, fibronectin, or laminin—with reversible properties, while maintaining the native architecture that enhances their function, is still a great outstanding challenge in the field. For example, can natural proteins like collagen be engineered so that the bundling and crosslinking can be controlled on-demand (e.g., by modifying them with protein, peptide, or DNA handles)? The large size, complexity, and intrinsic self-assembly properties of these proteins makes this a great challenge to the field, but one with potentially transformative effects on biomaterials science. Alternatively, designed molecules can be engineered to more effectively mimic natural proteins, like the hierarchical bundling of PA nanofibers with DNA^[53] to mimic collagen assembly.

However, based on the impressive work carried out to date by engineers, biologists, chemists, and physicists working together, we have no doubt that researchers will rise to meet this challenge, and create materials with the complexity of the natural environment, but with user-defined control over complex properties.

AUTHOR BIOGRAPHIES

Fallon M. Fumasi received her BS degree in Chemical Engineering from Oregon State University in 2014 and her MS degree in Chemistry from the University of Oregon in 2015. She is pursuing her PhD degree in Chemical Engineering at Arizona State University with Dr J.L.H. Her research focuses on developing a platform to spatiotemporally control biomolecule presentation to study cell signaling for orthopedic applications. Her research interests include materials design and characterization.

Nicholas Stephanopoulos is an assistant professor in the School of Molecular Sciences and the Biodesign Institute at Arizona State University. He obtained his PhD degree in Chemistry from the University of California, Berkeley in 2010. He then pursued postdoctoral studies at Northwestern University. The Stephanopoulos lab creates self-assembling nanomaterials from peptides, proteins, and DNA, including biomaterials with reversible properties for regenerative medicine.

Julianne L. Holloway is an assistant professor in Chemical Engineering at Arizona State University. She obtained her PhD degree in Chemical Engineering from Drexel University in 2012. She worked in the Department of Bioengineering at the University of Pennsylvania as a postdoctoral scholar. The Holloway Research Group studies the design of biomaterials that mimic the spatiotemporal complexity of the extracellular matrix for tissue engineering and regenerative medicine applications.